lakomski, gabriele (1991) beyond paradigms; coherentism and holism in research
TRANSCRIPT
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 1/97
INTRODUCTION
GABRIELE LAKOMSKI
The question of how best to research problems in the field of education continues to
occupy a central place in the work of educational writers. When the use of qualitative
versus quantitative methods in educational research is discussed, the debate commonly
turns on the issue of whether methods for studies in education should be borrowed from
the physical or the interpretive social sciences. In the first case, researchers argue that
empirical educational research requires quantitative-statistical methods to guarantee the
objectivity, i.e., scientific nature of results. On the other hand, proponents of qualitative
methods point out that exclusive attention to observable, social-educational phenomena
is too restrictive since it does not capture such inner phenomena as people’s intentions,
reasons, and their tacit knowledge which provide the “real” basis for social action.Hence, qualitative researchers, while not necessarily rejecting the view that educational
research is also explanatory and predictive, emphasize that their goal is an alternative
one which seeks to make explicit hidden meanings. From their perspective, educational
phenomena, studied in the scientific, i.e., “positivist” mode, are at best distorted and at
worst trivialized since educational issues with their “inner” dimension have merely been
described in their “surface” features.
While the qualitative-quantitative debate overtly emphasizes research methods, it
also presumes the existence of two different paradigms which are believed to be
underwritten by different epistemologies. The scientific paradigm is said to be that of
positivism, an umbrella term for a number of foundational theories of knowledge. Itsalternative, the interpretive, humanistic, or hermeneutic paradigm is premised on the
notion of verst ehen. Although there appears to be general agreement on. the existence
and validity of both paradigms, opinions differ on how to appraise such co-existence.
Either they are complementary, on the grounds that it takes more than one methodololgy- with its attendant epistemology - to answer relevant questions in education, or they
are oppositional. In the first view, increased attention to the practical procedures of
both forms of research is urged and fundamental, paradigmatic, and epistemological
differences tend to be de-emphasized. But according to the second view, these calls for
the blending of methods are countered by others who see such moves as obfuscating the
basic incompatibility of both paradigms. In this case, the possibility of reconciliation orsynthesis is denied, and the superiority of one or the other paradigm is advocated: they
remain oppositional.
Recently, there has emerged in the literature a third alternative on how to appraise
the co-existence of these two paradigms, the unity thesis, which denies epistemological
501
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 2/97
502 G. LAKOMSKI
diversity and argues that the paradigms view sanctioning it is false. Assuming a
non-foundational, coherentist theory of knowledge, the unity thesis argues that since
foundational theories of knowledge outrun their own resources to justify what they claim
are secure foundations, the epistemological foundations presumed by the quantitativeand the qualitative paradigms are consequently also unjustified. Furthermore, the very
conception of Kuhnian paradigms which underwrites the distinction is shown to be
incoherent, a result widely accepted in philosophy of science. The unity thesis rejects
the view that knowledge can be partitioned, believing it instead to be all of a piece. It
consequently disagrees with the paradigms view that different research methods can be
grouped under incommensurable paradigms.
The most important and far-reaching consequence of the unity thesis is that it can
provide an answer to what is a fundamental problem for the other two views. If one
assumes epistemological diversity as is the case in both forms of the diversity thesis,
then it is impossible rationally to integrate, render coherent, or even compare the
findings produced in either research tradition. In other words, growth in knowledge
cannot take place in the field of education. The enormous advantage of the unity thesis
is that - subject to its view of knowledge as non-foundational - it can offer criteria
and standards of justification for judging the respective merits of both traditions. They
can in fact be brought into a productive relation with one another when comparing them
according to coherence criteria which are routinely employed when judging scientific
theories. Knowledge in education grows by working out how much is shared between
theories - the development of touchstone - and how much is not, and making the
differences as clear as possible so that the nature of disagreements can be determined,
and possibly ruled out. In addition, for the unity thesis to work, it is not required that
researchers give up their commitments to whatever paradigm they adhere to, although
some might come to do just that when the checking of theories is concluded. It is initially
sufficient for them to be willing to engage in defending their accounts by using such
commonly accepted constraints on good theorizing as coherence, explanatory power,
comprehensiveness, and so on.
It is suggested that the unity thesis is the best available explanation of the
quantitative-qualitative distinction in educational research. The solution it offers to
the debate is novel, epistemologically sound, and practically superior since - in
emphasizing coherence and holism - it promotes the growth of knowledge in the
field and allows for methodological diversity.
Since Walker and Evers first presented their unity thesis in its most systematic and
comprehensive form in Keeves’ Educational Research Methodology, Measurement and
Evaluation: An I nternational Handbook, its growing importance and acceptance is
increasingly documented in educational research. It was given prominence in a recent
symposium between Husen and Keeves which appeared in the Spring 1988 issue of
Interchange under the title of “A Symposium on Educational Research: Unity or
Complementarity?” In addition, Keeves (1988, p. xvii), as editor of the International
Handbook, argues, referring to the ideas presented in the Walker-Evers thesis, that
a guiding theme of the volume is “that there is a unity in the field that arisesboth from the epistemological bases of inquiry into educational problems . . as
well as a coherence that arises from recent changes in social theory.” Furthermore,
there are available systematic applications of non-foundational epistemic justification
and coherence criteria in a number of fields. To mention just some more recent
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 3/97
Beyond Paradigms 503
examples: Evers and Lakomski’s (1991) Knowing Educational Administration, is the
first full-blown coherentist account in educational administration; Thagard’s (1988)
Computational Phil osophy of Science develops a coherentist justification for theory
choice of scientific theories, and Goldman’s (1988) book Moral Knowledge, applies
coherentist criteria to ethics.
The chapters collected in this Special Issue are a further contribution to this important
theoretical development. They specifically explore applications and implications of the
unity thesis in discrete areas of educational research. This is not to say that there
is equal acceptance of the unity thesis by all contributors. But, given that they are
united in their opposition to the relativist stance which currently characterizes much of
educational research, there is general acceptance of the importance of scientific realism,
naturalism, and coherentist justification. Together, these philosophical strands serve to
eschew dualisms and thus further the creation of unity in the study and ultimate solution
of educational problems.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 4/97
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 5/97
COHERENCE
CHAPTER 1
AND REDUCTION: rMPLICATIONS FOR
EDUCATIONAL INQUIRY
JAMES C. WALKER
Faculty of Education, University of Canberra, Australia
Abstract
This chapter outlines the consequences for educational inquiry of the view that human
knowledge is a “seamless web”, and that one major path to progress lies in the search
for coherence between theories and traditions of inquiry. An exciting example comes from
recent work in philosophy and cognitive science proposing a theory of the mind-brainwhich promises to unite, through intertheoretic reduction, cultural, psychological and
neuroscientific understandings of teaching and learning. This supports some appropriate
medium to long term strategies for educational research, such as interdisciplinary
specialization and the co-development of educational theories.
Introduction
Educational theory and research, while having their own characteristic interests and
problems, are influenced by developments in other disciplines, especially philosophy,
psychology and the social sciences. Over recent decades the form of influence hasbecome increasingly epistemological, as debates about validity and reliability, different
research paradigms, the organization of knowledge in the curriculum and the nature of
teaching and learning have thrown up different views about what we can know, how
we can come to know it and why our claims to knowledge are credible or justified.
Historical studies notwithstanding, educational inquiry tends to focus on the present,
because of its applied nature and pressures from current professional practice and
public interest. These pressures can make it hard to take a long term view of where
theory and research should be headed. Yet choices made now, about current projects,
affect our future directions and medium and long term options. More importantly, the
conceptualization and organization of educational inquiry as a whole, the relations
between its internal specializations, ‘and the relations between them and the social and
natural sciences, as well as educational policy and practice, need to be considered from
more than a short term perspective.
In this chapter I wish to consider some general philosophical issues whose bearing on
educational inquiry (Evers, 1987b; Walker, 1985b; Walker & Evers, 1984), although in
some respects long term, is closely related to the choices of the present. The issues
505
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 6/97
506 J. C. WALKER
themselves are not new. My aim in raising them is to draw attention to a strand
of contemporary philosophy which as yet has not been widely considered among
educational inquirers, although some of its intellectual antecedents have been considered
and, in fact, widely rejected. It is the materialist or physi~alist quest for a unified sciencewithin which educational inquiry, along with the social sciences generally, would sit
coherently with the natural sciences. This is presented as a working hypothesis about
the conditions for progress in the growth of knowledge, not for any other, ideological,
reasons.
One aspect of this strand of philosophy has had some currency in educational inquiry:
the view that all knowledge is of a piece, a seamless web. This “coherent&” epistemology
is due largely to the work of Quine. The particular way this coherentism is developed
in recent philosophy is even more novel in the educational field. The epistemology is
specified through a reductionist view of human nature exemplified in recent restatements
of physicalism by philosophers such as Clifford Hooker, Stephen Stich, and Patricia and
Paul Churchland.
The physicalism development in the 1940s and 1950s by the logical empiricists was
strongly associated, not just with their aspirations for a unified science, but also with
their interest in the history and philosophy of science, especially physics (Nagel,
1961; Oppenheim & Putnam, 1958). Contemporary physicalism, focusing sharply on
theories of the mind-brain, is associated with the growth of cognitive science, notably
the overlapping work in cognitive neurobiology, cognitive psychology and artificial
intelligence.
The unificationist quest runs against several popular beliefs: that there is a fundamental
epistemic divide between the social and natural sciences; that on each side of this divide
there are different modes of explanation and understanding including distinct views of
causation; that human learning and behavior are not reducible to physical entities or
relations; and that a materialist view of human persons and culture is inconsistent
with the moral aspect of human life, rooted in freedom of choice and responsibility
for actions. The quest for a unified science of education, then, conflicts with deeply
held epistemological, ontological and ethical beliefs. The challenge is to demonstrate
in principle, and if possible in practice, that the dualisms underpinning these beliefs
are neither necessary nor indeed our best options for the future (Walker, 1985a).
The Coherence of Natural Knowledge
Traditionally, epistemologists sought to show our knowledge could be certain, reliable
or warrantable by locating it on secure foundations. For classicaf empiricism the
foundations were observations or “sense data”, for classical rationalism the intuitions
of pure reason. Modern foundationalists offer more sophisticated variations on these
themes. If by sound methodology (such as valid inference) we coutd show our beliefs
to be based on these foundations, our claims to knowledge would be justified.
Foundationalism has lost support for many reasons, not least the impossibility of
identifying foundations without some prior theoretical framework itself contentious in
that it is not an instance of the foundational category. This is clear, for example, in the
now almost uncontroversial doctrine of the theory-ladenness of observation.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 7/97
Beyond Paradigms 507
The recognition that there is no one simple way of interpreting experimental results
undermines foundationalism. Neither confirmation nor disconfirmation will do on its
own. Taking up Duhem’s (1954) argument that hypotheses cannot be tested one by one
in isolation from the whole theoretical networks in which they are embedded nor from
the theoretical frameworks of research methodology, Quine (1953, 1960) points out
that neither experiment, logic, nor method alone can tell us which hypotheses should
be saved, revised or rejected. According to the Quine-Duhem thesis, what is being
tested in research is always a global theory or whole theoretical network. We move
from foundationalism to holism. We recognize that competition between theories and
hypotheses cannot be resolved simply by appeals to authoritative empirical evidence or
logic, both of which embody fallible theoretical commitments.
A consequence of foundationalism’s failure and the move to epistemological holism
is the recognition that there is no “first philosophy”, no way prior to knowing itself of
demonstrating any warrant for our claims to knowledge, no Archimedean point outside
our body of knowledge from which we can judge it. Quine (1960, p. 3) invokes Neurath’s
metaphor: “Science is like a boat, which we rebuild plank by plank while staying afloat in
it. The philosopher and the scientist are in the same boat.” Quine’s (1969) response to
this holistic outcome is the naturalizing of epistemology. We study knowledge itself as
a natural phenomenon, and learn from the successes and failures of attempts to acquire
it. The epistemic quest is understood as a problem solving process conducted by natural
organisms in the natural world. (For the physicalist, this includes the social world.)
Epistemology becomes continuous with natural science, viewed as one evolving whole.
What we can know will depend on what kind of creature we are, especially upon our
cognitive capacities, and upon the natural contingencies determining what we can learn
through addressing the problems in ourselves and our environment. The natural world
is a real world; it is not constituted by our knowledge, although we and our knowledge
are part of it; our naturalism is epistemologically realist. Inquiry is a search for “facts of
the matter” (Quine, 1977), notwithstanding that facts are perceived through theoretical
lenses and represented through theoretical frameworks. Holism, naturalism, pragmatism
and realism become the crucial features of our account of knowledge.
Coherence is easier to espouse than achieve. Given the theory-ladenness of observation
and the absence of foundations, we are left with the prospect of choosing between
competing theories with no recourse to an authoritative arbiter to prescribe a choice.Facing this, some theorists have quickly succumbed to a debilitating relativism, in which
there is no prospect of rational choice. The naturalism of the scientific realist, however,
cautions against such a move. What is needed is an account of theory competition in
which the naturalizing of epistemology guides us in selecting our best option.
One epistemological tradition influential among social scientists takes us halfway
there. Popper’s “evolutionary epistemology” presents theory competition as a process
analogous to the selective elimination of biological evolution (Popper, 1972). Theories
compete by addressing shared problems, developing as “theory-series” (Lakatos, 1970)
by a process of trial and error. Just as biological organisms - including whole species
- survive or die according to their inbuilt capacities to solve the problems imposedupon them by their environment, so human theories, including science, are assessable in
terms of their problem solving power, and competing theories can be judged as relatively
more or less powerful. The most systematic attempt to apply such an evolutionary
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 8/97
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 9/97
Beyond Paradigms 509
carefully, for example, to the ways in which the brain most effectively conducts its
cognitive enterprise, and to the fundamental features of the natural world as understood
through the physical sciences. Whereas in practice we may assume such systematic
coherence to be established in certain cases (there are no serious doubts about the
coherence of chemistry with physics), in other cases, notably relating to theories of the
mind and of learning, there is room for reservation: hence the present argument for a
long term perspective.
Respect for the intellectual or practical problem solving power of a theory implies a
pragmatic coherence theory of evidence, but not necessarily of truth. On this score,
while we endorse the earlier pragmatist (e.g., Deweyan) emphasis on problem solving
and epistemological holism, we reject its instrumentalist theory of truth. A Quinean
pragmatic realism employs the coherence tests of the superempirical virtues to sort out
the merits of competing theories, and permits us to use the resources of the resulting
preferred theory to spell out its relation to the world, to specify what constitutes, in its
case, truth as a correspondence relation between the sentences of the theory and the
facts of the matter. A coherence theory of evidence is distinct from and compatible
with a correspondence theory of truth (Quine, 1970). Epistemological pragmatism and
coherentism are partners with scientific realism (Evers, 1987b).
All this implies some conceptual and methodological common ground between
theories competing in any given field of knowledge (e.g., educational psychology),
and between (at least prima facie) non-competing theories in different fields (e.g.,
psychology and neurobiology). Minimally, all theories are amenable to application of
the criteria of superempirical virtue; maximally, given the reductionism expounded in
the next section, there will be, when coherence obtains, the possibility in principle of
intertheoretic reduction in which one theory can be explained within the framework
of another. Whereas long term reducibility to physical theory is a desirable outcome
in educational inquiry, an important short and medium term goal (on which the
long term achievement will depend) is a program of intra and interfield theoretical
co-evolution (P.S. Churchland, 1986 pp. 361-365). This program will involve attempts
both at construction of explicit cross conceptualizations and at the devising of research
methods and designs which enable comparison and integration of findings. Running
through this program is the attempt to discover and construct what, borrowing a term
from Lakatos (1970), we have called “touchstone theory” (Walker, 198.5~; Walker &Evers, 1982, 1988).
Touchstone is an intertheoretic set of concepts and methods and the empirical evidence
they generate and incorporate, geared to processing information and solving problems in
the natural world. It embraces logical as well as experimental practices. It provides the
resources for theory comparison and competition and for intertheoretically generated
theory refinement. It is what produces algorithmic and heuristic coherence or unified
problem solving power. As such, unlike the epistemic items posited by foundational
epistemologies, it is not fixed; as theories under consideration develop and change, so
may touchstone. It is part of our theory of the world, our web of belief, and may be
corrected. Touchstone is relative to the overall development of our epistemic whole,and changes as the development of theories generates new overlaps or displaces old
ones. Whether explicit or implicit, touchstone varies synchronically (if and when theories
are being compared) and diachronically (through evolution of theories). The point to
stress is that since criteria for evaluation are themselves embedded within theories,
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 10/97
510 J. C. WALKER
intertheoretic evaluative comparison depends on criteria being shared between theories.
As shared theory, then, touchstone gives field specific and interfield general content to
the coherentist account of the superempirical virtues.
The notion of touchstone theory needs to be understood in relation to the natureof theory competition. Given that theories are understood as proposed solutions to
problems, competing theories are alternative projected solutions to the same problem
or problems. For Tl to be in competition with T2, the theories must be addressing at
least one common problem. Granted some variability in their respective formulations
of the problem(s), if there is genuine competition there will be some semantical and
methodological commonality, and therefore implicitly at least commonality of theoretical
structure.
The interrelation of touchstone problems and touchstone between competing theories
provides the basis for identifying and distinguishing between fields of study, or disciplines
(Walker, 1985b; Walker & Evers, 1982). The recognition of closeness of competition
(with the limiting case being no competition - i.e., an unchallenged theory or body of
theory), naturalistically determined by the closeness of the range of problems addressed,
leads to the social-epistemic constitution of a more or less coherent discipline. The
looser the constitution, and the weaker the actual competition, the harder it is to
draw boundaries, which change as theories and their touchstone develop. Moreover,
the unevenness of theoretical development means that changes in one field, or even a
cluster of disciplines, may not be absorbed by intellectual relatives for some time, or
not at all, the cultural lag inhibiting epistemic progress - a problem for applied fields
such as educational research.
Competition, however, is not always patent, nor recognized by proponents of
competing theories. A common metatheoretical issue, evident not least in debates
over the mind-brain, is whether a competition is under way. Theories previously
thought compatible may be discovered to be incompatible. This typically emerges
from the recognition that there is indeed touchstone which can be brought to bear
on the respective theories; that theories of the mind and theories of the brain, for
example, are addressing the same set of problems. The discovery may be prompted by
developments within the fields housing the theories, or from discoveries elsewhere in
our web of belief. The debate may well proceed in terms of the adequacy of respective
problem formulations. A constructive stance towards the epistemic enterprise will notrule out the desirability of discovering new touchstone and, with it, hitherto undetected
competition. This applies particularly to fields such as educational inquiry which are
under pressure to focus on the short term and are not well geared to keep an eye on
intellectual developments across our global body of knowledge.
In then home territories and with an eye on their near neighbors, however,
educational researchers, like other social scientists, are increasingly recognizing not just
the desirability of rapprochement between adherents of different research traditions (as
in the quantitative/qualitative debate) but the methodological need for more coherent and
if possible integrated first order methodology within and across traditions. Educational
researchers have adopted multimethod approaches, particularly in sociological studies,emphasizing triangulation across data, investigators, theories and methods (Denzin,
1978). Although less widespread, perhaps, there is a similar awareness in psychological
research, especially since Campbell and Fiske (1959) (who coined the term “triangula-
tion”) first advocated multitrait/multimethod analysis in educational testing and measure-
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 11/97
Beyond Paradigms 511
ment, moved by the epistemological observation that construct validity requires both
convergent and discriminant validity. In the broader sweep of things, it can be argued
that a qualitatively conceived and researched framework is a fundamental precondition
for quantitative research. In the spirit of Quine’s comment that science is selfconscious
commonsense Campbell, discussing action research and program evaluation, concludes
that “quantitative knowing depends on qualitative knowing in going beyond it.” Both
the quantitative and the qualitative traditions need “cross validating additions”.
More than that, I have sought to remind my quantitative colleagues that in the successful laboratorysciences, quantification both builds upon and is cross-validated by the scientist’s pervasive qualitativeknowledge. The conditions of mass-produced quantitative social science in program evaluation are suchthat much of this qualitative base is apt to be lost. If we are to be truly scientific, we must reestablishthis qualitative grounding of the quantitative in action research ]Campbell, 1988. p. 3761.
It is well, too, to remember that educational policy makers and practitioners constantly
put together information of various kinds from diverse sources and produced by various
methods within a variety of theoretical frameworks. Their focus on practical problems
gives them a working context in which to find or construct the touchstone they need.
However much they might individually or in groups favor one kind of research over
another, practitioners and policy makers do not readily assume incommensurability
between approaches. Similar attention by researchers to the problem to be solved
rather than preoccupation with the method or theory to be defended is necessary
to construct touchstone to achieve the flexibility and open mindedness required for
a unified approach to educational theory and research.
The epistemic and social specialization and attendant ~agmenta~on of educationalinquiry - which affects modes of publication, conferences, appointments, funding and
the training of researchers as well as the conduct of particular research projects - tends
to work against the development of touchstone and a co-evolutionary framework. The
need for interdisciplinary work stands out, not just collaboration between workers in
different disciplines but, as Campbell has pointed out, for “interdisciplinary specialists”
who can work in “interdisciplinary space”. Reflecting on his own efforts in this
mode, he points out the obvious - but academically scandalous - implication: the
interdisciplinary specialist will have to comment on many fields in which they are not
competent. Owning up to such incompetence (and the present essay is another example),
Campbell dissipates the scandal by recalling Quine’s observation that both learning newlanguages and conceptual innovation are processes of trial and social correction. The
brave interdisciplinary specialist needs sympathetic support.
. the process of exploring this interdisciplinary niche requires corrective critical responses fromthose whose areas overlap you have to be willing to keep up the conversation with those whospeak your language imperfectly, patiently correcting their misconceptions while still encouraging
their efforts. It is at this point where our collective process so often fails, where bold explorers
of interdisciplinary space get no response, critical or otherwise, from the disciplines they overlap
and end in paranoid isolation if they persist at all. Proud scholars who refuse to talk with thosewho do not speak their language competently are neglecting their duty to . collective omniscience[Campbell, 1988. p. 4391.
Notwithstanding heartening breakthroughs, epistemic isolationism not only flourishes,but is a social phenomenon which has had some insistent theoretical backing in
modern epistemology. Clearly, our touchstone coherentism is incompatible with such
antiholist epistemologies as the doctrines of incommensurability between theories
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 12/97
512 J. C. WALKER
(Kuhn, 1970) and between forms of knowledge (Hirst, 1966). Arguments against
these positions have been presented elsewhere (against incommensurability of theories
or “paradigms”: Walker, 1985b; Walker & Evers, 1988; against incommensurability
of forms of knowledge: Evers, 1987a; Evers & Walker, 1983). Incommensurabilistdoctrines proclaim relativisms which outrun the doctrines’ explanatory resources.
Moreover, they are self contradictory. To argue incommensurability presumes that
some account (chiefly semantic) of a given theory can be provided showing lack of
common ground with another theory. But this account must itself be given in terms of
some theory, if not the second theory, then some third theory articulating semantically
with the other two. In other words, to argue incommensurability on semantic grounds
(Feyerabend, 1975) requires some semantic touchstone. The same goes for logical and
methodological arguments. This does not mean, of course, that all research methods
can be integrated at the substantive or first order level. It does mean, to the extent
that inquiry is progressing, that they can be rationally compared at the superempirical
level, and integrated into a broader, coherent framework. The relativity of touchstone,
and of alternative theories, is not a radical atomized relativity; it is relativity within a
developing epistemic whole which is embedded in the natural world.
Touchstone is relative, but far from thereby innocuous. Running through and unifying
all touchstone are the coherence constraints imposed by our naturalism, a naturalism
which forces theories into a testing relation with a real world, with the facts of the
matter. Therefrom comes its bite. As natural beings in the natural world, with survival
needs and evolutionary and cultural constitutions, we are not free arbitrarily to dispose
of a given theory we are employing to order our behavior and solve our problems, unless
we have an alternative theory or give up on the job. The exigencies of real practice,
including epistemic practice (e.g., science or teaching) demand theoretic response, and
where there are alternatives, choice. To make the choice, we have no way of deciding
(short of random response which in itself makes certain theoretical assumptions) other
than by comparing the alternatives according to criteria with which they are logically
compatible or which they share by way of logical commitment. Coherentism links a
theory of epistemic progress with our knowledge of the natural, or physical, world.
Epistemic progress is a natural process of which, if our naturalistic epistemology is
sound, a physical account is possible.
Reductionism
As presented here, physicalism is foremost an epistemological theory flowing from
an acceptance of strong coherence requirements, chiefly that psychology, social science
and educational inquiry be shown to be at least logically consistent with and preferably
reducible to physical theory. Hence theories of mental states, if correct, will be reducible
to neuroscience. The judgment in favor of physical theory derives from historical and
pragmatic considerations. In short, physical theory is our most powerful body of theory,
with by far the most sophisticated account of any relation to the real world so farprovided by any body of theory. This general claim has been powerfully argued by
P.M. Churchland (1979).
As such, our physicalism is to be distinguished from another, antireductionist,
position prevalent in cognitive science - functionalism - which can be interpreted
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 13/97
Beyond Paradigms 513
physicalistically. For functionalism, psychological theories of mental states are emergent
from but not reducible to neuroscience, since cognition involves semantic and logical
representations, whereas the study of neural structures deals only in causal relations.
Psychological states are functions of the causal role they play in the brain. Just asthe functional states of a computer are realized in electrical circuits, so mental states
are realized in the structures of the brain. Their emergent functions, however, cannot
be explained neurally (Dennett, 1978; Fodor, 1975; Pylyshin, 1984). As against this
espousal of the autonomy of psychology, the coherentist maintains that reduction to
physical theory is possible, and that to adopt it as a working hypothesis is the more
constructive alternative.
On the coevolutiona~ pi@re of science, psychotogy and neuroscience should each be vulnerable todisconfirmation and revision at any level by the discoveries of the other. And when inquiries converge
on a subject matter, as for example they do on Iearning, memory, attention and perception, eachshould be open to the discoveries of the other. The isolation of psychology from the disconfirmatoryevidence of neuroscience would be a mistake, because in general it is such susceptibility that keepsa science honest. Short run isolation of a science while it works up a head of steam is one thing,but isolation in the long run, isolation in princip le, is quite another. . . . The unity of science isadvanced as a working hypothesis . . because theoretical coherence is the “principal criterion of
belief-worthiness for epistemic units of all sizes from sentences on up” [P.M. Churchland, 19801.Once a theory is exempt from having to cohere with the rest of science, its confirmation ledgeris suspect and its credibility plummets [P.S. Churchland, 1986, p. 3761.
Our contention, then, is that the unit of science implies the reducibility of psychology
and the social sciences to physical theory. But what is meant by “reduction to physical
theory”‘? As P.S. Chur~hland comments, the word “reduction” has “a bewildering variety
of uses, many of which have connotations of insult and abuse.” (P.S. Churchland, 1986,
p. 278). Setting these aside, the most important point to make is that reduction is a
relation between theories, not between entities. Above all, we should be clear that
the physicalist reductionist is not saying that culture is reduced to behavior, nor the
mind to the brain, but that theories of culture are reduced to theories of behavior, that
theories of the mind are reduced to theories of the brain. In other words, reduction is
an epistemological, not an ontological, relation - though it has ontological implications.
It concerns relations between different sets of claims about what we know and what we
can know. For example, the theory of optics might be claimed to reduce to the theory of
electromagnetic radiation. So far as the human mind and human culture are concerned,then, the question is whether some theory of mental states is reducible to a theory
of the working of neuronal ensembles, or whether a theory of symbolic relations and
representations is reducible to a set of dispositions to behavior.
Intertheoretic reduction leads to two kinds of outcome generally considered advances
in science; explanatory unification and ontological simplification. The simultaneous
achievement of unity and simplicity, both dimensions of coherence, is a clear achieve-
ment. Here reduction can shade into another epistemically progressive step: the
elimination of one theory by another. Consider the fate of the caloric theory of
heat, superseded by the kinetic theory. The latter theory does not identify caloric
with molecular kinetic energy; its implication is that there is no such thing as caloric.The same goes for the phlogiston theory of combustion and the demonic possession
theory of nervous disorders.
Reduction is the tightest instance of theoretic co-evolution, and often the unexpected
outcome of it. The immediate methodological injunction of the reductionist is to set
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 14/97
514 J. C. WALKER
up a co-evolutionary framework. Even when reduction does not eventuate, this is an
epistemically progressive step. Speaking generally and in disciplinary terms, it is clear
enough that the histories of physics and chemistry, astronomy and dynamics, the theory
of infectious disease and microbiology, to name but a few, demonstrate the benefits of
co-evolution, as “discoveries at one level often provoke further experiments and further
corrections at the other level, which in turn provoke questions, corrections and ideas for
new explorations” (P.S. Churchland, 1986, pp. 363-364).
The contemporary physicalist account of intertheoretic reduction differs from its
logical empiricist predecessor in denying that the phenomena identified by the reduced
theory must be thoroughly correlated with those in the reducing theory. Logical
empiricists used various devices, especially bridge principles connecting the two sets
of phenomena, to achieve this correlation. Hooker (1981) points out that the history
of science provides no clear examples of this happening in practice. Rather, what is
reduced is not the old theory, but a corrected version of it, with the limiting case
being elimination. Whereas virtually nothing of the caloric theory of heat survived,
and a certain amount of classical dynamics, most of the theory of optics lives on.
Reductions, in other words, may be more or less “smooth”, or more or less “bumpy”.
This unevenness of reduction is a further demonstration of the need for insisting that it
is not phenomena, but theory, which is reduced - a point of significance for the prospect
of reducing mental to neural states and processes. It is also a point implied by scientific
realism: the world does not change just because our theories do, notwithstanding the
effect people’s theories can have in their interaction with phenomena of all kinds.
Intertheoretic reduction has an explanatory function. The new, reducing theory, by
advancing on the old, will explain the successes and failures of the old, preserving its
successes within a richer framework of understanding. This is illustrated vividly by
Hooker (1981, p. 49) in his account of thermodynamics and statistical mechanics.
As we have seen, the process of reduction is not necessarily an all-at-once affair. It
occurs through the co-evolutionary development of theories, in which extensions and
corrections are made to more than one theory as inquiry proceeds. P.S. Churchland points
out that this is already happening in the theoretic interanimation of neurobiologists,
psychologists and neurologists in research on memory, attention and learning, with
the possibility remaining open of behavioral observations from psychology reducing
to neurobiological hypotheses, but meantime both contributing to the wider programof understanding how the human information processing, storage and retrieval system
works. The field has a strikingly interdisciplinary character. It “has in the last twenty-odd
years become a classical exhibit of productive research on a nervous system capacity
at many levels at once” (P.S. Churchland, 1986, p. 268). Hence, granted that it would
be premature for physicalists to claim imminent reductions across the board, it would be
equally presumptuous for a believer in the autonomy of psychology to reject theoretical
co-evolution and the possibility of a reductionist outcome.
it would be simply boneheaded for a cognitive psychologist working on learning and memory
to refuse to care about animal models, pathway research, clinical cases, imprinting in chicks, and
habituation in Aplysia. We simply don’t know remotely enough yet to know what is not relevant[P.S. Churchland, 1986, p. 3731.
And, of course, science is always unfinished, even in cases of reduction. What is
reduced is one theory of phenomena as currently understood to another theory of
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 15/97
Beyond Paradigms 515
phenomena as currently understood. Since our epistemology is of a piece with our
theory of the natural world, and not a form of first philosophy, it too is a candidate
for continuing evolution, in concert with the rest of our web of belief, as well as for
intertheoretic reduction.It is from this perspective that we should understand the evolution of human
knowledge from commonsense or “folk” theory to science. We have moved from a
folk to a scientific physics, from a folk to a scientific biology. At present, however,
much educational inquiry is conducted within the framework of folk psychology and
sociology, describing and explaining human behavior in terms of beliefs and desires,
hopes, disappointments and fears. Intentions are explained by reference to belief-desire
combinations, and beliefs by reference to perceptions and inferences. Countless everyday
platitudes - generalizations about human nature and behavior - hang together as a
theory of internal states and their interaction with the environment, especially other
people. The crucial question is whether this body of theory is reducible to a scientificaccount. The considerable problems standing in the way of such a reduction - not
least the intentionality of basic concepts such as “belief” and “desire” (P.M. Churchland,
1981) - has led contemporary physicalists to be sceptical about the prospects of such
a reduction. Stich (1986) argues that the concept of “belief” has no place in cognitive
science, P.M. Churchland that the probability of elimination of folk psychology is high,
given the success of an explanatory neuroscience (P.M. Churchland 1988, pp. 46-47)
and P.S. Churchland that since folk psychology is so clearly amenable to scientific
improvement, “what will eventually reduce to neuroscience are generalizations of
scientific psychology that have evolved a long way from the home ‘truths’ of extant
folk psychology” (P.S. Churchland, 1986, p. 312),
Given our present dependence on folk psychology, it is helpful to approach the
issue from the other end of the epistemic scale, and ask what prospects there might
be for developing, not just an account of psychological theory, but of all theory, in
neurological terms. If there is a way forward - and an answer to functionalism -
here, we may be able to develop a co-evolutionary theoretical framework within which
the desired refinements or eliminations of folk theory emerge. There are signs that this
is a promising strategy.
