multimorphemic words in spoken production … · multimorphemic words in spoken production!! 1 ......
TRANSCRIPT
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
1
RUNNING HEAD: MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION
Towards a theory of multimorphemic word production: The Heterogeneity of Processing Hypothesis
Ariel M. Cohen-Goldberg
Department of Psychology, Tufts University
Address for Correspondence: Ariel M. Cohen-Goldberg Department of Psychology Tufts University 490 Boston Ave. Medford, MA 02155 USA Tel: (617) 627-3525 Fax: (617) 627-3181 [email protected]
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
2
Abstract
Theories of spoken production have not yet addressed the post-lexical processing of
multimorphemic words, that is, how a multimorphemic word’s phonological form is
prepared for production. This paper reviews what is known about how multimorphemic
words are represented in production at lexical and post-lexical stages as well as the
influence that lexical properties have on post-lexical processes. A proposal linking these
facts together is presented which predicts that post-lexical processes 1) should be weaker
when acting across morpheme boundaries and 2) should be influenced by the lexical
properties of each morpheme. Post-lexical processing is thus predicted to vary, or be
‘heterogeneous’, across a multimorphemic word. Phoneme competition (as indexed by
inhibitory effects of phoneme similarity) is compared within and across morphemes in
three analyses of oral reading latencies. Competition is found to be weaker across
morpheme boundaries, providing support for heterogeneity.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
3
Acknowledgements
I am grateful to Shira Cohen-Goldberg for her invaluable assistance with this project. I
wish to thank Naomi Berlove and Simon Fischer-Baum for the many theoretical
discussions relating to this work, the R-Lang mailing list for statistical assistance, and the
members of the Tufts Linguistic Seminar, the MIT Phonology Circle, and three
anonymous reviewers for comments on earlier versions of this paper.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
4
Introduction
Although there has been extensive psycholinguistic research investigating the
lexical aspects of multimorphemic word processing—examining, for example, whether
morphologically complex words are represented in a decomposed (morpheme-based) or
holistic (word-based) fashion (e.g., Butterworth, 1983; Caramazza, 1997; Fiorentino &
Poeppel, 2007; Marslen-Wilson & Zhou, 1999; Stockhall & Marantz, 2006; Taft, 2004;
Bybee, 1995; Elman, 2004; Kuperman, Schreuder, Bertram, & Baayen, 2009; Rubin,
Becker, & Freeman, 1979; Seidenberg & Gonnerman, 2000), which factors contribute to
productivity (e.g., Anshen & Aronoff, 1981; Baayen, 1994; Bauer, 2004; Cutler, 1980a,
1981; Hare & Elman, 1995; Hay & Baayen, 2003; Hay & Plag, 2004; Krott, Baayen, &
Schreuder, 1999; Marchman, Wulfeck, & Weismer, 1999; Plag, 1999), and whether
paradigmatic relations influence processing (e.g., Baayen, Levelt, Schreuder, & Ernestus,
2008; Ernestus & Baayen, 2007; Juhasz & Berkowitz 2011; Kuperman, Pluymaekers,
Ernestus, & Baayen, 2007; Milin, Kuperman, Kostic, & Baayen, 2009; Moscoso del
Prado Martin, Kostic, & Baayen, 2004)—relatively little is known about the post-lexical
processing of multimorphemic words in production. Collectively, post-lexical processes
are responsible for computing a word’s phonological form for production and comprise
phonological encoding, the process by which phonemes are retrieved and ordered,
grammatical phonological processing, where the phonological form of a word is adjusted
to conform to language-specific and universal well-formedness constraints, and
articulatory planning, where gestural scores and motor plans governing the movement of
the articulators are generated.
Current theories of spoken production, although they provide accounts of the
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
5
lexical processing of multimorphemic words, are silent as to how these words are
processed at post-lexical levels (Dell, 1986; Levelt, Roelofs, & Meyer, 1999). While the
default assumption could be that mono- and multimorphemic words undergo identical
post-lexical processing (e.g., there is no difference in the way that the phonological forms
of band and banned are computed), a number of studies have found differences in the
phonetic realization of mono- and multimorphemic words, indicating that differences
exist in the way they are processed at post-lexical levels (Cho, 2001; Frazier, 2006;
Losiewicz, 1995; Schwarzlose & Bradlow, 2001; Sugahara & Turk, 2009; Walsh &
Parker, 1983). Other studies have supported this assertion, finding that word-internal
morphemes are important planning units in phonological encoding (Roelofs, 1996a;
Roelofs & Baayen, 2002). To date, no general theory describing the way that post-lexical
processes operate over multimorphemic words has been proposed. Once a
multimorphemic word has been retrieved from the lexicon or assembled productively,
how is its phonological form computed? What effect, if any, does its lexical structure
have on post-lexical processing?
This paper represents a first attempt to bring what is known about lexical and
post-lexical processing in spoken production to bear on the issue of the post-lexical
processing of multimorphemic words. The rest of the Introduction is organized as
follows. First, evidence concerning the structure of multimorphemic words at lexical and
post-lexical stages of processing is reviewed. Then, psycholinguistic data concerning the
influence that lexical representations have on post-lexical processing is reviewed. These
facts are then brought together into a theory of multimorphemic word processing termed
the Heterogeneity of Processing Hypothesis (HPH). This hypothesis predicts that post-
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
6
lexical processing will vary across a multimorphemic word in two ways: 1) post-lexical
processes should apply more weakly across morpheme boundaries than to phonemes
within the same morpheme and 2) post-lexical processes should vary according to the
lexical properties of each morpheme. Evidence from three analyses of oral reading
latency data are then reported that support the first prediction of the HPH.
The lexical representation of morphologically complex words in production
Theories of spoken production hold that lexical representations are important
planning units in the production of words (in the rest of this article, the terms ‘morpheme’
and ‘lexical unit’ will be used interchangeably). Fundamentally, the morpheme is thought
to mediate between semantic and phonological processing—rather than directly
activating phonology, semantic representations activate lexical representations, which in
turn provide access to the word’s phonological form. This ‘two-step’ organization thus
divides word production into two main components: the selection of lexical items on the
basis of their fit to semantic and/or syntactic specifications and the retrieval and planning
of the word’s phonological form. This distinction has been successful in accounting for a
variety of experimental findings (e.g., Dell & O’Seagdha, 1992; Jescheniak & Levelt,
1994; Kempen & Huijbers, 1983; Levelt & Maassen, 1981; Levelt et al., 1991;
Schriefers, Meyer, & Levelt, 1990, Levelt, Roelofs & Meyer, 1999) as well as speech
errors in normal (e.g., Dell, 1986; Fromkin, 1971; Garrett, 1975, Butterworth, 1989) and
impaired individuals (e.g., Dell, Schwartz, Martin, Saffran, & Gagnon, 1997; Saffran,
Schwartz, & Marin, 1980, Bock, 1987) .
A major debate in both the comprehension and production literatures is whether
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
7
morphologically complex words are represented at lexical levels as whole words or in
terms of their component morphemes. In production, the question centers on what units
provide access to a word’s phonological form: ‘full-listing’ theories argue in favor of a
single representation (e.g., a single lexical representation provides access to all of the
phonemes in the word sunshine; Butterworth, 1983; Caramazza, 1997) while
‘compositional’ theories argue in favor of multiple representations (e.g., one
representation provides access to the phonemes /s/, /ʌ/, /n/ while another provides access
to /ʃ/, /aɪ/, /n/; Levelt et al., 1999).
Currently, a substantial body of evidence supports the notion that production
utilizes compositional, morpheme-based representations (see Bölte, Zwitserlood, &
Dohmes, 2004 for a review). Some of the evidence for this position comes from
morpheme-level speech errors in neurologically intact (Cutler, 1980; Fromkin, 1973;
Garrett, 1975, 1980; Melinger, 2003; Pillon, 1998; Stemberger, 1982) and impaired
individuals (Badecker, 2001; Badecker & Caramazza, 1991; Cholin, Rapp, & Miozzo,
2010; Cohen-Goldberg, Cholin, Miozzo, & Rapp, submitted; Delazer & Semenza, 1998;
Hittamair-Delazer, Andree, Semenza, De Bleser, & Benke, 1994; Miceli, Capasso &
Caramazza, 2004; Mondini, Luzzatti, Zonca, Pistarini & Semenza, 2004; Stemberger,
1985). Some studies of aphasic individuals have found effects of morpheme frequency in
compound production where high frequency constituents are more accurately produced
than low frequency constituents, providing detailed evidence for the existence of
component morphemes (Ahrens, 1977; Blanken, 2000; Rochford & Williams, 1965).
Other data come from chronometric studies that find morpheme-specific effects in
word production (Bien, Levelt, & Baayen, 2005), implicit priming (Roelofs, 1996a,
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
8
1996b; Roelofs & Baayen, 2002), and the picture-word interference task (Bölte, Dohmes,
& Zwitserlood, 2012; Lüttmann, Zwitserlood, Böhl & Bölte, 2011; Zwitserlood, Bölte, &
Dohmes, 2000, 2002). Interestingly, some of the studies that have found evidence for
component morphemes have found these effects in semantically opaque (e.g., Badecker,
2001; Lüttmann et al., 2011; Roelofs & Baayen, 2002) and morphologically irregular
words (Cholin, Rapp, & Miozzo, 2010), suggesting that morpheme-based representations
are not restricted to only the most productive words. Studies have also suggested that
written production makes use of morpheme-based representations, providing converging
evidence for the role of the morpheme in production (e.g., Badecker, Hillis & Caramazza,
1990; Badecker, Rapp & Caramazza, 1996).
The evidence is not completely decisive in favor of morpheme-based
representations, however. Some production studies have found evidence for both
morpheme-based and whole-word representations (e.g., Bien et al., 2005; Tabak,
Schrueder, & Baayen, 2010), suggesting that whole-word representations may influence
production in addition to morpheme-based representations (supporting ‘dual-route’
theories, e.g., Baayen, Dijkstra & Schreuder, 1997; Jackendoff, 1975). At the same time,
a handful of studies have failed to find evidence for morpheme-based representations
while finding evidence for whole-word representations (Chen & Chen, 2006, 2007;
Ernestus, Lahey, Verhees & Baayen, 2006; Janssen, Bi & Caramazza, 2008).
Despite some variation in the findings, there is strong evidence that
morphologically complex words are represented in a morpheme-based format (and
perhaps, in addition, as whole words) in production.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
9
The post-lexical representation of morphologically complex words in production
Despite the differences in their lexical representation, it is possible in principle
that mono- and multimorphemic words are represented identically at post-lexical stages
of processing (e.g., band and ban+ed = /bænd/). This state of affairs would be predicted,
for example, if post-lexical representations could only contain phonological information
(e.g., Hockett, 1942; Joos, 1964). There is reason to believe, however, that post-lexical
representations do in fact differ depending on a word’s morphological structure. In this
section, evidence is reviewed that post-lexical representations of multimorphemic words
contain 1) explicit information about morphological boundaries and 2) fewer/weaker
structural relations between the content of different morphemes.
As Roelofs (1996b) notes, grammatical phonological processes provide evidence
that morphological boundaries must be explicitly marked in post-lexical representations.
Phonologists have long noted that languages contain phonological rules that make
reference to word-internal morpheme boundaries. Some rules are triggered by the
presence of such boundaries (e.g., Finnish assibilation converts /t/→[s] when followed by
/i/, but only when spanning a morpheme boundary: /tilat+i/→[tilasi], not *[silasi];
Kiparsky, 1973) while others are blocked (e.g., the roots of a compound are independent
domains of syllabification in English, preventing syllabification from occurring across
the compound boundary: cf. ni.trate vs. night.rate). The existence of active, synchronic
rules of this sort indicates that morpheme boundaries are represented in some way in
post-lexical representations. Phonological theories have proposed that boundaries are
explicitly marked, either by special boundary symbols (e.g., + and #, Chomsky & Halle,
1968), indirectly through hierarchical prosodic structure (e.g., the phonological word;
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
10
Nespor & Vogel, 1986; Selkirk, 1984), or through morpheme affiliation labels (e.g.
morpheme ‘color’, van Oostendorp, 2005).1 Since morphophonological rules operate as
expected over novel, productively formed words (e.g., the novel compound sprite race
would be pronounced [spraɪt.reɪs], not *[spraɪ.treɪs]), the insertion of morphological
information must occur during the process of concatenating the individual phonological
representations.
