sziga - ch 12

19
Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015. Copyright © 2012. John Wiley & Sons. All rights reserved. What Is A Possible Language? Distinctive Features Languages differ from one an othe r w ithout limi t. Edward Sa pi r, La nguag e, 1921 In examining the phon et ic frameworks tested on a large n umbe r of languages ... one cannot fail to be impressed by the smaU number of features involved. Morris Halle, The Sound Pattern of Russ ian, 1971 , p. 20 Chapter outl i ne 12.1 lntrod uct' 1 on 254 12.1 .1 Pho nologica l un iversals 254 12.1 .2 Why bother with fo r ma l'lsm? 255 12.1.3 Some hypot h eses 256 12.2 Dist inctive features 257 12.2 .1 Background 257 12.2 .2 Major class and manne r fea tures 258 12.2 .3 Laryngeal feat ures 262 12.2 .4 Majo r place d ist inct io ns 262 12.2 .5 Subsidiary pl ace d ist in ctions 265 12.2 .6 Feat ures for vowels 267 12.3 How h ave ou r hypotheses fared? 270 The of Language: At1 Introduction to Phonetics and Ph onology, First Edition. Flizabeth C. Zsiga. © 2013 Flizabeth C. Zsiga. Publish ed 2013 by Blackwe ll Pub li shing Ltd.

Upload: alex-gurovich

Post on 19-Dec-2015

19 views

Category:

Documents


1 download

DESCRIPTION

Sziga - CH 12

TRANSCRIPT

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    What Is A Possible Language? Distinctive Features

    Languages differ from one another without limit. Edward Sapir, Language, 1921

    In examining the phonetic frameworks tested on a large number of languages ... one cannot fail to be impressed by the smaU number of features involved.

    Morris Halle, The Sound Pattern of Russian, 1971 , p. 20

    Chapter outline

    12.1 lntroduct'1on 254 12.1 .1 Phonological universals 254 12.1 .2 Why bother with formal'lsm? 255 12.1.3 Some hypotheses 256

    12.2 Distinctive features 257 12.2 .1 Background 257 12.2.2 Major class and manner features 258 12.2 .3 Laryngeal features 262 12.2 .4 Major place distinctions 262 12.2. 5 Subsidiary place distinctions 265 12.2.6 Features for vowels 267

    12.3 How have our hypotheses fared? 270

    The Sout~ds of Language: At1 Introduction to Phonetics and Phonology, First Editio n. Flizabeth C. Zsiga. 2013 Flizabeth C. Zs iga . Published 2013 by Blackwell Publishing Ltd.

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    254 WHAT IS A POSSIBLE LANGUAGE

    12 .1 introduction

    Chapter summary 271 Further reading 272 Review exercises 272 Further analysis and discussion 272 Further research 27 4 Go online 27 4 References 27 4

    T he preceding chapter provided a survey of common phonologica l alternations in the lan-guages of the world. As phonologists, however, we want to know more. We wan t to under-stand not only the set of things that do happen, bu t the set of things that can happen. We wan t to answer the questio ns: "What is a possible p ho no logica l contrast?" and "What is a possible phonological alterna tion?"

    T he chapte r begins with an excursus on why th is is an important question, and how formal represen tations help us to answer it. Section 12.2 then surveys one formal theory of how the possible contrasts and alternations in human language can be encoded: d istinctive feature theory. There is no t complete agreement among phono logists concerning the exact co ntent and enumeration of features - we will consider several hypo theses, and conclude the chapter with an evaluation of how our hypotheses have fared.

    1' 1 1 nhnno,ogical universals Imagine that you are the one to discover a language previously unknown to the world of the Linguistics Society. Traveling up a tributary of the Amazo n in Brazil, or descending into a hidden valley in Papua New Guinea, you come into contact with an isolated people, and have the opportunity to learn t heir language. (Stephen Pinker, in his book The Language lnstinct, provides a humorous but telling account of just such an encounter.) Do you have any idea what this new language will be like? Ca n you make any predictions about how the contrasts and alternations of th is language will be sim ilar to languages already known, or about the parameters along which they might vary?

    We wa nt to know whether there are any phonological universals, properties that are true of all languages. If there are universals, we want to know what they are, and why they occur. Edward Sapir hypothesized that languages co uld d iffer "without lim it." Later phonologists hypothesized tha t differences between languages were in fact very limited, and that all lan-guages were built on the same basic pattern, from the same building blocks provided by a Uni versa/ Grammar.