The first point to make is that it is possible to show how theories can be represented
naturalistically in neural terms, how a fully naturalized epistemology might be possible.So far, in presenting a coherentist epistemology, we have spoken of our knowledge
as “theories”, “hypotheses”‘, and “claims”, with the implication that these are stated
symbolically, for example in sentences. But if in our reductionist account knowledge
is in fact brain states, what does it mean to say that sentences exist in the brain, if
indeed they do? In short, how can we reduce a representational account of knowledge
to a physical account of brain processes, and how can we relate “knowing that”, or even
“knowing how”, to a causal theory of learning? Apparently we need a non-symbolic,
non-sentential account of knowledge.
Drawing extensively on cognitive science, particularly cognitive neurobiology, P.M.
Churchland argues that such an account is in prospect, (His ideas have been given aninitial educational application in Evers, [1990a].) Churchland points out that recent
work in neurobiology, such as the accounts of cognition in studies of parallel distributed
processing, is enabling us to understand knowledge as biological information processing,
in particular, pattern recognition. Salient instances are evident in the development of
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 16/97
516 J. C. WALKER
models of neural networks by researchers in artificial intelligence, which have tended to
supersede emphases on program writing as ways of understanding human information
processing. Such artificial neural networks model salient features of the brain’s neuronal
organization. Here I shall not duplicate the account of Churchland’s “connectionist”theory provided by Lakomski in her contribution to the present volume (p. 537), which
should be consulted before reading on. In particular, an understanding of the notion of
a prototype vector is presumed in the following exposition.
Evers (1990b) has pointed out that there could be a vast dividend for educational
theory if learning and cognition are matters of pattern recognition, pattern association
and pattern processing. To date the bulk of our theorizing about learning and cognition
has been conditioned by our linguistically (and so sententially) driven theories of
language as serial and logico-rational. This leads to a focus on serial algo~thms in
teaching mathematics and science, for example. Yet we know that there are people
who can multiply large numbers correctly in an instant and cannot be using serial
algorithms. Consider idiots savants. (See the movie Ruin Man.) Since the brain is
an enormous pattern associator, it is possible that such people have hooked into its
“machine language” for “direct processing”. The truly extraordinary possibility is that
we might one day be able to devise a pedagogy that allowed people’s learning to access
more directly the actual pattern processing features of brains, or restructure curricula
to reflect key patterns in knowledge so that learning within and between traditional
subjects is driven by considerations of pattern association, with logical structure being
a vital but derivative feature.
Or consider the nonsentential learning processes characteristic of much, probably
most, cultural transmission, and therefore of informal - and (embeddedly) formal- education. Typically, these are theorized mentalisticaliy by anthropologists and
qualitative educational researchers as symbol systems, in the manner of symbolic
interactionism, semiotics, and so on. The mentalistic predicates used to describe
cultural patterns are even less convincing devices than the sophisticated apparatuses
of linguistics for explaining learning.
The connectionist account provides Churchland with a way of naturalizing explanation.
First, activated prototype vectors constitute not just the creature’s recognition but its
understanding of the objective situation, which is reflected in its behavior.
Explanatory understanding consists in the activation of a specific prototype vector in a well trained
network in the apprehension of a problematic case as of a general type, a type for w&h the
creat ure has a det ai l ed and wel l i nformed rep~esent ut ~on. Such a representation allows the creature
to anticipate aspects of the case so far unperceived, and to deploy practical techniques appropriate
to the case at hand [P.M. Churchland, lYX9, p. 2101.
This not only enables us to account for depth and breadth of understanding by
reference to the degree of experience, practice and training, even though different
individuals may understand and classify a situation in the same basic way. More
importantly, it enables us to provide a unified theory of explanation.
One prominent fact, ill addressed by any existing account of explanation, is the variety of differenttypes of explanation. We have causal explanations, functional explanations, morat expianations,
derivational explanations, and so forth. Despite some procrustean analytical attempts, no one of
these seems to be the basic type to which all of the others can be assimilated. On the prototype
activation model, however, we can unify them all in the folio~~ing way. Explanatory understanding
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 17/97
Beyond Paradigms 517
is the same thing in all these cases: what differs is the character of the prototype that is activated
[P.M. Churchland, 1989, p. 2121.
The account is not just descriptive, either; but normative also. It addresses coherencecriteria, in particular simplicity - in supplying a reductive account of explanation - and
in explanatory power - in explaining the physical basis for all types of explanation. Thus
contemporary physicalism is not open to the charge leveled against the logical empiricists,
that they were hamstrung by a commitment to reducing all forms of explanation to one:
the deductive nomological model (Margolis, 1978; for a rebuttal see P.M. Churchland,
1980). Churchland’s reduction of all forms of explanation enables us to take a more
relaxed approach than the logical empiricists: we take whatever generalizations we
can get, and test them out against our coherentist criteria. Debates among educational
inquirers about the compatibility or incompatibility of logico-rational explanation and
interpretive understanding will need to be seen in a new light, a light which shouldencourage efforts to provide a unified account not only at the microphysical but at the
sentential-theoretical level. It is unlikely that if forms of explanation can be harmonized
at the former level they are incompatible at the latter. Good heart is given to those, like
Campbell, who assert their intellectual interdependence.
Reasons for scepticism about reduction abound. We have so far concentrated on
abstract epistemological discussion and on the individual mind-brain. Granted, however,
that education is a social and cultural as well as an individual process, it is as well to
consider the bearing of our coherentist and reductionist naturalism on our understanding
of culture and society. P.M. Churchland has considered two particular sceptical arguments
advanced from a cultural perspective. They emphasize the importance and complexity ofcultural contexts in our understanding of human nature, and the variability and plasticity
of human consciousness and behavior. First, it could be argued that in focusing on brain
structure and processes, our naturalism is limited to the microscopic level, whereas
much of what constitutes consciousness and behavior derives from the relations between
individual humans, and from cultural practices and institutions. Hence a reductionist
naturalism is explanatorily limited and deficient. Second, human beings are as varied
as the range of human cultures and the scope for human plasticity that cultures permit.
Part of this plasticity is to be explained by the reality of human self-determination,
creative activity and reflexive action. Moreover, cultures evolve, generating conditions
for further human variation.In contending that his naturalist connectionism - the network theory - can explain
these facts, P.M. Churchland shows that each argument rests on a misunderstanding
of the naturalist position. Human plasticity and the determination of consciousness
by the cultural surround are essential components of naturalism (P.M. Churchland,
1979). Furthermore, he points out that the network theory, interestingly, shares some
basic tenets with prominent anti-naturalist philosophy. There is touchstone with the
continental tradition’s insistence on the non-propositional or non-sentential character
of most human knowledge and the emphasis on human agency in continental and recent
analytic philosophy - notions shared widely in the social sciences and in educational
theory and research.
With the network theory of the brain, naturalistic reductionists can explain the
plasticity of human nature by delineating the sustaining underlying mechanisms, the
dimensions on which change is possible and the forces driving changes in cognitive
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 18/97
518 J. C. WALKER
configuration. Variation is possible because within the matrix, whose weights “determine
what features in the world one responds to, which values one embraces, and which range
of behaviors one commands” (P.M. Churchland 1989, p. 131). The sheer number of
neurons and synaptic connections in the brain yield a possible number of cognitiveconfigurations well in excess of the total number of elementary particles in the universe.
This is more than enough to explain human plasticity.
The cultural embedding argument emphasizes that explanations for human behavior
must include cultural features. We respond not just to light, sound and heat, but to
meaningful language, culturally significant facial expressions, moral judgments and
obligations, social customs, and so on - features of great subtlety and complexity. In
response, Churchland reiterates the powerful capacity of trained networks to recognize
and represent abstract and subtle features. What is important is the training, that there
is an appropriate “teacher” to shape representations of and responses to the cultural
environment. Examples are to hand of how networks can be trained to recognize visual,
logical and linguistic patterns.
It should therefore come as no surprise that a human infant comes to recognize and respond to
cultural features that resist definition in terms of notions like mass, charge, length and so forth,
because the most dominant “teacher” in the local environment is the culture into which the infant
is born. The set of weights that constitutes a child’s developing consciousness is continually being
shaped by the linguistic, conceptual and social surround. The developing brain comes to reflect the
elements and structure of that surround in great detail, for that is what networks do. What shapes
them is the stimuli they typically receive, and the subsequent corrections in their responses to which
they are typically subject. Small wonder that we become attuned to the categories of the culture that
raises us [P.M. Churchland, 1989, p. 1331.
It would be to miss the point to say that this concedes that the explanation of human
behavior must rely more on features of the cultural than of the microphysical level. To
pit the cultural against the physical is to substitute a false dichotomy for the relevant
distinction, which is between simple, context free features and complex, highly context
dependent features. The point is that it has already been demonstrated how a system
of physical elements can represent and respond to complex contexts, including cultural
contexts, and learn through so doing, its cognitive configurations developing in complex
and subtle ways through adjustment of weights. Naturalistic reduction explains how this
can be so.
Conclusion
If the drift of this essay is sound, educational researchers, especially as a community
with traditions, practices and organizations, should give some priority to attention to
two sets of epistemic-social relations. First, there should be a sustained emphasis on
strengthening relations between the various specializations and schools of thought in
education itself. Second, there is the exciting prospect of strengthened links with
basic research in cognitive science, and participation in the lively discussion of the
philosophical and methodological issues which are emerging. As basic research throws
up more discoveries with potential practical applications, educational inquirers have
much to contribute to our global knowledge of learning.
Our first task is to fasten onto existing practices of co-evolutionary research,
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 19/97
Beyond Paradigms 519
and to foster, through institutionalized practices as well as cultural change, a co-
evolutionary framework which will be recognized and respected by educational inquirers
generally. One part of this task could well involve taking up Campbell’s notion of the
interdisciplinary specialist, providing for the production of such people in our research
institutions and organizations and for the presentation of their contributions in our
conferences and journals. Closer contact with interdisciplinary specialists in basic fields
such as cognitive science would facilitate this.
To underpin these efforts intellectually we need a reflexive naturalistic epistemology,
with the unity of science as a constructive working hypothesis, uniting work on several
fronts simultaneously, contextualizing educational research within the total scientific
enterprise, and fostering a rehabilitated regard for science itself.
References
Campbell, D. T. (1988). Methodology and epistemology for social science. Chicago: University of Chicago
Press.
Campbell, D. T. & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-
multimethod matrix. Psychol ogical Bul l efi n, 56, 81-105.
Churchland, P. M. (1979). Scienti fi c reali sm and the pl asficif y of mind. Cambridge: Cambridge University
Press.
Churchland, P. M. (1980). Joseph Margolis: Persons and minds: The prospects of a non-reductive
materialism. Di al ogue, 19, 461-79.
Churchland, P. M. (1981). Eliminative materialism and the propositional attitudes. Journal of Phil osophy,
78(2).
Churchland, P. M. (1985). The ontological status of observables: In praise of the superempirical virtues. InP. M. Churchland & C. A. Hooker (Eds.), Images of science. Chicago: University of Chicago Press.
Churchland, P. M. (1988). M att er and consciousness: A contemporary int roduction to the phil osophy of
m i n d . Cambridge, MA.: MIT Press.
Churchland, P. M. (1989). A neurocomput at i onal perspecti ve: The natur e of mi nd und the strucmre of
science. Cambridge, MA.: MIT Press.
Churchland, P. S. (1982). Mind-brain reduction: New light from philosophy of science. Neuroscience, 7(5),
1041-1047.
Churchland, P. S. (1986). Neurophil osophy: Toward a unif ied science of the mind-brain. Cambridge, MA.:
MIT Press.
Dennett, D. C. (1978). Brai nstorms: Phil osophical essays on mi nd and psychology. Montgomery, Vt.:
Bradford Books.
Denzin, N. K. (1978). The research act: A t heoreti cal i ntr oducti on to sociol ogical methods (Second edition).
New York: McGraw-Hill.Duhem, P. (1954). The aim and structure of physical theory (Translated from the French second edition
(1914) by P. P. Wiener; first edition 1906). Princeton: Princeton University Press.
Evers, C. W. (1987a). Epistemology and the sructure of educational theory: Rehections on the
O’Connor-Hirst debate. Journal of Phi losophy of Education. 21(l), 3-13.
Evers, C. W. (1987b). Naturalism and philosophy of education. Educati onal Phil osophy and Theory, 19(2),
11-21.
Evers, C. W. (1990a). Educating the brain. Educati onal Phi l osophy and Theory , 22(2), 65-80.
Evers, C. W. (1990b). Personal communication.
Evers, C. W. & Walker, J. C. (1983). Knowledge, partitioned sets and extensionality. Journal of Philosophyof Education, 17(2), 155-170.
Feyerabend, P. K. (1975). Against method: Out li ne of an anarchisti c f heory of know ledge. London: NLB.
Fodor, J. A. (1975). The language of thought. New York: Crowell.
Hirst, P. H. (1966). Educational theory. In J. W. Tibble (Ed.), The Studyof
Education. London: Routledgeand Kegan Paul.
Hooker, C. A. (1981). Towards a general theory of reduction. 1. Historical and scientific setting; 2. Identity
in reduction; 3. Cross categorial reduction. Di al ogue, 20, 38-59; 201-236; 496529.
Kuhn, T. S. (1970). The structure of scienti f i c revolut ions (Second edition). Chicago: University of Chicago
Press.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 20/97
520 J. C. WALKER
Lakatos, I. (1970). Falsification and the methodology of scientific research programmes. In I. Lakatos &
A. Musgrave (Eds.), Cr~ficism and t he growt h of know ledge. Cambridge: Cambridge University Press.
Margolis, J. (1978). Persons and mi nds: The prospects of a non-r educti ve mat eri al i sm (Svnt hese Li brary . Vol.
125, Boston St udi es i n t he Phi l osophy of Science, Vol: 57). Dordrecht: Reidel. .Naeel. E. (1961). The str uctu re of science. New York: Harcourt. Brace & World.Opienheim, P. ‘& Putnam, H. (1958). Unity of science as a working hypothesis. In H. Feigl (Ed.), M innesota
studi es in t he phi l osophy of sci ence, 2, 3-36.
Popper, K. R. (1972). Obj ecti ve know l edge: An evohuionary approach. Oxford: Oxford University Press.
Popper, K. R. (1977). Part 1 of K. R. Popper & J. C. Eccles, The sel f and i ts brain. Berlin:
Springer-International.
Pylyshyn, Z. (1984). Computati on and cognit ion. Cambridge, MA.: MIT Press.
Quine, W. V. (1953). Two dogmas of empiricism. In W. V. Quine (1961). From a logical point of view.
Cambridge, Mass.: Harvard University Press.
Quine, W. V. (1960). Word and object. Cambridge, MA.: MIT Press.
Quine, W. V. (1969). Epistemology naturalized. In W. V. Quine, Ont ologi cal rel ati vi ty and other essays.
New York: Columbia University Press.
Quine, W. V. (1970). Phil osophy of logic. Englewood Cliffs, N.J.: Prentice-Hall.
Quine, W. V. (1977). Facts of the matter. In R. W. Shahan & K. R. Merrill (Eds.), American philosophy.Norman: University of Oklahoma Press.
Quine, W. V. & Ullian, J. S. (1978). The w eb of beli ef (Revised edition). New York: Random House.
Stich, S. P. (1986). From fol k psychol ogy t o cogni t i ve science: The case against bel i ef. Cambridge, MA.:
MIT Press.
Walker, J. C. (1985a) Materialist pragmatism and sociology of education. Bri ti sh Journal of Sociol ogy of
Educati on, 6( 1)) 55-74.
Walker, J. C. (1985b). Philosophy and the study of education: A critioue of the commonsense consensus.
The Austra l i an Journal of Educati on, 29(2), 101-114.
Walker. J. C. (1985~). The ohilosooher’s touchstone: Towards pragmatic unitv in educational studies.
Journal of Phi l osophy o f Educati on: 19(2), 181-198.. -
Walker, J. C. & Evers, C. W. (1982). Epistemology and justifying the curriculum of educational studies.
Bri t i sh Journal of Educati onal St udi es. 30(2), 213-229.
Walker, J. C. & Evers, C. W. (1984). Towards a materialist pragmatist philosophy of education. EducationResearch and Perspect i ves, 11(l), 23-33.
Walker, J. C. & Evers, C. W. (1988). The epistemological unity of educational research. In J. P. Keeves
(Ed.). Educati onal research methodol ogy, measurement and eval uati on: An int ernati onal handbook.
Oxford: Pergamon Press.
Biography
Professor James Walker is Dean of the Faculty of Education in the University
of Canberra, Australia. He has written numerous articles in the philosophy of
education, and has published in curriculum theory, educational policy and educationaladministration, and research methodology. His work in educational ethnography, a five
year study of transition from high school, is published in L outs and L egend s and related
articles. His current research interests include democratic philosophy of education, and
the relation between knowledge and practice in professional education.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 21/97
CHAPTER 2
TOWARDS A COHERENTIST THEORY OF VALIDITY
COLIN W. EVERS
School of Graduate Studies, Faculty of Education, Monash University, Clayton, Victoria,
3168, Australia
Abstract
Theorizing about validity has for a long time been dominated by epistemological assumptions
associated with logical empiricism. In the 1960s these assumptions were systematically
challenged by a paradigms construal of scientific knowledge. More recent educational
research methodology has been much influenced by this paradigm’s perspective and itsassociated epistemological relativism and pluralism, and there is now considerable debate
about how to understand the familiar terms of educational research appraisal within such
an epistemological setting. This chapter offers a coherentist epistemological perspective on
some features of this debate concerned with understanding validity.
Although subject to differing demands of practice, theory of research and theory
of validity have been linked by shared epistemological assumptions, especially those
assumptions that derive from the period of dominance of philosophy of science in
epistemology. In terms of its impact on educational studies, we may take this period as
beginning with logical empiricism in the late 194Os, through the paradigms era arising out
of the work of Kuhn and still dominating educational studies today, to recent attempts to
apply coherence theories of knowledge and justification to educational theory building
and adjudication.
In what follows, I shall trace some of these epistemological influences on ways of
understanding the validity of tests, experiments, and inquiry procedures. Although I
think the arguments mounted by Kuhn (1962), Hanson (1958), Feyerabend (1962), and
others against logical empiricism are decisive, I regard the resulting paradigms construal
of research and methodology in education as mistaken. These arguments are more a
reducti o ad absurdum of narrow empiricism than a prospective methodology in their own
right. Instead, I shall defend a coherence theory of justification and elaborate some of its
consequences for understanding research and validity. Finally, if both logical empiricism
and paradigms theory are false, then any successful applications of methodologies based
on those epistemologies must be drawing on coherence justification somewhere. I argue
that this turns out to be the case, implicitly in Cronbach’s work on validity, early and
late, and explicitly in Campbell’s more recent work on theory of research.
521
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 22/97
522 C. W. EVERS
Validity and Logical Empiricism
In 1954, a joint committee of the American Psychological Association, AmericanEducational Research Association, and the National Council on Measurements Used
in Education, produced a set of Technical Recommendat i ons for Psychological Tests and
Diagnostic Techniques. The “essential principle” behind the document is that
‘a test manual should carry information sufficient to enable any qualified user to make sound
judgments regarding the usefulness and interpretation of the test’ (Technical Recommendations,
1954, p. 202).
Since the usefulness of a test and its manual is partly a function of the degree to which
a test achieves its aims, the question of test validity is of prime importance. Roughly
speaking, validity in this context is a matter of the extent to which a test (or instrument,
or procedure) measures what it purports to measure. More generally, it is concernedwith the soundness of the inferences that can be made from test scores, or results.
The Technical Recommendations identifies four aims of tests and therefore four clusters
of possible inferences, or types of validity: content validity, which aims to measure
present performance by sampling an identified universe of performance; predictive
validity, concerned with future performance; concurrent validity, like predictive validity
but matched in the present rather than the future against some outside criterion; and
construct validity, where the trait or quality being measured is itself defined in terms
of the test. In the first revision of these recommendations, the 1966 Standards for
Educat ional and Psychological Test s and M anuals, predictive validity and concurrent
validity were collapsed into what was called “criterion-related validity”, thus yielding
three types of validity (Standards, 1966, pp. 12-13).
Focusing for a moment on tests, the interesting epistemological question is how we
can ever know whether our inferences from scores are sound. Strictly speaking, for a
score to count as a score, it must exist under some description. But descriptions are
comprised of words (or other symbolic tokens) which in turn must be meaningful in
order to sustain inferences. Ordinarily, this does not pose problems since most of the
words we use in everyday discourse are defined contextually, in terms of other words.
However, logical empiricism places severe restrictions on the adequacy of definitions.
Contextual definition is certainly part of the story, but eventually for meanings to be
known there must be some correspondence between some words and empirical evidence,
or observations (Feigl, 1950). Ostensive definition will do, but it seems to work best for
words with the most modest inferential connections. On the other hand, words that are
the richest in inferential structure, that are embedded in the most central parts of a
theory or theoretical context, seem to be least obviously connected to experience.
Empiricism’s compromise between inferential richness, or theoreticity, and empirical
adequacy is operational definition. Thus the theoretical term “length” would be
defined in terms of the sequence of observable operations used to carry out a certain
measurement procedure. Of course, if descriptions of operations are also theoretical
then we need to repeat the process until we reach observations sufficient for empirical
meaning. This is the problem with tests, which can actually be regarded as operational
definitions of scores. In the case of concurrent and predictive validity the empirical
content of the theoretical terms describing the scores is given by stipulating some
antecedently meaningful criterion. In the case of content validity, we are presumably
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 23/97
Beyond Paradigms 523
dealing with just some subset of an antecedently meaningful universe of examples. (See
Kerlinger, 1964, pp. 444449).
Even with these apparently simple cases, we now know that there are epistemological
difficulties. Take again the example of defining “length”. Presumably the operation takesplace at some particular time and place using some particular set of singularly specified
apparatus, including a rule,, Must every act of measurement use this particular rule
on penalty of yielding a different definition? If the answer is yes, then we are looking
at different, non-equivalent, definitions of length for every rule. If the answer is no,
then we must have some way of specifying an equivalence class on rules that preserves
sameness of operational definition. Something sufficient to permit us to say that rule,
= rule, = . . _ = rule, would do nicely. However, this amounts to the task of giving an
operational definition of “same length”. Such a task cannot be done for an indefinite n
unless we make use of some notion of “standard rule”. Standard rules do exist, of course,
but the considerations that go into their selection, namely those that will give generality
over time, place, and circumstance to the measurement of length, have long since outrun
the meagre resources of operational definition (Hempel, 1966, pp. 93-94).
The problem here is quite general, and has been noted by both Popper (1963,
pp. 44-45) and Quine (1957, p. 231). Namely, there is no such thing as a class of
similar objects. (For a proof of this theorem, see Watanabe, 1969 pp. 376-379.) As
Popper insists, similarity is always similarity-for-us. Since some similarity groupings
are essential for theorizing, given the weak naturalistic constraint that we have finite
learning capacities, operaticural definitions will reflect a prior theoretical decision to
group operations according to some weighting of features or saliences. In another
context, Popper calls these weightings “hypotheses”, and we can follow this usage
here. However, what this argument implies is a form of semantic holism. Observations
do not correspond one to one with theoretical terms to be defined, but instead distribute
their empirical content across the entire network of prior hypotheses and their inferential
contexts. Quine reaches this conclusion in his classic paper “Two Dogmas of Empiricism”
(1951). The upshot is that the epistemological demand for knowledge of the empirical
meaning of a term always outruns the resources posited for operational definition,
however simple the term. An absence of disagreement over the representativeness of
samples for content validity, or criteria with which predictive or concurrently tested
scores may be correlated is not a waiving of theoreticity so much as an indication ofshared, or t~~c~~t~~e, theory (Walker & Evers, 1988).
For construct validity, the Tec~nicQl ~ecu~~e~dutio~s document (1954) is less
sanguine; theory is acknowledged to intrude from the beginning:
To examine construct validity requires both logical and empirical attack. Essentially, in studies ofconstruct validity we are validating the theory underlying the test. The validation procedure involvestwo steps. First, the investigator inquires: From this theory, what predictions would we make regardingthe variation of scores from person to person or occasion to occasion? Second, he gathers data toconfirm these predictions [p. 2141.
The big advantage of construct validity, if it can be made to work, is that it promises
a way out of what Hempel (1965) calls the theoretician’s dilemma. Essentially, this isanother artifact of maintaining a sharp distinction between theory and observation. As
we noted earlier, where empiricist demands of definition can be met, theoretical terms
are invariably uninteresting. Where they enjoy extensive intertheoretic connections and
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 24/97
524 C. W. EVERS
enter into a wide range of deductive relations, they are hard to define. If we could
validate a whole theory, that is, show that the theory describes what it purports to
describe, we can have both empirical adequacy and the kind of inferential richness
needed to develop fine grained concepts suitable for the social sciences. An addedbonus would be that troubles with the first three (or perhaps two) types of validity
which, because of the intrusion of theory, become species of construct validity, might
admit of resolution.
Can whole theories be usefully validated? In their important paper on construct
validity, Cronbach and Meehl (1955) take up the challenge. They begin the task
of specifying the logic of construct validation by setting out their philosophy: “The
philosophy of science which we believe does most justice to actual scientific practice
will now be briefly and dogmatically set forth” (p. 78). Not surprisingly, they offer a
version of logical empiricism. They define a nomological network as a theory comprised
of an interlocking network of laws. A network relates observables to each other, to
theoretical constructs, and constructs to each other. To count as science a construct
must figure in a network some of whose laws involve observables, and so on (Cronbach
& Meehl, 1955, pp. 78-79). Validating a theory boils down to demonstrating that the
network is warranted by empirical evidence. To counter the . . .
“toughminded”, who fear that allowing construct validation opens the door to unconfirmable test
claims the answer is that unless the network makes contact with observations, and exhibits explicit,
public steps of inference. construct validation cannot be claimed [Cronbach & Meehl, 1955, p. 791.
As one might expect, worries over the relationship between evidence and meaning,
with an attendant shift to holism, apply equally to the justification of theories. The
business of exhibiting explicit, public steps of inference that were also epistemologically
compelling, came under great pressure in the 1960s and eventually led to the demise of
logical empiricism.
Epistemology and Logical Empiricism
For assessing its merits as an epistemology it is useful to see logical empiricism as an
example of foundationalism. Generally speaking, foundational justification proceeds firstby identifying an epistemically privileged subset of knowledge claims and then by arguing
that this subset somehow warrants all other justified knowledge claims. An early version
of foundationalism, which I would call “strict foundationalism”, was championed by the
empiricist philosopher David Hume in the eighteenth century, wherein knowledge was
reckoned as justified only if it was deducible from the privileged subset of sensory
experiences. With the arrow of deducibility going from a finite number of singular
sensory impressions Hume had no trouble showing that no general or lawlike empirical
claims are ever justified. To go from a finite set to an infinite set some principle of
induction is required. But the principle of induction must itself be warranted. We may
indeed have foundational evidence for such a principle - perhaps in the past it hasalways held - but we need the same principle to deduce that it will hold for future,
or unobserved, cases beyond the finite range of foundations. And such an argument is
circular.
This difficulty is known as the problem of induction and it renders problematical
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 25/97
Beyond Paradigms 525
all attempts to make unrestricted empirical generalizations from a finite observation
base. In Campbell and Stanley’s (1963) discussion of factors jeopardizing the validity
of experimental and quasi-experimental designs for research, they note its effect on
external validity by entering a caveat:
This caveat introduces some painful problems in the science of induction. The problems are painful
because of a recurrent reluctance to accept Hume’s truism that induction or generalization is never
fully justified logically. Whereas the problems of internal validity are solvable within the limits of the
logic of probability statistics, the problems of external validity are not logically solvable in any neat,
conclusive way. Generalization always turns out to involve extrapolation into a realm not represented
in one’s sample. Such extrapolation is made by assuming one knows the relevant laws [p. 171.
Three initial comments need to be made here. First, the external validity of
experiments is being contrasted with their internal validity. Basically, internal validity
is concerned with whether an experiment is significant in the production of someanticipated outcome, whereas external validity is concerned with the generalizability
of an experimentally produced effect. Second, from an epistemological point of view,
the justification of external validity for experiments is the same as the justification of
construct validity for tests. As Cronbach and Meehl(1955, p. 89) note, “the investigation
of a test’s construct validity is not essentially different from the general scientific
procedures for developing and confirming theories.” And finally, one might expect
the ubiquity of theory to blur the distinction between internal and external validity.
Logical empiricists, and Vienna Circle positivists before them, had a partial response
to Hume’s argument; namely, to alter the direction of deduction between knowledge
and its foundations. It is knowledge claims, grouped systematically into theories, or
networks, that imply privileged foundations, in this case observation reports, rather
than vice versa. For this sort of broad foundationalism, the relation of justification
between observation and theory is testability, where testability is thought to involve
two components. Observation reports that match those which may be deduced from
a theory are said to confirm the theory, and observation reports which fail to match
expectations fulsifv, or disconfirm, it. A theory may thus be regarded as validated to
the extent that it has been subject to many tests which have confirmed, but in no way
disconfirmed, it.
Of course, in practice the testing of theories is more complex, but it is the complexity
of practice that ultimately tells against logical empiricism. Consider confirmation and the
problem of induction. As there is only ever a finite number of confirming.observations,
theories will always be radically underdetermined by empirical evidence. We can fit an
arbitrary number of curves to a finite set of data points. Under these conditions the
notion of inductive support fails to have purchase as it is not clear which empirically
adequate but distinct theory is being supported. For example, Newtonian mechanics
enjoyed several hundred years of accumulated confirmations, yet it ultimately failed
to be validated not just on fine matters of detail but right through to its most central
theoretical categories. For all the evidence that confirmed it also confirmed relativity
theory. Cronbach and Meehl (1955, p. 87) are aware of the problem but end up running
together both cumulative inferential support and radical falsification: “Confidence in a
theory is increased as more relevant evidence confirms it, but it is always possible that
tomorrow’s investigation will render the theory obsolete.” There is also a puzzle over
what counts as relevant confirming evidence. In rigorous formulations of testability,
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 26/97
526 C. W. EVERS
the deductive relations between theory and observation are defined in terms of the
truth functional material conditional. Since it is the truth values of sentences that are
semantically important, a non-black non-raven, for example a leaf, logically can confirm
the hypothesis that all ravens are black (Hempel, 1965).
Campbell, who had read Popper (1959) and Hanson (1958), was aware of these
problems and hedged against them in Experi ment al and Quasi-Experi ment al Designs
fo r Research (see also Campbell, 1984). There, the important evidential relation was
falsification, not confirmation: “The task of theory-testing data collection is therefore
predominantly one of rejecting inadequate hypotheses” (Campbell & Stanley, 1963, p.
35). They claim that technically speaking, hypotheses are never confirmed; rather they
are “probed” by the results of experiments. On this account, the chief strategy behind
successful experimental design is to limit the number of plausible rival hypotheses about
the role of an experiment in producing some particular result. So, randomization was
thought to render implausible some eight alternative hypotheses that threatened the
internal validity of experiments. Let us suppose, for the moment, that this is so, although
we can note that in a later work Cook and Campbell (1979) demonstrate a complexity
even with internal validity by producing some threats not amenable to randomization.
The complexity of the social world, together with a paucity of true generalizations,
would make a similar methodology for external validity very difficult. This is because
whole theories, or networks, of hypotheses imply observations, and falsification
distributes its bad news only disjunctively across a network. As Quine (1951, p.
43) has claimed, and as Campbell (1986a, p. 508) has later acknowledged, we can
hold true any claim, come what may, if we are prepared to make drastic enough
revisions elsewhere in the network. The Duhem-Quine thesis, as this result is often
called, renders exceedingly problematical any purported evidential relationship between
a hypothesis and falsifying observations.
One response is to note that the thesis fails to distinguish plausible from implausible
hypotheses. If falsification is avoided only by invoking implausible rival hypotheses then
it is as good as falsification outright. However, note also that plausibility is not an intrinsic
property somehow embedded in some hypotheses rather than others. Plausibility is
an epistemic notion and is therefore imputed relative to the prior assumption of
some theory. In short, however bad the observational news may be for Newtonian
mechanics, fromthat
perspective, time dilation, mass increases due to velocity, andcurved space-time are just implausible.
Cronbach and Meehl (1955) do not use explicit plausibility judgments as a device
to limit the range of construct validity threatening alternatives arising from negative
evidence. Their advice is more diffuse:
The choice among alternatives, like any strategic decision, is a gamble as to which course of action
is the best investment of effort. Is it wise to modify the theory? That depends on how well the
system is confirmed by prior data, and how well the modifications fit available observations. Is it
worthwhile to modify the test in the hope that it will fit the construct? That depends on how much
evidence there is to support the hope, and also on how much it is worth to the investigator’s
ego to salvage the test. The choice among alternatives is a matter of research planning [p. X4].