In addition to containing explicit boundary information, a number of studies have
suggested that the post-lexical representations of multimorphemic words are weaker in
some fashion (or that phonemes are more ‘loosely’ joined together) at morpheme
boundaries. Using articulatory measurements of Korean homophones that differ only in
morphological structure (e.g., /nap+i/ vs. /napi/), Cho (2001) found that intergestural
timing was more variable when the gestures corresponded to heteromorphemic sequences
(e.g., /p+i/) than tautomorphemic sequences (e.g., /pi/). Hay (2003) reported broadly
similar results for multimorphemic words that vary in their lexical structure. In a
phonetic study of English suffixed words, Hay compared the rate of interconsonantal /t/-
deletion (a variable rule that deletes /t/s in the environment CtC, where ‘C’ indicates a
consonant) in words that are likely to have a morpheme-based representation (e.g., daftly,
which is less frequent than its root daft) and in words that are likely to be processed as
wholes (e.g., swiftly, which is more frequent than swift). Hay found that /t/-deletion was
weaker in the words that are likely to have morpheme-based representations and more
robust in the words likely to have whole-word representations.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1!See Kiparsky (1982) and Monahan (1982) for a different approach to the relationship between morphology and phonology. See also Scheer (2011) for an excellent overview of the various ways that phonological theory has encoded non-phonological information in phonological representations.!
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
11
Both of these studies suggest that post-lexical content from different morphemes
is less well-integrated than content from the same morpheme. Articulatory gestures
appear to not be as well-coordinated in time across a morpheme boundary and phoneme
sequences created from multiple morphemes appear less likely to trigger
phonological/phonetic processes. Consistent with this notion, Cho’s results have been
modeled within the framework of Articulatory Phonology by assuming that there are
fewer timing relationships between heteromorphemic than tautomorphemic gestures
(Nam & Saltzman, 2003).
One final piece of evidence suggesting that post-lexical representations contain
structural weaknesses at morpheme boundaries comes from the case of a brain-damaged
individual, WRG, who made discrete phonological repairs between word-internal
morphemes but not within morphemes themselves (Cohen-Goldberg et al., submitted). In
their investigation, Cohen-Goldberg and colleagues first determined that WRG’s stroke
impaired his lexical phonological processing while leaving his post-lexical processing
relatively intact. This dissociation was indicated by a stark difference between his
performance in picture naming (which requires lexical processing) and repetition (which
only requires post-lexical processing). When given drawings of objects with
monomorphemic names, WRG was only able to name 12% of the items but when he was
administered monomorphemic words for repetition, WRG was able to correctly repeat
89% of the items (97% accuracy by phoneme). WRG’s inability to name objects
indicates that he suffered from a severe lexical deficit while his ability to repeat words
reveals that his post-lexical phonological processing was relatively intact (he was still
able to plan and articulate the sounds of words).
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
12
Cohen-Goldberg and colleagues report that when producing multimorphemic
words in elicitation or reading, WRG frequently made insertion errors that improved (or
‘repaired’) the phonological structure of the word. For example, when producing words
suffixed with the past tense, WRG frequently inserted segments in between the root and
the suffix (e.g., walked → [wakɪt]) or after the suffix (e.g., slipped → [slɪptɪd]). Cohen-
Goldberg et al. demonstrated that WRG was significantly more likely to make these
insertion errors when the words contained phonologically marked obstruent-obstruent
coda clusters (insertion rate: 27%) than when they contained relatively less marked
sonorant-obstruent clusters (5%). The errors improved the phonological structure of the
coda by separating the coda consonants into different syllabic positions ([wa.kɪt]) or by
re-syllabifying them into different syllables ([slɪp.tɪd]). Intriguingly, WRG was able to
repeat the same words with near perfect accuracy, ruling out an articulatory locus for his
impairment.2
The key comparison came when WRG’s performance on coda clusters in suffixed
words (e.g., banned, walked) was compared to his performance on similarly structured
tautomorphemic environments (e.g., grand, fact). In contrast to his performance on
suffixed words, WRG’s insertion error rate on tautomorphemic coda clusters did not
differ between obstruent-obstruent (8%) and sonorant-obstruent codas (5%).
The fact that WRG’s errors were driven by phonological markedness coupled with
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!2 A similar pattern was observed in the case of stress, where WRG inserted segmental material in words containing stress clash (the dispreferred arrangement of stress on adjacent syllables, e.g., brísknèss → [brɪskɪdnɛs], uníquenèss → [junikɪdnɛs]; insertion rate: 32%) but not in words without stress clash (clévernèss; 0%). WRG never made these errors in repetition. As with the past tense-suffixed words, these insertions improved the phonological structure of the targets, in this case by separating the stressed syllables with an unstressed epenthetic syllable. !
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
13
the pre-articulatory locus for his impairment led Cohen-Goldberg and colleagues to
conclude that WRG’s phonological grammar was responsible for the insertions, which
were made in part to improve the words’ well-formedness. For the present purpose, the
contrasting pattern of repairs in tauto- and heteromorphemic environments supports the
notion that heteromorphemic phonemes are more loosely joined than tautomorphemic
phonemes. Cohen-Goldberg and colleagues argued that WRG’s lexical damage impaired
his ability to combine the phonological content of different morphemes, severely
weakening the strength with which phonemes from different morphemes are bound
together. This weakening allowed the phonological grammar to express itself through
phonological repairs at the morpheme boundary. In monomorphemic words—where
post-lexical representations do not need to be assembled from different morphemes—
phonemes were bound to each other sufficiently strongly to prevent the grammar from
making similar repairs in tautomorphemic environments.
A variety of evidence thus suggests that the post-lexical representations of
multimorphemic words differ from those of monomorphemic words in that they contain
explicit representation of the original morphological boundaries and structural
weaknesses at the morphological boundary. It seems likely that these weak points are
somehow related to the need to bind together on the fly phonemes from different
morphemes (Cho, 2001; Cohen-Goldberg et al., submitted).
Lexical influences on post-lexical processing
Having reviewed the representations of multimorphemic words at lexical and
post-lexical levels, one final element is needed in order to understand how
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
14
multimorphemic words are processed at post-lexical levels, which is to understand how
lexical representations influence post-lexical processing in the general case. As will now
be reviewed, spoken production research on the processing of monomorphemic words
has shown that the two-step organization of speech, by allowing lexical representations to
mediate access to phonological representations, has profound consequences for post-
lexical processing.
The fact that morphemes provide access to phonological representations affects
post-lexical processing in a variety of ways. The simplest consequence is that
morphemes influence the onset of post-lexical processing. Since morphemes provide
access to a word’s phonological form, the sooner a morpheme is accessed, the sooner its
phonemes will receive activation and the sooner post-lexical processing may begin. All
things being equal, for example, post-lexical processing will begin earlier for a high
frequency word than for a low frequency word. Morphemes similarly mediate the scope
of post-lexical processing, since all of a morpheme’s phonemes are simultaneously
activated during retrieval (see Cohen-Goldberg, 2012 and the references therein).
Beyond affecting the timing of processing, lexical representations have significant
effects on the computations performed by post-lexical processes. High frequency words
and words with many phonological neighbors are processed more accurately at post-
lexical stages, resulting in fewer sound-based errors in both neurologically intact (Dell,
1990; Harley & Brown, 1998; Stemberger & MacWhinney, 1986; Vitevitch, 2002) and
brain-damaged individuals (Goldrick, Folk, & Rapp, 2010; Knobel, Finkbeiner, &
Caramazza, 2008). These effects have typically been modeled via cascading activation
from lexical to phoneme levels, with more activation leading to faster and more accurate
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
15
processing (Knobel et al., 2008). Lexical properties such as frequency, neighborhood
density, and predictability have also been shown to influence the outcome of post-lexical
computations, affecting a wide variety of phonetic properties such as vowel space (Aylett
& Turk, 2006; Gahl, Yao, & Johnson, 2012; Munson & Solomon, 2004; Wright, 2004),
tone space (Zhao & Jurafsky, 2007), VOT (Baese-Berk & Goldrick, 2009), coarticulation
(Scarborough, 2004) and duration (Aylett & Turk, 2006; Bell, Jurafsky, Fosler-Lussier,
Girand, Gregory, & Gildea, 2003; Bell, Brenier, Gregory, Girand & Jurafsky, 2008; Gahl,
2008; Gahl et al., 2012). Lastly, morphemes have been theorized to provide access to
stored motor plans, allowing word-specific articulatory parameters to surface in speech
(Bybee, 2001; Goldrick, Baker, Murphy, & Baese-Berk, 2011; Pierrehumbert, 2003;
Yaeger-Dror, 1992, 1996).
To summarize: the lexically mediated organization of speech production has
significant implications for post-lexical processing, influencing its onset, accuracy, and
outcome, and the extent to which articulatory plans are retrieved vs. computed for the
word.
The post-lexical processing of morphologically complex words: the Heterogeneity of
Processing Hypothesis
The Heterogeneity of Processing Hypothesis (HPH), which will now be
described, brings together the empirical generalizations referenced above into a theory of
how post-lexical processing occurs for multimorphemic words. The HPH takes as its
starting place the fact that morphologically complex words are represented by multiple
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
16
lexical representations (morphemes).
The first consequence of this organization, termed Morpheme Assembly, is that
post-lexical representations must be assembled in some fashion during production. Since
each morpheme activates only its constituent phonemes, phonemes from different
morphemes must somehow be ‘bound’ together in order to create an integrated
phonological representation that can drive post-lexical processes. This assembly process,
though currently underspecified in theories of spoken production, likely involves—at the
very least—updating segmental position information to reflect the newly constructed
multimorphemic environment (e.g., the /k/ in cloth is no longer in word-initial position
when it appears as part of the compound tablecloth).! Stored articulatory plans, should
they exist for the morphemes in question, must similarly be joined together.
Based on the evidence described earlier, the process of assembling post-lexical
representations appears to leave boundary symbols and structural weaknesses at
morpheme boundaries in the post-lexical representation. This leads to the first prediction
of the HPH, which is that post-lexical processes will apply more strongly to
tautomorphemic phonemes than heteromorphemic phonemes. These ‘fault-lines’ at in the
post-lexical representations are proposed to cause each morpheme to act (to a certain
extent) as a processing domain for post-lexical processes, resulting in heteromorphemic
phonemes being less phonologically integrated with each other (e.g., less coarticulated)
than tautomorphemic phonemes.
The second consequence of the morpheme-based organization of morphologically
complex words, termed here Segregated Lexical Inheritance, is that phonemes in a
multimorphemic word will inherit the lexical properties of the morpheme they belong to.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
17
Given the two-step organization of speech production, each morpheme provides access to
only its constituent phonemes and as a result, each phoneme in turn is influenced by the
properties of its parent morpheme. This means that across the word, phonemes will b
influenced by different lexical properties: phonemes will receive different levels of
activation as a result of, e.g., their parent morpheme’s frequency and neighborhood
density and will inherit different articulatory parameters as a result of being associated
with morpheme-specific precompiled gestural scores. This leads to the second prediction
made by the HPH, which is that those aspects of post-lexical processing that are
influenced by lexical properties (e.g., duration, vowel space) will vary by morpheme
across a multimorphemic word.
Taken together, the Heterogeneity of Processing Hypothesis makes the crucial
prediction that, as a result of Morpheme Assembly and Segregated Lexical Inheritance,
post-lexical processing will vary across a multimorphemic word. That is, the post-lexical
processing of multimorphemic words is predicted to be heterogeneous.