    If there is such a thing as Universal Grammar, where does it co me fro m? Some linguists {pro minently, Noam Chomsky and his students, including Pinker) argue that the universal properties of language are programmed into our DNA. O n this view, the structure of human language and our ability to use it are part of our genetic endowment, in the same way that the structure of our eyes and our ability to produce liver enzymes are a product of our genetic endowment. just as we are genetically programmed to perceive light of a certain wavelength,

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 255

    and to digest food of a certain kind, we are genetically programmed to be sensitive to the difference between a sonorant and an obstruent (or a no un and a verb).

    Other linguists argue that where language is concerned our genetics give us only powerful brains that are good at classifying things and mouths of a certain shape and mobility, and that all the other purported universals of language come about because of a universal desire to communicate with each other, shared pathways for learning, and a common experience of the world. On this view, nouns and verbs are not pre-programmed mental constructs, but concepts that are learned through repeated interaction with actual objects and actions. The difference between sonorant and obstruent is not given, but is learned, as infants try out the effects of different vocal tract configurations in babbling. Since we all share the same basic anatomy and physiology, and experience more or less the same world, the same categories emerge every time.

    Both tJ1ese views emphasize the importance of the process of learning, b ut differ on what is innate and what is learned fro m the environment. A th ird view places the emphasis on language change rather tha n on hu man cognitio n. Th is view argues that humans could learn any language that the envi ro nment presented them with. Languages turn o ut to have a lo t in com mon simply beca use certain cha nges are more likely than others, which over time tend to produce languages with simila r properties.

    The debate over the existence and nature of a genetically-given Universal Grammar rages on, and wo n't be resolved here. We can, however, examine the nature and descriptio n of language universa ls: properties that do seem to hold across all 7,000-plus human languages tha t have been investigated, and are predicted to hold of the next human language to be discovered as well. In o rder to examine these properties, we will examine five hypotheses that have been argued to be true across aU languages. In formulating and testing these hypotheses, we will begin to use some more formaJ descriptions: terms and symbols that may have more in common with algebra and geometry than with English prose.

    12 1 2 uhu hnth~r with fo-mZ!Iism? Why bother with formal language? The question is a valid one. Why not just make our gener-alizations in English prose and be done with it? Why introduce possibly confusing technical terms and symbols? One reason is that we want our predictions to be as precise as possible.

    ln order fo r a hypothesis to be interesting, it must be falsifiable. A hypothesis is falsifiable if we can decide whether it is true or false based on observable data. And in order to decide if a statemen t is true or false, we must understand precisely what each of its term s mean. Co nsider, for example, a hypothesis tha t has been previously discussed:

    Hypothesis I: Pho nological alte rnations target natura l classes.

    ls this true o r fa lse? In order to test the hypothesis, we must examine alternat io ns in the lan-guages of the world, discover the classes of sounds tha t are ta rgeted, and dete rmine whether those classes are "natural" o r not. If we find a class of sounds that is not natural, we conclude that our hypothesis is no t supported. Our decision, however, depends o n the defin ition of"natural class." The English word "natural" has many different meanings. Some-. h "< d " d d d " d h times t e term means .oun 111 nature, as oppose to mvente or man-rna e, an t us

    linguists will speak of "natural languages" as opposed to computer languages. If "natural class" means a class of sounds found in any natural language, then our hypothesis is (unin-terestingly) always true: any set of sounds we might find in a natural language would be by definition a natural class. It would not matter what the data actually was, so it would be a waste of time to compile it. Or maybe we could define "natural" as "whatever feels easy and

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    256 WHAT IS A POSSIBLE LANGUAGE

    uncontr ived." Again, this is not a useful definition. What counts as easy, and to whom? We are all competent users of our native languages, so aU alternations are easy to the native speaker. Without a precise and useful definition of"natural," our hypothesis is unfalsifiable and thus completely uninteresting. In Section 12.2, we consider how to precisely define natural classes in terms of a proposed universal set of distinctive Jeat11res. The formal use and definition of these parameters lead us to clear hypotheses that can truly be tested against the data.