When thinking about gambling, investigators’ egos, and the nature of research
planning, remember that their paper was supposed to yield a research plan for making
explicit, public steps of inference necessary for claiming the validity of constructs.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 27/97
Beyond Paradigms 527
So far we have canvassed two familiar problems with logical empiricism’s account of the
justificatory relationship between theory and empirical evidence - underdetermination
problems with classical confirmation theory and complexity of test problems with
falsification. The final problem I want to raise challenges the whole point of foundationaljustification. Recall that foundationalism requires the identification of an epistemically
privileged subset of knowledge claims from which others derive their warrant. But what
guides the choice of such a subset? In the case of Hume’s classical strict foundationalism,
choice of foundation is guided by a theory of the powers of the human mind, notably
a theory of learning and cognition. Learning is occasioned by the receipt of sensory
impressions and cognition is partly a matter of the logical manipulation of these
impressions (Hooker, 1975). The trouble with such a theory is that it is not known
non-inferentially; it is not part of the foundations. Indeed it cannot be because it makes
general empirical claims about human learning, thus requiring inferential justification.
But if the selection of privileged knowledge claims depends on the use of non-privileged
theory, the structure of foundational justification collapses.
This argument also applies to broad foundationalism. For the choice of observational
evidence to test theories reflects theoretical beliefs about knowledge acquisition by
humans. By the same token, Campbell and Stanley’s plausibility judgments and
Cronbach and Meehl’s strategic decisions are likewise theory-laden. The correct solution
to this problem in my view, one defended by Quine (1969) and adopted with increasing
systematicity by Campbell, is to naturalize epistemology. If epistemology presupposes
theories of human learning and cognition why not just use the best theories available,
theories from science - psychology or cognitive neurobiology - rather than a priori,or armchair, theories? This sounds circular because the notion of “best” being employed
here is epistemic, so it will be one of the challenges of an alternative, non-foundationalist
epistemology to show that the circularity is not vicious.
Paradigms of Validity
All of the difficulties with logical empiricism we have canvassed have concerned
problems over the relationship between theory and empirical evidence. It would appear
that the matter of theory justification is not settled by evidence, however comprehensive.One conclusion drawn by a number of philosophers of science, for example Kuhn and
Feyerabend, is that if all the evidence there is for a theory is empirical kvidence, and
if empirical evidence can never be adequate for rational theory adjudication, then so
much the worse for the enterprise of rational theory adjudication. This is especially the
case where alternative theories are comprehensive enough to contain, or entail, theory
specific criteria for theory choice. What Kuhn calls “paradigms” provide a good example
of this:
In learning a paradigm the scientist acquires theory, methods, and standards together, usually in
an inextricable mixture. Therefore when paradigms change, there are usually significant shifts inthe criteria determining the legitimacy both of problems and of proposed solutions That
observation . provides our first explicit indication of why the choice between competing paradigms
regularly raises questions that cannot be resolved by the criteria of normal science . (scientists] will
invariably talk through each other when debating the relative merits of their respective paradigms. In
the partially circular arguments that regularly result, each paradigm will be shown to satisfy more or
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 28/97
528 c. w. EVERS
less the criteria that it dictates for itself and to fall short of a few of those dictated by its opponent
[Kuhn, 1962, pp. 109-1101.
Here we have an argument for the inc~~~e~~~ru~il i~y of paradigms, for epistemic
relativism between fairly comprehensive theories. Moreover, belief in paradigm specific
epistemologies has a direct relevance for research methodology. Orthodox logical
empiricist theorizing about research designs, validity, and reliability, for example, is
akin to “normal science”. But there are other, alternative ways of conceiving research
and inquiry. The question of which is best cannot rationally be decided because the
epistemic notion of “best” is relative to each paradigm.
Lincoln and Guba (I 985) offer a detailed version of the paradigms thesis of educational
research as part of their defense of naturalistic inquiry. (Note that this is a different
sense of “naturalism” to that employed by Dewey, Quine, or Campbell, when they
speak of epistemology.) Thus, consider their discussion of what it takes to establish thetrustworthiness of an inquiry. “Trustworthiness” is an epistemic notion to do with the
warrantability of an inquiry’s findings or inferences. As such, standards of justification
will be paradigm specific. To demonstrate this, they consider answers to the following
four research questions:
(1) How do we show the “truth” of the findings of a particular inquiry?
(2) To what extent are these findings applicable to other contexts?
(3) Can the inquiry be replicated?
(4) How can we establish that the results are independent of researcher biases and
perspectives? (See Lincoln & Guba, 1985, p. 290.)
Within the logical empiricist paradigm - what they call positivism - we have four
familiar answers. Establishing internal validity is crucial for the first; external validity
matters for the second; reliability is what the third is all about; and the fourth concerns
objectivity. But this cluster of answers draws on a common set of epistemological and
metaphysical assumptions. For example, that there is a world “out there” which can be
known which corresponds to true claims, that the knower can be separated from the
known, that events are relatively separable and independent, that different events have
different causes, and that what happens in the world can be known in a way free from
value assumptions (Lincoln & Cuba, 1985, p. 28).
However, using arguments from underdetermination of theory, complexity of tests,
and theory-ladenness of observation, Lincoln and Guba both challenge the truth
of logical empiricism and maintain that its epistemologi~al standards are distinct,
indeed orthogonal to some alternatives. Within the paradigm of naturalistic inquiry,
the above four questions would be answered as follows: establishing credibility, not
internal validity, is vital for the first; the second is a matter of transferability, not
external validity; the third requires a case for dependability, not reliability; and the
last involves confirmability, not objectivity (Lincoln & Guba, 1985, pp. 301-327). There
is, of course, a detailed epistemological story to be told about the adequacy conditions
for establishing each of these criteria for naturalistic warrantability. Suffice it to note that
these are supposed to be relative to the naturalistic inquiry paradigm and not the logical
empiricist paradigm, which is simply inapplicable - being a different paradigm. The
upshot is that the traditional notions of validity appear to have integrity, or definition,
only within logical empiricism.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 29/97
Beyond Paradigms 529
One puzzle for the paradigms approach is whether, for example, an ethnographic
study can be said to describe something “out there”, whether the non-positivist study
can be said to correspond with how the world really is. If logical empiricism has a
mortgage on correspondence truth, realism, and objectivity, then in what sense might
the inferences and findings of non-positivist research paradigms be regarded as “true”?
If we accept the relativism of the paradigms thesis then there is an equivocation over
the word “true”. Its various senses would be paradigm relative. This might satisfy those
researchers who are prepared to acquiesce in subjectivism, but it creates difficulties for
critics of logical empiricism who want to argue for social change or improvement - for
example defenders of critical theory, feminist research, or action research. These critics
want to say that there are realities “out there” that are oppressive independently of how
victims mistakenly see matters, that limit human potential regardless of how ideologically
content we may feel with our lot, that need to be changed. (See, for example, Foster,
1986; Bates, 1983, in the critical theory tradition.) On this view, radical subjectivism
becomes a political stance in de facto support of an existing distribution and exercise
of power. However, since logical empiricism is also thought to be part of the political
problem, where are the solutions?
Lather (1986, p. 65) describes this dilemma as being caught between a rock and a
soft place; between the “unquestionable need for trustworthiness in data generated by
alternative paradigms and . . . the positivist claim to neutrality and objectivity.” She
proposes a solution that is aimed at reconceptualizing validity within a postpositivist
framework of interpreted data and researcher commitment. Essentially, validity of
research, interpretation, and coordinating background theory, is a matter of ensuring
the presence of self-correcting research procedures and practices. Examples include
triangulation of methods, expanding construct validity to include accounts of how data
figured in the transformation of theory, require face validity to reflect participants’
reactions to inquiry, and propose guidelines for catalytic validity to require “that
respondents gain self-understanding and, ideally, self-determination through research
participation” (Lather, 1986, p. 67).
Various questions of detail could be asked about each of these proposals. However,
the puzzle for me is the point of the exercise, which appears to be aimed at removing
researcher bias through the provision of self-correcting research. For if positivism has a
lien on objectivity, then presumably “bias” and “correction” are also paradigm relative.The trouble is, these terms are clearest within the rejected logical empiricist paradigm.
But in the absence of a world “out there” that can be known by inquirers willing to
use some methods rather than others, it is difficult to know what to count as bias and
correction. On the other hand, if a clear meaning is established in some other paradigm,
it is difficult to see why there is dilemma at all over the rock and soft place. Why should
one research paradigm be obliged to meet the epistemological demands of another? Yet
the assumption of some such obligation appears to lie behind not just Lather’s discussion,
but Lincoln and Guba’s as well.
Consider, for example, the methodological virtue of triangulation. This is only an
epistemological virtue relative to the kind of inferences that are thought to be sanctionedif agreement occurs. But again, to ask the more basic question, why should agreement
be epistemologically more desirable than disagreement? I can suggest some metaphysical
baggage that would link agreement in triangulation with trustworthiness of inference, but
it will be uncongenial to paradigms theorists. It is, in a word: realism. The supposition of
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 30/97
530 C. W. EVERS
a real world in complex causal interaction with physical, thinking, acting, interpreting,
inferring humans (and whatever equally real causally discriminating apparatus they may
be using) certainly provides the basis for an economical account of the conditions under
which triangulation can be an epistemic virtue.
The structure of this argument is known as “inference to the best explanation”
(BonJour, 1985). Evidence for realism does not lie in some kind of direct sensory
experience of the furniture, or ontology, of the world; not even for the strictest
empiricist. Arguments for underdetermination, empirical test complexity, and theory-
ladenness tell decisively here. As Quine has remarked:
What are given in sensation are variformed and varicoloured visual patches, varitextured and
varitemperatured tactual feels, and an assortment of tones, tastes, smells, and other odds and
ends; desks are no more to be found among these data than molecules [Quine, 1960, p. 2501.
Yet the notion of “best” is still epistemic. Moreover, even if the realist hypothesis
is rejected as a way to defend the trustworthiness of certain inquiry procedures, any
argued replacement is “better” in an equally epistemic sense. This suggests that other
epistemological criteria are being employed, criteria that can be common to different
paradigms.
Coherence Justification
Let us explore the possibility of touchstone theory choice, and hence validity criteria
by re-examining our earlier arguments against logical empiricism. In each case the
structure of the argument amounted to showing that for some epistemological feature- objectivity, falsification, confirmation - the demands of the feature outran the
resources of empirical evidence. This argument structure is not saying that logical
empiricism is inadequate because it is not warranted by the data. Rather, the argument
is that logical empiricism is incoherent: it makes claims it cannot satisfy on its own
suppositions. But if this is the nature of the argument, then it will sustain another
conclusion; namely, there is more to evidence than empirical evidence. The reductio
argument against narrow empiricism only goes through if we suppose an equally narrow
view of evidence. However, if incoherence can function as evidence for the inadequacyof logical empiricism, then it ought to be applicable as a standard of evidence for the
adjudication of other theories. Churchland makes this point in a general way:
Since there is no way of conceiving or representing “the empirical facts” that is completely independent
of speculative assumptions, and since we will occasionally confront theoretical alternatives on a scale so
comprehensive that we must also choose between competing modes of conceiving what the empirical
facts before us are, then the epistemic choice between these global alternatives cannot be made
by comparing the extent to which they are adequate to some common touchstone, “the empirical
facts”. In such a case, the choice must be made on the comparative global virtues of the two global
alternatives, Tl-plus-the-observational-evidence-therein-construed, versus T2-plus-the-observational-
evidence-therein-(differently)-construed. That is, it must be made on superempirical grounds such
as relative coherence, simplicity, and explanatory unity [Churchland, 1985, pp. 41421.
Churchland’s point, together with the fact that paradigms theorists use the incoherence
of empiricism to argue that it is not correspondence true, suggests that the notion of
coherence evidence is compatible with correspondence truth. The strategy would be
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 31/97
Beyond Paradigms 531
to let coherence criteria grind out their story of which is the most warranted theory,
and then assume the existence of all objects presumed by that theory as constituting
the nature of the world - what the theory matches up with, or corresponds to. This
would even permit strong relativists to say that research is really (and correspondencetruly) paradigmatic in its structure. Such is the payoff from relocating certain narrow
empiricist notions of correspondence, realism, and evidence into a non-foundational,
coherent& context of epistemic justification.
Using this account of justification will permit us to solve a number of difficulties
raised earlier; for example the circularity problem created over what constitutes the best
theory of knowledge acquisition when these are embedded in the very epistemologies
under dispute. Requiring the theories to be coherent will force choices. It will also
select reflexiveness since a good epistemology should leave its subject matter and
itself learnable, and learnability will include the correction of error. Furthermore,
there will be a premium on naturalistic epistemologies if our most coherent theories
of humans reckon them to be part of the natural order. There are advantages of
parsimony in seeing knowledge as part of the natural order too, and in elaborating
accounts to include social factors in the production and distribution of knowledge. In
this way, a coherence theory of justification, with its emphasis on consistency, simplicity,
comprehensiveness, and explanatory unity, can function as a self-reflexive touchstone to
winnow rival epistemological alternatives without circularity.
Interestingly, Campbell defends the main claims being made here. His support of
naturalistic epistemology, in particular evolutionary epistemology, is long standing.
However, more recently, responding to Quine’s arguments, he has defended the
combination of coherence justification and correspondence truth: “My position is to
accept the correspondence meaning of truth and goal of science and to acknowledge
coherence as the major but still fallible symptom of truth” (Campbell, 1977, p. 445).
Needless to say, in drawing so heavily on scientific theory for epistemological details,
this view coheres well, not with naive realism, but with scientific realism. A challenge
for those who wish to produce a non-realist account of why agreement in triangulation
is an epistemic virtue would need to make it cohere with some view about the reality,
or otherwise, of knowing subjects and their environment.
Earlier, in the discussion about operational definition, we saw that empirical
evidence distributes its evidential support holistically. Nevertheless, there must besome provisional set of antecedent hypotheses that permit some observations to
count as more salient for the learning of causally contiguous (one word) sentences.
There must be some entering point for infant language learning to be possible. This
epistemological constraint of learnability will not save operational definition, but it
will create difficulties for the kind of semantic holism needed to support the thesis
that paradigms are incommensurable. For incommensurability trades on the idea that
the meaning of a term is a matter of its conceptual role in a theory. Systematically
different theories of leadership, intelligence, and education, for example, will leave
different uses of these terms orthographically identical but semantically distinct, so that
speakers will talk past each other. Now from an epistemological point of view, learningthe meaning of these terms requires a prior mastery of the theoretical network necessary
to provide conceptual roles. But theoretical networks are comprised of terms which in
turn have to be learned. So a regress threatens to make theories unlearnable. The correct
response to this problem is not to compromise semantic holism by reinstating a sharp
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 32/97
532 C. W. EVERS
theory/observation distinction. Rather, the learnability argument is best construed as
evidence for the existence of a certain amount of (shifting, provisional) touchstone
theory in the processing of experiences (Churchland, 1979, pp. 75-80; Papineau, 1979;
Walker & Evers, 1982, 1988). One might expect naturalistic learning theories to throw
some light on the nature of this touchstone in humans at the perceptual level.
The ubiquity of theory and the methodological appeal to touchstone coherence criteria
of theory justification suggest that varieties of validity are ultimately different kinds of
construct validity, and validating constructs is a matter of developing a most coherent
global theory. If this is so, it will explain the diffuseness that was first implicit and is
now more explicit in validity theory. Consider Cronbach’s recent reflections:
The positivists’ distinction between theory and observation can no longer be sustained and
there is no hope of developing in the short run the “nomological networks” we once grandly
envisioned Our best strategy is probably contextualism In brief, one offers a
generalization and then tries to locate the boundaries within which it holds. As the structureultimately becomes clumsy, someone will integrate most of the information into a more gracefulone. For scientists, this is a reminder that knowledge evolves slowly and indirectly, that one can be
prideful about contributing to the advance without the hubris of insisting that one has the “correct”
theory. For practical testers, this warns that an instructive program of construct validation - strong or
weak - is unlikely to reach the closure needed to defend a test that is already under fire [Cronbach,
1988, p. 141.
Messick (1988, p. 42); elaborating and defending the unified view of validity contained
in the 1985 Standards for Educati onal and Psychological Test i ng, identifies four bases for
test validity: (1) plausibility of interpretations as inductive summaries of evidence, (2)
the value implications of test interpretations, (3) the relevance of scores for particular
applications, and (4) the social consequences of proposed uses of tests. For Messick(1988) . . .
the heart of the unified view of validity is that appropriateness, meaningfulness, and usefulness
of score-based inferences are inseparable and that the unifying force is empirically grounded construct
interpretation [p. 351.
It would follow that the global theory to be validated, or shown more coherent than
rivals, would include ethical theory, as well as a theory of society and social causation.
Having seen earlier that Campbell accepted objections to logical empiricism and
embraced coherentist justification, I should note a qualification with regard to his
recent views on the distinction between internal and external validity. Campbell
(1986b) redraws this distinction as “local molar causal validity” and the “principle
of proximal similarity”. Roughly speaking, the former involves no generalization -
although Campbell qualifies this as an “exaggeration” - the latter does. The former
invites us to “back up from the current overemphasis on theory first” (Campbell, 1986b,
p. 70). Crucially, “in the new contrast, external and construct validities involve theory.
Local molar causal validity does not” (Campbell, 1986b, p. 76). The puzzle is that
Campbell knows that the whole enterprise of identifying local molar causes is laden with
theory - theory containing general terms. Identifying caused outcomes involves the use
of descriptions and terms that imply such outcomes are kinds. It is as though he is trying
to draw the observation/theory distinction at the internal/external validity level. Yet for
inferential purposes, the strongest distinction possible here is that between plausible and
implausible theories, and that requires a coherentist argument. Some remarks he makes
in another context on a parallel distinction - between the analytic and the synthetic -
may clarify this puzzle:
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 33/97
Beyond Paradigms 533
Thus while usually convinced by Quine when I read him, insofar as I can understand him, I havenot really stabilized the Gestalt switch which is required to abandon the analytic-synthetic distinction[Campbell, 1977, p. 441.1
So far, I have argued that the process of validation in general is coherent& and realist
rather than empiricist or paradigmatic, and that a coherence theory of justification is
compatible with a correspondence theory of truth. I want to conclude with some brief
remarks about judging hypotheses plausible, since that is of concern to both Cronbach
and Campbell. Let us suppose that we have two hypotheses up for consideration. Since
plausibility is relative to the quality of presumed background theory, let us locate each
within a theory, Tl and T2. Now whatever empirical tests we devise for adjudicating
between Tl and T2 - confirmation perhaps, or falsification - they will be insufficient.
We need to look at superempirical theoretical virtues. Work on these is at an early stage
in epistemology so it is not possible to give a rigorous account of their application. But
they can be made intuitively clear. For example, we should prefer Tl to T2 if Tl is simpler
than T2. If we construe simplicity in terms of number of assumptions, or axioms, then it
will be a virtue in this sense. Any theory can account for any phenomenon if we are
willing to just add assumptions. Indeed, without some premium on simplicity, the notion
of explanation loses its point if we can just add an assumption for every phenomenon
to be explained. We should prefer Tl to T2 if Tl explains more than T2. We should
prefer Tl to T2 if Tl is free from contradiction and T2 is not. Anything can be derived
from a theory that contains a contradiction. We should prefer Tl to T2 if Tl squares
better with what we already have reason to believe. We should prefer Tl to T2 if Tl
leaves behind fewer difficult unanswered questions. One question for each theory ishow can it ever be learned, or acquired. Coherence justification does not leave justified
theories floating above the evidence. Rather, the theory must cohere with some account
of how it could be learned (presumably from experience if naturalistic epistemology is
sound; but note that any epistemology will have to be self-referential, or reflexive) (see
Williams, 1980). Applying this condition, we may conclude that Tl coheres better with
an epistemology we have reason to believe, than T2. Herein lies the basis for plausibility
judgments among hypotheses.
There are other formulations of, and variations on, the above conditions which are
based on the work of Lycan (1988). For example, Evers and Lakomski (1991) have
undertaken a systematic application of coherence justification to competing theories ofeducational administration. And Thagard (1988) has attempted to explain the growth
of scientific knowledge, especially the replacement of one theory by another, using
coherence criteria. However, because of the global nature of coherence justification,
theory adjudication is a complex and often controversial business, even when simplified
by the acceptance of a great deal of shared background or touchstone theory. Judgments
of validity will be thus likewise controversial or provisional, mirroring the epistemic
status of the theories such judgments are about. Note that it is not the term “validity”
which is here being used equivocally; it is embedded within a coherentist realist
epistemology. Rather the uncertainty is over what knowledge, or theory, is most
likely to be true. But the answer to this question will depend in turn on improvingour understanding of the epistemology of theory choice, with gains for the coherentist
accruing to naturalistic approaches to human knowledge acquisition.
Perhaps the most interesting recent work on the task of naturalizing coherence
justification is in the area of cognitive neurobiological modeling. That most powerful
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 34/97
534 C. W. EVERS
of epistemic engines, or learning systems, the human brain, appears to learn by
applying holistic soft constraints on global neural network representations of knowledge,
suggesting there is much to be learned about knowledge and its justification from
neuroscience (Churchland, 1986; Thagard, 1989). Here, however, we need to look atnew ways of representing knowledge and its dynamics - not the familiar sentential
representations, but in terms of neural network geometries and connection strengths
between neurons. Much naturalistic epistemology in this vein is occuring in the study
of parallel distributed processes - the mathematical modeling of neural networks
(Rumelhart & McClelland, 1986; Muller & Reinhardt, 1990. For an introduction to
these ideas in education, see Evers, 1990).
Having recognized the limits of logical empiricism and more recently paradigms
theory, philosophers, methodologists, and cognitive scientists now have the opportunity
to study the microstructure of cognition, the fine details of epistemically progressive
global and local belief change. However, this is only the beginning of a full coherentist
account of theory validation. To steal a line from Donald Davidson, it’s nice to know
that we won’t run out of work.
References
Bates, R. (1983). Educati onal admini strat ion and the management of know l edge. Geelong: Deakin University
Press.
BonJour, I_. (1985). The struct ure of e~~~r ~e~~ now ledge. Cambridge, MA.: Harvard University Press.
Campbell. D. T. (1977). Descriptive epistemology: psycholo~~al, sociological and evolutionary. The ~iZZiamJames Lect ures. Harvard University. Cited as reprinted in Campbell (1988).
Campbell, D. T. (1984). Can we be scientific in applied social science? In R. F. Connor, D. G. Attman, & C.
Jackson (Eds.), Evaluat i on Studi es Revi ew Annual (pp. 2648). Cited as reprinted in Campbell (1988).
Campbell, D. T. (1986a). Science’s social system of validity-enhancing collective belief change and the
problems of the social sciences. In D. W. Fiske & R. A. Schweder (Eds.), M etut heory i n soci al science:
Pluralism and subjectivit ies (pp. 108-135). Chicago: University of Chicago Press. Cited as reprinted in
Campbell (1988).
Campbell, D. T. (1986b). Relabeling internal and external validity for applied social scientists. In W.
M. K. Trochim (Ed.), Adv ances in quasi -experi mental design and analy sis (pp. 67-77). San Francisco:
Jossey-Bass.
Campbell, D. T. (1988). M ethodol ogy and epist emal ogy for soci al science. Chicago: University of Chicago
Press.
Campbell, D. T. & Stanley, J. C. (1963). Experi mental and qi l asi- experj mental designs for research. Chicago:
Rand McNally.
ChurchIand, P. M. (1979). ~c~ent~ ~c eali sm and the pfast~cit ~l f mi nd. Cambridge: Cambridge University
Press.
Churchland, P, M. (1985). The ontological status of observables: In praise of superempirical virtues. In P.
M. Churchland & C. A. Hooker (Eds.), fmages ofscience.Chicago: University of Chicago Press.
Churchland, P. S. (1986). Neurophilosaphy. Cambridge, MA.: M.I.T. Press.
Cook, T. D. & Campbell, D. T. (1979). Quasi- experi mentat ion: Desi gn and analy sis i ssues for field sett ings.
Chicago: Rand McNally.
Cronbach, L. J. (1988). Five perspectives on the validity argument. In H. Wainer & H. I. Braun (Eds.),
Test val i d i tv (pp. 3-17). Hillsdale, N.J.: Lawrence Erlbaum.. . .Cronbach, L. J. & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bullet in,
52, 281-302. Cited as reprinted in C. I. Chase & H. G. Ludlow (Eds.) (1966). Readings i n educati onnl
and psychol ogical measurement . New York: Houghton Mifflin.
Evers. C. W. (19901. Educating the brain. Educati onal Phi l osophy and Theorv . 22C2). 65-80.
Evers, C. W. & Lakomski, G.71991). Know ing educati onal adzmimstrati on. Oxford: ‘Pergamon Press.
Feigl, H. (1950). Existential hypotheses. P~~~as~p~~~ fScience, 7(l), 3562.
Feyerabend, P. K. (1962). Explanation, reduction, and empiricism. M innesota Studi es in the Phi losophy of
Science, Vol. 3. Minnea~lis: University of Minnesota Press.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 35/97
Beyond Paradigms 535
Foster, W. (1986). Paradi gms and prom i ses. Buffalo, N.Y.: Prometheus Books.
Hanson, N. R. (1958). Patt erns of discovery. Cambridge: Cambridge University Press.
Hempel, C. (1965). Aspects of sci enti fi c explanat ion. New York: Free Press.
Hempel, C. (1966). Phil osophy of natural science. Englewood Cliffs: Prentice-Hall.
Hooker, C. A. (1975). Philosophy and meta-philosophy of science: empiricism, Popperianism and realism.Synrhese, 32, 177-231.
Kerlinger, F. N. (1964). Foundati ons of behavi oral research. N ew York: Holt, Rinehart and Winston.
Kuhn, T. (1962). The structure of scienufic revolutions. Chicago: University of Chicago Press.
Lather, P. (1986). Issues of validity in openly ideological research: Between a rock and a soft place.
I nt erchange, 17(4), 63-84.
Lincoln, Y. & Guba, E. (1985). Natural ist ic inquiry. Beverly Hills: Sage Publications.
Lycan, W. G. (1988). Judgemenr and j usti fi cati on. Cambridge: Cambridge University Press.
Messick, S. (1988). The once and future issues of validity: Assessing the meaning and consequences of
measurement. In H. Wainer & H. I. Braun (Eds.), Test val i d i ty (pp. 33-45). Hillsdale, N.J.: Lawrence
Erlbaum.
Muller, B. & Reinhardt, J. (1990). Neutral netw orks: an int roducti on. Berlin: Springer.
Papineau, D. (1979). Theory and meaning. Oxford: Clarendon Press.
Popper, K. R. (1959). The logic of scienti fi c discovery . London: Hutchinson.Popper, K. R. (1963). Conjecture and refutat ions. London: Routledge and Kegan Paul.
Quine, W. V. (1953). Two dogmas of empiricism. In W. V. Quine (1961), From a logical point of view
(pp. 20-46). Cambridge, MA.: Harvard University Press.
Quine, W. V. (1957). The scope and language of science. In W. V. Quine (1976), The way s of paradox and
other es.ruys (pp. 228-245) (second edition, enlarged). Cambridge, MA.: Harvard University Press.
Quine, W. V. (1960). Posits and reality. In W. V. Quine (1976), Th e w ays of paradox and ot her essays
(pp. 69-90). New York: Columbia University Press.
Quine, W. V. (1969). Epistemology naturalized. In W. V. Quine Ont ological relat iv it y and other essays
(pp. 69-90). New York: Columbia University Press.
Rumelhart, D. E. & McClelland, J. L. (Eds.), (1986). Parallel distributed processing, Vols 1 and 2.
Cambridge, MA.: M.I.T. Press.
Standards for educat i onal and psychol ogical t ests and manual s (1966). Washington, D.C.: American
Psychological Association.Technical r ecommendati ons for psychol ogical t ests and di agnosti c t echniques (1954). Psychol ogical Bul l et i n,
51: 201.-238.
Thagard, P. (1988). Computat ional phil osophy of sci ence. Cambridge, MA.: M.I.T. Press.
Thaaard, P. (1989). Exolanatorv coherence. Behavioral and Brain Sciences, 12. 435-502.
Watanabe, S: (1969). I ?now inghnd guessing. New York: John Wiley.
Walker, J. C. & Evers, C. W. (1982). Epistemology and justifying the curriculum of educational studies.
Bri t i sh Journal of Educati onal St udi es, 30(2), 213-229.
Walker, J. C. & Evers, C. W. (1988). The epistemological unity of educational research. In J. P. Keeves
(Ed.), Educati onal research, met hodol ogy and measuremem: An i nt ernat i onal handbook (pp. 28-36).
Oxford: Pergamon Press.
Williams, M. (1980). Coherence, justification, and truth. Review of M et aphysi cs, 34(2), 243-272.
Biography
Colin W. Evers holds degrees in mathematics, philosophy, and education, and teaches
in the School of Graduate Studies, Faculty of Education, Monash University, Australia.
His teaching and research interests are in philosophy of education, educational research
methodology, and educational administration.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 36/97
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 37/97
CHAPTER 3
POLICY ANALYSIS: PRACTICAL REASONOR EMPIRICAL SCIENCE?
GABRIELE LAKOMSKI
School of Education, The University of Melbourne, Parkville, Victoria 30.52, Australia
Abstract
This chapter examines William Dunn’s tr ansactio~ I model of argument which utilizes
prectical reason as its methodological framework for policy decision-making. It is argued
that this formal model, offered as an alternative to scientific policy analysis, fails subject
to its assumption of knowledge and its acquisition as stipulated by the folk psychologica1
sentential view of learning. Since recent empirical evidence does not support this view, amore defensible methodological framework for policy analysis is suggested based on the
causal account of knowledge, i.e., the way real brains acquire and process information.
The Methodological Dispute in Policy Analysis
methodological concerns have played a central role in public and educational policy
analysis for as long as these disciplines have existed as discrete fields of inquiry and
professional practices (Boyd, 1988; Garson, 1986; Mitchell, 1984). Amongst the many
problems still awaiting solutions none is more persistent than that of rationally selecting
“best policy” under conditions of imperfect knowledge and uncertainty. Traditionally,
Laswellian policy science (Lasswell, 1951; Lerner & Lasswell, 1951), or the ~y~o~~~c
~~u~~~~u~, ought to attack the problem by taking a global, historical view of the
policy context, and to study the conditions of social change with systems analysis
as its preferred theoretical framework and a narrow empiricism as its methodology
(Garson, 1986, p. 538).
While equally mindful of the complexity of social life, defenders of the anti-synoptic
tradition argued for an incremental, piecemeal approach to policy analysis (Braybrooke
& Lindblom, 1963) defended pluralism as its theoretical framework, and emphasized
case and contextual studies as appropriate methodologies. The major concern of
anti-synoptic writers was, and continues to be, however the alleged value-neutrality
of Laswellian policy science, a concern which has subsequently led to question the
status of the field as a science, as well as the role of policy analysis in a democratic
society*
One consequence of this development is that some analysts, motivated in part by
reports of the negligible and even negative impact of policy research (e.g., Dunn, Mitruff,
&. Deutsch 1981) simply declared the science question dead: “Resolved: Policy analysis is
537
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 38/97
538 G. LAKOMSKI
not a sci ence, is not scientific; indeed, scientific status is an inappropriate goal for policy
analysis” (Landsbergen & Bozeman, 1987, p. 625). Another consequence of the currently
prevailing anti-science mood is the rise of alternative models of policy analysis which
align themselves directly with postpositivist, interpretive social science (e.g., Callahan& Jennings, 1983; Jennings, 1983, 1987). Specifically, these anti-synopt ic writers argue for
the inclusion of political and ethical values, as well as the social context of policy analysis
generally (e.g., Dunn, 1983; Fischer & Forester, 1987; MacRae, 1976; Wildavsky, 1979)
and advocate that naturalistic approaches be adopted as the most appropriate to deal
with these issues (e.g., Lincoln & Guba, 1986). Combining many of the concerns of
both the older anti-synopt ic tradition as well as its contemporary interpretive expression
is the policy analysis model suggested by Dunn.
Dunn’s t ransact i onal model of argument (Dunn, 1981, 1982), based on Toulmin’s
practical logic (Toulmin, 1958), is a recent and explicit attempt at providing a
procedural model for rationally enabling policy choice. It represents a special case
of the methodological paradigm which largely characterizes the newer approaches to
policy analysis, that of practical reason with its emphasis on rhetoric, persuasion, and
legal reasoning. Since policy makers and analysts are meaning constructing agents, so
the argument goes, it is the epistemological feature of intentionality which provides
the real grounds to understand social complexity and thus facilitates superior policy
making. Unlike the “statistical empiricism” of Laswellian policy science, the alternative
methodological proposal of practical reason is believed to be antithetical to empirical
scientific methodology and to lend itself neither to the formulation of laws, nor to
causal explanation. It is normative, not descriptive, more concerned with the adequacy
and cogency of policy arguments than their truth.
But the claim of practical reason to represent human understanding in its essence
is a matter of the soundness of thp epistemology underwriting it. Following Quine’s
(1969; Quine & Ullian, 1978) “epistemology naturalized” argument, what makes for
the soundness of an epistemology is that it be learnable; that is, its embedded theory
of the mind must present an adequate account of human cognition and learning. An
epistemology, then, is only as valid as its implied theory of learning, and there is no
principled distinction to be made between the former and the latter.. Epistemology
is continuous with natural science (see Walker & Evers, 1982 who introduced this
argument into [philosophy of] education; Evers, 1987; Walker & Evers, 1988). In the
present context, this means that the account of learning embedded in practical reason
must be able to explain how we have come to know about such abstract objects as
purposes and beliefs which are taken to function in the explanation of behavior.