To illustrate these predictions, consider the compound word sunshine (Figure 1).
During retrieval, each morpheme will provide access to its constituent phonemes and
associated articulatory plan, should one be stored. Since each morpheme has its own
representation, binding processes must integrate them in some way: the segments from
each morpheme will have to be bound together as will the separate articulatory plans (this
binding is depicted graphically as grey bands). Further, since sun is far more frequent and
has many more neighbors than shine, its phonemes will receive more activation, indicated
in the diagram by darker borders. In contrast, consider the monomorphemic word
carbine. Since its lexical representation consists of a single unit, its phonological
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
18
representation will be stored and activated together as a unit, both at segmental and
articulatory planning levels. Further, all phonemes in the word will receive roughly the
same levels of activation and the same articulatory parameters since they are all
associated with the same lexical unit. As a result, post-lexical processing is predicted to
be relatively heterogeneous for sunshine and relatively homogenous for carbine.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
19
Figure 1. Depiction of the processing of multi- and monomorphemic words predicted by the Heterogeneity of Processing Hypothesis. Panel A depicts the processing of the compound word sunshine. Dark borders around the segments in sun represent greater activation due to sun’s relatively high frequency and neighborhood density. Grey bands indicate the need to bind together the segmental and articulatory representations of each morpheme. Panel B depicts the processing of the monomorphemic word carbine. In both panels, dotted lines indicate the possibility that lexical representations provide access to stored articulatory representations.!
SUN SHINE
/s/ /ʌ/ /n/ /ʃ/ /aɪ/ /n/
LIPST TIPT BODY
LIPST TIPT BODY
SUNSHINE
Lexical Representation
SegmentalRepresentation
ArticulatoryRepresentation
CARBINE
/k/ /a/ /r/ /b/ /aɪ/ /n/
LIPST TIPT BODY
Lexical Representation
SegmentalRepresentation
ArticulatoryRepresentation
A
B
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
20
Present Investigation
Study overview
The current study investigates the first prediction of the Heterogeneity of
Processing Hypothesis, specifically, that post-lexical processes apply more weakly across
morphemes than within. The investigation makes use of the phoneme similarity effect
reported recently by Cohen-Goldberg (2012). In a regression analysis of oral reading
reaction times, Cohen-Goldberg found that the phonological similarity of a word’s
consonants slowed production latencies. For example, subjects took longer to begin
uttering words like beep than weep, even after controlling for factors such as frequency,
length, orthographic consistency, letter similarity, etc. This inhibitory effect of similarity
was found for consonant pairs in onset clusters (trim), in the onset and coda (trim), and
crucially, even for onset and suffix consonants in verbs suffixed with –ed (trimmed). For
reasons that will be described below, Cohen-Goldberg concluded that the similarity effect
stemmed from phoneme competition during production. According to this proposal, the
phonemes within a word compete for selection during phonological encoding, with
similar phonemes competing more strongly than dissimilar phonemes.
The phoneme similarity effect can thus serve as a diagnostic to determine which
phonemes interact in the planning of a word’s phonological form, and to what extent. An
inhibitory effect of similarity indicates that two phonemes are competing for selection or
are otherwise inhibiting each other. The present study extends this approach to
investigate the post-lexical processing of multimorphemic words. According to the HPH,
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
21
post-lexical processing is weaker across morpheme boundaries than within, which
predicts that during phonological encoding, phonemes in different morphemes should
compete with each other less strongly than phonemes from the same morpheme.
Three studies are reported. Study 1 investigated the generality of the phoneme
similarity effect in multimorphemic words. Although Cohen-Goldberg (2012) found
effects of root-suffix similarity in words suffixed with the past tense, it is possible that the
past tense suffix, being a single phoneme that integrates into the same syllable as the root,
constitutes a special case, and that similarity effects are not found in multimorphemic
words more generally. To this end, Study 1 investigated whether the phoneme similarity
effect can be found in the production of words with more typical suffixes, namely –ing
and –er. Studies 2 and 3 then tested the prediction made by the HPH, which is that post-
lexical processing is weaker across morpheme boundaries than within a single
morpheme. Study 2 compared the strength of the phoneme similarity effect in words
containing the suffix –ed to monomorphemic words ending in /d/ (e.g., bayed vs. bed).
Study 3 compared the strength of the similarity effect in words suffixed with –y to
monomorphemic words ending in /i/ (e.g., baby vs. flaky).
Methods
The methods used in the present analyses are identical to the methods used in
Cohen-Goldberg (2012) and will be reviewed here. The data for the analyses were
obtained from the English Lexicon Project (ELP), which collected oral reading latency
times for over 40,000 English mono- and multimorphemic words (Balota, Yap, Cortese,
Hutchison, Kessler, Loftis, Neely, Nelson, Simpson & Treiman, 2007). In that study,
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
22
words were presented one at a time in uppercase and naming latency was measured using
a voice key. Only correct responses (as self-reported by the subjects) were included in
the present analyses.
For the analyses reported here, all correct trials corresponding to the relevant
words (e.g., words suffixed with –er) were retrieved from the ELP website.3 Linear
mixed-effects models were used to analyze the data and were run with the lmer package
(Bates, Maechler, & Dail, 2008) of the statistical package R. Each model contained 1) a
set of control variables to control for factors known to affect oral reading processes 2) a
set of random effects variables to allow the results to generalize to new subjects and
items, and 3) the variables of interest (e.g., consonant similarity). The dependent variable
in each was log naming latency in milliseconds. The control variables, which control for
factors such as word frequency, length, neighborhood density, orthographic consistency,
etc., were drawn from Baayen, Feldman & Schreuder (2006) for Studies 1 and 2 and Yap
& Balota (2009) for Study 3. These variables were shown in these studies to significantly
predict oral reading latencies in mono- and multisyllabic words, respectively. In addition
to these variables, several other variables were included in order to control for various
aspects of processing; these variables will be described in each study and are listed in
Appendix A. In all analyses, a variable encoding the letter similarity of the critical letters
(the letters whose corresponding sounds were of interest) was included. This variable
was computed using the uppercase similarity norms of Boles and Clifford (1989) and was
included to make sure that effects of phonological similarity were not due to letter
similarity. Each model contained full random effects structure, with random intercepts
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!3 http://elexicon.wustl.edu
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
23
for subjects and items and random slopes for the variables of interest (including
interaction terms) for both subjects and items.
Consonant similarity was calculated using the feature-based metric proposed by
Bailey and Hahn (2005). This measure defines similarity in terms of the number of
features shared by two phonemes and ranges from 0-4 (Voicing, Place, Manner,
Sonority). Similarity values for singleton consonants were obtained from the appendix of
Bailey and Hahn (2005). In cases involving more than 2 consonants, similarity was
calculated by averaging the similarity of all pairs. For example, the similarity of the
onset and coda of the word crowd was calculated by averaging SIM(/k/,/d/) and
SIM(/r/,/d/). In Study 3, vowel transcriptions were obtained from the Hoosier Mental
Lexicon (Nusbaum, Pisoni, & Davis, 1984) and vowel similarity was calculated using the
features Height (values: high, mid, low), Backness (front, central, back), Rounding
(round, unround), and Tenseness (tense, lax). Vowel similarity thus also ranged from 0-4.
Each analysis was performed the same way: log reaction times were first fit by the
full model and data points with a standardized residual greater than ± 2.5 were considered
outliers and removed. The model was then refit to the data and interpreted. Since it is
not currently possible to calculate p-values for linear mixed-effects models (Baayen,
Davidson, & Bates, 2008) and since Markov chain Monte Carlo simulations used to
estimate p-values have not yet been implemented for models with random intercepts,
model comparison was used to determine the significance of the variables of interest.
Differences in the likelihood of nested models approximately follow a χ2 distribution
(Agresti, 2002), and so a significant chi-square test of these models indicates that the
variable of interest significantly improved the fit of the model. Model comparisons were
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
24
used to test the significance of all variables of interest in the present studies. Tables
containing the results of the analysis for each fixed effect are reported for each study.
Lastly, it should be noted that collinearity among the control variables was not
controlled. Since beta weight, sign, and standard error estimates are only reliable when
collinearity is minimal, the results for these variables are provided to facilitate
interpretation of the models but should not be used to draw conclusions about them.
Collinearity was always controlled for the variables of interest, however, making their
estimates reliable.
Suitability of oral reading latency data
Before continuing, an important issue that must be addressed is whether oral
reading is an appropriate task for investigating the phonological processing of
multimorphemic words. One concern is that oral reading involves many different
processes and as a result phonological effects may not necessarily reflect the processes
involved in spoken production. Another concern is that oral reading can be accomplished
in a sub-lexical manner and thus oral reading data may not necessarily reflect the
processing of multimorphemic representations. Each of these concerns will be addressed
here.
For oral reading reaction time data to be relevant to the theoretical questions of
the present study, they must reflect how phonological representations are processed
during spoken production and ideally should reflect a pre-articulatory stage of processing.
Cohen-Goldberg (2012) reported three findings indicating that the phoneme similarity
effect observed in the ELP data possesses just these properties. First, although Cohen-
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
25
Goldberg found significant effects of phonological similarity in oral reading reaction
times, no such effects were found in an analysis of visual lexical decision reaction times.
This suggests that although phonological representations may be generated during
orthographic comprehension (e.g., Meyer, Schvaneveldt, & Ruddy, 1974; Rubenstein,
Lewis, & Rubenstein, 1971) the phonological similarity effect observed in the oral
reading data originates at a point in the reading process following orthographic and
semantic processing, i.e., spoken production. Second, Cohen-Goldberg analyzed oral
reading latencies of words suffixed with the past tense. Although in monosyllabic words
the past tense suffix has two allomorphs—[t] and [d]—the regression models obtained a
significantly better fit when the suffix was assumed to be [d]. This indicates that the
similarity effect arises at a relatively early stage of phonological processing, before the
final phonetic form of the suffix has been generated (e.g., during phonological encoding).
Finally, for the similarity effect to be informative about the processes involved in spoken
production, it is important to establish that it is truly a general component of speech
planning and is not specific to the reading process. To this end, Cohen-Goldberg
conducted an analysis of picture naming latencies and also found an inhibitory effect of
consonant similarity. Together, these findings indicate that the oral reading reaction times
analyzed in Cohen-Goldberg (2012) and used in the present investigation do in fact
reflect general properties of post-lexical phonological processes and that the similarity
effect is not merely an artifact of the reading process.
The second concern pertains to whether the oral reading data are at all informative
about multimorphemic word processing. For this to be the case, it is essential that the
subjects in the ELP study processed the morphologically complex stimuli as being truly
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
26
multimorphemic and not simply as a string of letters. Most theories of reading posit that
there are two main ways in which words can be read aloud. Words can be read
‘lexically’, whereby the pronunciation of a known word/morpheme is retrieved from
long-term memory and words can also be read ‘sub-lexically’, whereby the pronunciation
of a word is assembled based on the common spelling-sound mappings of the language
(e.g., TH → /θ/). According to these theories, morphologically complex words will only
be processed as being multimorphemic if they are processed through the lexical pathway:
the lexical route will process complex words in terms of their component morphemes
while the sub-lexical route, which is hypothesized to only operate over graphemes, has no
information about morphological structure and will treat morphologically complex words
as simple letter strings. The concern then is that if subjects in the English Lexicon Project
were to have read the morphologically complex words in a sub-lexical fashion, the words
would have been processed as if they were monomorphemic and the resulting reaction
time data would not be informative about the phonological processing of
multimorphemic words.