    In addition to wanting our hypotheses to be precise, we want them to be constmined. The m ore specific a hypothesis is, the more things it rules out, the more interesting the hypothesis and the theory behind it is. Imagine that I have a theory of weather patterns that leads me to make the prediction: "The high temperature in Washington D.C. on July 4 will be between 20 and II 0F." The prediction is falsifiable: it could in principle reach I II 0 or drop down to 19, but even in the event the prediction is proven true, it doesn't mean I have a very good theory of weather. My theory is too unconstrained; it allows too ma ny possibi lities; it says little if anything about what July in Washington is really like. If, on the other hand, I make the highly constrained prediction that on July 4 it will reach a high of 94 at 2 p.m., and that predictio n is co nfirmed, you m ight conclude t hat my theory of weather patterns is a pretty good one.

    Every linguistic hypothesis is part of a theory of what human la nguage is like .. T he more constrained ou r hypotheses are, the more possibilities we can rule o ut, the more we under-stand what rea l human languages are actually like. For example, I might hypothesize that our newly-discovered language will use sounds generated by the voca l tract. This is fa lsifiable - it is logically possible that this group of people comm unicates using co mplicated foo t stomping -but once the hypothesis turns out to be true (yes, they do use spoken language), we don't have much more to say. A more constrained hypothesis might predict, for example, that "this language will syllabify VCV sequences as V-CV not VC-V." If that hypothesis turns out to be true, we can ask interesting follow-up questions about why syllabic organization is universal, and why syllables of CV form are preferred. Following up o n o ur more constrained hypoth-eses, we may learn something really interesting about possible human languages.

    For the purposes of hypothesis testing, English prose descriptio ns are both too vague and too powerful: Too vague in that English words are not necessarily precisely defined, and too powerful in that our prose can describe anything and everything. Instead of describing alternations in English prose, we will propose a constrained set of descriptors, and consider the hypothesis that these, and only these, parameters are available for human language pho-nology. Every proposed feature or feature set is a hypothesis about what phonological gram-mars can manipulate. We will see that not all of the phonetic pa rameters and relationships that the mouth can produce, the ear can perceive, or the analyst can describe turn o ut to be part of phonology.

    12 .1.3 some hypotheses Here are the hypotheses we wi ll consider. They formed the sta rting poin t of the book The Sound Pattern of English (Chomsky and Halle, 1968), which laid out the program of genera-tive phonology.

    Hypothesis 2: Phonological contrasts, classes, and alternations can be defined in terms of a closed set of phonetic parameters that is the same across all languages: the set of distinctive Jeat11res.

    Hypothesis 3: Features are defined in terms of binar)' (+/- ) oppositions that are specified for every segment (e.g., every segment is either [+voice) or [-voice]).

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 2,57

    Hypothesis 4: Every lexical item has a unique underlying representation (UR), which is defined in terms of combinations of distinctive features.

    Hypothesis 5: Underlying representations are converted into contextually-varied surface representations (SR) by the operation of a set of feature-changing rules.

    Hypotheses I, 2, and 3 concern the nature of distinctive features in codifying contrasts and alternations. Hypotheses 4 and 5 concern the role of features in mediating underlying and surface representations. We begin in Chapter 12 with hypotheses 1- 3 and the discussion of the definition of distinctive features. Chapter 13 continues with URs, SRs, and hypotheses 4 and 5.

    12.2 distinctive features

    12.2 1 background The idea of a set of distinctive pho no logical features rests on the tenet that language is d is-crete and combinatorial. Sentences are made up of words, words are made up of segments, and segments are made up of features that define the categories to which the segments belong, and determine the al ternations in which they participate.

    The idea of binary dist inctive features began with the Structuralist idea that all objects must be understood as part of a system: the nature of any sound in a given language can only be understood in terms of its opposition to other sounds in the system: [p) holds a certain place in the linguistic system because it contrasts with [b] . Nikolay Trubetskoy, in his work Grund:zuge der Phonologic (Foundations of Phonology, 1939}, suggested a large set of possible dimensions of contrast. Trubetskoy's student Roman Jakobson further defined and developed the idea in 1mportant ways: it was Jakobson who proposed that featural opposi-tions should be binary, that the set of features is universal, and that every feature should be unambiguously defined in both articulatory and acoustic terms.