It is the purpose of this chapter to argue that the epistemology underlying Dunn’s
procedural model, the classical Justified True Belief (JTB) account, is unsound by way
of arguing that its implied view of learning and cognition, characterized by the “folk
psychological” (Stich, 1983) propositional attitudes, is mistaken. Showing this requires
demonstrating first that folk psychology is an empirical theory, and that it subsequently
can be assessed in the manner of all theories (P.M. Churchland, 1981, 1988a, 1989b;
P.S. Churchland, 1989). When evaluated as such, folk psychology’s view of knowledgeturns out to be sentential. But according to the neurosciences whose business it is
to explain brain functioning, there is no support for the assumption that the brain
functions primarily as a linear “sentence-cruncher”, to use P.S. Churchland’s (1989)
colorful phrase. The consequence for Dunn’s model is, then, that policy choice as an
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 39/97
Beyond Paradigms 539
example of intelligent, rational action has little to do with overt linguistic behavior,
regardless of the form this takes. If so, it would seem advisable to hold the folk
psychological categories of belief and desire at arms length and concentrate on how
we in fact process information as a precondition for understanding policy choice as a
specific example of cognitive action.
Policy Analysis as Reasoned Argument
In his introduction to the “Symposium on Social Values and Public Policy” (Dunn,
1980-81), Dunn states that the policy sciences need to be based on a perspective which is
neither value-neutral nor value-committed. The policy sciences must not be value-neutral
since value-neutrality would deprive policy of political significance and take it back to
the canons of positivism and the “empiric0-analytic sciences”. To be value-committed,
on the other hand, is equally undesirable because this would deny the policy sciences
of “reasoned ethical discourse” which is Dunn’s central concern. In his view, policy
knowledge can only grow through open critical discourse which, in turn, presupposes
rules or standards of assessment that enable the participants in policy making to examine
rival ethical claims without anyone thus dominating the outcome (Dunn, 1980-81, p.
519). These rules or standards are spelt out in the t r ansact ional m odel ofargum ent which
is Dunn’s proposal for operationalizing pr actical r eason directed towards establishing the
adequacy and cogency of claims to knowledge rather than their truth. (Dunn, 1981, 1982)
Following Toulmin, there are no context-independent criteria of judging the merits of an
argument, and this is as should be, according to Dunn, given that policy analysis, as a
practical science, is carried out in specific contexts.
Dunn offers his model as a reply and an alternative to the kind of social-scientific
experimentation advocated by Campbell (1969) in his “experimenting society”. Given
that reforms are “symbolically mediated and purposive social processes”, Dunn believes
they are more akin to arguments than to quasi-experimentation, or scientific experi-
mentation generally, which directs questions to nature directly. The strength of the
t ransact ional model, as Dunn sees it, consists in broadening the range of standards
thought suitable for challenging and assessing knowledge claims, including a number
of specific tests which help determine their adequacy, cogency and relevance. Thesetests also have the function of “plausible rival hypotheses” to the knowledge claims
under examination. The model which best represents the way in which agents settle
competing claims in social contexts is that of jurisprudential reasoning. Its standards
include, amongst others, rules for making valid causal inferences. As a consequence,
and unlike the replication of experiments, argumentation leads “toward a pragmatic and
dialectical conception of truth . . . Knowledge is no longer based on deductive certainty
or empirical correspondence, but on the relative adequacy of knowledge claims which
are embedded in ongoing social processes” (Dunn, 1982, p. 94).
Of central importance in the t r ansact i onal model is the distinction between analytic
and substantial arguments which Dunn takes over from Toulmin (1958). Toulmin arguesthat formal logic is too narrow to be of any use in the practical assessment of arguments.
Our claims to knowledge are sound when our supporting arguments are adequate. What
is to count as an adequate argument is dependent upon the field from which it is taken:
“validity is an intra-field, not an inter-field notion” (Toulmin, 1958, p. 255).
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 40/97
540 G. LAKOMSKI
The merits of a procedural schema, based on the above described assumptions and
proposed to guarantee rational policy choice if followed conscientiously, are said to
be such that it allows (1) for the visual representation of the systematic structures
of competing arguments; (2) for the critical analysis of the frames of reference orideologies of those contesting knowledge claims; and (3) for the development of “truth”
and “utility” tests by means of which knowledge claims can be tested. Finally, and most
importantly, it allows for democratic knowledge creation via the critical exchange and
challenge of knowledge claims in communicative action. The model is composed of
six elements which are data (D), claim (C), warrant (W), backing (B), rebuttal (R),
and qualifier (Q). The first three elements parallel those of the classical syllogism, to
be supplemented by the second set. Backing (B) denotes additional data, claims or
arguments introduced in case the warrant is in doubt (Dunn, 1982, p. 96). Warrants
provide reasons for the acceptance of a claim (Dunn, 1981, p. 42). Rebuttal (R)
specifies conditions under which the adequacy or cogency of a knowledge claim can
be challenged. Policy claims and rebuttals together “form the substance of policy issues,
that is, disagreements among different segments of the community about alternative
courses of government action” (Dunn, 1981, p. 42). Qualifier (Q) serves to indicate the
degree of cogency or force of a claim.
Following the de~nition suggested by Weiss and Bucuvalas (1980), Dunn (1982)
considers truth tests to be “decision points concerning evidence; the grounds for
accepting or rejecting truth claims include . , . empirical as well as formal rational tests”
(p. 100). Relevance or utility tests, on the other hand, “are decision points concerning the
delineation of an appropriate domain of inquiry or action”, a definition Dunn expands to
include “the explicit or implicit purposes of knowledge claimants or their challengers”
(p. 101). Truth tests which appraise the adequacy of a knowledge claim and challenge
its causal assumptions are considered more problematic. Dunn suggests a classification
of truth tests which takes account of alternative modes of explanation (von Wright),
different knowledge-constitutive interests (Habermas), and competing standards for
assessing ethical claims (MacRae). Specifically, Dunn (1982, p. 105) proposes that
there should be different truth tests for different “knowledge transactions”, presumably
based on the assumption that there are different kinds of knowledge to be transacted
which have their corresponding purposes. These are e~pirico”a~aZytic, concerned with
logical consistency of laws, etc. and/or their correspondence to observed regularities;interpretive, directed toward human purposes, reasons, and motives; pragmatic denoting
effective past action; authori tat ive, relating to the well-established status of those
producing knowledge as well as its general acceptance, or the use of approved methods;
lastly, crit ical in the Habermasian sense of liberating human actors from unexamined
doctrines. According to Dunn, we learn what counts as an adequate argument under
specific conditions in specific professions, knowledge is fallible and corrigible. This is
quite in keeping with Toulmin’s epistemological prescription to ignore scepticism and
moderate our ambitions (Toulmin, 1958, p. 248).
Appl yi ng t he Procedure
It is easy to see why Toulmin’s applied logic seems attractive to policy writers
such as Dunn (and also Mason & Mitroff, 198CrSl) since here we appear to have
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 41/97
Beyond Paradigms 541
a sound rationale for including the ordinary arguments of practitioners and those
of the policy experts and scientists thus satisfying the democratic ideal central to the
anti-synoptic tradition. The model is all the more attractive since it is also combined
with an assurance that the transactional model is not only context-sensitive, but alsovalue-critical and intersubjectively valid. That is, it avoids the problems associated with
straight-out partisan approaches without having to compromise on the right (democratic)
values. Furthermore, we are also presented with a formal set of procedures by means of
which we could settle any kind of claim or argument and thus guarantee policy choice.
But can we?
Dunn acknowledges that stakeholders, subject to their different worldviews,
ideologies, and frames of reference, are bound to challenge the presuppositions
used to back up a warrant, that is, the reasons provided for a claim. He does not
consider such challenges a disadvantage, but points out that in the empirico-analytic
and hermeneutic sciences such questioning is not even possible, and it is this special
feature which makes his model unique. While the potential for publicly challenging
knowledge claims may be considered an advantage (accepting for the sake of the
argument Dunn’s claim regarding the alleged inability of both natural science and
hermeneutics to challenge assumptions), whether or not it is of value depends on
the model’s ability to settle the challenge, to determine what is to count as the
more adequate argument, given its own theoretical resources. Since policy analysis
in Dunn’s view aims “t o produce and t ransform poli cy-relevant informat ion that may
be ut il i zed in poli t i cal set t i ngs t o resolv e poEi cy problems” (Dunn, 1981, p. 35) the
model’s claims stand or fall with the accuracy of its account of knowledge production
and the practical performance it is said to guarantee.
Some problems are readily apparent. It is a commonplace to note that in the
empirical world power, learning, and linguistic differences are distributed unevenly,
disadvantaging some from the outset, and that gender, race, as well as age play a
significant role in who gets to talk in social situations, and who is listened to, and
who is included in the deliberation process in the first place. In addition, the implied
assumption that all are committed to rational debate, that all share the same willingness
to declare their ideologies or standpoints honestly, and - by implication - that none
are thereby disadvantaged, is quite unrealistic.
For his model to work, Dunn must presume that power and inequality are suspendedin order for “good will” to prevail; that our standards and levels of learning and policy
relevant knowledge are even; that we are all equally capable of putting our points of
view; and that gender differences, race and age are irrelevant; in other words, that
we live in a perfectly rational world. Here Dunn’s model resembles Habermas’ ideal
speech situation which suffers from similar problems (Lakomski, 1988). Central to the
assumption of perfect rationality is the belief that human agents are in fact capable
of “knowing their own minds”: their beliefs, thoughts and intentions which comprise
self-knowledge, and secondly, that this self-knowledge is fundamentally linguistic. Only
then is it reasonable to assume that rational deliberation could work at all. But before we
consider the relation between beliefs and behavior directly, let us first examine whetherthe decision procedure suggested by Dunn permits a policy choice to be made following
its own prescriptions. Consider the following example.
According to recent research (Gill, 1988), girls achieve better academically when
they are in single-sex rather than co-educational classrooms (D). Therefore, we might
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 42/97
542 G. LAKOMSKI
conclude that it is definitely better (Q) to educate girls in single-sex classrooms or
even schools (C). Our warrant or reason for (C) is that the separation of boys and
girls for purposes of academic instruction caused higher academic achievements for
girls than had been the case when they were taught in co-educational settings. Wecould further back our warrant by saying that equality of opportunity is a fundamental
principle which demands single-sex classrooms in the interests of academic achievement
of females. We could further argue that whatever increases the academic achievements
of any underprivileged group ought to be done, females being one such group. Recall
Dunn’s point that the empirical evidence is rarely in doubt, but that the underlying
reasons and assumptions are challengeable. Suppose, then, that a group of parents
opposes the claim that it is better to educate girls in single-sex classrooms on the
assumption that short-term gains (higher academic achievements) do not outweigh
longer-term social consequences such as presumed inadequate socialization. The parent
group thus advocates co-education, rather than single-sex education, on the assumption
that learning together is more important in terms of ultimately overcoming sexism than is
segregation, even if such segregation yields better academic performance for the girls in
the short term. We thus have two competing claims and must choose between competing
policy prescriptions. How do we settle the conflict?
On Dunn’s own account, the relevance and cogency of any person’s or group’s
knowledge claim is always context-dependent. In the present example, we have two
different “contexts” in terms of two different feminist theories. Following the logic of
Dunn’s model, what is relevant and cogent to a separatist feminist policy maker may not
at all be relevant and cogent to a parent of a different feminist persuasion who believes
that separatism is a mistaken strategy for overcoming sexism and its practices. On their
own accounts, each theory is equally valid. But since Dunn is interested in producing
and transforming “poli cy-relevant informat ion t hat may be uti l i zed in poli t ical sett ings to
resolve policy problems” the question of which claim is the more “adequate” must be
settled. This is to be done by truth tests, by introducing “different sets of assumptions
and underlying presuppositions” (Dunn, 1982, p. 103) which might challenge a claim’s
causal assumptions. But truth tests themselves have, as we saw earlier, their own
relevant contexts. These are ultimately the different kinds of knowledge presumed
to underlie the different purposes for “knowledge transactions”. But if truth tests are
thus tied to their own respective spheres of influence, then, by definition, there areinfinitely many, equally true arguments or theories. But since not all claims can be
true, and if we cannot decide between true and false claims, we cannot decide the
status of any claim. Knowledge is impossible and we end up with incommensurable
theories. For if competing arguments, as well as relevance and cogency tests, are
dependent on their contexts, and if there are specific truth tests for specific types of
“knowledge transaction” based on different kinds of knowledge, then the standards of
appraisal cannot possibly serve to arbitrate between competing claims. Both arguments
and standards of appraisal are based on their respective knowledge foundations, and
there is in principle no difference between the arguments to be assessed and that which
is supposed to be doing the assessing. Not only can competing claims not be settled sinceeach is restricted by its own epistemic sphere of influence (the principle of “intra-field
validity”), but claims within the same epistemic community face difficulties in terms
of the support or evidence they themselves can marshal1 since whatever warrant or
backing may be brought up is only as valid as the presumed conception of knowledge
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 43/97
Beyond Paradigms 543
underwriting them, and its truth or validity cannot be presumed a priori. This means in
the present example that no amount of new reasons or arguments offered, either for or
against the above feminist claims, could in principle settle the issue of which educational
policy to adopt. As a consequence, the t ransact i onal model of argument makes policydeliberation both relativist and arbitrary. Given the example presented here, efforts toovercome sexism and its practices become a matter of subjective preference which can
hardly rate as a satisfactory policy prescription. The inability to make a choice is the
direct result of the model’s epistemological assumptions which are made more explicit
in the following section.
The Regress of seasons
As a procedure of justification, Dunn’s model is faced with the threat of an infiniteregress of reasons, characteristic of the JTB account of knowledge (Armstrong, 1981;
Williams, 1977, 1980), which in Evers and Lakomski’s (1991) definition is expressed as
follows:
Person x knows that p, where p is some particular claim to knowledge, if and only if
(i) p is true
(ii) x believes that p, and
(iii) x is justified in believing that p for the reason q (p. 5).
Displaying the justificatory structure of JTB shows how neatly Dunn’s decision modelparallels it. Of particular significance in the transactional model, as well as the JTB
account, is condition (iii) because for q to be a justifying reason, it must itself be an
item of knowledge, just as backings and warrants must be to support an argument. But
if q is already an item of knowledge, then JTB’s definition is circular since it already
contains an appeal to knowledge in the definiens. But more importantly, an infinite
regress threatens since q can only be known by virtue of knowing something else. Faced
with the problem of circularity as well as the infinite regress, the solution has been sought
in the distinction between derived and immediate knowledge. JTB, then, is an account of
derived knowledge, of claims known in virtue of being implied by further knowledge. The
chain of implications is not infinite, however, as it stops at knowledge that is immediate
or underived. It follows on this view that all our derived knowledge must rest on, or be
derivable from, some foundation of immediate knowledge, traditionally the so-called
sense-data, first person sensory reports, and observation statements. Dunn’s acceptance
of a plurality of foundations is a variation on the theme that knowledge is believed to be
in need of foundations, albeit different ones. This multi-foundationalist view has strong
similarities with the paradigms’ view (Walker & Evers, 19SS), although Dunn does not
employ that concept.
In line with many interpretivist social scientists and philosophers who followed Kuhn’s
and Feyerabend’s lead in raising decisive objections against logical empiricism, Dunn
accepts that (social) reality is a matter of interpretation (with physical reality being a
matter of direct observation), and that (social) theories are always underdetermined by
data or evidence (Dunn, 1982, p. 104). As a consequence, he also accepts that no theory
can ever be justified evidentially which leads to the conclusion that all theories are equally
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 44/97
544 G. LAKOMSKI
plausible or acceptable. There is thus no such thing as a true theory in the social sciences,
and objectivity becomes unobtainable. But this conciusion only follows if evidence equals
empirical evidence only. In addition to the fact that theories are always underdetermined
by the available evidence, we also know that what is taken as empirical evidence is alwaystheory-laden. So in comparing theories we cannot simply rely on empirical evidence as
the neutral arbiter, as our “touchstone” (Walker & Evers, 1988). Since we are now in
the business of actually comparing theories (observation being as theoretical as anything
else), and since we know what makes for better or worse theories, the solution is that
we use their traditional virtues, i.e., relative coherence, simplicity, and explanatory unity
- P.M. Churchland’s (1985) “superempirical virtues” - to argue for a global assessment
of theories which includes empirical evidence as one criterion amongst several. On this
account, then, the best theory is one which coheres best with our interpreted experience
(BonJour, 1985; Evers & Lakomski, 1991, pp. 37-44; Williams, 1980). Additionally,
Dunn’s model relies on a theory of meaning in which terms are entirely determined by
their embedding conceptual framework, an impossible position to hold. For if meaning
were entirely determined by its conceptual role, then we could never have learnt either
of the feminist theories mentioned above. Nor could we have learnt about such abstract
objects as intentions, thoughts, and reasons. Theories are complex networks of sentences
which we do not learn ail at once. We must have begun by learning simple terms or
expressions first, but in order for us to understand these and their role in the theory,
knowledge of the whole theory is already presupposed or we would not be able to assign
them the function we do. Here we see that JTB promises more than it can deliver given
the theoretical tools it has at its disposal. In presuming learning capacities - a theory of
the mind - which postulate antecedents to learning which are themselves unlearnable,
it comes out unknowable on its own account: it fails to be self-referential.
The Theoretical Character of Belief and Desire
In specifying conditions for claims to count as knowledge, an epistemology implicitly
presumes a theory of the mind. What we can know crucially depends on the requisite
perceptual and cognitive capacities which we developed as a species. And we have
to learn from infancy onwards what these capacities are. We do not just “have”observational judgments, they have to be learnt beginning with the prior learning of
making complex perceptual discriminations, as Campbell (1974) and Popper (1981), for
instance, have long argued. We also have to learn the requisite linguistic or propositional
system within which we constitute our beliefs, desires, and thoughts. Since we first have
to acquire that, it cannot itself be the medium of learning, leading to the conclusion that
there has to be a type of learning prior to that involving the manipulation of sentences
(P.M. Churchland, 1989b, p. 155). And here our commonsense understanding of
ourselves fails to provide an answer. In order to demonstrate where folk psychology falls
short, it is necessary to make explicit what has been implied in the preceding discussion,
namely, that it is an empirical theory, and one which - on the current evidence - mightneed to be replaced. Here the Churchlands’ work provides the most telling arguments.
The propositional attitudes, so called because they express a distinct attitude towarda specific proposition (P.M. Churchland, 1988b, p. 63), i.e., the thought that (gum trees
are lovely; the belief that (humans act rationally); t he fear t hat (the Middle East crisis
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 45/97
Beyond Paradigms 545
will turn into war) have long been taken to play a causal role in human behavior
and are central to folk psychology. Human conscious intelligence, according to folk
psychology, is said to consist of the rational manipulation of these propositions by
means of deductive inference. But rather than being the “sort of super-causal logical
relation” (P.M. Churchland, 1988a, p. 213) they are alleged to be, the propositional
attitudes display relationships typical of all theoretical explanations, as can be seen
by comparing them to the logical pattern formed by physical laws. If the latter can
be characterized by the “numerical attitudes”, both sets can be described as follows
(P.M. Churchland, 1988b, p. 64; 1981, pp. 70-71):
Propositional atti tudes Numerical atti tudes
. . . believes that p . . . has a length, of n
. . . suspects that p . . . has a kinetic energyj of n
Either “attitude” can be completed by putting a proposition in place of p or a number
in place of n. Only then do we have a determinate predicate. The logical relations
holding between numbers and numerical atti tudes also hold between propositions and
propositional atti tudes. Most importantly, where these relations hold universally, we are
in a position to state laws - for the latter set of relations as much as for the former,
numerical ones. In other words, we utilize the abstract relations which hold in the domain
of certain abstract objects such as numbers, vectors, or propositions in order to help us
state the empirical regularities between real states and objects, and that includes those
between various kinds of mental states. Summing up, the full-blooded intentional idiom,
contrary to popular opinion, possesses the same complex logical structure as the rest of
our scientific theories.
An important consequence of folk psychology’s empirical character is that it might
actually be false. This possibility, however, seems quite counterintuitive because we
appear to get by rather well by relying on its categories in terms of explaining and
predicting everyday behavior with a fair degree of success. Even if folk psychology
were false, would it matter for our practices, including those which comprise “policy
analysis”‘? And how would we then explain our behavior? These are difficult questions
which cannot be answered at present. But to see that they are real, and that much
depends on the answers, let us consider some of the fundamental problems for folkpsychology which lend support to the claim that it might be a false empirical theory.
Probl ems f or t he Sent ent i al Paradi gm
In the following section, I shall briefly note two major problems of the view that our
cognitive activity consists in the manipulation of sentences (for the original definition
of the sentential view see P.S. Churchland, 1980, 1989. See also Hacking’s 1975 accountof the reasons for “lingualism’s” favored philosophical status). The first is what Stich
(1983, p. 214) and P.S. Churchland (1989, p. 388 onwards) call the “infralinguistic
catastrophe”: much intelligent behavior is displayed by organisms who do not have any
overtly linguistic capacity, and that includes behavior by human infants, adults suffering
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 46/97
546 G. LAKOMSKI
from left-hemisphere lesions, as well as deaf mutes (see P.S. Churchland, 1983). To
explain this, one could of course argue that this behavior is not really “cognitive” in
the required sense. But this would beg the question since it is already assumed that
intelligent behavior is identical with linguistic behavior. Secondly, while granting that the
above is intelligent behavior although it does not show itself linguistically, one prominent
suggestion maintains that it is based on a “thought language”, Mentalese, in which
organisms devoid of overt linguistic behavior reason and solve problems (Fodor, 1975).
The assumption that organisms possess such a language of thought which, according to
Fodor, is a proper language with all relevant characteristics seems far-fetched, and its
existence in principle impossible to ascertain since it is radically unlike the linguistic
capacity we display as adults. Its categories can be reached only via those of our
language, making the point of independently establishing them moot. (For a fuller
discussion of Mentalese see P.S. Churchland, 1980; also Stich, 1983, pp. 187-197.) But
the more interesting and important issue here is that of how infants learn our language
if Mentalese is assumed to be the template. If a child acquires, for example, German
by matching a German word with the appropriate word in Mentalese, the child can only
learn those German words for which there are words in Mentalese. If matching is the
process, then one must conclude that the child does not in fact learn any new concepts
at all in addition to those pre-existing in Mentalese. More absurd still is the further
implication that all the new concepts gained through scientific discoveries must have
already been present in Mentalese from the beginning, terms such as quarks, electrons,
atoms, neurons, etc. (P.S. Churchland, 1980). The existence of a “language of thought”,
then, would seem to stretch the bounds of credibility too far. Given these problems
which have been noted albeit briefly, it is more reasonable to assume that cognitive
activity does not equal “sentence-crunching”, an assumption which, inter alia, enjoys
the theoretical virtue of simplicity.
The second problem for the sentential view relates to that of accessing knowledge.
For an organism to survive, it is essential that it have speedy access to knowledge
relevant to respond to “the four F’s: feeding, fleeing, fighting, and reproducing”
(P.S. Churchland, 1987, p. 548). This goes as much for the cat’s recognition of a mouse
as food as it does for a driver’s recognizing a red light as signalling “stop”. Essential
in either situation is that the relevant stimuli are recognized instantly. But how does the
relevant information processing system know which bits of information to call up? Whatis its mechanism of sorting so that the correct response follows? The problem is far from
innocent, notwithstanding the fact that we manage to respond by and large correctly
and instantly to such stimuli in our everyday commerce with the world. Since this is
so, calling up our mental store of sentences/beliefs in order to find those which may
apply in a specific situation, and systematically eliminating those which do not by the
relevant logical rules, would present a sure recipe for evolutionary disaster since such
sorting would require a lot of time. Besides, the sentential/belief view presumes that we
do have immediate access to our mental states, an assumption not born out by research
in social and cognitive psychology (for relevant research see Nisbett & Wilson, 1977).
Borrowing Hanson’s “there is more to seeing than meets the eye-ball”, we may now saythat there is more to cognition than is expressed in sentences. But if the structures of
knowledge are not sentences, then what are they? And if the propositional attitudes do
not really explain what goes on inside our heads, and hence explain our behavior, what
kind of an explanation of behavior could we possibly advance?
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 47/97
Beyond Paradigms 547
Neural Networks and Nonsentential Representation
The beginnings of an answer to both questions are suggested by computational
neuroscience (for a brief overview see Sejnowski, Koch, & Churchland, 1988) inwhat is termed “Connectionism”: the mind/brain’s capacity for the parallel distributed
processing of information, or PDP models of brain functioning (for an introduction
see P.M. Churchland, 1988, Ch. 7; P.S. Churchland, 1989, Ch. 10; a full account is
Rumelhart & McClelland, 1986. A first application of the network account in education
is given by Evers, 1990). Secondly, given the PDP account, a most promising alternative
to the folk psychological sentential explanation of behavior is the prototype activation
model developed by P.M. Churchland (1989a). In its simplest form this means that
instead of calling up endless lists of sentences, i.e., rules or laws, the mind/brain “calls
up” a prototype of some cognitive situation (P.M. Churchland, 1988, pp. 217-218).
Churchland acknowledges that the idea of prototypes is neither new nor uncontroversial.
In the present context, it poses the particularly difficult problem of how it is represented
in cognitive creatures such as ourselves if it is not represented linguistically. Here recent
research into the functional properties of neural networks provides an answer. These
are artificial networks which have been constructed to simulate essential features of the
neuronal organization of the brain. What is so remarkable about them is that they have
been extremely successful in learning assigned tasks such as differentiating mine from
rock echoes, recognizing complex visual features, and transforming written text into
speech, NETtalk, to mention just some (P.M. Churchland, 1988, p. 156 onwards).
A simple network has three specific architectural features. It possesses (1) (neuron-
like) processing units, (2) connections between these units, and (3) connection weights
which are the differential strengths of connections between the processing units
(P.S. Churchland, 1987, pp. 550-553; P.M. Churchland, 1988, pp. 156-165). The means
of communication between the processing units are signals such as (neuronal) firing
rate which are numerical rather than symbolic. The bottom, input, layer of units are
something like sensory units directly receptive to environmental input; then follows an
intervening layer, the so-called “hidden units”, and finally the output layer. (In real
brains there are between five and fifty intervening layers.) Each bottom unit emits an
output through its own “axon”. The strength of this output is a function of the unit’s
level of stimulation. Each axon branches out into a number of terminal branches and
sends a copy of that output to each and every “hidden unit”. The bottom units thus
make a variety of synaptic connections with each of the intervening units. The strength
each connection possesses is called its weight. What happens is that the set of activity
levels (input vector) induced by stimulating the input units is transported upward to
the hidden units, changing in the process by the influence of the output function of
the bottom cells, the pre-existing pattern of synaptic weights and the summing activity
within each of the hidden units. A pattern of activations is thus produced at both the
input level and another one at the hidden unit level. What kind of pattern results from
this activity, for a given input, depends entirely on the configuration of synaptic weights
which hit the hidden units (P.M. Churchland, 1989a). The process characterizing activityfrom bottom to hidden layers is repeated from hidden to top layers. The output vector,
then, similarly is the result of the activation pattern generated by the hidden units. This
description makes it clear that the most important characteristic of networks is their
complete interconnectivity. It is possible to construct model networks with whatever
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 48/97
548 G. LAKOMSKI
number of input units, hidden units, and output units we wish. And we can also begin to
appreciate the processing power embedded in the (modest) two-tier arrangement. Most
important of all: the synaptic weights in the overall system can be modified in order to
obtain the vector-to-vector transformation we want, networks can be trained to learn, asthe example of Sejnowski and Rosenberg’s (1987) NETtalk shows. The most remarkable
thing to remember is that networks are not given any rules, laws, or generalizations
prior to or during the learning process. The crucial factor in learning is the value of the
synaptic connection weights which, in turn, are determined by a learning algorithm called
“back-propagation of error” or the general i zed del t a rul e (P.M. Churchland, 1988b, p.
159). Basically, the strategy exploited by the algorithm is “the calculated error between
the actual values of the processing units in the output layer and the desired values, which
are provided by a training signal” (P.S. Churchland, 1987, p. 551). The error signal of
the output layer is fed back to the input layer and is used to adjust each weight in the
network. This process is done countless times (via programing a conventional computer
to act as “teacher”) by feeding the network with diverse examples of Fs, for instance,
and by the network looking for a configuration of weights which will turn the neurons
at the hidden level into a set of complex feature detectors to which the output units will
respond in their turn. In being thus trained up the network slowly learns as the weights
are adjusted after each back propagation of the previous error signal. Learning thus
consists in minimizing the mean squared error over the training set of words, or Fs, or
whatever else was fed into the system. In this manner, the network actually generates
a set of internal representations for the relevant features to be recognized.
Policy Analysis as Naturalized Science
The most important results of studying the functions of artificial neural networks and
PDP for our purposes are (1) t hat l earni ng i s not a mat t er of t he li near processi ng of
symbol s and rul es but is rat her a gl obal, or netw ork, affai r, and (2) t hat consequent l y,
represent at i ons i n t he hi dden uni t s are not symbol s ei t her but are neuronal pat t erns
of act ivat ion. Although neural network modeling is relatively young, leaving many
questions still to be answered, it does suggest that human understanding and the
grounds for action are not to be sought in sets of stored generalizations such as thepropositional attitudes but are rather located in one or more prototypes as these were
earlier described.
If natural language appears to be no more than “a surface abstraction of much richer,
more generalized information processes in the cortex, a convenient condensation fed
to the tongue and hand for social purposes” (Hooker, 1975, p. 217), what follows
for policy analysis? The first result is that reliance on folk psychological categories
as explanatory of policy behavior, since mistaken, ought to be given up. We might
need to continue using them since there is as yet no other medium of representation,
but this communicative function must not be confused with its purported explanatory
function. Indeed, what the preceding thumbnail sketch of brain functioning does show isjust why commonsense or folk psychology is as successful as it is: it allows us to recognize
problematic situations instantly because of the work done at the sublinguistic, neuronal
level, i.e., given appropriate prototypes. Another consequence is that we can concede
that a good policy analyst, as has been observed many times in the literature, may well
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 49/97
Beyond Paradigms 549
be distinguished by his or her “insight” or “intuition” since such expressions can now be
seen as place-holders for the complex computational processes of the brain. We know
indeed more than we can tell, and we are beginning to explain why this is so.
Will knowledge of causal, neurological, detail shape and change the practice ofpolicy analysis in view of the fact that natural language is the only medium we have
for communication? Initially perhaps not. We appear to have to continue using an
idiom not up to the task of representing our mental processes, or representing them
falsely. But what we learn from the development of neural networks, PDP and the
prototype activation model is that there is an alternative, empirical account of how
cognitive creatures in fact acquire knowledge of their environment. And if this is the
causal story, then it follows that this will eventually include policy relevant knowledge
as a possibly quite specific configuration of synaptic connection weights. We learn even
more specifically that such knowledge generation is global given the network character
of brain functioning, and that any policy decision made is the result not of rational
deliberation or prachd reumn but of the successful relevant prototype being activated.
And we may comment in passing that the distinction between analytic and substantive
arguments, believed to be all important by Dunn (and Toulmin), simply evaporates. It
is of no relevance in the neurological story of information processing.
How does the brain learn to recognize and subsume a specific problematic situation
under a relevant policy prototype? Unlike an artificial network which is trained up by
a “teacher”‘, for us, our global environment steps in as the “teacher”. In one sense,
this amounts to saying that a pohcy analyst, for instance, needs to gather as many and
as varied policy experiences as possible (leaving unspecified for the present what such
an experience might be) since successful subsumption of policy relevant information is
predicated upon already existing prototypes. “Goodness” of policy choice would then
be a function of how well trained a relevant neural network is, i.e., it is a function of
the richness of existing prototypes.
This raises the question of what function such sentential structures as our best scientific
theories have in relation to the practice of policy analysis in view of the fact that we
appear to be learning without the aid of sentential structures. The short answer is that
theories, even those we judge true by our coherentist standards, do not seem necessary
in order to learn a complex practice (Evers, 1990). This is not to say that true theories are
irrelevant to learning, only that since networks do not partition the inputs they receiveinto categories such as “true” or “false”, or “entailment”, or “implication”, they do not
possess any kind of privileged position. Since, on the other hand, true scientific theories
do explain the way the world is, they must have some functional importance. But this is
a big question in need of research. Tentative and speculative as these suggestions have
been, and counterintuitive as the elimination of folk psychology might appear, the way
forward for policy analysis lies in the more detailed, empirical work of the neurosciences
since they hold the key for the explanation of all cognitive behavior including that of
deciding which policy option is best.
References
Armstrong, D. M. (1981). Bel i eJ tr uth and know l edge. Cambridge: Cambridge University Press.BonJour, L. (1985). The strucfure ofempi ri caf know l edge. Cambridge, MA: Harvard University Press.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 50/97
550 G. LAKOMSKI
Boyd, W. L. (1988). Policy analysis, educational policy, and management: Through a glass darklv? In N.