Two factors suggest that the participants in the English Lexicon Project did in fact
rely on lexical processing to read the morphologically complex words. The first relates to
the fact that the pronunciation of many multimorphemic words is not transparent in the
orthography. For example, the voicing of the plural suffix –s varies despite always being
spelled with the letter S. The past tense suffix –ed undergoes similar voicing changes and
is not pronounced with a vowel (in monosyllabic words) despite being spelled with the
letter E. In other cases, the morphological structure of a word is essential to the proper
parsing of letters into digraphs (without knowledge of the compound boundary, the word
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
27
pothole would be pronounced with a /θ/). Finally, suffixes such as –ese and –ity cause
stress and vowel changes that are not marked in the orthography (Japán~Jápanese,
mort[']l~mort[æ]lity). The sub-lexical reading procedure, since it relies on common
orthography-phonology mappings, will not generate the correct pronunciation for any of
these words (e.g., passed → *[pæsɛd]). Although we cannot independently verify the
accuracy of the subjects’ responses (they were all self-reported as accurate), it is
reasonable to assume that these words were in fact produced correctly, thus indicating
that subjects were utilizing the lexical reading procedure.
The second factor relates to the fact that the experimental design used to collect
the data for the ELP likely biased subjects to favor lexical over sub-lexical processing. It
is known that the oral reading process can be biased by task context, causing subjects to
favor or rely on one processing route over the other. For example, it has been found that
reading a number of non-words in a row can cause subjects to favor sub-lexical
processing while reading a number of orthographically irregular words can bias subjects
in favor of lexical processing (e.g., Reynolds & Besner, 2005; Zevin & Balota, 2000).
Given that all of the words used in the ELP’s reading tasks were real words, and given
the fact that the vast majority of these words were multimorphemic, irregularly spelled,
or both, it seems likely that responses were dominated by lexical processing. Putting all
of these considerations together, it seems likely that the reaction time data analyzed here
reflect post-semantic, pre-articulatory processes acting over lexically generated
morphological representations, making them suitable for examining how
multimorphemic words are produced in normal speech.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
28
Study 1
In order to test the generality of the cross-morpheme phoneme similarity effect
described in Cohen-Goldberg (2012), oral reading reaction times for words suffixed with
–ing and –er were analyzed. These suffixes were chosen based on their representation in
the English Lexicon Project—both were found in over 300 word types, ensuring
sufficient power for the analysis. Other affixes were judged to appear in too few word
types and trials to provide sufficient power: (un–: 77 word types; re–: 56; –ly: 186; –ness
98; –less 130; –ful 77). Similarity was calculated for the root-final consonant and suffix
consonant. For example, for the word baking, the similarity of /k/ and /ŋ/ was used; for
the word baker, the similarity of /k/ and /r/ was used. To ensure that the effects of
similarity were not due to competition among the phonemes of the root (e.g., /b/ and /k/
in baker), variables encoding the letter and phoneme similarity of these consonants were
included in the model.
Results
Separate analyses were conducted for the –ing and –er datasets. After outliers
were removed, the –ing dataset consisted of 21,897 trials from 831 words (457 trials were
removed as outliers). The similarity of the root coda and suffix consonant was found to
have an inhibitory effect on reaction times (β = .011; s.e. = .002; t = 4.6; χ2(1) = 21.0, p <
.00001). Since Root Coda-Suffix Similarity was correlated with Root Coda-Suffix Letter
Similarity (r = -0.585), the former was residualized against the latter and the analysis was
rerun. The residualized predictor was still found to be inhibitory, confirming the stability
of the effect in the model (β = .011; s.e. = .002; t = 4.6).
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
29
After outliers were removed, the –er dataset consisted of 8554 trials from 327
words (185 trials were removed in this fashion). The similarity of the root coda and suffix
consonant was found to have a significant inhibitory effect on reaction times (β = .016;
s.e. = .004; t = 3.9; χ2(1) = 15.6, p < .0001), an effect that persisted once root coda-suffix
similarity was regressed against two predictors with which it was moderately correlated
(Consistency PC2: r = .31; Root Coda-Suffix Letter Similarity: r = .38) and reentered into
the model (β = .016; s.e. = .004; t = 3.9). The results of the –ing and –er analyses are
provided in Table 1 and Table 2, respectively.
In order to be sure that these effects were truly related to phonological similarity
and not due to the processing of repeated letters (Schoonbaert & Grainger, 2004), words
containing identical letters in the relevant positions were removed. For the –ing analysis,
all words containing root-final letters N, G, and NG were removed, resulting in the
removal of 83 words and 2158 trials. For the –er analysis, all words containing root-final
R were removed (e.g., bearer), resulting in the exclusion of 4 words and 90 trials. The
models were refit and significant inhibitory effects were again found for both suffixes (–
ing: β = .010; s.e. = .003; t = 4.1; χ2(1) = 17.1, p < .0001; –er: β = .016; s.e. = .005; t =
3.5; χ2(1) = 13.1, p < .001). This confirms that the effect is truly related to phonological
similarity and not simply the existence of repeated letters.
Summary
The effect of phoneme similarity across morpheme boundaries was investigated in
words suffixed with –ing and –er. In both cases, similarity had a significant inhibitory
effect on reaction times. These results replicate and extend the findings of Cohen-
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
30
Goldberg (2012), confirming that phonemes in different morphemes interact with each
other during phonological processing.
Table 1 Result summary: coefficient estimates β, standard errors SE(β), and t-values for all predictors in the Study 1 analysis of words suffixed with –ing.
Predictor Coeff. β SE(β) t
(Intercept) 6.544 0.051 128.2
Initial Segment Frication-Frication 0.044 0.005 8.7
Initial Segment Frication-Short -0.005 0.006 -0.8
Initial Segment Voicing-Voiceless 0.020 0.006 3.3
FrequencyInitialDiphoneSyllable 0.006 0.001 4.3
FrequencyInitialDiphone -0.017 0.003 -5.3
Root Consistency PC1 0.000 0.001 -0.4
Root Consistency PC2 -0.006 0.001 -4.1
Root Consistency PC3 -0.015 0.003 -4.8
Root Log Frequency -0.005 0.002 -2.8
Root Written-Spoken Frequency Ratio -0.002 0.002 -0.9
Root Orthographic Neighborhood Density 0.001 0.001 1.5
Root Inflectional Entropy -0.010 0.006 -1.6
Word Root Log Frequency -0.017 0.002 -10.3
Word Orthographic Neighborhood Density -0.002 0.001 -2.0
Word Letter Length 0.016 0.004 4.3
Root Onset-Coda Letter Similarity 0.015 0.004 3.5
Root Onset-Coda Phoneme Similarity 0.007 0.002 2.9
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
31
Root Coda-Suffix Letter Similarity 0.015 0.005 3.4
Root Coda-Suffix Phoneme Similarity (residualized) 0.011 0.002 4.6
Table 2 Result summary: coefficient estimates β, standard errors SE(β), and t-values for all predictors in the Study 1 analysis of words suffixed with –er.
Predictor Coeff. β SE(β) t
(Intercept) 6.383 0.063 102.0
Initial Segment Frication-Frication 0.034 0.008 4.2
Initial Segment Frication-Short -0.005 0.010 -0.5
Initial Segment Voicing-Voiceless 0.017 0.009 1.8
FrequencyInitialDiphoneSyllable 0.011 0.003 4.4
FrequencyInitialDiphone -0.003 0.003 -1.0
Root Consistency PC1 -0.001 0.002 -0.6
Root Consistency PC2 -0.006 0.002 -2.6
Root Consistency PC3 -0.001 0.002 -0.2
Root Log Frequency -0.006 0.002 -2.7
Root Written-Spoken Frequency Ratio 0.008 0.003 2.3
Root Orthographic Neighborhood Density 0.000 0.001 -0.1
Root Inflectional Entropy -0.016 0.008 -2.1
Word Log Frequency -0.013 0.002 -7.1
Word Orthographic Neighborhood Density 0.001 0.002 0.4
Word Letter Length 0.020 0.006 3.3
Root Onset-Coda Letter Similarity -0.005 0.007 -0.7
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
32
Root Onset-Coda Phoneme Similarity 0.001 0.004 0.4
Root Coda-Suffix Letter Similarity -0.003 0.007 -0.4
Root Coda-Suffix Phoneme Similarity (residualized) 0.016 0.004 3.9
Study 2
Having shown in Study 1 that phonemes from different morphemes compete with
each other during phonological processing, the strength of this competition within and
across morphemes was investigated in Study 2. A key prediction of the Heterogeneity of
Processing hypothesis is that post-lexical processing is weaker across morphemes than
within due to the need to bind together representations from different morphemes. This
means that phoneme competition—indexed by the inhibitory effect of phoneme
similarity—should be weaker for pairs of phonemes in different morphemes than for
pairs within the same morpheme. To test this prediction, the similarity effect was
compared in words suffixed with –ed to the effect in monomorphemic words ending in
/d/.
All monomorphemic monosyllabic words ending in /d/ were extracted from the
ELP as were all monosyllabic bimorphemic words containing the suffix –ed. To ensure
that the mono- and multimorphemic words had similar syllabic structure, only words with
a VC rhyme structure were included in the analysis. This meant that all monomorphemic
words contained a single coda consonant (e.g., crowd, bread, slide) while all suffixed
words consisted of a vowel-final root plus the past-tense suffix (e.g., glowed, tied,
weighed). This criterion also ensured that the past-tense morpheme was always
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
33
expressed as the /d/ allomorph. The multimorphemic words in this set were found to be
slightly but significantly longer by phoneme than the monomorphemic words (mean
length mono: 3.43; multi: 3.55; t(3703) = -5.73, p < .0001). Analyses indicated that if all
trials of the 3 multimorphemic words containing 3 onset consonants (screwed, sprayed,
strayed) were removed, the two sets would not differ in length (mono: 3.43, multi: 3.44;
t(3626) = -0.33; p = .74). To ensure the generalizability of the results, the analyses were
run with and without these words.
For this analysis, the model consisted of the baseline control variables as well as
the following variables of interest: Number of Morphemes, Onset-Coda Phoneme
Similarity (range: 1-4), and the interaction term Onset-Coda Phoneme Similarity x
Number of Morphemes (see Appendix for all of the variables included in the model). To
facilitate interpretation, Number of Morphemes was contrast coded (one morpheme = -1,
two morphemes = 1) and Onset-Coda Phoneme Similarity was centered (centered range: -
1.4–2.8). The model also contained the interaction term as a random slope for both
subjects and items (the model failed to converge when similarity and number of
morphemes were also included, so it was decided to remove them in favor of the
interaction term).
Following the removal of 84 outliers, the dataset consisted of 2586
monomorphemic trials (from 99 word types) and 1035 multimorphemic trials (from 42
types). The regression analysis revealed a significant inhibitory effect of phoneme
similarity (β = .020; s.e. = .075; t = 2.7; χ2(1) = 8.1, p < .01) and a trend for an inhibitory
effect of the number of morphemes (β = .013; s.e. = .008; t = 1.7; χ2(1) = 3.1, p = .08).
Crucially, the effect of phoneme similarity interacted with number of morphemes (β = -
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
34
.017; s.e. = .006; t = -3.0; χ2(1) = 9.8, p < .01), with the effect of similarity diminishing in
the suffixed words relative to monomorphemic words (the results of this analysis are
presented in Table 3). The interaction term was still significant when the suffixed words
containing 3 onset consonants were excluded (β = -.017; s.e. = .006; t = -2.9; χ2(1) = 9.3,
p < .01). Figure 2 presents a graph of the interaction effect. In this figure, Number of
Morphemes (centered) is on the x-axis, the various levels of Onset-Coda Similarity
(centered) are represented by the various dashed lines, and raw RT in milliseconds is on
the y-axis (the log transformation used in the analysis was undone in the graph to
facilitate interpretation). The changing spread of the similarity variable depicts how the
effect of similarity on RT differs in mono- and multimorphemic words.
A couple of measures were taken to rule out possible confounds in this analysis.