    Jakobson's work on phonology spanned most of the twen tieth century. jakobson wrote extensively on markedness, language acquisition, and language typology (tha t is, types of contrasts and alternations that are co mmon or u ncom mon cross- linguistically). The clearest exposition of jakobson's distinctive feature system is foun d in his co-authored work with Gunnar Fant, a phonetician, and Morr is Halle, Jakobson's colleague and former student. Jakobso n, Fant, and Ha lle p ub lished Preliminaries to Speech Analysis: The Distinctive Features and Thei r Correlates in 1952. Whi le the se t of accepted distinctive features has u ndergone sign ifican t revisions in the decades since, the features currently in common use grew o ut of those p roposed by jakobson, Fant, and Halle.

    ...

    Jakobson hved from 1896 to 1982. and was aaive in research. writing. and teaching up until his death. He began his career as a student of Trubetskoy in Moscow in 1915 and became a prominent member of the influential Prague linguistics Circle in the 1 920s. He fled across Europe and eventualo/ came to New York Cny during World War II. and ended his career as colleague to Morns Halle and Noam Chomsky at MIT.

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    258 WHAT IS A POSSIBLE LANGUAGE

    In order to test hypotheses about distinctive features, phonologists consider the different systems of contrast and alternation in the languages of the wo rld, with the goal of determin-ing whether a particular set of features correctly characterizes the required classes. It is to this consideration that the following sections now turn.

    1., ., ., rn:.ior class and manner features

    We have seen that phonological alternations often reference large classes of sounds: conso-nants vs. vowels, sonorants vs. obstruents, stops vs. fricatives, nasal consonants, approximant consonants, lateral consonants. Any set of distinctive features must include parameters that distinguish these classes of sounds. A widely-accepted set of such features is defined below. As o riginally proposed by Jakobson, we define each dimension in terms of a binary(+/- ) contrast, and define each parameter in both acoustic and articu latory terms.

    [+/-syllabic] \ ,, ulll" y defin ition: [+syllabic] sou nds form the nucleus of a syllable. \u>t' tc defi ni tion: [+syllabic] sounds have higher amplitudes than neighboring so unds. l'vptL.ll natural classes and alterna tions: vowels are [+syllabic]. as opposed to conso-nants which are, in the default case, [-syllabic]. Glides are also [-syllabic] . Co nso nants may become [+syllabic] if they form the nucleus o f a syllable on their own, as in [bA?J;t], for example. We will fi nd this feature to be useful in defining these natural classes at this point, but will revisit the need for a [+/-syllabic] feature in Chapter IS, on syl-lable structure.

    [ +/- consonantal] ' definition: (-cons] = having no significant vocal tract constriction

    Ac 1 definition: [-cons] = high amplitude resonance natural classes and alternations: Vowels and consonants are often targeted

    separately: alternations often apply to just o ne or the other, or are conditioned by just one or the other. For example, vowel harmony skips consonants, place assimilation applies to just consonants. Vowels and glides are [-consonantal]. consonants are [+consonantal].

    [ +/- sonoran t] \ c d 11<

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 259

    Table 12 Major classes or sounds. Plosives Fricati'es Nasals Approximants Laterals Glides Vowels

    Syllabic +

    Consonantal + + + + +

    Sonorant + + + + +

    Continuant + + (+) + + Nasal + -(or+) Lateral -(or+) +

    com mon patterning with oral s tops, are [- continuant]. [Con tinuant] is the active feature in many alte rnations of len ition and fortition .

    [+/-nasal) Aru, ul1 "1 1 defin itio n: [+nasal] sounds are articulated with an o pen velum. Acou,tlc defin itio n: [+nasa l] sounds have nasal resonance. 1 " I natu ral classes and a lternatio ns: Assimilation o f nasa li ty is com mon; nasa l stops a re often targeted as a class. For example, nasal stops assimi la te in place of articulation, most commonly to ano ther stop.

    [ +/- lateral) definition: vocal tract closed at the center line, open at the side.

    definition: a particular resonance pattern. natural classes and alternations: !-sounds often form a class unto themselves

    in assimilations (as in Catalan, Arabic, and Komi).