J. Boyan (Ed.), Handbook of research on educati onal admi nistr ati on. New York: Longman.
Braybrooke, D. & Lindblom, C. E. (1963). A strat egy for decision. London: Collier-Macmillan.0. T___I__^ n , I .I
Campbell, D. T. (1969). Reforms’as experiments. American Psychol ogi st, 24: 409329.
Campbell, D. T. (1974). Evolutionary epistemology. In P. A. Schilpp (Ed.), Thephilosophy of Karl Popper.
La Salle, IL: Open Court.
Churchland, P. M. (1981). Eliminative materialism and the propositional attitudes. The Journal of
Phil osophy, LX XVI II (2). 67-90.
Churchland, P. M. (1985). The ontological status of observables: In praise of the superempirical virtues. In
P. M. Churchland & C. A. Hooker (Eds.), Images of science. Chicago: University of Chicago Press.
Churchland, P. M. (1988a). Folk psychology and the explanation of human behaviour. The Aristote l ian
Society, Supplementary Volume LXII, 209221
Churchland, P. M. (1988b). Mat ter and consciousness (Revised edition). Cambridge, MA: M.I.T. Press.
Churchland, P. M. (1989a). On the nature of explanation: A PDP approach. In P. M. Churchland A
neurocomput at i onal perspecti ve: The natur e of mi nd and t he struct ure of sci ence. Cambridge, MA:
M.I.T. Press.
Churchland, P. M. (1989b). On the nature of theories: A neurocomputational perspective. In P. M.Churchland, A neurocomput at i onal perspecti ve: The nat ure of mi nd and the struct ure of sci ence. Cambridge,
MA: M.I.T. Press.
Churchland, P. S. (1980). Language, thought, and information processing. Nous, 14, 147-170.
Churchland, P. S. (1983). Consciousness: Transmutation of a concept. Pacifi c Phi losophical Quart erl y, 64,
80-95.
Churchland, P. S. (1987). Epistemology in the age of neuroscience. The Journal of Phi l osophy, 84(10),
544-553.
Churchland, P. S. (1989). Neurophil osophy: Toward a unif ied sci ence of the mi ndlbrai n. Cambridge, MA:
M.I.T. Press.
Dunn, W. N. (Ed.) (198C-81). Symposium on social values and public policy. Policy Studies Journal,
9(4).Dunn, W. N. (1981). Public policy analysis. An introduction. Englewood Cliffs, NJ: Prentice-Hall.
Dunn, W. N. (1982). Reforms as arguments. In E. R. House, S. Mathison, J. A. Pearsol, & H. Preskill(Eds.). Evaluation studies review annual. Vol. 7. Beverlv Hills. CA: Sage.
Dunn, W. N. (Ed.) (1983). Val ues, ethi cs, and the pract i ce of pol i cy6naly sis. Lexington, MA: D. C.
Heath.
Dunn, W. N., Mitroff, I. I., & Deutsch, S. J. (1981). The obsolescence of evaluation research. Evaluation
and Program Pl anni ng, 4, 207-218.
Evers, C. W. (1987). Naturalism and philosophy of education. Educati onal Phil osophy and Theory, 19(2),
11-21.
Evers, C. W. (1990). Educating the brain. Educati onal Phi l osophy and Theory , 22(2), 65-80.
Evers, C. W. & Lakomski, G. (1991). Knowing educational administration. Oxford: Pergamon Press.
Fischer, F. & Forester, J. (1987). Confronti ng values in policy analysis. Beverly Hills, CA: Sage.
Fodor, J. A. (1975) The l anguage of t hought. New York: Crowell. (Paperback edition (1979) Cambridge,
MA: Harvard University Press)
Carson, G. D. (1986). From policy science to policy analysis: A quarter century of progress. In Dunn, W.N. (Ed.), Symposium on social values and public policy, Pol i cy Studi es Journal , 9(4), 535-544.
Gill, J. (1988). Whi ch w ay t o school ?: A revi ew o f the evi dence on t he singl e sex versus coeducati on debat e
and an annot at ed bi bl i ography of t he research. Commonwealth Schools Commission: Canberra.
Hacking, I. (1975). Why does l anguage mat ter to phil osophv? Cambridge: Cambridge Universitv Press
Hooker, C. .A. (i975).-Philosophy and metaphilbsophy’of science: Empiricism, Popperianism’and realism.
Synt hese, 32, 177-231.
Jennings, B. (1983). Interpretive social science and policy analysis. In Callahan, D. & Jennings, B. (Eds.),
Ethi cs, th e social sci ences, and pol i cy anal ysi s. New York: Plenum.
Jennings, B. (1987). Policy analysis: Science, advocacy, or counsel. In S. Nagel (Ed.). Research i n publ i c
poli cy analy sis and management. Greenwich, CT: JAI Press.L, I
Lakomski. G. (1988). Critical theorv. In J. P. Keeves (Ed.). Educational research. methodoloav. and. ~ I “,measurement. An internati onal handbook. Oxford: Pergamon Press.
Landsbergen, D. & Bozeman, B. (1987). Credibility logic and policy analysis. Know l edge: Creati on,Di ffusion, Ut i li zati on, S(4), 625-649.
Lasswell, H. D. (1951). The policy orientation. In Lerner. D. & Lasswell, H. D. (Eds.), Thepolicy sciences.
Stanford: Stanford University Press.
Lerner, D. & Lasswell, H. D. (Eds.) (1951). The policy sciences. Stanford: Stanford University Press.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 51/97
Beyond Paradigms 551
Lincoln, Y. S. & Guba, E. E. (1986). Research, evaluation, and policy analysis: Heuristics for disciplined
inquiry. Poli cy Studi es Revi ew , 5(3), 54&S&.
MacRae, D. (1976). The saci a~fu ncti on afsaciuZ science. N ew Haven: Yale University Press.
Mason, R. 0. & Mitroff, I. I. (1980-81). Policy analysis as argument. In Dunn, W. N. (Ed.), Symposium
on social values and public policy. Policy Studi es Journal , 9(4), 579584.
Mitchell, D. E. (1984). Educational policy analysis: The state of the art. Educational Administration
Quarterly, 20(3), 129-160.
Nisbett, R. E. Bc Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes.
Psychological Review, 84(3), 231-259.
Popper, K. R. (1981). Obj ecti ve know fedge. An evol uti onary appranch (Revised edition reprinted with
corrections and a new appendix 2). Oxford: The Clarendon Press.
Quine, W. V. 0. (1969). Epistemology naturalized. In Quine, W. V. 0. Ont ological relat iv it y and other
essays. New York: Columbia University Press.
Quine, W. V. 0. & Ullian, J. S. (1978). The w eb of bel i ef (Second edition). New York: Random
House,
Rumelhart, D. E. & McClelland, J. L. (1986). Paral lel distr i buted processi ng: Explorat i ons i n the
microstructure ofcognition. Vol. I. Foundations. Cambridge, MA: M.I.T. Press.
Sejnowski, T. J. & Rosenberg, C. R. (1987). Parallel networks that learn to pronounce English text,Complex Systems, 1, 145-168.
Sejnowski, T. J., Koch, Ch.. & Churchland, P. S. (1988). Computational neuroscience. Science, 241,
1299-1306.
Stich, S. (1983). From f ol k psychol ogy t o cogni t i ve science: The case against bel i ef. Cambridge, MA:
M.I.T. Press.
Toulmin, S. (1958). The uses of argument. Cambridge: Cambridge University Press.
Walker, 3. C. & Evers, C. W. (1982). Epistemology and justifying the curricutum of educational studies.
Bri t i sh Journal of ~ducat i anai St udi es, 30(2), 21S229.
Walker, J. C. & Evers, C. W. (1988). The epistemological unity of educational research. In Keeves. J. P.
(Ed.), Eyducat i onal research, methodol ogy, and measurement . Oxford: Pergamon Press.
Wildavsky, A. (1979). Speaki ng tr uth to power: The art and craft of poli cy anal ysis. Roston: Little,
Brown.
Weiss, C. H. & Bucuvalas, J. (1980). Truth tests and utility tests: Decision makers’ “frames of referencefor social science research”. American Soci ol ogical Revi ew , 45, 302-312.
Williams, M. (1977). Groundless belief. Blackwell: Oxford.
Williams. M. (1980). Coherence justification and truth. Review of ~efup~y sics. 34(2), 243-272.
Biography
Gabriele Lakomski is a senior lecturer in the Institute of Education, The University
of ~eibourn~, Australia. Her most recent publication is C. W. Evers and G. Lakomski
(1991) Knowing Edu~utio~al Ad~~n~stratio~, published by Pergamon Press. She serves on
a number of editorial boards including Curriculum Inquiry and the Journal of EducationalAdministration and is Section Editor of the second edition of the International
Encyclopedia of Education. She writes about philosophical problems in educational
(administration) research, research methodology, educational policy, and feminist issues.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 52/97
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 53/97
CHAPTER 4
HERMENEUTICS: A THREAT TO SCIENTIFIC SOCIALSCIENCE?
DENIS C. PHILLIPS
Professor of Education and Philosophy, Stanford University, California, U.S.A.
Abstract
In the past few decades, numerous commentators who have been influenced by the
Continental hermeneutical tradition have argued that the social sciences, and related
applied fields such as educational research, are closer in certain logical or epistemological
respects to the humanities than to the natural sciences. This case is outlined; and then itis argued that, although many of the points made by hermeneuticists are both sound and
important, the conclusions that are drawn go too far. Two sets of questionable conclusions
are dissected: (a) those embodying claims about the epistemology of the social sciences
and related fields; and (b) those that describe the nature and limits of the social sciences
and educational research.
Introduction
Considering only the last three hundred years, from about the time of Vito onwards, a
massive literature on hermeneutics has accumulated. But in the past two decades there
has been a veritable avalanche of material - a poor academician can be driven to the
edge of bankruptcy trying to keep pace with the new books.
Unfortunately this recent material is more a repository of enthusiasm than of
enlightenment. There are differing accounts of the nature of the key issues, although
what comes shining through is the fact that hermeneuticists manage to reach (via
difficult and sometimes nearly impenetrable prose) some far-reaching and important
conclusions about the nature of the social sciences (and, by default, about related
areas such as empirical educational research). To the skeptical eye, the literature is
full of claims, but the arguments are left sketchy or unclear (or both); and there isa dearth of concrete examples - Stegmiiller (1988, p. 109) laments that “analysis of
examples is totally absent”. To add insult to injury, some writers (without much by
way of supporting argumentation) extend the scope of hermeneutics - so that, like
the Scarlet Pimpernel, hermeneutical issues are claimed to be everywhere.
The following discussion will attempt to bring some order to this complex domain.
First, there will be a distillation of the hermeneuticist case, especially in the form in
which it is advanced as a criticism of traditional empirical social science and related
553
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 54/97
554 D. C. PHILLIPS
fields; then the discussion will focus upon the resulting claims that are made. In general,
the center of interest will be the image held by hermeneuticists of both social science
and research in applied fields such as education, and their epistemologies.
However, there is one more preliminary matter. Throughout the discussion the terms
“hermeneutical” and “interpretive” will be used as synonyms; their use will be varied
simply as a stylistic device to maintain the reader’s interest. One word derives from
Greek, the other from Latin; but they mean the same and even, in their classical usages,
refer to the same winged messenger of the gods (whose function was to communicate
the wishes of the deities in a form that mere mortals could understand).
The Interpretivist Case
The interpretivist case runs as follows, although it must be stressed that what follows
is a general account and of course individuals disagree about many of the details.
According to interpretivists, physical scientists deal with objects by explaining their
behavior either in terms of external forces or in terms of inner processes that result from
their physico-chemical microstructure. Notions of force, energy, causation, and natural
law are central; and the.methods by which knowledge is built-up are observational and
experimental. The underlying epistemological premise is a form of empiricism (many
interpretivists would say it is a form of logical positivism). Until very recently, social
science (especially in the U.S.A.) has proceeded by mimicking the physical science
approach - behaviorism in psychology being one example among many, though alsothe most notorious.
On the other hand, hermeneuticists would argue, humans are not mere physical
objects; people are impelled by ideas, knowledge, and hopes and desires. They harbor
intentions. And these things depend upon the use of symbols, as in language; as Gadamer
puts it (1977, p. 29), “language is not only an object in our hands, it is the reservoir
of tradition and the medium in and through which we exist and perceive our world.”
Symbols and language, of course, are impossible without societies. Furthermore, many
actions undertaken by individuals are actually constituted by public meanings, socially
adopted rules, conventions, and the like; thus, to take a fairly trivial example, one cannot
understand the game of tennis, let alone play it, unless one understands the rules andconventions that define the valid activities of the game (a serve that is “in”, an acceptable
placement of a return of the ball, etc.). But activities as diverse as participating in a
dinner party, consulting a physician, writing a philosophical paper, and giving evidence
in court, are no less constituted by socio-cultural rules and conventions.
It seems to follow from all this that to explain the actions of a person (as opposed to
the behavior of a physical object), an investigator must uncover the understandings of
the actor - how the actor interprets the situation he or she is in, how the mores and
beliefs of the society in which the actor is located are influential, what the actor sees
as being the possible responses that are open (given the social beliefs the actor holds),
and the symbolic meaning of the forms of behavior that are open to the actor in thatparticular setting. As the social scientist Bauman put it (1978, p. 12):
Men and women do what they do on purpose. Social phenomena, since they are ultimately acts of men
and women, demand to be understood in a different way than by mere explaining. Understanding them
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 55/97
Beyond Paradigms 555
must contain an element missing from the explaining of natural phenomena: the retrievaf of purpose, ofintention, of the unique configuration of thoughts and feehngs which preceded a social phenomenonand found its only manifestation, imperfect and incomplete, in the observable consequences of actionTo understand a human act, therefore, was to grasp the meaning with which the actor’s intention
invested it; a task, as could easily be seen, essentially different from that of natural science.
This is the heart of the hermeneuticist or interpretivist position, and it is this to which
reference is being made by these labels in the subsequent discussion.
It should be noted in passing that a rift occurs at this point between several “schools”
of interpretivists. Some believe that it is necessary - as a corollary of the points
outlined above -to pursue the subjective understandings of actors; these can be labeled
“phenomenologically oriented hermeneuticists”. Others eschew this subjective approach
and focus instead upon the public meanings together with the observable actions of, and
interactions between, people in social settings. There is an accompanying disagreement,
therefore, about the methods of inquiry that are appropriate. It is probably true to saythat mainstream American social scientists tend to look askance at methods that smack
of subjectivism; and philosophers in the English-speaking world have a similar attitude
towards theories of meaning that focus upon the “pictures”, ideas, or intentions internal
to the individual. The dominant contemporary philosophical approach to meaning
focuses instead on the public realm, on how people operate with language - a view
that has clear implications for methodology. Although dominant in North America, this
latter orientation is not confined to it (see, for example, Apel, 1977, p. 301). However,
it is not the purpose of the present paper to pursue issues concerning the methodology of
interpretive studies; the focus here is unabashedly the theoretical arguments and claims
made on behalf of the importance of hermeneutics for the social sciences.To return to the exposition of the interpretivist case: human action, according to
the general position being expounded here, is a type of text (albeit an unwritten
one) - for a text is nothing more than a collection of symbols expressing meaning
(or even layers of meaning), although this meaning itself may be expressed in terms of
metaphors or complex cultural symbols. Hence it is possible to use the discipline that has
developed over many centuries to interpret texts - the discipline of hermeneutics, with
its central notion of the hermeneutic circle - to interpret and throw light upon human
action. This extension of hermeneutics to cover the non-written realm began with the
nineteenth-century figures Schleiermacher and Dilthey (see Palmer, 1969, Part 11); but
it reached its apogee with the work of Taylor (1977) and Ricoeur in the early 1970s.Thus, in his essay “The Model of the Text: Meaningful Action Considered as Text” first
published in 1971, Ricoeur wrote (1977, p. 316):
Now my hypothesis is this: if there are specific problems which are raised by the interpretation of textsbecause they are texts and not spoken language, and if these problems are the ones which constitutehermeneutics as such, then the human sciences may be said to be hermeneutical (1) inasmuch astheir objecr displays some of the features constitutive of a text as text, and (2) inasmuch as theirmethodology develops the same kind of procedures as those of . . text interpretation.
Many in the hermeneutics camp have gone on to point out that human societies are
full of the “objectifications” of meaning (as Gadamer, Betti and others term it) - notonly written texts, but social institutions, practices and rituals, and physical artifacts.
Hermeneuticists generally go further than this, however, and stress that interpreters
who are attempting to grasp the meaning of an actor or to grasp meaning that has
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 56/97
556 D. C. PHILLIPS
been objectified in some way, have their own understandings shaped by the fact that
they themselves are members of a particular culture at a particular historical moment.
Interpretation, in other words, is not an act in which a “disembodied” investigator is
trying to decipher the (pre-established) meaning of a culturally and historically situated
actor or institution; rather, the interpreter, too, must become hermeneutically aware of
his or her own historicity (or “preunderstanding”, as some writers term it). As Linge puts
it, in his editor’s introduction to Gadamer’s Phifosophicd Hermeneutics (1977, p. xiv):
This methodological alienation of the knower from his own historicity is precisely the focus of
Gadamer’s criticism. Is it the case, Gadamer asks, that the knower can leave his immediate
situation in the present merely by adopting an [interpretive] attitude? An ideal of understanding
that asks us to overcome our own present is intelligible only on the assumption that our own
historicity is an accidental factor. But if it is an onto log ica l rather than a merely accidental and
subjective condition, then the knower’s own present situation is already constitutively involved in
the process of understanding.
Interpretivists sometimes use examples such as the following: in the physical sciences,
the behavior of objects is explained by bringing to bear physical laws - such as when the
orbit of a planet is explained by deducing its behavior from Newton’s or Kepler’s laws
(together with a statement of the initial conditions). On the other hand, the action of
Julius Caesar in crossing the Rubicon is not explained by bringing it under a law - for
there are no laws of nature pertinent to the voluntary actions of Roman generals standing
on the banks of particular rivers such as the Rubicon; rather, Caesar’s action is explained
in terms of his intentions, and in terms of the symbolic importance of that particular river
(which marked the border between divisions of the Roman empire). Caesar’s action was
not the product of laws of nature (despite the fact that his body was a physical object), but
it was voluntary - a result of his consciously reaching the decision to carry out a revolt (a
revolt being, of course, a social phenomenon). Furthermore, our attempts to understand
Caesar’s action are mediated by the historical/cultural milieu in which we, as interpreters,
are located; so, as hermeneuticists, we are struggling to understand ourselves at the very
same time that we are struggling to understand Caesar.
Some Far-Reaching Conclusions
Before the discussion proceeds it should be acknowledged that there is much in the
interpretivist position, as just outlined, that is compelling. Humans are not mere physical
objects; and to understand or explain why a person has acted in a particular manner,
the meaning (or meanings) of the action have to be uncovered - and to do this the
roles of language and of social symbolisms and values have to be taken into account.
Furthermore, it is clear that every society contains many “objectifications” of meaning
in its rituals, symbolisms, institutions, and so forth. (These words are being written in
the U.S.A. on July 4th amid the festivities, which serves to drive the point home.)
What shall be disputed are some of the very wide-ranging conclusions about research
in education and the social sciences that are drawn by hermeneuticists, conclusions
that stray well past what is warranted by the preceding position. These questionable
conclusions can be clustered into two major groups.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 57/97
Beyond Paradigms
Epistemological Conclusions
557
The first set of wide-ranging conclusions can be introduced via reference to Charles
Taylor. In his now classic essay “Interpretation and the Sciences of Man” first publishedin 1971, Taylor reveals himself to be a powerful spokesman for the view that the
epistemological foundations of empirical science are an unsatisfactory base on which
to erect a “science of man”. Taylor refers disparagingly to the ingredients that make
up the “epistemological bias” of empirical social science, and he writes that
many, including myself, would like to argue that these notions about the sciences of man are
sterile, that we cannot come to understand important dimensions of human life within the bounds
set by this epistemological orientation [Taylor, 1977, p. 1061.
Along what seems to be similar lines, Macdonald and Pettit argue in their Semantics
and Social Science that the epistemology of the social sciences is close to that of
the humanities: “Social science, insofar as its concern is the explanation of human
behaviour, begins to look like a discipline which belongs with the humanities rather
than the sciences” (Macdonald & Pettit, 1981, p. 104). This is a view which must
come as something of a shock to empirical educational researchers; and the shock
is exacerbated by the fact that Macdonald and Pettit are not alone. Thus, somewhat
less pithily, Rabinow and Sullivan assert, in their I nt erpr et i ve Socia l Science: A Reader
(1979, p. 13), that
Interpretive social science has developed as the alternative to earlier logical empiricism as well as
the later systems approaches, including structuralism, within the human sciences. It must continueto develop in opposition to and as a criticism of these tendencies. Here interpretive social science
reveals itself as a response to the crisis of the human sciences that is constructive in the profound
sense of establishing a connection between what is studied, the means of investigation, and the ends
informing the investigators. But at the same time it initiates a process of recovery and reappropriation
of the richness of meaning found in the symbolic contexts of all areas of culture.
So, then, the first set of wide-ranging conclusions that are drawn are epistemological;
and yet detailed epistemological arguments are in short supply in this literature - for
example, it has not been shown in any detailed way how it is that hermeneuticists
actually know, that is, how the products of their interpretive endeavors are warranted.
(Once again Stegmuller’s remark comes to mind; he states that philosophers of science
customarily support their claims about the epistemology of science by detailed analyses
of examples of scientific work, but hermeneuticists do not do the same. in their own
fields.) The issues here will be taken up again later.
A different but obviously closely related form in which the epistemological claims
surface is in terms of the relation between the human sciences, the natural sciences, and
the humanities. The issue can be phrased as a question: is hermeneutical or interpretive
social science really a science, or is it a branch of the humanities? As Connolly and
Keutner put it in the introduction to their edited volume Hermeneut i cs versus Science?
(1988, p. l), “do the hermeneutical disciplines . . differ in some important way from the
natural sciences, i.e., are those disciplines “autonomous”?” And Schutz put it extremelyclearly when he wrote (1962, p. 34):
There will be hardly any issue among social scientists that the object of the social sciences is human
behavior, its forms, its organization, and its products. There will be, however, different opinions
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 58/97
5.58 D. C. PHILLIPS
about whether this behavior should be studied in the same manner in which the natural scientists
studies his object or whether the goal of the social sciences is the explanation of the “social reality”
as experienced by man living his everyday life within the social world.
A spatial analogy might help clarify this second form taken by the epistemological
claims of the hermeneuticists. (This is meant only as a preliminary heuristic device;
obviously it is hard to locate specific theorists precisely, for their thought is usually
complex and defies simple accurate categorization.) The humanities, the social sciences,
and the natural sciences can be visualized as arranged - in that order - along a
continuum. With respect to this continuum, several schools of thought exist: (1)
Some scholars have wanted to drive a wedge between the humanities and the rest,
by insisting upon the “autonomy” of the humanities; typically, this has been done by
stressing the nature of the humanities as interpretive disciplines, in which hermeneutics
(and especially the hermeneutic circle) has a central position (see Stegmuller, 1988). (2)
Others have hammered at the same wedge, by insisting that the sciences are demarcated
from the humanities by having a logical character accurately described by the logical
positivists. (3) A number of scholars have wanted to remove the wedge entirely. One
group has tried to do this by insisting that al l inquiry, to be genuine inquiry aimed at
producing warranted knowledge, must have the same underlying epistemology; usually,
the epistemology of science is taken as the model. On some readings, Dewey, and
perhaps Popper, belong to this group. (It should be stressed that in taking science as
the paradigm case of knowledge, these thinkers are not necessarily advocating a narrow
positivistic view of knowledge; in fact both Dewey and Popper have a fairly liberal view
of the nature of science - a topic on which there shall be more discussion later.) (4) Adifferent group has wanted to remove the wedge entirely by stressing that all knowledge
contains a strong interpretive element. Heidegger and Gadamer, according to some
of their remarks, ought tentatively to be classified as members of this group. Thus
Gadamer (1977, p. 38) writes that “Hermeneutical reflection fulfills the function that
is accomplished in all bringing of something to conscious awareness. Because it does,
it can and must manifest itself in all our modern fields of knowledge, and especially
science.” (5) Others, in particular writers such as Taylor (1977), Macdonald and Pettit
(1981), and Dilthey (1976) wish at l east to drive a wedge into the continuum between
the natural sciences and the social sciences, so that the social sciences end up being
grouped with the humanities. Typically, as discussed earlier, the argument is that thesocial sciences, like the humanities, must give a central place to interpretive methods.
It can be seen, therefore, that there is a degree of overlap between the views of those
in groups (4) and (5); but (4) offers a somewhat more radical position than (5).
Conclusions Concerni ng t he Nat ure of Social Science
The interpretivist case as outlined earlier also embodies within it certain views about
the nature and purpose of the various social and human sciences. Humans live in
societies, and societies are rife with objectifications of meaning; and it is with the
elucidation of these that the social sciences are centrally concerned. As Dilthey (1976,p. 192) put it,
Here the concepts of the human studies is completed. Their range is identical with that of
understanding and understanding consistently has the objectification of life as its subject-matter.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 59/97
Beyond Paradigms 559
Thus the range of the human studies is determined by the objectification of life in the external
world. Mind can only understand what it has created. Everything on which man has actively
impressed his stamp forms the subject-matter of the human studies.
But is it altogether clear that Dilthey is right? And even if the answer to this is in
the affirmative, does it follow that the central methods of the “human studies” must be
hermeneutical?
Skeptical Commentary
These two groups of far-reaching conclusions both require careful scrutiny. There
is some overlap between them, of course, so the discussion of each cannot be keptabsolutely water-tight. It will make sense to build up to the central issue concerning
epistemology, so for want of a better arrangement the discussion will proceed in reverse
order.
Commentar y: The Nature of Social Science
In the view of the interpretivists, the social sciences or “human studies” (together
with related fields such as educational research) are almost entirely concerned withmeaningful human action together with the objectifications of meaning that are to be
found in human societies. (Dilthey, 1976, pp. 163-7, did allow that a study of nature
was also relevant, insofar as natural events are frequently the stimuli for human action,
and form the focus of mankind’s attempts to develop knowledge.) But the fact of the
matter is that the social sciences are not so centrally concerned with hermeneutical
matters as has been supposed by supporters of the interpretivist position. To make this
case, it need not be denied that some sort of interpretive activity is required in some of
the social sciences; the point is that there is much else besides.
(1) In general, it may be true that the social sciences study phenomena that are social;
and social phenomena, as the interpretivists claim, are constituted by the use of language
and by other symbolic interaction - and thus cry out for hermeneutical.analysis. But
the “in general” marks an important caveat. The qualification is required because there
are many social sciences and they do not constitute a “natural kind”; the category is
human-made and is of necessity somewhat vague. The point is that there are some
social sciences where hermeneutical activity does not appear to be central - witness
various branches of psychology, and much of economics.
According to some accounts psychology is a member of the social (and certainly of
the human) sciences; and it is clear that psychology includes within its domain the
study of mechanisms, such as the cognitive and emotional ones, that underlie individual
human performance. Mechanisms like these can be studied in a manner that is as little
hermeneutical as is, say, biological research. Cases in point are the use, by cognitive
psychologists, of nonsense syllable experiments, and designs that focus on performance
on non-verbal tests such as Raven’s “Progressive Matrices”.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 60/97
560 D. C. PHILLIPS
And then there is economics, which is usually regarded as a clearcut member of the
social sciences - yet much of it can hardly be claimed to be hermeneutical. Some
branches of this “dismal science” certainly study the effects of social choices (as in
market phenomena), but these choices are conceptualized as being the mathematical
aggregate of individual choices. And it is crucial to note that, in general, the individual
is treated in the manner of an “ideal type” in physics: the individual is presumed to
be fully rational, fully knowledgeable, and to have a clear prioritization of needs and
desires. Mathematical modeling plays an important role here, but not hermeneutics. (All
that this adds up to is merely that economics is not, in essence, a descriptive discipline
which aims to discover by historical and interpretive methods why individuals make the
economic choices that they do.)
(2) However, even in those social sciences that do focus upon social phenomena- cultural anthropology, political science, and sociology are typical cases - there
is something more to study than human actions driven by motives, reasons, and
socially-determined understandings and interactions. Human actions have consequences
(both intended and unintended), and the study of these might not always require
hermeneutical interpretation. Theorists such as Popper place a great deal of emphasis
on the unintended consequences of human behavior; indeed, these consequences are
seen as a major driving force in history and are part of the reason that it is impossible
accurately to predict the future. Popper writes, in italics no less, that “only a minority of
social institutions are consciously designed while the vast majority have just “grown”, as
the undesigned results of human actions” (1961, p. 65). A little later (p. 67) he stresses
the “unavoidable unwanted consequences of any reform.” Sometimes Popper uses a
simple economics example to illustrate his point: a person who decides to buy a house
does not want the market suddenly to go up, but it will be the unintended consequence
of his or her entry into the housing market that prices indeed will rise. The study of
the laws of economics, and of how much the price of a commodity will change as the
number of individuals in the market changes, seems to be a non-hermeneutical scientific
activity. (The broader implications of Popper’s insight here will be discussed shortly.)
Consider a non-economic example: a political party in power in a country might
adopt a foreign policy for a set of reasons that requires interpretive understanding;
but unintended consequences of this policy might be that citizens resident overseas
have to return, gasoline shortages might break out as a consequence of disruption ofoverseas supplies, and there could as a result be a rise in the unemployment rate, which
in turn might differentially affect members of minority groups, leading to race riots and
the eventual overthrow of the party in power! All of these things can be documented,
correlated, and studied without use of hermeneutical methods. (This is not to deny, of
course, that some of these issues could be studied, for other purposes, using interpretive
methods. The point is that they also can be studied, and are studied in the social sciences,
using non-interpretive methods.)
At this point, if not earlier, an objection is likely to surface: The supporter of
hermeneutics is likely to protest that, contrary to the claim I have made, of course
all these research activities inescapably do require the use of interpretive methods!It is hard to resist the conclusion that, in making this counter-claim, hermeneuticists
have changed the meaning of the key term involved. Perhaps the point can be made
in terms of a distinction between a weak (and almost trivial) sense of “hermeneutics”
or “interpretation”, and a strong sense. In the weak sense, all endeavors that use the
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 61/97
Beyond Paradigms 561
medium of language involve interpretation-from following the directions in a recipe, to
understanding an advanced lecture in an academic speciality, the language of the writer
or speaker must be comprehended. In this weak sense, hermeneutics is like the Scarlet
Pimpernel. Unfortunately, though, this is a useless point for advancing the inte~retivists’strong case; for it does not ~~~~0~ that because physicists or chemists or sociologists
use (and must understand) language, the epistemology of their disciplines is somehow
suspect or weaker than they thought it to be, or that they must use strong hermeneutic
methods as their core methodology. The strong hermeneutical program arises, not simply
because of the universal human use of language, but because of special problems within
this umbrella - the problem of understanding written records of human thought or
action (or other objectifications of these things, such as monuments or social practices
or rituals) from ages or cultures that are different from the interpreter’s own. Indeed,
the strong hermeneutical program only makes sense on the assumption that interpreters
do (in general) understand their own culture (or at least their own sub-culture) and their
own language; for otherwise there would be descent into a self-referential nightmare in
which an interpreter might not understand (in the weak sense) his or her own actions or
writings or thoughts of a prior moment (not to mention the fact that children would not
understand - and therefore could not talk with, or learn from - their parents)! If this
was the case, of course, all inquiry would instantly grind to a halt, and humans would
be ossified at the social level of sea anemones (if, indeed, these organisms can be said
to be ossified).
(3) To pick up the main thread of the argument: even where the center of attention
in a social science is an issue that clearly involves interpretation (in the strong sense),
there are many related issues that are non-hermeneutical (in this sense). For example,
members of a population might vote in a surprising way at an election, and their
actions may require culturally-sensitive interpretation (in the strong sense) in order to
be understood. (This is the sort of thing that is done, or that is attempted, by “TV
experts” on election night.) But other issues arise in understanding elections - such
things as the influence of the weather on the turnout of voters, the party preferences
of younger versus older voters, and the turnout of members of various ethnic groups.
To gather information on matters such as these, no strong hermeneutical activity has
to be engaged in. Certainly on some matters, the voters might have to be asked for
information (for example, in an exit poll of young voters to see which candidates theyvoted for), but what takes place here is quite unlike the strong hermeneutical activity
carried out by literary experts interpreting the meaning of Hamlet’s soliloquy or by
historians trying to understand some action of Julius Caesar.