First, to ensure that the effect of similarity was not being driven by the inhibitory effect
of processing repeated letters (Schoonbaert & Grainger, 2004), all words containing an
initial letter D (e.g., dread, died) were removed and the analysis was rerun. The
interaction was still found to be significant (β = -.018; s.e. = .006; t = -3.3; χ2(1) = 11.6, p
< .001), indicating that the effect is truly related to phonological similarity. Second, it
was observed that long and short vowels were not equally distributed across the two
groups of words. All of the multimorphemic words contained a long vowel while the
monomorphemic words contained both short and long vowels. If the strength of the
phoneme similarity effect is influenced by the length of the vowel, this numerical
imbalance could lead to a weaker effect of similarity in the multimorphemic words. An
analysis conducted on the monomorphemic words found that vowel length did not
influence reaction times (Vowel Length: β = .004; s.e. = .005; t = 0.7; χ2(1) = 0.4, p = .5)
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
35
and did not interact with phoneme similarity (Onset-Coda Phoneme Similarity x Vowel
Length: β = .001; s.e. = .004; t = 0.3; χ2(1) = .1, p = .8), ruling it out as a confound in the
overall analysis.
Summary
The post-lexical processing of multimorphemic words was investigated by
comparing the inhibitory effect of phoneme similarity in suffixed words relative to form-
matched monomorphemic words. A significant effect of similarity was found, indicating
that phonemes from the same and different morphemes compete with each other during
phonological encoding. The similarity effect was reduced when the final phoneme
corresponded to the past tense suffix, suggesting that phonemes in different morphemes
do not compete as strongly with each other as phonemes located within the same
morpheme. These results support the Heterogeneity of Processing Hypothesis.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
36
Figure 2. Results of Study 2. Number of Morphemes (centered values: -1, 1) is depicted on the x-axis while Phoneme Similarity (centered values: -1.37–2.79) is represented by the dashed lines (raw RT is on the y-axies). The interaction between phoneme similarity and number of morphemes is illustrated by the change in the spread of the lines: the effect on RT is larger in monomorphemic words (coded as Number of Morphemes = -1) than disyllabic words (coded as Number of Morphemes = 1).
Table 3 Result summary: coefficient estimates β, standard errors SE(β), and t-values for all predictors in the Study 2 analysis of words suffixed with –ed and monomorphemic words ending in /d/.
-1.0 -0.5 0.0 0.5 1.0
640
660
680
700
720
NumberOfMorphemes.Centered
RT
(ms)
-1.370.020.480.711.412.79
PhonemeSimilarity.Centered
Predictor Coeff. β SE(β) t
(Intercept) 6.509 0.073 89.2
Initial Segment Frication-Frication 0.070 0.012 6.1
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
37
!
!
Study 3
As a follow-up to Study 2, the strength of the phoneme similarity effect was
Initial Segment Frication-Short 0.066 0.018 3.7
Initial Segment Voicing-Voiceless 0.064 0.014 4.6
FrequencyInitialDiphoneSyllable 0.004 0.004 1.0
FrequencyInitialDiphone -0.011 0.004 -2.4
Root Consistency PC1 0.001 0.003 0.3
Root Consistency PC2 0.007 0.003 2.2
Root Consistency PC3 0.003 0.003 1.2
Root Log Frequency -0.008 0.004 -2.1
Root Written-Spoken Frequency Ratio -0.002 0.004 -0.6
Root Orthographic Neighborhood Density -0.001 0.002 -0.6
Root Inflectional Entropy -0.017 0.008 -2.2
Word Log Frequency -0.011 0.003 -3.1
Word Orthographic Neighborhood Density 0.002 0.002 0.8
Word Letter Length 0.026 0.009 3.0
Onset-Coda Letter Similarity -0.015 0.012 -1.2
Onset-Coda Phoneme Similarity (centered) 0.020 0.007 2.7
Number of Morphemes (centered) 0.013 0.008 1.7
Onset-Coda Phoneme Similarity x
Number of Morphemes (centered) -0.017 0.006 -2.9
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
38
evaluated in words ending in the adjective-forming suffix –y and monomorphemic words
ending in /i/. This dataset provides an opportunity to test the generalizability of the cross-
boundary weakening observed in Study 2: the analysis involves a different suffix, a
different syllable structure (disyllabic words), a different similarity metric (vowel
similarity), and a different set of regression variables (to accommodate the visual
processing of disyllabic words).
All disyllabic words ending in /i/ were extracted from the English Lexicon
Project. Words containing the diphthongs /aɪ/, /oɪ/, and /aʊ/ were excluded, as were
words with r-colored vowels. Words were classified as either monomorphemic or
containing the adjective-forming suffix –y. Words that could be construed as containing a
bound root (e.g., silly, happy, burly) were categorized as monomorphemic.4 Since the
effect of vowel similarity is likely to diminish with distance, only words containing a
single phonological consonant between the two vowels were included in the analysis
(e.g., body, soapy).
The variables used in the analysis were based on the work of Yap and Balota
(2009), who investigated the factors that influence the processing of multisyllabic words.
Since not all of the variables that Yap and Balota found to be significant predictors of
oral reading reaction times have been collected for multimorphemic words, a subset of
their predictors were included in this analysis. In addition to these variables, two
variables encoding properties of the root word were included (root frequency and root
orthographic neighborhood density) as was a variable encoding the similarity of the
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!4 Preliminary analyses investigating whether these words behave as if they contain the suffix –y, despite containing a bound root, were inconclusive. As such, they were treated as monomorphemic.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
39
vowel letters (e.g., the similarity of OA and Y for soapy). Vowel similarity was
calculated using the procedure described in the methods section above. The full set of
variables is provided in Appendix A.
Following the removal of 176 outlier trials, the dataset consisted of 3698
monomorphemic trials (147 types) and 4114 multimorphemic trials (159 types). The
analysis found a marginal inhibitory effect of vowel similarity (β = .007; s.e. = .004; t =
1.9; χ2(1) = 3.6; p = .056) and a significant advantage for multimorphemic words relative
to monomorphemic words (β = -.015; s.e. = .005; t = -2.9; χ2(1) = 8.0; p < .01). The
interaction between similarity and number of morphemes came close to but did not reach
significance, with the effect of similarity diminishing in multimorphemic words (β = -
.006; s.e. = .004; t = -1.7; χ2(1) = 3.1; p = .076). A summary of the results can be found
in Table 4.
Summary
The results of Study 3 broadly support the Heterogeneity of Processing
hypothesis. A marginal inhibitory effect of vowel similarity was found, suggesting that
vowels compete for selection during phonological encoding in much the same way as
consonants, as was a significant facilitatory effect of morphological structure.
At the moment, it is unclear why the effect of similarity is weaker here than the
effect of similarity observed for consonants in Study 2. Shattuck-Hufnagel (1986) reports
that spontaneous speech errors involving vowels occur less frequently than errors
involving consonants, suggesting that vowels may not compete with each other as
strongly during phonological encoding as do consonants. Another possibility is that
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
40
segment competition decreases across syllable boundaries, causing the vowels analyzed
in here to interact with each other more weakly than the consonants analyzed in Study 2.
Crucially, a facilitatory interaction between vowel similarity and number of
morphemes was found that approached significance, suggesting that competition between
phonemes is reduced across the morpheme boundary. These results are broadly similar to
those observed in Study 2 and thus provide some support for the notion that
morphological boundaries act as processing boundaries in spoken production.
Phonological processing, indexed here by vowel competition, appears to be weaker
across morphemes than within.
Table 4 Result summary: coefficient estimates β, standard errors SE(β), and t-values for all predictors in the Study 3 analysis of words suffixed with –y and monomorphemic words ending in /i/.
Predictor Coeff. β SE(β) t
(Intercept) 6.718 0.076 88.2
Initial Segment Voicing -0.009 0.005 -1.8
Initial Segment Frication 0.038 0.008 4.5
Syl1 Onset FF Consistency -0.056 0.026 -2.1
Syl1 Rhyme FF Consistency 0.005 0.016 0.3
Syl2 Onset FF Consistency -0.029 0.018 -1.7
Syl2 Rhyme FF Consistency -0.209 0.062 -3.4
Syl1 Onset FB Consistency -0.068 0.023 -3.0
Syl1 Rhyme FB Consistency 0.004 0.013 0.3
Syl2 Onset FB Consistency 0.052 0.017 3.0
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
41
Syl2 Rhyme FB Consistency -0.041 0.029 -1.4
Root Log Frequency -0.001 0.002 -0.4
Root Orthographic Neighborhood Density 0.000 0.001 0.1
Word Log Frequency -0.022 0.002 -9.7
Word Orthographic Neighborhood Density 0.002 0.002 1.1
Word Length in Letters 0.037 0.006 6.0
Vowel-Vowel Letter Similarity 0.012 0.009 1.4
Vowel-Vowel Similarity (centered) 0.007 0.004 1.9
Number of Morphemes (centered) -0.015 0.005 -2.9
Vowel-Vowel Similarity x Number of Morphemes (centered) -0.006 0.004 -1.7
General Discussion
In this paper, the character of post-lexical processing for multimorphemic words
was considered. Drawing on findings from the linguistic and psycholinguistic literature
showing that 1) multimorphemic words are represented in terms of their component
morphemes at lexical levels, 2) have structural weaknesses at morpheme boundaries in
their post-lexical representations, and the fact that 3) post-lexical processes are influenced
by lexical properties, the Heterogeneity of Processing Hypothesis was proposed which
predicts that post-lexical processing is heterogeneous across a multimorphemic word. In
particular, post-lexical processing is predicted to be 1) weaker for heteromorphemic than
tautomorphemic sequences and 2) to vary according to each morpheme’s lexical
properties.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
42
The first prediction was tested by comparing the strength of phoneme competition
during phonological encoding across and within morphemes. Analyses of oral reading
latencies showed that phoneme competition—as indexed by an inhibitory effect of
phoneme similarity—was significantly weaker in words suffixed with –ed compared to
matched monomorphemic words ending in /d/. An analysis of words suffixed with –y
showed a similar pattern that approached significance. These results indicate that
phonological encoding, the stage when phonemes are retrieved, selected, and ordered,
applies more weakly for heteromorphemic than tautomorphemic phonemes, providing
support for the HPH.
The status of affixes in spoken production
In an examination of words suffixed with –ed, Cohen-Goldberg (2012) reported
significant effects of onset-suffix similarity on oral reading reaction times, suggesting
that root and suffix phonemes compete with each other during post-lexical processing.
The present study replicated these findings (Study 1: –er, –ing) and demonstrated that
this competition is weaker than the competition that occurs among root phonemes (Study
2: –ed, Study 3: –y). One question these results raise is: what is the strength of root-
suffix phoneme competition? Although significant effects of onset-suffix similarity were
found in Study 4 of Cohen-Goldberg (2012) and Study 1 here, inspection of the beta
weights for the interaction terms in Studies 2 and 3 (see Tables 3 and 4 above) suggest
that the effect of similarity is nearly or completely diminished in multimorphemic words.
The clearest comparison is Cohen-Goldberg’s Study 4 and Study 2 of the present
investigation, which both examined competition in words suffixed with –ed. Although it
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
43
is possible that Study 2 represents a failure to replicate the heteromorphemic similarity
effect, what seems more likely is that cross-boundary competition occurs but is rather
weak and requires a large number of trials to observe. Cohen-Goldberg’s study, which
found a significant effect of onset-suffix similarity, consisted of 15,896 trials from 620
words while in the present work, Study 2 contained 1035 trials from 42 words. Future
work should attempt to quantify the size of the competition effect within and across
morphemes (and across different affixes) using sufficiently large datasets.
While the reduction in the phoneme similarity effect observed in Studies 2 and 3
is consistent with the claim that post-lexical processes are weaker when applying across
suffix boundaries, it is important to consider an alternative possibility which is that the
suffix phonemes may simply have had less activation—and thus competed more
weakly—than the equivalent non-morphemic phonemes.
Although it is not possible at this time to definitively say one way or another,
there is some evidence that affix phonemes are not weaker than root phonemes and may
in some ways have a more robust representation at post-lexical stages of processing.