    Table 12.1 summarizes how these features distinguish these major sound classes. Note that every column in Table 12.1 is unique: each class of sounds is defined by a unique

    combination of features. Plosives and fricatives share m ost of their features, but are distinct in t hat plosives are [-

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    260 WHAT IS A POSSIBLE LANGUAGE

    Another questio n is how to handle the continuant value of affricates. Because of the dif-ficu lty of the question, I have left affr icates out of Table 12.1 entirely. In terms of ar ticulation, affricates are both (- cont inuant] and [+continuant] : [-continuant] during their closure phase and [+continuant] during their release phase. However, it would go against Jakobson's conception of distinctive features to say that they are both plus and minus continuant. He considered the plus/minus distinction to indicate mutually exclusive categories to which a segment must be assigned. To say that a segment is both +F and-Fat the same time would be, for him, like saying a number is both positive and negative, a switch is both on and off, or a direction is both East and West: it's a logical contradiction. Crucial, however, is the observation that affricates are not minus and plus continuant at the same time, but in sequence. Jakob son's system, however, had no way of accounting for changes in the middle of a segment.

    Here, provisionally, we will assign the affrica tes to the [-conti nuan t] class, on the basis of our defin ition (they do have a complete closure) and on ou r prelimina ry observatio n that they seem to pa tte rn with the stops, making the [-continua nt] segments a na tural class. (In English palatalization, for example, (t] alternates with (\f] wh ile [s j alte rnates with [JJ , and in Zoque we saw that nasals assim ilate to both stops and affricates but dele te befo re frica-tives.) How then are stops and affricates to be distinguished? In English, and in many other languages, affrica tes and sto ps are distinct in place of a rticulatio n: [ tl is alveolar and [\f] is pos t-alveo lar. T here are many languages, however, where there are stops and frica tives at the same place o f art iculatio n, so another feature is needed. We will adopt the solutio n o f stating that affricates have all the same features as stops, except that affricates are marked as [+delayed release] .

    [ +/- delayed release] definition: [+delayed release] = opening from a stop into a fricative, rather

    than being fully released into an open vowel articulation. definition: [+delayed release] = period of frication upon release.

    natural classes and alternations: affricates are [+delayed release]; stops and all other segments are (- delayed release].

    This works for distinguishing stops and affricates, but it introduces an interesting asym -metry: we may often fi nd reason to refer to the class of sounds that are affrica tes, but we never seem to have occasion to refer to the class of all the sou nds that are not affr icates. The same holds true for the featu res [nasal] and [lateral]: the class of ]+nasa l] and [+lateral] are very useful. For example, nasals assim ilate in place of articulation to a fo llowing stop, or late rals assimilate in place of a rticulatio n to a fo llowing coronal. Bu t the class of "sounds that are no t nasal" o r "sounds that are not lateral" do n't seem to be a classes at all, b ut more like an elsewhere case.

    T his is in contrast to o ther fea tures, where both + and- values are equa lly useful. We refer equally o ften to the class of [- sonorant] obstruents as to the class of so nora nts, and equally o ften to the class of [- conso nantal] vowels and glides as to the class o f conso nants. T hat is, some contrasts, like [+/ - sonorant], [+/- consona ntal] and [+/- continuant] seem to be t ruly bi nary, while o thers like [nasal], (lateral] and [de layed release] seem mo re one-sided. Recognizing this asymmetry, m ost linguis ts have departed from the requirement of b inary symmetry (and fro m our Hypoth esis 3) and have argued that featu res like [nasal] and [lateral] do not have plus a nd minus values. They are unar;' rather than binary. If a sound has an open velum, it is designated (nasal], sometimes abbreviated [N], otherwise, no specification for this feature is given.

    Note the differing predictions that are made by the use of binary vs. unary features. If [-nasal] is among the set of features universally available, then we would expect that ian-

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 261

    Table 12 Revised feature values for the major classes of sounds. including affricates and some unary features.

    Plvcs Fricatives Affricates Nasals Approx. Laterals Glides Vowels

    Syllabic +

    Consonantal + + + + + +

    Sonorant + + + + +

    Continuant + + (+) + + Nasal N

    Lateral L

    Delayed rei DR

    guages wou ld make use o f it. If in fact they do not (as seems to be the case), we shou ld remove it from the se t, maki ng our prediction about possible human languages mo re co n-strained. Given our current state of knowledge, most phonologists work with in a system that has some binary feature values and so me unary feature values. Here, we will mark some features as b inary and so me as unary, recognizi ng, however, tha t these are hypotheses subject to further testing and refinemen t (Table 12.2) .