(4) Finally, it should be noted that in many sciences different levels of phenomena
are distinguished - as when physical scientists distinguish between the sub-atomic level,
the atomic level, the molecular level, and so on. The relationship between such levels
is a highly debated matter: can phenomena at one level be “reduced to” (i.e., explained
in terms of) phenomena and laws at a “lower” level? Although the issues here are
exceedingly complex, it seems clear that explanatory principles used at one levei do
not always apply at higher or lower levels.The same holds true in the social sciences and educational research; and it seems that
supporters of the interpretivist position (that is, of course, the strong position) would
be wise to consider the possibility that the use of hermeneutics might be appropriate at
some levels but not at others - leaving at least some phenomena to be dealt with by
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 62/97
562 D. C. PHILLIPS
non-hermeneutic social inquiry. It has already been seen that economics is an example
where the focus of attention is often (at least} on the group or aggregate level rather
than on the level of the individual human actors - and at the aggregate level there
seems to be a place for non-hermeneutical activity.Schelling (1978, Chapter 1) gives a simple example that can be adapted here for
illustrative purposes. (Artistic license has been exercised, and a different moral has
been drawn from the one Schelling highlights - he is interested in the question of the
fruitful co-ordination of the indi~~idual and group levels.) When audience members enter
large lecture halls, they seat themselves according to their own individual preferences.
That is, the choice of seating is an individual action, and the sitter’s knowledge, beliefs,
desires, and so forth may all play a role; and the choice of seating may also be a symbolic
act, such as one of defiance. Furthermore, a person’s choice is affected by the choices
made by people who entered the hall earlier. To understand why an individual chose a
particular seat, some sort of interpretive inquiry might be appropriate. And yet, if oneleaves the individual level of analysis and moves to a ‘*higher” level - the level where
audiences in halls rather than individuals become the “unit of analysis” - then it might
be apparent that there is a generalizable pattern to the filling of halls, the knowledge of
which could be helpful to designers of lecture halls, safety experts, and so on. And to
discover this pattern, no hermeneutical methods might have to be used; indeed, it can
be put even more strongly - hermeneutical methods could hinder the discovery of the
pattern rather than help. (&helling argues, quite rightly, that the motives of individuals
might have to be considered if any attempt is going to be made to change future seating
patterns; but if the aim is not to change the pattern but to use knowledge of it in future
planning, then understanding the “micromotives” is not necessary for the comprehension
of the pattern in the “macrobehavior”.)
Another way to phrase the point just made is that plot all of the patterns that are found
at the macrolevel in society are “objectifications” of meaning. And reference to Popper
can bolster the point: his argument that often what are most important in social affairs
are the unintended consequences of action is, in effect, making the point that there are
aspects of society that are not objectifications of meaning (for, by definition, unintended
consequences do not embody anyone’s intentions or meanings). Hermeneuticists who
assume that all social phenomena are objectifi~ations, have a distorted view of social
phenomena - and at the very least they owe us an argument to justify their position.
If they concede the point, they still owe us a discussion of the criteria that can be used
to distinguish those phenomena that are objectifications of meaning from those that are
not (a debt which, up to the present, they have seemed reluctant to discharge).
The conclusion that must be reached, then, is that although many social science
inquiries need to involve (strong) interpretive methods, many - very many - do
not have to. For it appears that the image of social science held by interpretivists is
too narrow; it is a view that is colored and limited by their own enthusiasms. This
conclusion is as far-reaching as those reached by the interpretivists, and it has important
implications: it weakens the remaining set of conclusions of the interpretivists. Given that
their view of social science is recognized as unduly narrow, it becomes more difficult toinsert a wedge between social science (as it really is) and the natural sciences and thereby
to group the social sciences with the humanities; and thus it becomes more difficult to
sustain an epistemological onslaught. But it is to this remaining broad set of far-reaching
epistemological conclusions of the interpretivists that the discussion now must turn.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 63/97
Beyond Paradigms 563
Commentar y: The Epistemol ogical Conclusions
There are two elements that require discussion here. In the first place, hermeneuticists
often attack the epistemology of traditional social science, which they regard ascrudely empiricist, or worse, as a form of positivism. Second, there is the matter
of the epistemology of (strong) hermeneutical social science itself - that is, what do
intepretivists want to put in place of the present “inadequate” epistemology, and is their
alternative adequate?
To deal with first things first; clearly it is a travesty to regard all of mainstream
social science, even just in the U.S.A., as being neo-behaviorist in spirit. (What of the
recent developments in cognitive science, social psychology, ethnomethodology and
anthropology, linguistics, political science, organizational theory, etc.?) In the late
twentieth century it is abundantly clear that neither the natural nor the social sciences
have to be viewed as being based on logical positivism; an image of science has emerged
over the past few decades according to which it is a more open and more speculative
endeavor than had previously been thought (see the discussion of the work of Kuhn,
Lakatos, Popper, Feyerabend and others in Phillips, 1987, Part 1). It is not stretching
the truth to suggest that when hermeneuticists attack the epistemology of mainstream
social science, what they have in mind is what Popper has called “misguided naturalism”
(Popper, in Adorn0 et al ., 1976, pp. 90-91). In effect they are attacking what by now
is recognized widely to be a straw man.
At least one alternative analysis of the epistemology of science has been offered by
Popper himself; hermeneuticists such as Taylor do not discuss it, for it seems immune
from the sort of charges they offer of positivism. (This is not to say that Popper has all
of the answers, or even some; his work remains a source of controversy - but in some
respects it is clearly an advance over positivism.) Popper denies that human knowledge
(including, of course, scientific knowledge) is certain by virtue of the fact that it is erected
upon unshakeable foundations. His books develop the case for a non-foundationalist
epistemology - see, for example, his Logi c of Scient i fi c D i scovery, Conjectur es and
Refutations, and Obj ect iv e Know l edge - although it should be stressed that Popper
is not alone among twentieth-century epistemologists in regarding foundationalism as
outdated. (For an example of a psychologist who holds this epistemology, see Weimer,
1979.) That is, Popper and many others do not approach the problem of knowledge
in terms of seeking the “rock-bottom” and indubitable foundations upon which the
certain knowledge of science (and of everyday life, so far as it has certain knowledge)
is built by a process of induction. Instead, these luminaries stress that no knowledge is
unshakeably certain, and that there are no absolutely sound foundations for knowledge.
Human knowledge is speculative, it projects tentatively into the future; whatever reason
we have to believe the things we do believe, it is not because they are based on absolutely
sound foundations! Our beliefs, and the considerations that led us to hold them, are
always open to the possibility of revision. (Popper has offered an account of the issues
surrounding the “rationality of scientific belief” that arise here, but it is not clear that
his resolution of the problems is acceptable. See Newton-Smith, 1981.)
Acceptance of a non-foundationalist approach to epistemology, in which all knowledge
is regarded as tentative, has an additional virtue: it allows a softening of the predicament
highlighted by Gadamer (who did not see it as a predicament so much as a too-often
neglected fact of life), namely, the fact that interpretations must inevitably emerge
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 64/97
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 65/97
Beyond Paradigms 565
(Betti, 1980, especially pp. 68-69). The application of this distinction to literature leads
to some controversies - Fish, for example, would deny that Shakespeare’s intentions
(the “meaning” of Hamlet) is important, for what is relevant in literature is what readers
can impose or construct for themselves (the “significance”, which of course Fish wouldprefer to label as “meaning”). However, even in the contentious realm of literary theory
the distinction itself is useful and provides terminology that serves to highlight the issues
at stake.
In those areas of the social sciences where the strong hermeneutical program seems
appropriate, even greater light is shed by Betti’s distinction. Consider a person in a social
setting who performs some act that draws the attention of a social scientist. (The study
of individual human action, it will be recalled from the preceding discussion, is one area
where hermeneutical methods do seem appropriate in the social sciences.) Betti would
have us, in effect, recognize two sets of issues: (a) what did the actor intend (that is,
what was the meaning of the act), and (b) what can the social scientist or interpreter
say about the act (that is, what is the significance of the act). This is not to say that on
every occasion Gove of these matters are of interest, but both are possible concerns. Of
course Betti is not alone in pointing to these two things; the interpretivist Winch (most
notably in his well-known dispute with Jarvie) aho makes this point - and Winch argued
that the issue having priority was the identification or description of the act, which by
conceptual necessity involves the determination of the actor’s intentions. Only after the
act has been identified, Winch suggested, might the social scientist be able to go on and
say something about it in terms of his or her own disciplinary perspective (assuming that
this second phase is relevant to the particular inquiry) (Winch, 1970, pp. 249-259; see
also Winch, 1958).
What can be said, then, about the epistemological unde~i~nings of this two-stage
interpretive process? The second stage is less problematical, relatively speaking. After
an act or document is interpreted as being an instance of X (for example, an expression
of jealousy), then it is relatively straightforward to judge if it falls within the domain
of some theory T. To be an acceptable theory, T would need to have a warrant that
is appropriate - if T is a theory of literary criticism, then it would need whatever
warrant is required for reputable status within that field, whereas if T is from sociology
or economics then different types of warrants would be appropriate. Within the social
sciences, there is a degree of agreement - although it is far from universal - aboutsuch matters as whether a theory T is well-warranted (or if not, why not), and whether
phenomenon X fails within the domain of T. (This sounds simple enough on the surface,
but of course there are many complexities; these, however, are subject to lively debate
and investigation within the traditional academic domains. Whether or not one judges
the epistemological program sketched here to be reasonable depends upon whether one
regards epistemology as a total field as viable, or as dead. If the reader judges it to be
dead, there is not much more to be said, except that this essay should have ceased being
of relevance long ago!)
The epistemological difficulties of the second phase pale into insignificance, however,
when compared to the problems faced by the first. How does an interpreter know that heor she has correctly identified the intentions of an actor (or has understood the meaning
that has been objectified in some social institution)? Neither Winch, Betti, Gadamer,
Dilthey, Taylor or the rest of the hermeneutical horde has made much headway here,
although many of them certainly espouse the ideal of settling on correct interpretations.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 66/97
566 D. C. PHILLIPS
Betti is an illuminating figure here, for he explicitly wants to establish an “objective”
position; he writes (1980, p. 57) of “the demand for objectivity: the interpreter’s
reconstruction of the meaning contained in meaning-full forms has to correspond to
their meaning-content as closely as possible”, and this requires “honest subordination”
(i.e., subordination of the interpreter to the “other” whose meaning is being deciphered).
Betti criticizes Gadamer’s book Truth and M ethod on the ground that (unintentionally) it
undermines the quest for objectivity, which Gadamer also espouses. Yet the best that
Betti can do himself is to argue that objectivity arises through the strenuous subjective
efforts of the interpreter intuitively or emphathetically to understand the meaning of the
other! This hardly seems an adequate means to achieve the goal he set out in the form
of a methodological canon which he labels, somewhat grandiosely, as “the canon of the
hermeneutical autonomy of the subject”:
By this we mean that meaning-full forms have to be regarded as autonomous, and have to beunderstood in accordance with their own logic of development, their intended connections, and in
their necessity, coherence and conclusiveness; they should be judged in relation to the standards
immanent in the original intention: the intention, that is, which the created forms should correspond
to from the point of view of the author . [Betti, 1980, p. 581.
From the point-of-view of the present writer, this canon is fine, but the epistemological
resources with which Betti wants to operationalize it are, to say the least, deficient.
A case can be made - although it can only be sketched here - that for the purposes
of social science, meanings and intentions can be investigated using traditional scientific
methods. That is, it can be argued that there is no epistemological difference in kind
between gaining knowledge about the other objects of science and gaining knowledgeabout meanings and intentions. Many branches of science can provide cases where the
objects of interest are not directly observable or measurable, but where their presence
(and their nature) is inferred from what is observable. This process is hypothetical, and
it is not guaranteed to be successful; but it is self corrective - by a bootstrapping
process involving testing and elimination of errors (which is itself a tentative business),
the warrants for the claims that are made about such objects become stronger (though,
many would argue, never so strong that matters become completely settled). Again,
what is sauce for the scientific goose is sauce for the hermeneutic gander: intentions and
meanings can be investigated in the same way. Tentative hypotheses can be checked,
if somewhat indirectly; empirical evidence can have a bearing on hermeneutical issues;
and hermeneuticists can - and do - use the hypothetico-deductive method common
across the sciences (Follesdal, 1979).
Conclusion
The net conclusion is that although there are some areas of social science and
educational research where the strong hermeneutical program is important, these
neither exhaust the scope of the social sciences and educational research nor offer
any serious grounds on which to hold that the social sciences and related applied fields
are more closely allied with the humanities than with the natural sciences. Those who
hold the contrary view, and claim that there is a similarity in kind with the humanities,
or that empirical social science is completely misconceived, need to offer more detailed
arguments and examples.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 67/97
Beyond Paradigms 567
Nevertheless, the sometimes exaggerated claims of the hermeneuticists have served
a very useful purpose: these claims have forced the adherents of traditional “pure” and
“applied” social science to broaden their view of the nature of persons - instead of
treating people on a par with inanimate objects they have been forced to regard personsas actors located within social and historical webs of meaning. And this constitutes a
watershed. (For an example of how “traditional” researchers have come to accommodate
the hermeneutical position, see Gage, 1989. Critics of Gage’s earlier works have often
labeled him as a neo-positivist.)
Acknow ledgements - The author has profited greatly from conversations with Deborah Kerdeman,
although she does not fully endorse some of what is written herein. Henry Alexander, Ron Glass, Ray
McDermott, Gabriele Lakomski, Colin Evers, and Harvey Siegel have also given helpful feedback on
earlier versions.
References
Apel, K.-G. (1977). The a priori of communication and the foundation of the humanities. In F. Dallmayr
and T. McCarthy (Eds.), Understandi ng and soci al inqui ry . Notre Dame, IN: University of Notre Dame
Press.
Bauman, Z. (1978). Hermeneut i cs and social science. New York: Columbia University Press.
Betti, E. (1980). Hermeneutics as the general methodology of the gei stesw i ssenschuft en. In J. Bleicher
(Ed.), Contemporury hermeneurics. London: Routledge.
Connolly, J. & Keutner, T. (Eds.) (1988). Hermeneuti cs versus science? Notre Dame, IN: University of
Notre Dame Press.
Dilthey, W. (1976). Di l they: Selecfed w ri ti ngs. Cambridge: Cambridge University Press.
Fish, S. (1980). Is there a text in t his class? Cambridge, MA: Harvard University Press.
Follesdal, D. (1979). Hermeneutics and the hypothetico-deductive method. Dialectica, 33, 319-336.
Gadamer, H.-G. (1977). Philosophical hermeneutics (Tr. D. Linge). Berkeley: University of California Press.
Press.
Gage, N. I,. (1989). The paradigm wars and their aftermath. Educational Researcher, 18,4-10.
Macdonalcl, G. & Pettit, P. (1981). Semant i cs nnd social sci ence. London: Routledge.
Newton-Smith, W. (1981). The rationality ofscience. ondon: Routledge.
Palmer, R. (1969). Hermeneutics. Evanston, IL: Northwestern University Press.
Phillips, D. C. (1987). Phil osophy, science and social i nquiry . Oxford: Pergamon Press.
Popper, K. (1961). The poverty of historicism. London: Routledge.
Popper, K. (1976). The logic of the social sciences. In T. Adorn0 ef al. (Eds.), The posit iv ist dispute in
German sociology. London: Heinemann.
Rabinow, P. & Sullivan, W. (Eds.). (1979). I nt erpr efi ve social science: A reader. Berkeley: University of
California Press.
Ricoeur, I’. (1977). The model of the text. In F. Dallmayr & T. McCarthy (Eds.), Linderstanding and
social inquiry. Notre Dame, IN: University of Notre Dame Press.
Schelling, T. (1978). M icromot i ves and macrobehavi or. New York: Norton.
Schutz, A (1962). The probl em of social reali t y: Coll ected papers 1. The Hague: Martinus Nijhoff.
Stegmtiller, W. (1988). Walther von der Vogelweide’s lyric of dream love and quasar 3C 273. In J. Connolly
& T. Keutner (Eds.), Hermeneuti cs versus science? Notre Dame, IN: University of Notre Dame Press.
Taylor, C. (1977). Interpretation and the sciences of man. In F. Dallmayr & T. McCarthy (Eds.),
Understandi ng and social inqui ry . Notre Dame, IN: University of Notre Dame Press.
Weimer, W. (1979). Notes on the methodology of scientific esearch. N ew Jersey: Lawrence Erlbaum.
Winch, P. (1958). The idea of a social science. London: Routledge.
Winch, P (1970). Comment. In R. Borger & F. Cioffi (Eds.), Explanati on i n the behavi oral sci ences.
Cambridge: Cambridge University Press.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 68/97
568 D. C. PHILLIPS
Biography
D. C. Phillips, an Australian, moved to Stanford University in California in 1974,
where he is currently Professor of Education and Philosophy, and Chairperson of
the Program in Research and Evaluation Methods in Education. He is the author
or co-author of a number of books including Phi l osophy, Science, and Social I nquir y
(1987), Vi sions of Chi l dhood (1986)) Perspecti ves on Learni ng (1985), and Toward
Reform of Program Evaluat i on (1980). He is co-editor of the 1991 NSSE Yearbook
Evaluati on and Educati on: At Quart er-Century . During 1990-91 he was President of the
Philosophy of Education Society.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 69/97
CHAPTER 5
META-ANALYSIS, METHODOLOGY AND RESEARCH
INTEGRATION
BRIAN D. HAIG
Department of Psychology, University of Canterbury, Christchurch, New Zealand
Abstract
This chapter presents a critique of meta-analysis by focusing on its underlying rationale.
It is argued that drawing a distinction between scientific and evaluative inquiry, where
meta-analysis is depicted as a methodology for quantitatively combining the results of
evaluation research, presupposes an unacceptable and largely empiricist view of both
forms of inquiry. Against prominent advocates of meta-analysis, it is claimed that a tenable
view of scientific research will be inherently evaluative. Additionally, it is argued that
the alleged incommensurability of theoretical frameworks does not prevent inter-theoretic
communication and integration, and that Glass’ meta-analytic rationale offers a deficient
account of the nature of methodology and policy research. It is contended further, that the
inherent deficiencies of Fisherian outcome studies render meta-analysis a worthless research
exercise. As a result it is concluded that serious attempts to integrate worthwhile knowledge
claims cannot be achieved statistically through the meta-analyses of outcome studies, but will
instead require “qualitative”, densely reasoned efforts to construct postulational and global
theories. Finally, the chapter gives some attention to a neglected perspective on research
integration by outlining an explanatory coherentist theory of justification and showing how
this can be implemented procedurally in social science research.
Introduction
The recent adoption and widespread use of meta-analysis procedures to integrate the
results of outcome studies in many areas within the social sciences stands as one of the
most striking research developments of the last decade. Meta-analysis is an approach to
data analysis that involves the quantitative analysis of data analyses of extant empirical
studies. Hence the term “meta-analysis” coined by Glass (1976). Meta-analysis is
concerned with the statistical analyses of the results of data analyses from many
individual studies in a given domain for the purposes of integrating or synthesizing
those research findings.
Meta-analysis comes in a variety of forms (Bangert-Drowns, 1986), but in its simplest
form meta-analysis requires computation of the average effect size for a group of studies.
For Glass, the effect size measure is the standard score obtained by subtracting the mean
of the control group from that of the treatment group and dividing this difference by the
standard deviation of the control group. This is done for each of the relevant dependent
569
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 70/97
570 B. D. HAIG
variables in each study. The effect sizes are then summed and divided by the total number
of effects to obtain the average effect size.
In the face of burgeoning and fragmented research literatures displaying conflicting
results, meta-analysis is offered as a rigorous and objective alternative to the customaryunsatisfactory integration methods of narrative literature reviews and vote-counting of
significance test outcomes. Narrative reviews have been criticized as casual, severely
selective and unable to portray accumulated knowledge (Light & Smith, 1971), while
vote-taking from box-score tallies of significance test outcomes has been faulted for
its faiture to acknowledge the methodological asymmetry between confirmation and
refutation, and for its bias in favour of large-sample studies for which the significant
outcomes are largely a function of statistical power (Mechl, 1978).
Despite the claimed advantages of meta-analysis, a number of different types of
criticism have been leveled against it. For example, Slavin (1984) has argued that the
actual use of meta-analysis procedures in education constitutes a retrograde step in theart of research integration. Others (e.g., Bruno & Ellett, 1988; Cook & Leviton, 1980;
Erwin, 1984) have pointed out serious methodological limitations of the approach; while
at a m~ta-theoretical level Eysenck (1984) has argued that the meta-analytic enterprise
is unscientific, and constitutes “an abuse of research integration”.
However, to date evaluations of meta-analysis and its applications have shown little
regard for the accompanying rationale provided by Glass (Glass, 1972; Glass & Kliegl,
1983). Glass himself rightly claims that many misunderstand meta-analyses of outcome
research because they fail to take cognizance of the rationale. This failure is offered by
Glass as the reason for the widespread misunderstanding of Smith, Glass, and Miller’s
(1980) original meta-analysis of psychotherapy outcome studies.
My purpose in this chapter is to provide a critique of meta-analysis by focusing
on the conception of inquiry embodied in its underlying rationale. It is argued that
drawing a distinction between scientific and evaluative inquiry, where meta-analysis is
depicted as a procedure for combining the results of evaluation research, presupposes
an untenable and essentially empiricist view of both types of inquiry. Additionally, it is
claimed that there are a number of deficiencies in Glass’ perspective on methodology,
and that meta-analysis is premised on a deficient account of policy research. Finally,
some general suggestions are made about desirable future directions for our research
integration efforts. These involve adopting a coherentist theory of justification to help
construct and evaluate postulational and global theories.
Scientific and Evaluative Inquiry
The core of the rationale for Glassian meta-analysis involves drawing a distinction
between scientific and evaluative inquiry. Glass’ (1972) position is that researchers
as scientists are concerned to satisfy their curiosity by seeking truthful conclusions in
the form of theories comprising explanatory laws. By contrast, evaluators undertake
research on behalf of a client which is aimed at producing useful decisions based ondescriptive determinations of the worth of particular products or programs. For Glass
the meta-analysis of outcome studies properly involves the integration of the products
of evaluative research only,
GIass differentiates scientific from evaluative inquiry in respect of nine basic contrasts.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 71/97
Beyond Paradigms 571
I shall critically consider the most important of these contrasts, arguing that none of
them plausibly distinguish scientific and evaluative inquiry. This will serve to undermine
seriously an essential part of the rationale he provides for meta-analyzing outcome
studies, and at the same time help provide a more defensible characterization of the
research process.
Motivation of the Inquirer
According to Glass scientific inquiry is undertaken largely to satisfy the curiosity of
the researcher, and to this end will involve the construction of theories. By contrast,
the researcher’s basic concern in conducting evaluative inquiry is to help solve a client’s
practical problem.But trying to distinguish between the two types of inquiry in this way won’t do. For
one thing, science, being a human activity, is pursued for a multiplicity of reasons, both
personal and epistemic. Also, given that science aims at valuable truth, then the use of
a theory to help solve a practical problem counts as an epistemic virtue and contributes
to the overall excellence of that theory. Further, it is desirable to conceive of scientific
inquiry itself as a problem oriented endeavor where the concern is to formulate better
ill-structured problems so that we might solve them (Haig, 1987; Nickles, 1981).
Laws and the Particular
Glass briefly invokes the popular distinction between nomothetic and idiographic
research to differentiate further scientific and evaluative inquiry. For him scientific
inquiry involves the search for laws understood as statements of relationship among
variables or phenomena, whereas evaluation just involves the description of the value,
or values, of a particular thing.
This contrast between nomothetic and idiographic forms of inquiry is clearly based
on the widely held view that causal laws are universal, or generally applicable, empirical
regularities. However, it is more defensible to think of causal laws as the causally
necessary activity of generative mechanisms rather than their conditions of activation
or expressions of effect (Bhaskar, 1978). On this view it is a contingent matter whether
the mechanisms happen to be in a closed system like an experiment in which they can
produce empirical regularities. A law does not cease to exist in an open system just
because its empirical manifestations are absent. It is just that these latter are typically
altered or checked by the work of other causal mechanisms in an open system.
Now, by taking causal laws to involve the natural necessity of causal mechanisms
rather than the generality of empirical regularities, we can effectively dismiss Glass’
use of the popular distinction between nomothetic and idiographic inquiry for wrongly
holding causal laws and claims about the particular to be alternatives. Because laws arecentrally a matter of necessity, nomothetic inquiry can be either idiographic or universal
in nature. A science of the particular is a perfectly proper project, as for example in
the ethogenic study of individual lives using autobiographical methods. Moreover, to
endorse a study of the particular is not to foreclose the possibility that the future
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 72/97
572 B. D. HAIG
comparative study of individual lives may well reveal deep structural universals (cf.
De Waele & Harre, 1979).
The Role of Explanation
According to Glass science involves the continual search for subsurface explanations of
surface phenomena. Evaluative inquiry, on the other hand, does not seek explanations.
“A fully proper and useful evaluation can be conducted without producing an explanation
of why the product or program being evaluated is good or bad or how it operates to
produce its effects . . . [It] is usually enough for the evaluator to know that something
attendant upon the [product or program] is responsible for the valued outcomes.”
(Glass, 1972, pp. 5-6) Glass’ position seems to be that, even though program
treatments can be causally responsible for their measured outcomes, it matters little
that knowledge of this gleaned from evaluation studies does not tell us how programs
produce their effects, because such knowledge is not needed for policy action.
Glass is surely correct in asserting that scientists are centrally concerned with the
construction of causal theories to explain phenomena, for this is the normal way in
which they achieve understanding of what they study. However, he is wrong to insist that
proper evaluations can deliberately ignore knowledge of underlying causal mechanisms.
The reason for this is that the effective implementation and alteration of social programs
requires knowledge of the relevant causal mechanisms involved (Gottfredson, 1984),
and strategic intervention in respect of these is the most effective way to bring about
social change. I grant that orthodox realism is wrong to insist that the relevant causal
mechanisms will always be unobserved, but it is the case that appeal to knowledge of
covert mechanisms will frequently be required for understanding and change.
Truth and Social Utility
This is probably the major contrast for Glass. He asserts that scientific inquiry
characteristically attempts to assess the truth of knowledge claims, whereas evaluative
inquiry attempts to gauge the worth of things. Glass takes truth to comprise the empiricalvalidation and logical consistency of knowledge claims, while worth is understood as
social utility. He acknowledges that truth is highly valued and worthwhile, but, this
point aside, he insists that the contrast between truth and utility effectively helps to
distinguish science from evaluation.
It is clear that Glass’ position amounts to drawing a sharp fact/value distinction in
which theoretical knowledge is claimed to be value free. But I shall show in a moment
that this is a distinction that cannot be sustained. At the same time, it should be
appreciated that Glass fails to make an important epistemic distinction that is crucial
to a satisfactory understanding of science: in identifying truth with empirical adequacy
and logical coherence Glass has conflated the epistemic notions of truth and acceptance.Truth is best understood as (causal) correspondence with reality where it functions as
a guiding ideal for science. As such it is a highly valued, though unattained, goal that
helps us make sense of science as an attempt to represent and intervene in the world.
However, truth is only accessible indirectly by way of the various criteria we use to
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 73/97
Beyond Paradigms 573
evaluate and accept theories (Hooker, 1987). Empirical adequacy and logical coherence
are in fact two such criteria. They do not constitute truth itself, but instead function as
surrogates for truth.
By separating fact from value Glass clearly believes that the theoretical knowledge ofscience is free from value commitments. But this is definitely not the case. As just noted,
excellence of theory is determined by a multiplicity of related epistemic criteria. These
typically include explanatory power, existential depth, internal and external consistency,
fertility and practical application. Criteria such as these are in effect the good-making
features of theories, and as such they provide us with the means for judging theories to
be of value. The employment of such a range of values is an expression of the idea that
realist science pursues theoretically interesting or valuable truth, not truth simpliciter.
Contrary to Glass’ view of the matter, scientific inquiry deliberately strives to marry the
true and the good.
The point being made here is not just that science is influenced by value commitments,
but that it is actually constituted by values. That values are an inextricable part of science
can be seen dramatically when we realize that science is helping us turn our world into an
artifact which, because it involves the realization of human designs, necessarily combines
fact and value (Hooker, 1987). The old empiricist fact/value bifurcation has never been
true to science, and therefore cannot be used as a sound basis for distinguishing between
scientific and evaluative inquiry.
None of Glass’ major contrasts plausibly differentiate scientific and evaluative inquiry.
The post-positivist sketch presented here takes science to be a value laden, problem-
oriented human endeavor that tries to construct valuable causal explanatory theories ofboth particulars and universals. This is a view of science that emphatically rejects Glass’
empiricist view of both scientific and evaluative research. As we shall see later, it is a view
of science that assigns little importance to research integration through meta-analysis.
The Nature of Methodology
Met hodology QS Empiri cal
An important part of the rationale for Glassian meta-analysis involves adopting aconception of methodology as a substantive empirical discipline. According to Glass,
critics have often misunderstood meta-analysis because they have failed to appreciate
that it embodies a methodology of this sort. He claims that for any given empirical
domain a methodology combines with an object field and a taxonomy to give that
domain its basic structure. None of these three components are given to us a priori
as a product of logic. Instead, they are chosen for both arbitrary and historical reasons.
Methodologies, for instance, are selected and developed partly as a response to the
structure and pragmatic needs of society. In this regard, Fisherian experiments are said
to embody principles that grow out of the demand for control made by a technological
society. MethodoIogical assumptions are not established a priori. Rather, they aregenuinely refutable conjectures. Indeed, it should be emphasized that one of the main
functions of Glassian meta-analysis is to undertake the empirical investigation of such
assumptions as part of its own object field.
Glass believes that the most serious criticisms of meta-analysis lose their force when
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 74/97
574 B. D. HAIG
they are examined from the standpoint of empirical methodology. The two major
criticisms he examines concern the quality of studies included in meta-analyses, and the
incommensurability involved in combining such studies. I shall consider these problems
in turn and suggest that Glass’ view of each is far from satisfactory.
The Quality of Study Problem
Glass has strongly criticized the traditional practice of excluding from the review
process all those studies deemed to be methodologically unsound. He objects that
judgments of exclusion frequently reflect the particular subjective biases of the
researchers involved and that such judgments are made a priori involving the treatment
of methodological principles as dogma. With his empirical approach to methodology,Glass adopts liberal criteria for the inclusion of studies in a meta-analysis. Faced with
the predictable “garbage in, garbage out” complaint, Glass defends the inclusion of
methodologically poor studies on the grounds that it is an empirical question whether
threats to internal validity affect the results of studies. He contends that meta-analysis
can answer this question by ascertaining the relation between these threats and effect
sizes.
However, as Erwin (1984) has correctly noted, Smith et aZ .‘s (1983) meta-analysis
of the psychotherapy outcome literature arbitrarily excludes all studies with non-
experimental controls. Having criticized other reviewers for the a priori exclusion of
“bad” studies, Smith et al. fail to give non-experimental studies a chance to justify
empirically their inclusion in the meta-analysis.
More serious than this inconsistency is the point noted by Erwin that Glass and
his co-researchers cannot justifiably assume that a randomized experiment with two
treatments is sufficient to conclude that a psychotherapeutic treatment causally produced
beneficial effects. This is because such experiments generally are incapable of ruling
out all plausible rivals to the hypothesis under test. This methodological limitation
of outcome experiments is but part of a deeper worry I have that such experiments
are vitiated by further serious methodological shortcomings. I believe that Fisherian
experiments are broken-backed and that, because such experiments dominate meta-
analyses, there will not be a sufficient number of well controlled studies to determine
how methodological quality affects results. Indeed, I believe we are led to the conclusion
that the meta-analysis of Fisherian outcome studies is not really a worthwhile research
exercise.
Fisherian experiments are characterized by their employment of the procedure of
randomization to control for unwanted nuisance variables, the use of analysis of variance
to partition causes, and the appeal to statistical significance tests to help evaluate the
causal hypotheses in question. Unfortunately, each of these three features does not do its
intended job, thus seriously undermining the Fisherian approach to experimentation.
The procedure of randomization is employed in Fisherian experiments for two reasons:
it provides a justification for significance tests by guaranteeing the correct probability
distribution under the null hypothesis; and, as just noted, it is introduced in order to
control the influence of nuisance variables. But neither of these reasons is plausible. As
I shall note shortly, Fisherian significance tests are flawed and should be abandoned.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 75/97
Beyond Paradigms 575
This removes one of the reasons for appealing to randomization. Also, as Urbach
(1985) has argued, randomization is both unworkable and unnecessary. Randomization
is unworkable because an infinite number of randomizations will be needed to match
the possible sources of error; not all possible influences are randomized out by a singlerandomization. Randomization is unnecessary because we can identify the most potent
nuisance variables and match groups on these. (Here I ignore the fact that matching
may introduce its own problems; cf. Meehl, 1970).
Fisherian experiments make use of analysis of variance (ANOVA) procedures in order
to fathom the relative contributions that various factors make to producing an outcome.
However, ANOVA is seriously deficient as an analysis of causes, and is often a poor
guide to causal structure in the social sciences. ANOVA is frequently used as a method
for trying to discern the separate causes of variation, but as Lewontin (1974) and Sober
(1984) have shown, partitioning by ANOVA cannot separate causes that occur in an
interactive world. Rather than provide us with an analysis of separate causes, ANOVA,
with its linear model, provides us with a tautological partitioning of total variance among
observations into main effects and interactions of various orders. To turn ANOVA’s
linear model into a contingent one relating the values of variables we would need general
statements about functions. But the ANOVA model is a local analysis giving results
dependent on the distributions of the particular populations sampled. As Lewontin puts
it, ANOVA confuses the local analysis of variance with the global analysis of causes.