First, although /t/ and /d/ are frequently deleted in word-final position in English,
especially when followed by another consonant, it has been found that this deletion
process is less likely to occur when the final segment corresponds to the past-tense suffix
–ed than non-morphemic /t/ and /d/ (e.g., Neu, 1980; Roberts, 1997). Furthermore, it has
been shown in the laboratory that the plural and past tense suffixes are pronounced with
greater duration than their non-morphemic counterparts (Losiewicz, 1995; Schwarzlose
& Bradlow, 2001; Walsh & Parker, 1983;). Finally, work by Kuperman et al. (2007) on
Dutch interfixes suggests that affixes may receive paradigmatic support from the lexicon,
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
44
increasing the robustness of their representation and the duration of their phonetic
realization. It thus seems reasonable to conclude that the suffix phonemes examined in
the present investigation were not weaker—and in fact may be stronger from a
phonological perspective—than the latter.
Towards a functionalist theory of morphophonology
The fact that phonemes interact with each more weakly when in different
morphemes may be important in accounting for a number of morphophonological
phenomena in the world’s languages. It has been noted that there is a cross-linguistic
tendency for phonological complexity to correlate with morphological productivity:
words with productive morphology often contain marked phonological sequences
(Burzio, 2002; Carlson & Gerfen, 2011; Hay & Baayen, 2003; Zuraw, 2009). One
example of this general organization may be seen in Turkish vowel harmony: while
vowels in root-suffix combinations must harmonize, the vowels in compounds do not.
Similarly in English, although geminates are banned from monomorphemic words
(*spaghe[tt]i) and words containing less productive suffixes (e.g., in-: i[n]umerable,
*i[nn]umerable), they are allowed in words containing more productive affixes and
compounds (e.g., un-: u[nn]ecessary; boo[kk]eeper).
This pattern has played a pivotal role in some theories of phonological grammar,
primarily the theory of Lexical Phonology and Morphology proposed by Kiparsky (1982)
and Monahan (1982). Although a review of this theory is beyond the scope of this paper,
the critical element for present purposes is that productive morphology is assumed to
form a lexical stratum (‘Level 2’ morphology) that is subject to fewer phonological rules
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
45
than a stratum containing less productive morphology (‘Level 1’). This organization
caused certain phonological rules affecting less productive morphology to be absent in
productive morphology (e.g., degemination). While this organization furnished the
theory with descriptive adequacy, it crucially failed to provide it with explanatory
adequacy—the correlation between morphological productivity and phonological
inertness was merely stipulated and not derived from more basic principles.
The heterogeneous quality of post-lexical processing argued for here may provide
part of a functional explanation for this correlation. While the origin of phonological
grammars remains hotly debated, there is growing belief that many phonological rules
derive from phonetically-motivated sound patterns (e.g., Blevins, 2004). For example, it
has been claimed that vowel harmony rules arise when phonetic vowel-to-vowel
coarticulation becomes learned as a phonological process (e.g., Benguerel & Cowan,
1974; Bell-Berti & Harris, 1979; Purcell, 1979; Fowler, 1981). In multimorphemic
words, the fact that post-lexical processes apply more weakly across morphemes than
within may serve to resist to some degree the types of phonetic changes that seed
grammatical rules. For example, vowel-to-vowel coarticulation may be reduced across
compound boundaries, producing fewer phonetic traces that could be grammaticalized
into a compound vowel harmony rule. The fact that each morpheme has somewhat
different phonetic properties (e.g., different degrees of vowel space expansion) as a result
of different levels of cascading activation and differences in stored motor plans may
further disrupt these diachronic processes. In contrast, a disyllabic monomorphemic
word, which constitutes a homogenous domain for post-lexical phonological processes,
receives far more consistent levels of activation, and is associated with a single motor
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
46
plan, is likely to be more vulnerable to these sorts of phonetic processes and as a result, a
more likely target for grammatical rules.5
The results in Study 2 suggest a similar effect may be behind the distribution of
geminates in English mono- and multimorphemic words. The inhibitory effect that
consonant competition exerts in production has been proposed to be one of the factors
that may lead to the various ‘Obligatory Contour Principle’ bans that languages have on
similar and geminate consonants within morphemes (Berg, 1998; Cohen-Goldberg, 2012;
Frisch, 2004). The fact that competition is reduced across morpheme boundaries, as
demonstrated in Study 2, suggests that this functional pressure is eased in productively
formed words (the effect of similarity has a less deleterious effect in processing) and
perhaps contributes to the appearance and tolerance of geminates in English productive
morphology.
Lexicalization
A similar process, but in reverse, may be behind morphophonological patterns
relating to lexicalization, the process by which historically morphologically complex
words come to be represented by a single morpheme. The lexicalization process is
frequently accompanied by shifts towards monomorphemic phonology (e.g., Brinton, &
Traugott, 2005). For example, the word cupboard diachronically underwent the
phonological change [kʌ́p.bɔ̀rd] → [kʌ́.br̩d]. The original pronunciation contains
phonology that is only found in compounds (adjacent syllables that bear stress, adjacent
stop consonants that disagree in voicing) while the contemporary pronunciation has
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!5 This analysis bears some similarity to, and indeed may provide a functional explanation for, output-output theories of the phonological grammar (Benua, 1997; Burzio, 1994).
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
47
eliminated these patterns in favor of those found in monomorphemic words (trochaic
stress, no voicing disagreement in adjacent stops). Initially, the influence of having two
distinct roots in the lexical representation (cup, board) may have weakened the
phonological constraints that would have otherwise repaired its form (e.g., the sequence
/pb/). As the meaning of the word drifted, however, the representation may have shifted
from comprising multiple lexical representations to consisting of a single unit, thereby
increasing the strength of post-lexical processes across the word. This would have led,
over time, to the word taking on a more monomorphemic phonological profile.
The Heterogeneity of Processing hypothesis thus may provide a processing
account of lexicalization and of lexical strata. Lexicalization (and shifts from later to
earlier lexical strata, e.g., Level 2 to Level 1) could be defined at lexical levels as the shift
towards unitary lexical representations and at post-lexical stages as the homogenization
of processing. As phonological representations across morphemes become more unitary
and as variation in activation levels are reduced, post-lexical processing may become
increasingly homogenous. Cross-boundary interaction should increase—phonemes in
different morphemes will have a stronger influence on each other’s pronunciation—and
phonemes across the word should come to inherit increasingly similar levels of activation
and articulatory parameters.
Future directions
The work reported here constitutes an important but preliminary step towards
understanding the post-lexical processing of multimorphemic words. More work is
needed to confirm the results reported here, ideally using other post-lexical processes
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
48
(e.g., coarticulation). The second prediction of the HPH—that processing will vary
across a multimorphemic word as a function of each morpheme’s lexical properties—
similarly needs to be investigated.
On the theoretical side, the HPH needs to be elaborated to account for how post-
lexical processing occurs when multimorphemic words have whole-word representations
in addition to their morpheme-based representations. The exact direction this elaboration
will take will depend in part on lexical theories of morphological processing. The
simplest case is if whole-word representations compete with morpheme-based
representations in a horse race, with only one or the other being used to access post-
lexical representations. For example, both <afternoon> and <after>+<noon> may exist
as lexical representations for afternoon in the phonological output lexicon but at any
given time, only one is utilized to produce the word. In this case, the morpheme-based
representation would be processed as described by the HPH above while the whole-word
representation would likely be processed (more) similarly to a monomorphemic word. A
more complex situation is if whole-word and morpheme-based representations both
contribute in the processing of a word. This could happen if a whole-word lexical
representation provides access to its component morphemes (<afternoon> →
<after>+<noon>) or if both representations compete for selection and send cascading
activation to post-lexical levels. In this case, the strength of the morpheme boundary and
the exact lexical properties inherited by phoneme may be a blend of the two
representations (e.g., Hay, 2003).
One promising outcome of this program of research is that the HPH, should it find
support, could potentially serve as a linking hypothesis allowing post-lexical processing
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
49
data to be used to inform theories of lexical processing. To date, nearly all studies
examining the lexical structure of morphologically complex words have utilized tasks
thought to tap into lexical stages of processing (e.g., examining the effect of morpheme
frequency on reaction times). The HPH, however, since it explicitly links lexical and
post-lexical processing, may allow inferences about lexical processing to be made on the
basis of post-lexical data. If the second claim of the HPH (that phonemes inherit the
lexical properties of their parent morpheme) can be verified, for example, it could be used
to adjudicate between morpheme-based and whole-word theories of lexical
representation. If morphologically complex words are represented in terms of their
component morphemes, each root in a compound should have different phonetic qualities
that correlate with each root’s lexical properties. If morphologically complex words are
instead represented by a single lexical unit (e.g., Caramazza, 1997), each root should
have the same phonetic qualities and these qualities should correlate with the lexical
properties of the entire word.
Another extension of the theory would be in the investigation of gradient theories
of morphological processing. Although the discussion of lexical representation in the
Introduction was cast in terms of a binary word-based/morpheme-based distinction,
recent theories have argued that the distinction between morpheme- and word-based
representations is gradient, rather than categorical. Some approaches have suggested that
morphemes can be joined together with different strengths (Hay, 2003) while others have
suggested that a variety of factors, including each morpheme’s paradigmatic/information
theoretic support from the lexicon will affect a word’s structure (e.g., Kuperman, Betram
& Baayen, 2008, 2010; Moscoso del Prado Martín, Kostić, & Baayen, 2004). Apart from
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
50
Hay’s (2003) study discussed in the Introduction, the consequences of this structure on
post-lexical processing have not yet been investigated. It seems likely that the HPH
could be extended fairly straightforwardly to make predictions for gradient structure (e.g.,
some words have stronger cross-boundary processing than others) and investigation along
these lines could not only shed light on the influence of gradient lexical structure on post-
lexical processing but potentially provide a point of converging evidence for gradient
theories of morphology.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
51
Appendix A Fixed-effects variables used in the regression analyses. Study 1 Study 2 Study 3 Initial Segment Frication Initial Segment Voicing FrequencyInitialDiphoneSyllable FrequencyInitialDiphone Root Consistency PC1 Root Consistency PC2 Root Consistency PC3 Root Log Frequency Root Written-Spoken Frequency Ratio Root Orthographic Neighborhood Density Root Inflectional Entropy Word Root Log Frequency Word Orthographic Neighborhood Density Word Letter Length Root Onset-Coda Letter Similarity Root Onset-Coda Phoneme Similarity Root Coda-Suffix Letter Similarity Root Coda-Suffix Phoneme Similarity
Initial Segment Frication Initial Segment Voicing FrequencyInitialDiphoneSyllable FrequencyInitialDiphone Root Consistency PC1 Root Consistency PC2 Root Consistency PC3 Root Log Frequency Root Written-Spoken Frequency Ratio Root Orthographic Neighborhood Density Root Inflectional Entropy Word Log Frequency Word Orthographic Neighborhood Density Word Letter Length Onset-Coda Letter Similarity Onset-Coda Phoneme Similarity Number of Morphemes Onset-Coda Phoneme Similarity x Number of Morphemes
Initial Segment Voicing Initial Segment Frication Syl1 Onset FF Consistency Syl1 Rhyme FF Consistency Syl2 Onset FF Consistency Syl2 Rhyme FF Consistency Syl1 Onset FB Consistency Syl1 Rhyme FB Consistency Syl2 Onset FB Consistency Syl2 Rhyme FB Consistency Root Log Frequency Root Orthographic Neighborhood Density Word Log Frequency Word Orthographic Neighborhood Density Word Length in Letters Vowel-Vowel Letter Similarity Vowel-Vowel Similarity Number of Morphemes Vowel-Vowel Similarity x Number of Morphemes
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
52
!References Agresti, A. (2002). Categorical data analysis (2nd Edn.). New York, NY: John Wiley &
Sons. Ahrens, R. (1977). Wortfindungsstöngen für zusammengeste Worte (Nomina Composita)
bei Aphasien. Archiv für Psychiatric and Nervenkrankheiten 224, 73-87. Aronoff, M. & Anshen, F. (1981). Morphological productivity and phonological
transparency. Canadian Journal of Linguistics, 26, 63-72. Aylett, M., & Turk, A. (2006). Language redundancy predicts syllabic duration and the
spectral characteristics of vocalic syllable nuclei. Journal of the Acoustical Society of America, 119, 3048–3058.