As mentioned earlier, significance testing gets incorporated into the Fisherian
conception of experimentation in order to evaluate empirically the hypotheses under
test. Unfortunately, significance tests have a number of shortcomings which have long
been discussed in the literature (cf. Oakes, 1986). The basic problem with significance
tests is that they provide very weak support for a hypothesis or theory. It is acknowledged
by many professional statisticians that the point-null hypothesis is (quasi) always false in
the social sciences. This being the case, reasonable sample size makes the achievement
of statistical significance the likely outcome of an experiment. Even if we make the
achievement of statistical significance contingent upon the rejection of a directional null
hypothesis, we still have about a fifty-fifty chance of obtaining a statistically significant
result with a “good” experiment (Meehl, 1967).
Glass and Kliegl (1983) express deep dissatisfaction with the practice of outcome
research, particularly in the area of psychotherapy. Nonetheless, they still thinkmeta-analysis can winnow out something of value from such studies. In this section of
the Chapter I have argued that the methodological flaws inherent in Fisherian outcome
studies makes them irredeemable.
The Problem of I ncommensurabil it y
Research integration through meta-analysis has also been criticized for lumping
together studies that are fundamentally incomparable. Glass handles this criticism by
drawing a distinction between practical and theoretical commensurability. He claims the
more common criticism - that meta-analysis mistakenly mixes “apples and oranges”
- has to do with practical commensurability; and that, by comparison with theoretical
commensurability, it poses little problem. For Glass theoretical commensurability “. . . is
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 76/97
576 B. D. HAIG
a long-standing point of debate in the philosophy of science, and the best that can be
said of progress toward the solution of the problem is that there has been little” (Glass
& Kliegl, 1983, p. 39).
However, Glass’ treatment of commensurability is grossly inadequate. In separatingpractical from theoretical commensurability, he implausibly presupposes a strong
observation/theory distinction. But post-positivist philosophy of science makes it
abundantly clear that such a distinction is not to be had. Observation statements are
theory laden and are assembled into networks with other classes of theoretical statements
to form scientific theories. This holds for both observational and postulational theories.
As a result, any meaning variance will hold, not only across different postulational
theories, but also across different observational theories such as outcome studies. There
is only theoretical commensurability.
Glass quickly lays the spectre of theoretical incommensurability by declaring that
there has been minimal philosophical progress on this issue (really, cluster of issues).
He does not say anything about the (different) views of Kuhn (1970) and Feyerabend
(1978) on this problem. Nor does he indicate that there are a number of ways of coping
with incommensurability. For one thing, it is a common understanding of Kuhn’s position
that incommensurability means that rival theories cannot be comparatively appraised,
whereas Kuhn only meant to suggest that the appraisal cannot be effected by a neutral
set of procedures and facts. Further, the real problem here is incommensurability of
meaning, which is said to arise because a term gets its meaning from its place in a
theory, with the result that it will mean something different for competing theories.
It is commensurability of meaning that Feyerabend dispairs of achieving when we are
confronted with the task of comparing rival theories.
In maintaining that the meaning of a scientific term is determined entirely by its role
in the network of statments that comprise the theory, proponents of incommensurability
such as Feyerabend adopt an implausibly extreme view of conceptual role semantics.
Not only does such a view of meaning fail to square with the history of science (pace
Feyerabend; see Nersessian, 1984), but it makes incommensurable theories unlearnable.
As Walker and Evers (1988) point out, we cannot possibly understand any part of
a theory unless we properly comprehend the entire theory, but such comprehension
itself is impossible for us until we have learned the various parts. For conceptual role
semantics to be viable, it must permit the meaning of a term to be determined only partlyby its conceptual role in a theory. When this is the case we can avoid the unattractive
consequence of incommensurability that rival theories are not comparable in terms of
some shared criteria.
A different way of avoiding the problem of meaning incommensurability involves
adopting a theory of meaning that shifts its focus from the idea of meaning to that
of reference. Putnam (1979) has developed one such promising theory which provides
for continuity of reference while admitting incommensurability of meaning. Although
not without its problems, this theory does concern itself with “. . meaning that is
pretty natural for a wide range of linguistic practices, and which does not invite talk of
incommensurability. It is the kind of theory that scientific realists about entities need”(Hacking, 1983, p. 91).
Yet a different strategy for avoiding the problem of incommensurability has been
fashioned by Nersessian (1984). She believes that the post-positivist tendency to
conceptualize and solve the problem of incommensurability in terms of the philosophy
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 77/97
Beyond Paradigms 577
of language does not square with meaning change in actual scientific practice. From
a cognitive-historical examination of meaning changes in various electromagnetic field
theories that span different paradigms she concludes that, while the meaning of
“electromagnetic field” differs significantly across theories, each concept shares partof the meaning of its predecessors. This suggests that there is meaning variance, but
with a significant amount of commensurability.
It would appear, then, that philosophers have made sufficient progress in solving
the problem of meaning commensurability not to dispair of constructing, evaluating
and integrating different and possibly competing theories. There is no need to seek a
meta-analytic haven from the spectre of incommensurability.
Met hodology as Theory
As noted above, Glass rejects the influential conception of methodology as an a priori
enterprise and maintains it is an empirical endeavor. I believe he is right to criticize
a priori methodology, but wrong to suggest that methodology is just an empirical
enterprise. Viewing methodology as a priori knowledge is dubious because the notion of
a priori knowledge is itself highly questionable. The a priori categories of analytic truth,
synthetic a priori truth, logical truth and mathematical truth have all been subjected to
serious criticism within philosophy (e.g., Haack, 1974; Kitcher, 1983; Quine, 1953). But
while one can accept Glass’ claim that methodological statements are genuinely refutable
conjectures, it does not follow that methodological assertions are evaluated solely on
empirical grounds. The reason for this is that, in science, procedural knowledge, no less
than substantive knowledge, has the status of warranted conjectural theory and that,
broadly speaking, both kinds of knowledge are validated using the methods of science
(cf. Walker & Evers, 1988). Because substantive scientific theories are underdetermined
by the relevant data, they are additionally evaluated on superempiri~al dimensions such
as explanatory power, systemic worth and fruitfulness. We should not expect it to be
any different with our methodological theories. To be sure, empirical evidence will have
some bearing on assessing the soundness of methodological rules, but these assessments
will be highly inconclusive without invoking appropriate superempirical criteria. I have
already argued that the deeply flawed nature of Fisherian outcome studies prevents Glass
from being able to carry out sensibiy his empirical scrutiny of the merits of those studies
deemed methodologically suspect on a priori grounds. Additionally, I note that, where
Glass admonishes researchers for engaging in what he thinks are a priori methodological
debates, it may very well be the case that some of the disputes are really a posteriori
inter-theory debates about contingent matters.
Glass and his associates (Glass, McGaw & Smith, 1981; Smith et al., 1980) have
repeatedly emphasized that meta-analysis recommends itself over traditional review
procedures because of its objectivity. This is said to be achieved by proscribing judgment
strategies in meta-analysis which will prevent biases entering into the results it produces.
Here Glass appears to saddle himself with the untenable empiricist doctrine of logicism,which maintains that knowledge claims can be produced by making use of data and logic
only. However, logicism has been shown to fail (e.g., Maxwell, 1975) precisely because
human judgments have been found to be an important component in the production
of worthwhile knowledge. In fact, appeal to the various superempirical criteria in
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 78/97
578 3. D. HAIG
appriasing theories is one striking way in which human judgments necessarily enter
into the knowledge production process.
bet a-analy st and Pol i cy
Glass and Kliegl (1983) maintain it is naive to believe that rational policy decisions
must be based on relevant knowledge from well established theories. They echo Meehl’s
(1978) judgment that “most so-called “theories” in the soft areas of psychology (clinical,
counseling, social, personality, community, and school psychology) are scientifically
unimpressive and technologically worthless” (p. 806). However, by reinterpreting
such theories as the modest products of evaluative research, and submitting them to
meta-analysis where appropriate, Glass and Kliegl believe that useful knowledge can be
provided for decision makers. In this way they believe they can overcome researchers’
habitual tendency to engage in “partisan squabbles and theoretical hot-dogging when
attempting to inform policy makers” (p. 35). However, meta-analysis has generally
failed in its attempt to establish clear judgments of pragmatic worth for policy makers.
Different meta-analyses in the same subject area have often produced different results.
For example, the constructive replication of the initial Smith et al . (1980) meta-analysis
of psychotherapy outcome studies by Prioleau, Murdoch and Brody (1983) produced
discrepant conclusions. Because meta-analyses are unavoidably replete with human
judgments over which researchers will differ, it is only to be expected that they will
be unable to provide clients with unambiguous messages.
Also concerning policy, it is worth noting briefly that Glass and Kliegl make
inappropriate use of Habermas. They claim “Habermas (1971) argued convincingly that
the knowledge-constitutive interests that determine, in part, the selection of a certain
methodology for science can be derived from the structure and pragmatic needs of the
society in which the science exists” (p. 35). However, Habermas’ (1971) critical-theoretic
analysis of cognitive interests relies uncritically on an inappropriate empiricist theory
of science (a point Habermas himself now concedes). Relatedly, Habermas’ insistence
that our knowledge-constitutive interests are somehow transcendental and a priori is
implausible, and clearly should be anathema to Glass and Kliegl.
This point aside, it is important to stress that methodologies and social institutionsdo relate to each other in mutually supporting ways. As part of an empiricist conception
of inquiry meta-analysis helps to serve as a prop for our extant social institutions by
providing them with the conceptual resources that help maintain the status quo. One way
meta-analysis reinforces the status quo stems from its acceptance of, and reliance upon,
fragmented literatures spawned by narrow specialist research. An appreciation of the
need for action seldom eventuates from fragmented knowledge. With its characteristic
focus on outcome studies, m&a-analysis is simply not capable of giving us broad,
coherent perspectives.
Meta-analysis further reinforces the status quo by restricting its attention to outcome
studies which focus on phenomenal appearances and refrain from considering underlyingcauses. This willingness to stop short of attempting to tell decent causal stories
contributes to a general inability to regard educational programs and social institutions
more generally as structurally problematic, and results in an absence of knowledge of the
relevant causes which would be the objects of strategic social change.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 79/97
Beyond Paradigms 579
A third way in which meta-analysis reinforces the status quo stems from the fact that
it is not critically aim-oriented. By willingly accepting clients’ goals, evaluation research
employs meta-analysis as part of an instrumental rationality concerned to devise and
follow efficient means to clients’ ends. As such, meta-analytic methodology affords usneither the inclination nor the ability to challenge the goals of clients, programs or social
institutions.
Glass and Kliegl’s (1983) remarks on research policy are consistent with the prevailing
view, which arises from a coupling of empiricist epistemology and liberal political
theory (cf. Unger, 1975). According to this view research policy is an essentially
political reaction to the demands of pressure groups and to the “externalities” of
economic activity. It operates, therefore, as an expedient and instrumentally rational
endeavor. What is needed, by contrast, is a prudent research policy that accommodates
the cognitive dynamics of good science. Such a research policy would be critically
aim-oriented, cognizant of the considerable time often needed to produce well-developedtheories, and mindful of the fact that the application of mature theories often involves
further and specific basic mediating research. Of course, a satisfactory realization of
science transformed along these lines would depend crucially upon changes to our
specifically scientific and broader social institutions.
Theory and Integration
It is important to realize that meta-analysis does not integrate research findings in
the constructive sense of combining and systematizing parts into larger wholes. Rather,
it functions as a reductionist enterprise, and in two related ways: it either ignores
explanatory theories that purportedly refer to hidden causal mechanisms, or it factors
out the specifically postulational component of such theories (thus helping to present
the empiricist picture of strict cumulative progress). Clearly, serious attempts to integrate
worthwhile knowledge cannot be achieved statistically through meta-analysis, but instead
will require “qualitative” densely reasoned efforts to construct suitable theories. For, it
is well-structured theories that are the real bearers of significant integrated knowledge.
With its empiricist “anti-theory” bias, meta-analysis neglects two of the most important
vehicles we have for conveying worthwhile knowledge: global theories and postulational
theories.
Global Theories
Cook and Campbell (1979) claim modern philosophers of science such as Popper,
Kuhn and Feyerabend have exaggerated the role of comprehensive theory in advancing
scientific knowledge. However, I believe such an emphasis is part of a general
post-positivist attempt to rectify empiricism’s tendency to concentrate on small-scale
observational theories as cognitive entities separated from other parts of the scientificendeavor such as methodology and metaphysics. The idea of a global theory, as
developed by Hooker (1987), suggestively articulates the structure that fits a good
number of our scientific theories. Global theories (e.g., quantum theory, evolutionary
theory, radical behaviorism) are large-scale theories that are typically composed of a
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 80/97
580 B. D. HAIG
partial world view, a methodology, and a theory of instruments. They also specify
what is observable, provide a language for data reports, and will often make use of
additional theories deemed necessary for their application. Good global theories exhibit
the systemic feature of integrating their cognitive components into tight conceptualstructures. They are, therefore, important realizations of integrated knowledge.
Interestingly, radical behaviorism stands as one of our best examples of a well-
structured global theory. It contains a world view comprising a deterministic, physicalistic
and monistic metaphysics; it also includes an ethic of persons denied freedom, dignity
and moral responsibility and a vision of an ideal society in Skinner’s novel Walden
Two. Its methodology comprises an inductive account of method and an instrumentalist
conception of theories, while its theory of instruments justifies experimental use of
the conditioning chamber and cumulative recorder. Radical behaviorism also specifies
that operant and respondent behavior and their controlling stimuli are the legitimate
observables, and that the language for data reports about such observables is constrained
by an operationally defined terminology. Finally, radical behaviorism makes some appeal
to additional theories in its applications, as for example, the theory of placebos in
evaluating behavior therapy. Although Skinner’s brand of behaviorism will strike
many as unacceptably empiricist, it does, nevertheless deserve to be regarded as a
well-structured global theory with a high degree of internal coherence. What it does
lack, however, is the important virtue of explanatory coherence, a notion which I will
come to shortly.
Clearly, one better diagnostic alternative to evaluating the quality of outcome studies
through meta-analysis would involve attempts to assess how effectively global our extant
theories are. I expect that many of them would turn out to be minimally global with the
various relevant components being related by little more than conjunction. Attempting
to make such theories effectively global would help us fathom the extent to which the
conjuncts mutually cohere. I have already indicated that I think Fisherian experimental
methodology does not consistently square with the interactive nature of the social reality
it helps us investigate. That we have failed to appreciate this point stems in good part
from the fact that our methodological and ontological commitments largely remain
uncoupled in weakly structured global theories.
Po.~tu~atio~a~Theories
Postulational, or deep-structural, theories are those theories that purport to refer
to hidden generative mechanisms. Much of the world’s furniture is hidden from our
view; so, if we want to know how things operate rather than settle for an account of
their surface features, then we must fashion deep-structural or postulational theories.
These are the theories that realism advocates and that we need to understand and
effect change. Postulational theories recommend themselves to us because they have
more epistemic virtues than theories about observables, including outcome studies.
According to empiricism, empirical adequacy, or factual support, is the mark of a
good theory. However, scientific realism seeks to determine the general excellence of
theories, and to this end will invoke the various superempirical criteria that have been
mentioned above. Of these, explanatory power, internal and external consistency and
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 81/97
Beyond Paradigms 581
systemic worth in particular, combine to give us some measure of a theory’s explanatory
coherence. In the next section I want to take up the idea that explanatory coherence is
central to theory justification.
Integration as Explanatory Coherence
Foundationalism is the historically influential theory of epistemic justification that
claims we are justified in believing those theories that are appropriately related to a
privileged source. Empiricist researchers characteristically take observational data to be
that source for the empirical sciences and rely almost exclusively on empirical adequacy
as the measure of good theory. Foundationalist epistemologies are now widely rejected
and coherentist justification has emerged as an atttractive alternative (BonJour, 1985;
Williams, 1977). Briefly, coherentism maintains that a belief is justified in virtue of
its coherence with other accepted beliefs. One contemporary version of coherentism,
explanationism, asserts that coherence is determined by explanatory relations and
that all justification aims at maximizing the explanatory coherence of belief systems
(Lycan, 1988).
Today the major challenge to explanationism comes from reliabilism, which asserts
that a belief is justified to the extent that it is acquired by reliable methods (Goldman,
1986). However, I believe with Thagard (1989) that both explanationism and reliabilism
can fruitfully be combined within a broad coherentist theory of justification and that such
a theory has positive application in empirical research contexts. In this final section of
the chapter I want to show how such a theory of justification can operate in social science
research. The attainment of explanatory coherence is an important, but neglected part
of research integration which simultaneously contributes to the unity and justification
of knowledge.
Retr oducti ve M eth od
In order to show how explanatory coherence can be implemented at a methodological
level, I need to outline the theory of scientific method that serves as my orientingframework. This theory of method is essentially retroductive in character and claims
that science often does, and should, proceed through a number of phases: regulated
by a developing problem comprising a set of empirical and conceptual constraints
(Haig, 1987; Nickles, 1981), relevant observed data are obtained, and are then
analyzed for potentially interesting patterns (Tukey, 1980). Once established, these
data patterns are explained by postulating the existence of an underlying causal
mechanism, through a retroductive, theoretical reasoning process (Curd, 1980). From
a positive judgment of the initial plausibility of such an existential hypothesis, attempts
are then made to elaborate on the nature of that mechanism, often by way of constructing
plausible models from an appropriate and familiar source (Harre, 1976). Partly becausetheories are underdetermined by the relevant data, subsequent attempts to test the
developed theory against its competitors will not in general be decisive (cf. Harding,
1976) and theory appraisal will have to be undertaken on dimensions in addition
to empirical adequacy. To repeat, these superempirical features will include initial
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 82/97
582 B. D. HAIG
plausibility, fertility, explanatory depth, unifying power, internal and external coherence
and practical efficacy (Churchland, 1985; Kuhn, 1977). It should be noted that this
retroductive account of method gives explicit attention, and assigns major importance
to, the three contexts of theory generation, theory development and theory appraisal.
Coherence Justi fi cati on and Theory Generat i on
In indicating how a broad conception of coherence justification can operate in the
prosecution of social science research I shall largely confine myself to the retroductive
method’s context of theory generation. As I have just noted, the basic goal in this
context is to generate plausible theory to explain significant data patterns. The data
analysis phase itself is appropriately viewed as a two-stage compound affair where the
patterns thrown up by exploratory data analysis are critically checked through the use of
confirmatory data analysis procedures (Tukey, 1980). Exploratory data analysis involves
descriptive, and frequently quantitative, detective work designed to reveal patterns or
structure in the data sets under scrutiny, such data often being displayed visually in
(semi-) graphical form. It is important that researchers give extended attention to this
exploratory phase, because, in securing a heavy information yield from our data, they
are more likely to throw up the provocative data patterns that occasion the need for
explanatory theory.
Although carefully conducted exploratory analyses may, in addition to suggesting data
patterns, actually carry some measure of validity (thus calling into question the standard
exploratory/confirmatory contrast), it will normally be appropriate to check on our
emergent data patterns through use of confirmatory data analysis procedures. Computer
intensive resampling methods such as the jackknife, the bootstrap and cross-validation
(Diaconis & Effron, 1983; Effron & Gong, 1983) constitute one important set of
confirmatory procedures that fits well within a coherentist framework. By exploiting
the massive computational power of the modern high-speed computer, these methods
free us from the restrictive assumptions of normal statistical theory and permit us to
gauge the reliability of chosen statistics by making something like a billion calculations
on, say, 50 data points. The jackknife, for example, is an important attempt to establish
the accuracy of a computed estimate of some quantity of interest such as a mean, or a
standard deviation, or a correlation. It proceeds by removing one observation at a time
from the original data set and recalculating the statistic of interest for each of the reduced
data sets. The variability of the statistic across all the truncated data sets can then be
described by giving us an empirically obtained measure of the reliability or stability of
the original estimate.
The important point to be made about reliability checks like those afforded by the
jackknife is that they can be made part of a coherentist approach to justification, where
the coherence is provided by consistency of test outcomes. Here the reliability checks
on our suggested data patterns constitute a validating strategy. Our willingness to acceptthe results of such data analyses is in accord with what Thagard (1989) calls “the principle
of data priority”. This principle asserts that statements about observational data have a
degree of acceptability on their own. Such claims are not indubitable, but they do stand
by themselves better than claims justified solely in terms of what they explain. What
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 83/97
Beyond Paradigms 583
justifies the provisional acceptance of data statements is that they have been achieved
by reliable methods. What justifies our provisional belief in the patterns thrown up by
exploratory data analysis is their reliabilist confirmation through use of computer based
consistency tests.
It is important to stress here that the justification of belief in such patterns is
heuristic and forward looking: we provisionally accept such beliefs for purposes of
constructing a theory that will explain the data patterns, but our knowledge of the
data patterns will receive the further required justification, if and when, they enter
into the explanatory relations of the theory that contains them. According to our
coherentism which links explanationism and reliabilism, data statements are not fully
acceptable without being linked to plausible explanatory propositions. Here I should
state that, even though reliabilism embraces the principle of data priority, it is not a
modest foundationalist position. This is because foundationalism insists that justificatory
relations are uni-directional (going from basic propositions to nonbasic propositions),
whereas the strategy of explanatory justification just outlined is bi-directional and
mutually enhancing.
The analysis and confirmation of striking data patterns provides a natural stimulus to
the generation of new theory which helps explain why the data pattern as they do. Now,
the type of ampliative inference involved in the generation of new explanatory theory
is not inductive, but retroductive. Retroductive reasoning may be characterized briefly
as follows: some observations are encountered which are surprising because they don’t
follow from any accepted hypothesis; we come to notice that those observations would
follow as a matter of course from the truth of a new hypothesis in conjunction with
accepted auxiliary claims; we therefore conclude that the new hypothesis is plausible
and thus deserves to be seriously entertained and further investigated. This typical
depiction of retroductive reasoning focuses on its logical form and, as such, is of
limited value unless it is conjoined with a set of regulative principles which enable
us to view retroduction as a pattern of inference, not just to any explanations, but to
the most plausible explanations.
These principles, which function as constraints within our developing research
problem, will be variously empirical, metaphysical and methodological in nature. In
judging the initial plausibility of our new theories we are in effect making prospective
judgments about their pursuit-worthiness, and we do this in the first instance byfathoming whether our theories are the products of sound retroductive reasoning.
Here I want to press the point that explanatory coherence is central to the.judgments of
the initial plausibility of our retroductively obtained theories. According to the program
of explanatory coherence that I have adopted, the explanatory component of a theory
coheres with the data component if the former explains the latter. In this regard, our
explanatory theories will normally be expected to explain more than one data pattern or
piece of empirical evidence. A theory which explains two different data patterns or types
of data is said to be consilient, and a theory which explains three classes of facts is judged
more consilient, and so on (Thagard, 1989). In this way consilience captures the idea of
explanatory breadth, which is an important aspect of explanatory coherence. In fact,the explanatory coherence account of justification I have adopted employs the notion
of consilience to capture the important methodological principle that a theory is better
supported, the wider the variety of evidence for it. And, in general, the more consilient
a theory is, the greater its initial plausibility will be.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 84/97
584 8. D. HAIG
In showing how explanatory coherence justification applies in the context of theory
generation, I have ignored the contexts where theories are developed and subsequently
subjected to concerted appraisal. It should be made clear, however, that Thagard’s
(1989) program for implementing judgments of explanatory coherence works throughinference to the best explanation as it applies to the evaluation of mature theories.
“Inference to the best explanation” is a useful expression that covers the protean
inference involved in the multi-criteria1 appraisal of mature theories. By focusing on
theory generation my concern has been with inference to one (or more) plausible
explanations, not inference to the best explanation. Finally in this section I note that, to
be consilient, a theory must be postulational. That is, it must invoke causal mechanisms
in order to have a chance of explaining the various data displays it attempts to unify.
Because empiricist observational theories seldom embrace the causal mechanisms that do
the crucial explanatory work, they are generally unable to achieve an acceptable level of
consilience. Outcome studies, meta-analyzed or not, also suffer from this deficiency.
Conclusion
The currently popular practice of meta-analyzing outcome studies has been called into
question in this chapter. My critique has focused on Glassian meta-analysis because,
in addition to being widely used in educational and psychological research, it boasts
a better developed methodology than other approaches to meta-analysis. However,
Glass’ methodology is deficient in many respects, principally because it embraces
an unacceptable empiricist conception of both science and evaluation. I have alsosuggested that the poverty of Fisherian outcome studies renders their meta-analyses
by whatever form a pointless research endeavor. The alternative thesis of this paper is
that well-structured theories, rather than meta-analyzed domains, are the real bearers
of significant integrated knowledge. Consistent with Walker and Evers’ unity of research
thesis, I have sought to show how coherence justification can fruitfully be applied in
the course of generating empirically-based explanatory theories. The brand of scientific
realist philosophy adopted in this chapter takes explanation and unity to be cardinal
epistemic goals. With this in mind I have recommended linking these twin virtues in an
explanatory coherentist theory of knowledge justification.
References
Bangert-Drowns, R. L. (1986). R eview of developments in meta-analytic method. Psychological Bullet in.
99, 388-399.
Bhaskar, R. (1978). A reali st rheory of scienceSecond edition). Sussex: Harvester Press.
BonJour, L. (1985). The structure of empiri cal know l edge. Cambridge, MA: Harvard University Press.
Bruno, J. E. & Ellett, F. S. (1988). A core analysis of meta-analysis. Quali ty and Quanti ty , 22, 111-126.
Churchland, P. M. (1985). The ontological status of observables: In praise of superempirical virtues. In
P. M. Churchland & C. A. Hooker (Eds.), Images of science (pp. 3547). Chicago: University of
Chicago Press.
Cook, T. D. & Campbell, D. T. (1979). Q uasi-experimentation. Chicago: Rand McNally.
Cook. T. D. & Leviton, L. C. (1980). Reviewing the literature: A comparison of traditional methods with
meta-analysis. Journal of Pers~naIi ~y, 48, 449--l ‘j 2.
Curd, M. V. (1980). The logic of discovery: An analysis of three approaches. In T. Nickles (Ed.), ~c~e~~~~c
d~c~very. logic and rati onali~ (pp. 201-219). Dordrecht: Reidel.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 85/97
Beyond Paradigms 585
Diaconis, P. & Efron, B. (1983). ~omouter”intensive methods in statistics. Scienti fi c American, 248.
De Waele, J.-P. & Ha& R. (1979). Autobiography as a psychological method. In G. P. Ginsburg (Ed.),Emerging str at egies i n social psycholoni caZ research (pp. 177-209). Chichester: Wiley.
Efron, 6. & Gong, G. (1983). -A-leisurely look at the Gdotstrap, the jacknife, and cro&-validation. American
Stat i sti cian, 37, 3&B.
Erwin, E. (1984). Establishing causal connections: Meta-analysis and psychotherapy. M idw est Studies in
Phil osophy, Vol. 9 (pp. 421-436). Minneapolis: University of Minnesota Press.Eysenck, H. J. (1984). Meta-analysis: An abuse of research integration. Special Educution, 18, 41-59.
Feyerabend, P. K. (1978). Against method. London: Verso.Glass, G. V. (1972). The wisdom of scientific inquiry on education. Journal ofResearchn Sci ence Teaching,
9, 3-18.Glass, G. V. (1976). Primary, secondary and meta-analysis of research. Educat i onal Researcher, 5, 3-8.
Glass, G. V., McGaw, B., & Smith, M. L. (1981). M et& analy sis i n social research. Beverly Hills: Sage.Glass, G. V. & Kliegl, R. M. (1983). An apology for research integration in the study of psychotherapy.
Journal qf Consult i ng and Cli nical Psychology, 51, 28-41.
Goldman, A. I. (1986). Epistemology and cognit ion. Cambridge, MA: Harvard University Press.
Gottfredson, G. D. (1984). A theory-ridden approach to programme evaluation. American Psyc~o~og~t,39, 1101-1112.
Haack, S. (1974). Deviant logic. Cambridge: Cambridge University Press.Habermas, J. (1971). Know l edge and human i nt erest s. Boston: Beacon Press.Hacking, I. (1983). Represen~ng and i nt erv eni ng. Cambridge: Cambridge University Press.Haig, B. D. (1987). Scientific problems and the conduct of research. Educati onal Phif osophy and Theory,
19, 22-32.
Harding, S. G. (Ed.) (1976). Can theories be refuted? Dordrecht: Reidel.Harr$, R. (1976). The constructive role of models. In L. Collins (Ed.), The use of model s i n the social
sciences (pp. 16-43). London: Tavistock.Hooker, C. A. (1987). A reali sti c theory of sci ence. Albany: State University of New York Press.Kitcher, P. (1983). The nature of mathemat ical know l edge. Oxford: Oxford University Press.Kuhn, T. S. (1970). The structure of sci enti j c revol uti ons (Second edition). Chicago: University of Chicago
Press.Kuhn, T. S. (1977). The essential tension. Chicago: University of Chicago Press.Lewontin, R. C. (1974). The analysis of variance and the analysis of causes. Ameri can Journal of Human
Geneti cs, 26, 400-411.
Light R. J. & Smith, P. V. (1971). Accumulating evidence: procedures for resolving contradictions amongdifferent research studies. Harvard Educational Review, 41 429-471.
Lycan, W. G. (1988). Judgement and just if icat ion. Cambridge: Cambridge University Press.Maxwell, G. (197.5). Induction and empiricism: A Bavesian-freQuentist alternative. M innesota Studi es i n
the Ph~l o.~op~y of Science, Vol . 6 (pp. ‘106-165). Minneapolis: cniversity of Minnesota Press.
Meehl. P. E. (1967). Theory-testing in psychology and physics: A methodological paradox. Philosophy of
Science, 34, 103-11.5.Meehl, P. E. (1970). Nuisance variables and the ex post facto design. M innesota Studi es i n the Phil osophy
of Science, Vol. 4 (pp. 373-402). Minneapolis: University of Minnesota Press.
Meehl, P. E. (1978). Theoretical risks and tabular asterisks: Sir Karl, Sir Ronald, and the slow progressof soft psychology. fournal of Consult i ng and Cli nical Psychology, 46, 8Of& 334.
Nersessian, N. J. (1984). Faraday to Einstein: constructi ng meaning i n scienti fi c theories. Dordrecht: MartinusNijhoff.
Nickles, T. (1981). What is a problem that we might solve it? Synthese, 47, 85-118.Oakes, M. (1986). Statistical nference. Chichester: Wiley.Prioleau, L., Murdock, M., & Brody, N. (1983). An analysis of psychotherapy versus placebo studies. The
Behavi oral and Brai n Sciences, 6, 275-310.
Putnam, H. (1979). Phi losophical papers, Vol. 2. Cambridge: Cambridge University Press.Quine, W. V. (1953). From a logical point of vi ew . Cambridge, MA: Harvard University Press.Slavin, R. E. (1984). Meta-analysis in education: How has it been used? Educational Researcher, 13,
6-15.
Smith, M. L., Glass, G. V., & Miller, T. I. (1980). The benefi t s of psychot herapy . Baltimore: Johns Hopkins
University Press.Sober, E. (1984). The nature ofsefection. Cambridge, MA: MIT Press.Thagard, P. (1989). Explanatory coherence. The Be~a~~oru~and Brai n Sciences, 12, 435-502.
Tukey, J. W. (1980). We need both exploratory and con~rmatory. Ameri can Sfat~ t i c~un, 34, 23-25.
Unger, R. M. (1975). Know ledge and poli ti cs. New York: The Free Press.Urbach, P. (1985). Randomization and the design of experiments. Phifosophy ofScience, 2, 256-273.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 86/97
5% B. D. HAIG
Walker, J. C. & Evers, C. W. (1988). The epistemological unity of educational research. In(Ed.), Educufi onai research, methodol ogy and measurement : an i nt ernat i onal handbook
Oxford: Pergamon Press.
J. P. Keeves
(pp. 28-36).
Williams, M. (1977). Groundless befief. Oxford: Blackwell.
Biography
Brian D. Haig is Senior Lecturer in the Department of Psychology at the University
of Canterbury, New Zealand. His research interests cover Research Methodology,
Science Education, and Philosophical Psychology. His work appears in journals such
as Educat i onal Ph~i osophy and Theory, the Journal of Research in Science Teaching;
Phi l osophy of Social Science, and the ~u~i e~~n f Peace Proposals.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 87/97
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 88/97
588 D. T. CAMPBELL
other. I shall come back to this later. What I first want to note is that there are also
hermeneutic doctrines of method, and that these doctrines overlap heavily with the
actual practices of the physical sciences, and with doctrines of method and epistemology
in non-foundationalist post-positivist (but science-admiring) philosophers of science such
as Quine and Hesse.
To make this point, one must distinguish between validity-seeking hermeneutics and
what I will call “ontologically nihilistic” hermeneutics. Let me characterize these (with
a confession that I joust best with over-clarified straw men or, as Weber called them,
“ideal types”). The ontologically nihilist hermeneutists worship creative novelty in
interpretation as an end in itself, banning disputation as to which are more valid.
The validity-seeking hermeneutists regard such creative novelty as a means to better
interpretations, to be achieved by disputation within an interpretive community, using
coherence arguments and other hermeneutic principles. The hermeneutic ontological
nihilists argue with great sophistication that the concept of truth is incoherent,
and on this basis have concluded that the goal of truth, and argumentation about
comparative plausibility of competing interpretations, should be given up. I do not
mean to ridicule this epistemology. Traditional empiricist/logicist philosophers of
science are also in agreement that one cannot compare beliefs with the referents
of those beliefs (e.g., the Ding an Sich) as a truth test. Instead, one must compare
beliefs with beliefs (perhaps presuming privileged status for perceptually generated
beliefs). This also holds for non-foundationalist post-positivists such as Quine and
Popper, who advocate a correspondence meaning for the concept of “truth”, but
who also emphasize the unavailability of correspondence as a truth test for specific
beliefs (e.g., Quine/Duhem equivocality). They also concede to the skeptics that for
referential descriptive knowledge, the requirements of “fully justified” and “known-to-be
true”, belief cannot be met. But they, and the validity-seeking hermeneutists, advocate
continuing the presumptive, fallible, dialectical, search for validity, a search which also
characterizes ordinary perception and learning.
As validity-seeking hermeneutists, I identify Schleiermacher, Dilthey, Weber,
Habermas (I have found a few paragraphs and a footnote in his 1983, p. 251-61, and
note 8, particularly useful), and Geertz (1973, 1983). From my frustrating samplings,
I have judged Gadamer and Ricoeur as, in net, ontological nihilists, but am not
prepared to cite chapter and verse. (Perhaps this is an ideal type with no occupants,not even the “paradigms theorists” of education and the social sciences.)
The hermeneut i c ci r cl e. The methodological tactic most recurrently cited is the
“hermeneutic circle” (or cycle, or spiral), which I have tried to make sense of as
“part-whole iteration” (Campbell, 1988, pp. 478, 505-507), developed in the fallibilist,
conjectural domain of deciphering ancient texts for which no dependable dictionaries
were already available. A guess as to the purpose and meaning of the whole guides
guesses as to the interpretation of specific words, which guesses, if tentatively trusted,
lead to a revision of the guess at the whole, which again leads to revisions of translations
of the parts, etc. The non-foundationalism of this process is dramatized when an
interpretative community decides that a particular word is a copyist’s error, and that they,the interpretive community, know better the original author’s intention than does the
copyist’s text, or even that the author made a slip of the pen in a hypothetical “first” text.
The hermeneutic circle, and all hermeneutic methods, are specific forms of coherence-
maximizing strategies, properly heralded in the present essays as the new consensus-
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 89/97
Beyond Paradigms 589
principle of mainstream Anglo-American post-positivist philosophy of science. I believe
that it has emerged from a dialectical process of mutual criticism within that tradition,
uninfluenced by continental philosophy. But the continental heirs of Hegel had this
emphasis first.
It was through long ago conversations with my philosopher friend at Northwestern,
Errol Harris, that I learned the Hegelian emphasis on coherence, and recognized it as
what Quine (in the last 15 paragraphs of “Two Dogmas”, 1951; Campbell manuscript,
1990) and I (1966) were employing. But for Harris’s Hegel it also went beyond a strategy
for knowing to a definition of “truth” with the contrast pair being a coherence definition
of truth versus a correspondence definition. For those cited by Evers and Walker (see
also Lehrer, 1974, 1990), the contrast pair is coherentism versus foundationalism. While
it is the foundationalists who most clearly represent a correspondence definition of
truth (and perhaps even the foolish faith that it can be implemented as a truth test
for some specific beliefs), it seems to me that OUT coherentist tradition is compatible
with a correspondence goal of truth (and hence a correspondence definition of the
meaning of the term “truth”) even though our epistemology assures us that we will
never meet it. Certainly defenders of Tarski’s version of the correspondence definition,
such as Popper and Quine, have never assumed its availability as a truth-test for specific
beliefs. We accept a quasi-Hegelian surrogate goal of increasing coherence even if we
regard this as merely our best available symptom of truth (particularly if the cumulated
evidence has been produced by scientific disputation involving competing hypotheses
about “reality”. Similarly for van Fraassen’s “empirical adequacy” as a goal [Paller &
Campbell, 19891).
I feel sure that in that vast literature there are many other hermeneutic principles, but
I do not recollect encountering such a list. (This is a request for help.) I will provide a
few more here. Congruent with Evers’ observations, I will embed these in texts from
Quine. But I regard these Quine quotes as brilliant ambivalent outbursts from his faith
in logic (a faith which logic itself has undermined), and feel sure that, overall, they are
fully consistent with the older, validity-seeking hermeneutics. And where I differ from
Quine (the 10 per cent of Quine which I have read), I feel that the direction of my
difference is in a quasi-hermeneutic direction. (Not being erudite about hermeneutics, I
am in danger of over-identifying myself with it. One hermeneutic tendency I clearly reject
is a “text-foundationalism”, a tendency to limit ones interpretative goals and permissible
resources to a specific text, excluding beliefs about the author, his audience, and his
times. I am using the term metaphorically for principles in validity-seeking scholarship,
shared by humanities and sciences.)
Omni fa l l ib i l i st t rust . Our predicament as knowers is such that, in improving the validity
of our beliefs, we have no other strategy available than to trust the great bulk of our
beliefs while we revise a small subset of them. But none of our beliefs are foundational.
All are potentially open to revision. What substitutes for foundations is the bulk of our
other relevant beliefs, none of them individually foundational. Our goal in revisions is
to increase coherence.
Quine’s repeated reference to Neurath’s boat (or raft) which must be repairedwhile afloat at sea epitomizes this perspective. “Our boat stays afloat because at
each alteration we keep the bulk of it intact as a going concern” (Quine, 1960, p.
3). And from the precious last 15 paragraphs of “Two Dogmas”, some illustrative
fragments: “Our natural tendency [is, and should be] to disturb the total system as
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 90/97
590 D. T. CAMPBELL
little as possible. ” “No statement is immune to revision.” “Even . . . the logical law
of the excluded middle.” I have epitomized this Quinean message as the “doubt-trust
ratio” (Campbell, 1988, pp. 363-365, 477-482) estimating usually to be 99% trust to 1%
doubt (1988, p. 318), although dropping to 85% or 90% trust in a scientific revolutionin astronomy or physics (1988, p. 482). (The slogan “omnifallibilist trust” has been
introduced in Cook & Campbell, 1986.)
Pattern matching. All of our observations are both fallible in principle and have some
inevitable imperfection in practice. The beliefs and theories we relate to our observations
are incomplete and oversimplified. The impressive degree of congruence between them
found in the best of science cu~~~~ have been achieved by a perfect, foundational
relationship at any point. Instead it is achieved by a pattern matching, spreading the
fringe of imperfection over all of the presumed points of contact between observation
and belief or theory (Campbell, 1966). Identification by “feature detection” employs a
local pattern matching, and does not negate this universal hermeneutic principle.
I ncreasing cor respondence w i t h ncreasing scope. In the hermeneutics of translations, a
word in isolation is less confidently read than a word embedded in a sentence. Still better,
if it be embodied in a paragraph, or a book or a literature. All words are equivocal,
polysemous, indexical to some extent. Context is needed, and the larger the context the
better. From recent Bible translation comes the example of a Hebrew word which occurs
only once in the Pentateuch, and whose reading has been under dispute. In the last 50
years, there have been discovered cuneform libraries of a pre-Hebraic closely related
Semitic language, Ugaritic, in which a cognate word is frequently enough used to have
provided enough context so that the scholars have achieved a working consensus on its
interpretations. This consensus is now being used to achieve a consensus on the orphan
word in the Biblical text.
To Quine is attributed a “holism” which amounts to total suspension of consensus
formation on the relative validity of beliefs. This is both criticized, and used to legitimate
an ontologically-nihilistic relativism. I find this a wrong reading. Here is the crucial
paragraph from “Two Dogmas” (Quine, 1951, p. 39):
RusselI’s concept of definition in use was . . an advance over the impossible term-by-term empiricism
of Locke and Hume. The statement, rather than the term, came with Russell to be recognized as
the unit accountable to an empiricist critique. But what I am now urging is that even in taking the
statement as unit we have drawn our grid too finely. The unit of empirical significance is the wholeof science.
My gloss on this, which Quine accepts, is as follows: in practice, science is never going
to be mapped in an apodictic or foundational way to empiricism, nor to observations
(whether in sense data terms or a language of ordinary objects) nor to “stimulations”.
Nonetheless the fallible and underdetermined match between science and the empirical
steadi l y i ncreases as science advances, and as broader samples of science are employed
in the context of interpreting specific scientific statements. The empirical accmmtabil i ty
of science improves as one moves from “term by term” to “statement by statement”,
i.e., the “statement as the unit of accountability” was an improvement over the
“term-by-term empiricism”, but still not an encompassing-enough unit. Larger units
of science (theories for specific domains, multi-domain integrated theories, etc.) would
be still better and, by metaphoric extension, “the whole of science” best of all. This
metaphoric extension refers (a la Peirce’s conceptualization of truth) to “the whole of
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 91/97
Beyond Paradigms 591
an eventually perfected, totally comprehensive, integrated science”, i.e., to something
Quine does not expect ever to be completely realized, and certainly not now available
to the “more thorough pragmatist”. That metaphoric asymptote: “the whole of science” is
overly emphasized by Quine’s critics. What is important is the directionality it dramatizes:i ncreasing corr espondence w i t h i ncreasing scope. This principle applies whether that
correspondence be interpreted as between scientific beliefs and “stimulations” or
between beliefs and “observations of objects and events”. Indeed, it is a very general
principle of cognition: the stimulation of a single retinal cell is nearly totally equivocal as
to correspondence with an object posit. A pattern of stimulation of a hundred thousand
retinal cells may leave such little equivocality that only a skeptical philosopher would
notice. The normative rule might be made explicit as: to improv e the ji t of beli ef to
[st imul at i ons] [observ at i ons] [t he ext ernal w orl d], expand t he scope of both bel i efs and
[observations].
Parti al, proximal revi sion. The web of belief, or the coherence of texts, is not so
tightly related that revision of one part requires the revision of all. We do, and should,
revise those nearest in the web of belief, nearest in terms of implication steps. I do
not have space here to document Quine’s recognition and normative recommendation
of piecemeal revision in those 15 paragraphs, but will share with those interested my
unpublished 1990 “Exegesis”. I have tried to capture this aspect of his recommended
strategy under the concept of “the ramification-extinction of plausible rival hypotheses”
(1988, pp. 518-519). That is, we consider the extended implications in our network of
beliefs for each of the rival interpretations of a specific set of data, and find “implausible”
and hence “rule out” those hypotheses whose ramifications are discordant with other
beliefs beyond the immediate focus which are we inclined to trust.
Fall i bil ist pri vi l eging of observat ions and core. The above hermeneutic principles, and
the “principle of charity” to follow are, I believe, clearly compatible with the validity-
seeking hermeneutic tradition, as well as being Quine’s recommendations as to how to
proceed in science. The present one may be specific to science. All through those last 15
paragraphs of “Two Dogmas” are references to a “periphery” of empirical observations,
stubborn or “recalcitrant” ones which motivate readjustments in the “interior”. While
none of these are infallible (we may interpret them away as hallucinations or the meter
being out of callibration, etc.), it is clear that Quine, as an empiricist, advises us to be
more reluctant to discount them than the theories which relate them. While the textual
evidence is less clear, I believe that he regards very central beliefs, as the belief in
knowable order or the laws of logic, also relatively privileged. Normatively;we should be
less willing to revise these (periphery and core) than the intermediate theoretical beliefs.
I suspect that the coherence theorists here assembled share this privileging of center and
periphery, and recommend putting most of the coherence-improving readjustments into
the intermediate theoretical beliefs.
The Principle of Charity. This hermeneutic principle advises us to attribute to the
author of the text a shared humanity. We prefer that tentative translation which makes
most of his or her beliefs “true” or “rational”, qualified by consideration of the
different environment and historical period in which the author lived. Especially clearlyin Habermas, we privilege the assumption that the author is trying to communicate, and
is honest (except for partisan biases we too share, or corrupting social predicaments
that preclude an “ideal speech community”). This particular name for this principle
has been popularized by Quine’s great hermeneutics of translation, Word and Object
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 92/97
592 D. T. CAMPBELL
(1960) (he attributes it to Wilson). It is a mainstay of Donald Davidson’s philosophy,
but he mistakenly denies any frame-of-reference relativity, attributing to all others his
own perfect rationality and implicitly the same background beliefs. Mary Hesse (1980,
pp. xviii, 160,161) has specifically identified the principle of charity with the hermeneutic
tradition.
We can envisage physical scientists as part of a hermeneutic community using all of
the principles but “charity” in their interpretation of laboratory data about inanimate
objects. They do, however, use the principle of charity in their efforts to understand the
cryptic and elliptical writings of their fellow scientists. We social scientists too should
apply this principle to the writings of our fellow scientists studying human social behavior.
We should apply it also in trying to understand what the hermeneutists reviewed by
Phillips are trying to say. I go further and join the hermeneutists in recommending that
we systematically extend this principle to those fellow humans we study. Would Denis
Phillips disagree? I believe that we can devise experiments and data collection procedures
to test the principle of charity in this application. By speaking about the meanings of
objects in our shared experiential space we can induce in fellow humans’ behavioral
dispositions which we, as behavioral researchers, might not be able to distinguish
from non-verbally induced conditioned responses. And if we asked “what have you
learned” after the conditioned-response experiments, might not human subjects reply in
terms of meanings about objects in their experiential environments, rather than reciting
which muscle to twitch after which sense-receptor stimulation? (See Campbell, 1963,
1988, pp. 94-146, for more detail on this perspective.) Tolman applied the principle
of charity to his white rats. He, and Watson’s renegade student Lashley, produced the
research which Merleau-Ponty (1963) cites against behaviorism, but wrongly arguing
against doing psychology in a natural-scientific perspective. In contrast, I (1953) have
summarized this same literature as a series of crucial experiments showing that most
animals, and most of one sample of humans, learn about objects in the world rather
than which muscles to contract when. Cannot we similarly operationally “delineate”
(not “define”) the claims about human beings Phillips’ hermeneutists have made, and
put them to scientific test?
There is another convergence between Quine and the “paradigms theorists” or
‘“oniologically-nihilist hermeneutists” which must be noted. Quine has a slogan about
the radical translation situation: “There is no truth of the matter.” Not only is it
possible that we will not be able to choose between two translations, we may be wrong
in assuming that there is a singular correct translation to strive for. Similarly for theories
of physics. We should anticipate the possibility of irresolvable rivalry between complete
theories. (One can get into his writings on such issues in his recent, 100 page Pursuit of
Truth, 1990.) But note that for Quine, these are asymptotic projections of fundamental
“indeterminacies” or “underdeterminations” (he uses both terms, we might prefer the
latter). Overall, he clearly recommends that we proceed with the task of trying to find
empirical predictions on which our competing hypotheses of translation or scientific
theories differ. He believes that we have made progress toward resolving these in
the past in both fields. But as a scrupulous logician, he wants us to remember that
completion of the task of eliminating all rival theories is not guaranteed, and that proof
that we have done so is logically impossible.
For me (and possibly for Quine), there is a double-hermeneutics for the translation
situation which gives a special justification to the “no truth of the matter” slogan.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 93/97
Beyond Paradigms 593
The conditions of language learning for children, and for adults as new words are
introduced into the language, precludes perfect consensus on word meanings within
the core linguistic community. Ostensive illustrations are essential, but these are
far from definit ive, instead are equivocal as both Quine and Wittgenstein have noted.The ostensive-instance-sets of the language learner are never sufficient to insure
conversion to identical meanings (Campbell, 1973). This means for me that Quine
should acknowledge a fourth indeterminacy, in an individual’s learning of the meanings
of the words, to recognize that this too (although very well done) is also an indeterminate,
underdetermined, hermeneutic process. (Instead, his renewed emphasis on linking of
“stimulations” to “holophrastic sentences” in Pursuit of Truth, seems to me too
much like the earlier atomistic foundationalism of “sense data”, which even the
logical positivists themselves gave up for “protocol sentences in the ordinary object
language. “)
Quine’s setting-out these limitations to complete knowledge, this support for the
technical arguments of the skeptics, is not accompanied by a paralyzed defeatism
about the achievement of improved belief or unified theory. Instead, he describes
(particularly in those last 15 paragraphs of “Two Dogmas”) how we do, and how we
should, proceed in spite of our predicament as knowers. He is an advocate of science’s
past achievements and future prospects. This, as I see it, is akin to the stance of the
validity-seeking hermeneutic tradition.
Paradi gm, Li nguist i c, and Cult ural Soli psi sms. The hermeneutic processes of knowing
so far described support a relativism of cultures, languages, and paradigms. Of course,
if groups start out with differing sets of beliefs they will be prone to making different
coherence-enhancing belief revisions, and will doubt the validity and rationality of others
and the consensuses of other believing communities. In science, for example, Britain’s
most eminent physicist of the 1870s Kelvin, could not exclude his Christian faith from
the coherence set he used in evaluating Darwin. Scrupulously using the very best physics
of the day, he proved that the earth was too young for the processes Darwin described
to have taken place. Since then, the limited interpretive horizon provided by physics,
geology, and biology have changed, and the most devoutly Christian among the most
eminent physicists no longer reject Darwin in their coherence-maximizing efforts.
Anthropologists place Kelvin and Darwin in “the same” culture. If there was this much
perspectival relativism in spite of their overwhelmingly shared culturally-given belief-
sets, how much more should be expected, and is indeed found when anthropologists
have studied more exotic cultures. (Up to my limited investment, I am a card-carrying
cultural relativist [Campbell, 1972; Segall, Campbell, & Herskovits, 19661.) Now, had
these anthropologists been complete-incommensurabilists akin to the stereotype of the
“paradigms theorists” controverted by the present essays, they would loyally have
reported that each exotic culture was utterly incomprehensible. Instead, they uniformly
reported that the longer they lived with and studied an exotic people, the more humanly
reasonable they seemed. And they convinced readers of the lesson of cultural relativism
by plausibly presenting exotic world views in comprehensible and sympathetic terms.
Kuhn himself, paradigmatic “paradigms” theorist, has been clear in recent writings(1976, 1983, 1990) that he never intended to deny that members of one paradigm could
learn another, nor to deny that in the history of science they had regularly done so,
and made reasonable (if unproven) choices between paradigms they comprehended.
(One would, of course, expect first-paradigm biases, akin to phonetic, grammatical,
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 94/97
594 D. T. CAMPBELL
and semantic “accents” on the part of second language learners.) Kuhn in his original
presentation (1962) also acknowledged sub-specialty and individual differences in the
“shared” paradigm.
There are perhaps some in the relativist, social-constructionist sociology of science(“sociologists of scientific knowledge”, in the “sociology of knowledge” tradition) who
approach total paradigms incommensurability, or “paradigm solipsism”. They endorse
a linguistic solipsism, an exaggerated version of the Whorf-Sapir-Cassirer-Borges
hypothesis that our ordinary entification of the perceptual world into objects and
events is totally determined by the arbitrary categories our language provides us.
I have added Borges to the list in honor of Collins’ (1985) excellent book, which
uses a Borges story as an introduction. In contrast, I (1973, 1989) have argued (and
have cited Quine in at least partial support) that the perceptual reification of external
objects as existing independently of our perception of them is shared by many animals,
is developmentally and phylogenetically prior to language learning, which gets startedwith equivocal ostentations, which would not work were not our perceptual reifications
of middle-sized objects in our visual-tactual field highly similar from person to person,
and because of the ostention requirement, this allows the “way the world is” to edit the
sorts of words that can usefully become socially shared. (My version of this pre-linguistic
perceptual background differs from Quine’s in that I add to his list a fifth indeterminacy,
in the pre-linguistic reidentification of an “object” as “the same”.)
These “ostensionable” (Campbell & Paller, 1989) objects and events are available to
both ingroup child and anthropologist (Horton, 1982) and make possible some degree
of exotic language learning. They do not provide foundational, definitive translation. In
my own best hermeneutic achievement (1964), one of our comprehension checks was a
Muller-Lyer illusion figure with a 700% discrepancy. If the other-culture respondents
reported a different choice on this item than would the anthropologist, we scored them
as not understanding. If they reported differently on the Mtiller-Lyer items with -5% to
+50% discrepancy (and were Guttman-scale consistent), we scored them as perceiving
differently. In spite of this profound epistemological equivocality, we plausibly claimed
to find cultural differences in perception, and reported on direction of difference
and amount. (Had the differences in perception been profound, we could not have
confirmed that we were communicating!) Campbell and Paller (1989) have noted that
it is “ostensionables” at this level that are called for in the “demonstration” of results
in the ideology of the scientific revolution.
From this hermeneutic perspective, it is unlikely that any of the educational research
“paradigms theorists” are claiming total incomprehensibility across paradigms. We
and they are from the same culture, including social science subculture. We share
enough “horizon of interpretation”, or “interpretive framework” to achieve meaningful
disagreement, operationally illustratable (without, of course, definitional operations).
We must enter into hermeneutic, dialectical disputation with them on rival descriptive
claims, even if only imperfectly translated.
In the above, I have approved of, and built upon, Phillips’ contribution, but by
shifting to hermeneutic methodology rather than assumptions about human nature,
have identified the coherentism shared by all four essays with the hermeneutic tradition,
and have joined Walker and Evers in denying paradigm incommensurabiIity between
us and the “paradigms theorists”. I have affirmed, over-elaborated, and qualified the
relationship to Quine which Evers and Walker recognize that I share with them. I have
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 95/97
Beyond Paradigms 595
echoed Evers on the epistemological relevance of children’s language learning. What
remains for these last few paragraphs are more specific comments on the papers.
In the course of affirming what he and I share, Evers quite reasonably chides me for
trying to draw an observation/theory distinction at the level of internal/external validity.His point is well-taken, and the issue is one I am prepared to reverse myself on, but
for the present will treat it as a messy loose-end for future resolution. I suspect that
some version of the fact-theory distinction will survive in our post-positivist consensus.
Stegmtiller (1976) in formalizing Kuhn, makes such a distinction, in that the “anomalies”
and retained “facts” are “laden” with other theories than the one under test. While I have
emphasized (1987) that the neural connections involved in vision embody “theoretical”
anticipations as to objects in the world, and object-permanence, such use of “theory”
leaves us needing another term distinguishing between such pan-human unconscious
theorizing and the explicit theories of science. Quine’s periphery/interior distinction
discussed above, and my emphasis on hermeneutic usefulness of “ostensionable” objectssupports such a distinction. But I recommend to all taking Evers’ call for revision of my
point of view very seriously.
Lakomski, Walker, and Evers show a great respect for the neurologizing of
epistemology by Churchland, Churchland, and Stich, and the parallel distributed
processing approach to artificial intelligence of Rumelhart and McClelland. I, of
course, endorse the coherentism they cite this group as sharing. I applaud that aspect
of P.D.P. which sets out to achieve object recognition in spite of differing peripheral
“stimulations” on each exposure, and the talk of “patterns” or “configurations” in this
process. But I reject P.D.P.‘s programmatic peripheralism, and the militant avoidance
of simulating conscious experiences in addition to overt responses. (Edelman, 1987, has
a version of P.D.P. much superior to this regard.) If “folk psychology” be interpreted
only as that Anglo-American philosophical tradition of epitomizing beliefs (and hence
that subset “knowledge”) as sentence or propositions, I, too, reject it. (“Maps” are a
better metaphor than sentences.) But if they want to reject the folk concept of beliefs in
non-sentential form, and folk concepts such as goals, purposes and perceived objects, I
vigorously disagree. My hermeneutic coherentism requires fallibilist continuity with folk
knowing, and folk concepts about the knowing process (Campbell, 1988, Chapter 14).
In this, I reject the overall program of “eliminative materialism”, shared by Churchland,
Churchland, and Stich.
Haig’s explanatory coherentist theory of knowledge justi~cation is one I am happy
to endorse. But Glass’ achievements, and the Fisherian tradition of random assignment
to treatments, fit well within the coherence framework. In Whigish (but plausible)
retrospect, all the talk about plausible rival hypotheses or threats to validity in the
quasi-experimental design tradition, can be seen as relative coherence arguments,
devoid of logical or empirical “proof”, made on human discretionary bases. Random
assignment to treatments greatly reduces the plausibility of some rival hypotheses, often
otherwise very plausible ones. Moreover, we are not likely to ever achieve such a
complete coherence that the possibility of the results being due to “mere chance” in
small experiments can be ruled out. The “bootstrapping” methods Haig favors shareFisher’s world view. (Before modern computers, assumptions about normal distributions
and sampling from infinite universes greatly simplified the presumptive estimates of
plausibility.) Were the diverse experiments Glass pools totally incommensurable, the
consistent findings Glass discovers would not plausibly have been found. Our coherentist
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 96/97
596 D. T. CAMPBELL
perspective explains how we can use research results from others with differing theories
and epistemologies. Thus Haig can use Glass’ results without his epistemology.
Along with Haig, I believe that coherentists can also frequently find quasi-experiments
validly interpretable, and that the whole context of many threats to validity (or nuisance
variables) must be considered together, rather than singularly. My students will be
surprised to learn that I can even allow the possibility of unbiased matching. But I
must reject Haig’s statement (p. 575) that “Randomization is unnecessary because we
can identify the most potent nuisance variables and match groups on these.” Because all
the nuisance variables will be measured with partial unreliability and partial invalidity
(or irrelevance), where one really needs matching to produce plausible pre-treatment
equivalence, one is bound to undermatch for these reasons. (My best teaching on these
issues is in Campbell & Boruch, 197.5; but see also Campbell & Stanley, 1963/66 for
the threat of “regression artifacts”, and Cook & Campbell, 1979, Chapter 4 by Charles
Reichardt, and Chapter 7.) LISREL measurement models are addressed to this problem
(and are much better than co-variance adjustment or a beta weight for the treatment as
a dummy variable), but usually have to settle for implausible co-measures of a common
latent variable. But I do agree with Haig that significance tests control for only one
threat to validity, and are often mistaken as controlling for all.
But as I indicated at the beginning, these are minor notes on a set of essays
foreshadowing the post-positivist consensus on a coherentist philosophy of science. I
am happy to join in recommending this perspective to all educational researchers.
References
Campbell, D. T. (1953). Operational delineation of “what is learned” via the transportation experiment.
Psychological Review, 61, 167-174.
Campbell, D. T. (1963). Social attitudes and other acquired behavioral dispositions. In S. Koch (Ed.),
Psychol ogy: A st udy of sci ence Volume 6 (pp. 94-172). New York: McGraw-Hill. Reprinted in D. T.
Campbell, 1988.
Campbell, D. T. (1964). Distinguishing differences of perception from failures of communication in
cross-cultural studies. In F. S. C. Northrop & H. H. Livingstone (Eds.), Cross-cultural understanding:
Epist emol ogy in anfhr opol ogy(pp. 308-336). New York: Harper & Row.
Campbell, D. T. (1966). Pattern matching as an essential in distal knowing. In K. R. Hammond (Ed.),
The psychology of Egon Brunswik (pp. 81-106). New York: Holt, Rinehart and Winston. Reprinted inH. Kornblith (Ed.) (1985), Nafural izi ng epislemology (pp. 49-70 ). Cambridge, MA: MIT Press.
Campbell, D. T. (1972). Herskovits, cultural relativism, and metascience. In M. J. Herskovits, Cultural
re lat iv ism (pp. v-xxiii). New York: Random House.
Campbell, D. T. (1973). Ostensive instances and entitativity in language learning. In W. Gray & N. D.
Rizzo (Eds.). Uni ty through diversit y (pp. 1043-1057). New York: Gordon & Breach.
Campbell, D. T. (1987). Neurological embodiments of belief and the gaps in the fit of phenomena to
noumena. In A. Shimony & D. Nails (Eds.), Nat ural i sti c epist emol ogy: A symposium of IWO decades
(pp. 165-192). Dordrecht, Holland: D. Reidel Publishing.
Campbell, D. T. (1988). (E. S. Overman, Ed.) M ethodol ogy and epi stemology fo r sociul sci ence: Selected
papers. Chicago, IL: University of Chicago Press.
Campbell, D. T. (1989). Models of language learning and their implications for social constructionist analyses
of scientific belief. In S. L. Fuller, M. DeMey. T. Shinn. & S. Woolgar (Eds.). The cognitive turn (pp.
153-158). Boston, MA: Kluwer Academic Publishers.Campbell, D. T. (1990). Exegesis on fifteen famous paragraphs from Quine. Unpublished manuscript.
Campbell, D. T. & Boruch, R. F. (1975). Making the case for randomized assignment to treatments by
considering the alternatives: Six ways in which quasi-experimental evaluations in compensatory education
tend to underestimate effects. In C. A. Bennett & A. Lumsdainc (Eds.), Evaluation and experiments:
Some cri t i cal issues i n assessi ng social pr ograms (pp. 195-296). New York: Academic Press.
7/27/2019 LAKOMSKI, Gabriele (1991) Beyond Paradigms; Coherentism and Holism in Research
http://slidepdf.com/reader/full/lakomski-gabriele-1991-beyond-paradigms-coherentism-and-holism-in-research 97/97
Beyond Paradigms 597
Campbell, D. T. & Paller, B. T. (1989). Extending evolutionary epistemology to “justifying” scientific
beliefs: A sociological rapprochement with a fallibilist perceptual foundationalism? In K. Hahlweg & C.
A. Hooker (Eds.), Issues in evol uti onary epist emol ogy (pp. 231-257). Albany, NY: State University of
New York Press.
Campbell, D. T. & Stanley, J. C. (1963/66). Experimental and quasi-experimental designs for research
on teaching. In N. L. Gage (Ed.), Handbook of research on t eaching (pp. 171-246). Chicago, IL:
Rand McNally. Reprinted as (1966) Experi mental and quasi- experi mental designs for research. Chicago,
IL: Rand McNally.
Collins, H. (1985). Changing order: Repli cati on and inducti on in scienti fi c practi ce. Beverly Hills, CA: Sage
Publications.
Cook, T. D. & Campbell, D. T. (1979). Q uasi-experimentation: Design and analysi s for fi el d sett ings.
Chicago, IL: Rand McNally.
Cook. T. D. & Campbell, D. T. (1986). The causal assumptions of quasi-experimental practice. Synt hese,
68, 141-180.
Edelman, G. M. (1987). Neural D arw ini sm: The theory of neuronal group sel ecti on. N ew York: Basic
Books.
Geertz, C. 111973). The int erpretati on of cult ures. N ew York: Basic Books.
Geertz, C. 1,1983). Local know l edge. N ew York: Basic Books.Habermas, J. (1983). Interpretive social science vs. hermeneuticism. In N. Haan, R. N. Bellah, P. Rabinow,
& W. M. Sullivan (Eds.), Social science as moral i nqui ry . New York: Columbia University Press.
Hesse, M. (1980). Revol uti ons and reconstructi ons i n the phil osophy of sci ence. Bloomington, IN: Indiana
University Press.
Horton, R., (1982). Tradition and modernity revisited. In M. Hollis & S. Lukes (Eds.), Rat ional i ty and
relat iv ism (pp. 201-260). Cambridge, MA: MIT Press.
Kuhn, T. S. (1962). The structure of sci enti fi c revol uti ons. Chicago, IL: University of Chicago Press.
Kuhn, T. S. (1976). Theory-change as structure-change: Comments on the Sneed formalism. Erkenntnis,
10, 179-199.
Kuhn, T. S. (1983). Commensurability, comparability, communicability. In. P. D. Asquith & T. Nickles
(Eds.), PSA 1982, Volume 2. East Lansing, MI: Philosophy of Science Association.
Kuhn, T. S. (1991). The road since structure. In A. Fine, M. Forbes, & L. Wessels (Eds.),PSA 1990,
Vol. II. E:ast Lansing, MI: Philosophy of Science Association.Lehrer, K. (1974). Know l edge. Oxford: Clarendon Press.
Lehrer, K. (1990). Theory of know l edge. Boulder, CO: Westview Press.
Merleau-Ponty, M. (1963). The structure of behavior. Boston, MA: Beacon Press.
Paller, B. T. & Campbell, D. T. (1989). Maxwell and van Fraassen on observability, reality, and justification.
In M. L. Maxwell & C. W. Savage (Eds.), Science, mi nd and psychology : Essays i n honor of Grov er
Maxwell (pp. 99-132). Lanham, MD: University Press of America.
Quine, W. V. (1951). Two dogmas of emuiricism. Phi l osophical Revi ew . 60. 2S-43. Reurinted in W. V.
Quine (1963); From a logicalpoint of vi e& (pp. 2& 46). Nkw York: Harper Torchbooks.’
Quine. W. V. (1960). Word and object. New York: Wilev.I J
Quine, W. V. (1990). Pursuit of truth. Cambridge, MA: Harvard University Press.
Segall, M. H., Campbell, D. T., & Herskovits, M. J. (1966). The i nfl uence of cult ure on vi sual percept i on.
Indianapolis, IN: Bobbs-Merrill.
Stegmiiller, W. (1976). The struct ure and dynami cs of theories. New York: Springer-Verlag.