Baayen, R. H. (1994). Derivational productivity and text typology. Journal of Quantitative Linguistics, 1(1), 16-34.
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59, 390–412.
Baayen, R. H., Dijkstra, T., & Schreuder, R. (1997). Singulars and plurals in Dutch: Evidence for a parallel dual route model. Journal of Memory and Language, 37, 94–117.
Baayen, R. H., Feldman, L., & Schreuder, R. (2006). Morphological influences on the recognition of monosyllabic monomorphemic words. Journal of Memory and Language, 55(2), 290–313.
Baayen, R. H., Levelt, W., Schreuder, R., & Ernestus, M. (2008). Paradigmatic structure in speech production. Proceedings of the Chicago Linguistics Society, 43(1), 1-29.
Baese-Berk, M., & Goldrick, M. (2008). Mechanisms of interaction in speech production. Language & Cognitive Processes, 24(4), 527-554.
Badecker, W. (2001). Lexical composition and the production of compounds: Evidence from errors in naming. Language & Cognitive Processes, 16(4), 337-366.
Badecker, W., & Caramazza, A. (1991). Morphological composition in the lexical output system. Cognitive Neuropsychology, 8, 335-367.
Badecker, W., Hillis, A., Caramazza, A. (1990). Lexical morphology and its role in the writing process: Evidence from a case of acquired dysgraphia. Cognition, 35, 205-243.
Badecker, W., Rapp, B., & Caramazza, A. (1996). Lexical morphology and the two orthographic routes. Cognitive Neuropsychology, 13, 161-175.
Bailey, T., & Hahn, U. (2005). Phoneme similarity and confusability. Journal of Memory and Language, 52(3), 339–362.
Balota, D. A., Yap, M. J., Cortese, M. J., Hutchison, K. A., Kessler, B., Loftis, B., et al. (2007). The English Lexicon Project. Behavior Research Methods, 39(3), 445- 459.
Bates, D., Maechler, M. & Dai, B. (2008). The lme4 Package (Version 0.999375-40): linear mixed-effects models using S4 classes. Available (November 2011) at http://cran.r-project.org/web/packages/lme4/lme4.pdf.
Bauer, L. (2004). The function of word-formation and the inflection-derivation distinction. In H. Aertsen, M. Hannay & R. Lyall (Eds), Words in their Places. A Festschrift for J. Lachlan Mackenzie (pp. 283-292). Amsterdam: Vrije Universiteit.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
53
Bell, A., Jurafsky, D., Fosler-Lussier, E., Girand, C., Gregory, M., & Gildea, D. (2003). Effects of disfluencies, predictability, and utterance position on word form variation in English conversation. Journal of the Acoustical Society of America, 113, 1001–1024.
Bell, A., Brenier, J. M., Gregory, M., Girand, C., & Jurafsky, D. (2009) Predictibility effects on durations of content and function words in conversational English. Journal of Memory and Language, 60(1), 92-111.
Bell-Berti, F. & Harris, K. (1979). Anticipating coarticulation: some implications of a study of lip rounding. Journal of the Acoustical Society of America, 65, 1268-1270.
Benguérel, A.-P. & Cowan, H. (1974). Coarticulation of upper lip protrusion in French. Phonetica, 30, 41-55.
Berg, T. (1998). Linguistic structure and change: An explanation from language processing. Oxford: Clarendon Press.
Bien, H., Levelt, W. J. M., & Baayen, R. H. (2005). Frequency effects in compound production. Proceedings of the National Academy of Sciences, 102(49), 17876-17881.
Blanken, G. (2000). The production of nominal compounds in aphasia. Brain and Language, 74, 84-102.
Blevins, J. (2004). Evolutionary phonology: the emergence of sound patterns. Cambridge: Cambridge University Press.
Bock, J. K. (1987). An effect of accessibility of word forms on sentence structures. Journal of Memory and Language, 26, 119-137.
Boles, D., & Clifford, J. (1989). An upper-and lowercase alphabetic similarity matrix, with derived generation similarity values. Behavior Research Methods Instruments and Computers, 21(6), 579–586.
Bölte, J., Zwitserlood, P. & Dohmes, P. (2004). Morphology in experimental speech production research. In T. Pechmann & C. Habel, (Eds.) Trends in linguistics: Multidisciplinary approaches to language production (pp. 431–471). Berlin: Mouton de Gruyter.
Bölte, J., Dohmes, P., & Zwitserlood, P. (2012). Interference and facilitation in spoken production: Effects of morphologically and semantically related context stimuli on picture naming. Journal of Psycholinguistic Research, 1-26.
Brinton, L. J. & Traugott, E. C. (2005). Lexicalization and Language Change. Cambridge: Cambridge University Press.
Burzio, L. (2002). Missing players: Phonology and the past-tense debate. Lingua, 112(3), 157-199.
Butterworth, B. (1983). Lexical representation. In B. Butterworth (Ed.), Language production (pp. 257–294). New York: Academic Press.
Butterworth, B. (1989). Lexical access in speech production. In W. Marslen-Wilson (Ed.), Lexical representation and process (pp. 108-135). Cambridge, MA: MIT Press.
Bybee, J. (1995). Regular morphology and the lexicon. Language and Cognitive Processes, 10(5), 425-455.
Bybee, J. (2001). Phonology and language use. Cambridge: Cambridge University Press. Caramazza, A. (1997). How many levels of processing are there in lexical access?
Cognitive Neuropsychology, 14, 177-208. Carlson, M. T. & Gerfen, C. (2011). Productivity is the key: Morphophonology and the
riddle of alternating diphthongs in Spanish. Language, 87(3), 510-538.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
54
Chen, T-M., & Chen, J-Y. (2006). Morphological encoding in the production of compound words in Mandarin Chinese. Journal of Memory and Language, 54, 491-5.
Chen, J-Y., & Chen, T-M. (2007). Form encoding in Chinese word production does not involve morphemes. Language and Cognitive Processes 22(7), 1001-1020.
Cho, T. (2001). Effects of morpheme boundaries on intergestural timing: Evidence from Korean. Phonetica, 58(3), 129-162.
Cholin, J., Rapp, B., & Miozzo, M. (2010). When do combinatorial mechanisms apply in the production of inflected words? Cognitive Neuropsychology, 27(4), 334-359.
Chomsky, N. and Halle, M. (1968). The sound pattern of English. New York: Harper and Row.
Cohen-Goldberg, A. M. (2012). Phonological competition within the word: Evidence from the phoneme similarity effect in spoken production. Journal of Memory and Language. http://dx.doi.org/10.1016/j.jml.2012.03.007
Cohen-Goldberg, A. M., Cholin, J., Miozzo, M., & Rapp, B. (submitted). The interface between morphology and phonology: Exploring a morpho-phonological deficit in spoken production.
Cutler, A. (1980a). Productivity in word formation. Papers from the Sixteenth Regional Meeting, Chicago Linguistic Society, 45-51.
Cutler, A. (1980b). Errors of stress and intonation. In V. Fromkin (Ed.), Errors in linguistic performance: Slips of the tongue, ear, pen, and hand (pp. 67–80). New York: Academic Press.
Cutler, A. (1981). Degrees of transparency in word formation. Canadian Journal of Linguistics, 26, 73-77.
Delazer, M., & Semenza, C. (1998). The processing of compounds: A study in aphasia. Brain and Language, 61, 54–62.
Dell, G. (1986). A spreading-activation theory of retrieval in sentence production. Psychological Review, 93(3), 283–321.
Dell, G. S. (1990). Effects of frequency and vocabulary type on phonological speech errors. Language and Cognitive Processes, 4, 313–349.
Dell, G. S. & O'Seaghdha, P. G. (1992) Stages of lexical access in language production. Cognition, 42, 287-314.
Dell, G., Schwartz, M., Martin, N., Saffran, E., & Gagnon, D. (1997). Lexical access in aphasic and nonaphasic speakers. Psychological Review, 104(4), 801–838.
Elman, J. L. (2004). An alternative view of the mental lexicon. Trends in Cognitive Sciences, 8(7), 301-306.
Ernestus, M. & Baayen, H. (2007). Paradigmatic effects in auditory word recognition: The case of alternating voice in Dutch. Language and Cognitive Processes, 22(1), 1-24.
Ernestus, M., Lahey, M., & Verhees, F. (2006). Lexical frequency and voice assimilation. Journal of the Acoustical Society of America, 120(2), 1040-1051.
Fiorentino, R., & Poeppel, D. (2007). Compound words and structure in the lexicon. Language and Cognitive Processes, 22(7), 953-1000.
Fowler, K. (1981) Production and perception of coarticulation among stressed and unstressed vowels. Journal of Speech and Hearing Research, 46, 127-139.
Frazier, M. (2006). Output-Output Faithfulness to Moraic Structure: Evidence from American English. In C. Davis, A. R. Deal & Y. Zabbal (Eds.), Proceedings of NELS
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
55
36 v. 1 (pp. 1-14). Frisch, S. A. (2004). Language processing and segmental OCP effects. In B. Hayes, R.
Kirchner & D. Steriade (Eds.), Phonetically based phonology. Cambridge, UK: Cambridge University Press.
Fromkin, V. A. (1971). The non-anomalous nature of anomalous utterances. Language, 47, 27-52.
Gahl, S. (2008). "Time" and "thyme" are not homophones: Word durations in spontaneous speech. Language, 84, 474–496.
Gahl, S., Yao, Y., & Johnson, K. (2012). Why reduce? Phonological neighborhood density and phonetic reduction in spontaneous speech. Journal of Memory and Language, 66, 789-806.
Garrett, M. E. (1975). The analysis of sentence production. In G. H. Bower (Ed.), The psychology of learning and motivation (pp. 133-175). San Diego: Academic Press.
Garrett, M. F. (1980). Levels of processing in sentence production. In B. Butterworth (Ed.), Language production (Vol. 1, pp. 177-210). London: Academic Press.
Goldrick, M., Baker, H. R., Murphy, A., Baese-Berk, M. (2011). Interaction and representational integration: Evidence from speech errors. Cognition, 121, 58-72.
Hare, M. & Elman, J. L. (1995). Learning and morphological change. Cognition, 56, 61-98.
Harley, T. A., & Bown, H. E. (1998). What causes a tip-of-the-tongue state? Evidence for lexical neighbourhood effects in speech production. British Journal of Psychology, 89, 151–174.
Hay, J. (2003). Causes and consequences of word structure. New York and London: Routledge.
Hay, J. B. & Baayen, R. H. (2003) Phonotactics, parsing and productivity. Italian Journal of Linguistics, 1, 99–130.
Hay, J. & Plag, I. (2004). What constrains possible suffix combinations? On the interaction of grammatical and processing restrictions in derivational morphology. Natural Language & Linguistic Theory, 22(3), 565-596.
Hittmair-Delazer, M., Andree, B., Semenza, C., De Bleser, R., & Benke, T. (1994). Naming by German compounds. Journal of Neurolinguistics, 8(1), 27-41.
Hockett, C. (1942). A system of descriptive phonology. Language, 18, 3-21. Jackendoff, R. S. (1975). Morphological and semantic regularities in the lexicon.
Language, 51, 639–671. Janssen, N., Bi, Y., & Caramazza, A. (2008). A tale of two frequencies: Determining the
speed of lexical access for mandarin Chinese and English compounds. Language and Cognitive Processes, 23, 1191-1223.
Jescheniak, J. D. & Levelt, W. J. M. (1994) Word frequency effects in speech production: Retrieval of syntactic information and of phonological Form. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(4), 824-843.
Joos, M. (1964). A chapter of semiology in the English verb. In C. Stuart (Ed.) Report of the 15th Annual (1st International) Round Table Meeting on Linguistics and Language Studies (pp. 59-72). Washington, D. C.: Georgetown University Press.
Juhasz, B. J. & Berkowitz, R. N. (2011). Effects of morphological families on English compound word recognition: A multitask investigation. Language and Cognitive Processes, 26(4-6), 653-682.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
56
Kempen, G., & Huijbers, P. (1983). The lexicalization process in sentence production and naming: Indirect election of words. Cognition, 14, 185-209.
Kiparsky, P. (1973). Phonological representations. In O. Fujimura, (Ed.) Three dimensions in linguistic theory. Tokyo: TEC.
Kiparsky, P. (1982) Lexical Phonology and Morphology, in I.S. Yang ed. Linguistics in the Morning Calm, 3-91. Seoul: Hanshin.
Knobel, M., Finkbeiner, M., & Caramazza, A. (2008). The many places of frequency: Evidence for a novel locus of the lexical frequency effect in word production. Cognitive Neurosychology, 25(2), 256-286.
Krott, A., Baayen, R. H., & Schreuder, R. (2001). Analogy in morphology: Modeling the choice of linking morphemes in Dutch. Linguistics, 39, 51-93.
Kuperman, V., Bertram, R., & Baayen, R. H. (2008). Morphological dynamics in compound processing. Language and Cognitive Processes, 23, 1089–1132.
Kuperman, V., Bertram, R., & Baayen, R. H. (2010). Processing trade-offs in the reading of Dutch derived words. Journal of Memory and Language, 62, 83-97.
Kuperman, V., Pluymaekers, M., Ernestus, M. and Baayen, H. (2007). Morphological predictability and acoustic duration of interfixes in Dutch compounds. Journal of the Acoustical Society of America, 121(4), 2261-2271.
Kuperman, V., Schreuder, R., Bertram, R., & Baayen, R. H. (2009). Reading of multimorphemic Dutch compounds: towards a multiple route model of lexical processing. Journal of Experimental Psychology: Human Perception and Performance, 35, 876-895.
Levelt, W. J. M., & Maassen, B . (1981). Lexical search and order of mention in sentence production. In W. Klein & W. J. M. Levelt (Eds.), Crossing the boundaries in linguistics: Studies presented to Manfred Bierwisch (pp. 221-252). Dordrecht: Reidel.
Levelt, W., Roelofs, A., & Meyer, A. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences, 22, 1–75.
Levelt, W.J.M., Schriefers, H., Vorberg, D., Meyer, A.S., Pechmann, T., & Havinga, J. (1991). The time course of lexical access in speech production: A study of picture naming. Psychological Review, 98, 122-142.
Losiewicz, B. L. (1995). Word frequency effects on the acoustic duration of morphemes. Journal of the Acoustical Society of America 97, 3243.
Lüttmann, H., Zwitserlood P., Böhl, A. & Bölte, J. (2011). Evidence for morphological composition at the form level in speech production. Journal of Cognitive Psychology, 23(7), 818-836.
Marchman, V. A., Wulfeck, B., & Weismer, S. E. (1999). Morphological productivity in children with normal language and SLI: A study of the English past tense. Journal of Speech, Language, and Hearing Research, 42(1), 206–219.
Marslen-Wilson, W., & Zhou, X. L. (1999). Abstractness, allomorphy, and lexical architecture. Language and Cognitive Processes, 14(4), 321-352.
Melinger, A. (2003). Morphological structure in the lexical representation of prefixed words: Evidence from speech errors. Language and Cognitive Processes, 18(3), 335-362.
Meyer, D. E., Schvaneveldt, R. W., & Ruddy, M. G. (1974). Functions of graphemic and phonemic codes in visual word-recognition. Memory & Cognition, 2, 309-321
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
57
Miceli, G., Capasso, R., & Caramazza, A. (2004). The relationships between morphological and phonological errors in aphasic speech: Data from a word repetition task. Neuropsychologia, 42, 273-287.
Milin, P., Kuperman, V., Kostić, A., & Baayen, R. H. (2009). Paradigms bit by bit: an information-theoretic approach to the processing of paradigmatic structure in inflection and derivation. In J. P. Blevins & J. Blevins (Eds.), Analogy in grammar: form and acquisition (pp. 214-252). Oxford: Oxford University Press.
Miozzo, M., & Caramazza, A. (1999). The selection of determiners in noun phrase production. Journal of Experimental Psychology: Learning, Memory, & Cognition, 25, 907-922.
Monahan, K. P. 1982. Lexical phonology. Ph.D. dissertation, Massachusetts Institute of Technology.
Mondini, S., Luzzatti, C., Zonca, G., Pistarini, C., & Semenza, C. (2004). The mental representation of Verb–Noun compounds in Italian: Evidence from a multiple single-case study in aphasia. Brain and Language, 90, 470-477.
Moscoso del Prado Martín, F., Kostić, A., & Baayen, R. H. (2004). Putting the bits together: An information theoretical perspective on morphological processing. Cognition, 94, 1-18.
Munson, B., & Solomon, N. (2004). The effect of phonological neighborhood density on vowel articulation. Journal of Speech, Language, and Hearing Research, 47(5), 1048.
Nam, H., & Saltzman, E. (2003). A competitive, coupled oscillator model of syllable structure. Proceedings of the International Congress of Phonetic Sciences XV (pp. 2253–2256).
Nespor, M., & Vogel, I. (1986). Prosodic Phonology. Dordrecht: Foris. Neu, H. (1980). Ranking of constraints on -t, d deletion in American English: A statistical
analysis. In W. Labov (Ed.) Locating language in time and space. New York: Academic Press.
Nusbaum, H. C., Pisoni, D. B., & Davis, C. K. (1984) Sizing up the Hoosier mental lexicon: measuring the familiarity of 20,000 words. Research on Speech Perception Progress Report, No. 10. Bloomington: Indiana University, Psychology Department, Speech Research Laboratory.
Pierrehumbert, J. (2002). Word-specific phonetics. In C. Gussenhoven & N. Warner (Eds.), Laboratory phonology VII (pp. 101–140). Berlin: Mouton.
Pillon, A. (1998). Morpheme units in speech production: Evidence from laboratory-induced verbal slips. Language and Cognitive Processes, 13, 465–498.
Plag, Ingo (1999). Structural constraints in English derivation. Berlin: Mouton de Gruyter.
Purcell, E. (1979). Formant frequency patterns in Russian VCV utterances. Journal of the Acoustical Society of America, 66, 1691-1702.
Reynolds, M., & Besner, D. (2005). Contextual control over lexical and sublexical routines when reading English aloud. Psychonomic Bulletin & Review, 12(1), 113-118.
Roberts, J. (1997). Acquisition of variable rules: A study of (-t, d) deletion in preschool children. Journal of Child Language, 24, 351-372.
Rochford, G. & Williams, M. (1965). Studies in the development and breakdown of the use of names. Journal of Neurosurgery and Psychiatry, 28, 407–413.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
58
Roelofs, A. (1996a). Serial order in planning the production of successive morphemes of a word. Journal of Memory and Language, 35, 854-876.
Roelofs, A. (1996b). Morpheme frequency in speech production: Testing WEAVER. In: Booij, G.E., van Marle, J. (Eds.), Yearbook of Morphology. Kluwer Academic Publishers, Dordrecht, pp. 135–154.
Roelofs, A. & Baayen, H. (2002). Morphology by itself in planning the production of spoken words. Psychonomic Bulletin & Review, 9(1), 132-138.
Rubenstein, H., Lewis, S. S., & Rubenstein, M. A. (1971). Evidence for phonemic recoding in visual word recognition. Journal of Verbal Learning and Verbal Behavior, 10, 645-657.
Rubin, G. S., Becker, C. A., & Freeman, R. H. (1979). Morphological structure and its effect on visual word recognition. Journal of Verbal Learning and Verbal Behavior,
18(6), 757-767. Saffran, E. M., Schwartz, M. F., & Marin, O. S. M. (1980) Evidence from aphasia:
Isolating the components of a production model. In B. Butterworth, (Ed.), Language production (Vol. 1, pp. 221-242). New York: Academic Press.
Scarborough, R. A. (2010). Lexical and contextual predictability: Confluent effects on the production of vowels. In C. Fougeron, B. Kühnert, M. D’Imperio, & N. Vallée (Eds.). Laboratory Phonology (Vol. 10, pp. 557–586). Berlin: Mouton.
Scheer, T. (2011). A Guide to Morphosyntax-Phonology Interface Theories. Berlin: Walter de Gruyter.
Schriefers, H., Meyer, A.S., & Levelt, W.J.M. (1990). Exploring the time course of lexical access in language production: Picture-word interference studies. Journal of Memory and Language, 29, 86-102.
Schoonbaert, S., & Grainger, J. (2004). Letter position coding in printed word perception: Effects of repeated and transposed letters. Language and Cognitive Processes, 19(3), 333-367.
Schwarzlose, R. & Bradlow, A.R. (2001). What happens to segment durations at the end of a word? Journal of the Acoustical Society of America, 109, 2292.
Seidenberg, M. S., & Gonnerman, L. M. (2000). Explaining derivational morphology as the convergence of codes. Trends in Cognitive Sciences, 4(9), 353-361.
Selkirk, E. (1984). Phonology and syntax: the relation between sound and structure. Cambridge, MA: MIT Press.
Stockhall, L., & Marantz, A. (2006). A single route, full decomposition model of morphological complexity: MEG evidence. The Mental Lexicon, 1, 83-123.
Stemberger, J. (1985). Bound morpheme loss errors in normal and agrammatic speech: One mechanism or two? Brain and Language, 25, 246–256.
Stemberger, J. P., MacWhinney, B., (1986). Frequency and the lexical storage of regularly inflected forms. Memory & Cognition, 14, 17–26.
Sugahara, M. & Turk, A. (2009). Durational correlates of English sublexical constituent structure. Phonology, 26, 477-524.
Taft, M. (2004). Morphological decomposition and the reverse base frequency effect. Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, 57(4), 745-765.
van Oostendorp, M. (2005) Derived Environment effects and Consistency of Exponence. Workshop on Freedom of Analysis, University of Tromsø.
MULTIMORPHEMIC WORDS IN SPOKEN PRODUCTION!!
59
Vitevitch, M. S. (2002). The influence of phonological similarity neighborhoods on speech production. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 735–747.
Walsh, T. & Parker, F. (1983). The duration of morphemic and non- morphemic /s/ in English. Journal of Phonetics, 11, 201–206.
Yaeger-Dror, M. (1996). Phonetic evidence for the evolution of lexical classes: The case of a Montreal French vowel shift. In G. Guy, C. Faegin, J. Baugh, & D. Schiffrin (Eds.) Towards a social science of language (pp. 263–287). Philadelphia: John Benjamins.
Yaeger-Dror, M. and Kemp, W. (1992). Lexical classes in Montreal French. Language and Speech, 35, 251–93.
Yap, M. J., & Balota, D. A. (2009). Visual word recognition of multisyllabic words. Journal of Memory and Language, 60(4), 502-529.
Zevin, J. D., & Balota, D. A. (2000). Priming and attentional control of lexical and sublexical pathways during naming. Journal of Experimental Psychology: Learning Memory and Cognition, 26(1), 121-135.
Zhao, Y. and Jurafsky, D. (2007), The effect of lexical frequency on tone production. In Proceedings of the 17th International Congress of Phonetic Sciences, Germany.
Zuraw, Kie (2009). Frequency influences on rule application within and across words. Proceedings of CLS (Chicago Linguistic Society) 43.
Zwitserlood, P., Bölte, J., & Dohmes, P. (2000). Morphological effects on speech production: Evidence from picture naming. Language and Cognitive Processes, 15(4/5), 563-591.
Zwitserlood, P., Bölte, J., & Dohmes, P. (2000). Where and how morphologically complex words interplay with naming pictures. Brain and Language, 81, 358-367.