    Finally, consider how these features work to defin e specific natural classes. The set of sonorant consonants is defined as the set of sounds that are [+so no rant, +consonantaJI: all and only the sonorant consonants share these t\vo features. The set of fricatives is [ +continu-ant, -sonorantl: all and only the fricatives share these t\vo features. Using the table above, which features would you use to define the natural class of glides? Try to create your speci-fications using as few features as possible.

    One might ask, at this point, how is it that [+continuant, - sonorant) is better than "frica-tive"? Isn't saying "fricative" simpler? It is, but it is also less predictive. By using distinctive features, we are making a specific, testable hypothesis that all natural classes can be defined in terms of these parameters (for the moment, those in Table 12.1.)

    For example, imagine a phonologist is working on language similar to Zoque (call it A), and she discovers that the segments [m, n, IJ) delete before the sounds [f, s, J, v, z, 3] in this language. She will immediately note that these lists of sounds are not random, but constitute a class of sounds that have certain p honetic parameters in common. Further, these param-eters are among those given by the list of distinctive features: the sou nds [m, n, IJ] const itute the class of sou nds in this language tha t are [nasa l, +consonantal], and the sounds [f, s, J, v, z, 3] constitute the class of so unds in this language that are [+conti nuant, -so no-rant]. All is well: her hypotheses about the way phonology works are suppo rted. This pho-nological alte rn ation targets na tural classes defi ned in terms of the p roposed distinctive features. (Of course, if the language's phonetic inventory were slightly d iffe rent, the set of sounds specified by these features would be different as well: if the language happened to have [x] and [Y], they also would be included in the set. However, there is no need to wo rry about excluding so unds that do not occur in the language u nder study. For language A, the features [+continuan t, -sonoran t] designate the set of all the fricat ives that occur in A.)

    Now imagine that, in the next city, the linguist fi nds a different dialect (B) which has a somewhat different alternation: nasal consonants delete before [f, s, J, v, z, 3, p, bl, but not [t, k, d, g, m, n, IJ]. That is, nasals delete before fricatives and labial oral stops. I can certainly describe this class in English prose, but I cannot describe it in terms of a single set of distinc-tive features. (Try it.} There is no set of features that define all and only the fricatives and labial stops. Thus, by hypotheses I and 2, Dialect A is a possible human language, and Dialect

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    262 WHAT IS A POSSIBLE LANGUAGE

    Table 1 ~ Laryngeal features.

    Voiceless Voiced Ejectives Creaky Voiceless Voiced aspirated aspirtlted and? vorce

    VoiCe + + +

    Spread glottis SG SG

    Constricted glo ttis CG CG

    B, which targets the set [f, s, J, v, z, 3, b, p) within a single rule, is not. T he set is not a na tural class. Should such a rule actually t urn out to exist, we will have to revise our hypotheses.

    12 .2 .3 laryngeal features La nguages often make use of cont rasts and alternations involving la ryngea l distinctions. La nguages such as French may contrast [p] and [b), la nguages like Tha i may co nt rast [p] a nd [p11 ], languages like Quechua may contrast plain [p] a nd glotta lized or ejec tive [p' ], and languages like Hi ndi m ay make a four-way contrast between plai n vo iceless [p), voice less aspira ted [ph), plain voiced [b], and b reathy voiced (b6]. Finally, so me languages, such Hausa, may contrast plain voice vs. creaky, or glottalized, vo ice, usually o n sonoran ts. Features suf-ficient for describ ing these co ntrasts are defined below, and shown in Table 12.3. It is gener-ally accepted that the features [spread glo ttis] and [constricted glottis] are unary. Some linguists argue for a unary analysis of [voice) as well, arguing that "plain" voiceless stops are just that: they have no laryngeal specification at all. Unary-feature analyses of voicing assimi-lation would take us beyond the scope of this book, however, so the feature [voice] is here assumed to be binary.

    [+/- voice) I t definition: sounds that are (+voice) have vibrating vocal folds.

    Aco definition: [+voice) sounds are periodic (have a regularly-repeating pattern) . ,1 natural classes and alternations: Many languages have alte rnations involving

    voicing, especially among obstruents.

    [spread glottis] .\ ltc tll , y definition: la rge glottal opening gesture t\col s defin ition: extended period of vo icelessness upon release h pte,, natu ral classes and alternatio ns: aspira ted stops have the feature [spread glo ttis].

    (constricted glottis] \ d l l \ definition: constr ic ted vocal folds t\v definition: irregular glo ttal pulses I\ I'' natural classes and alternations: ejectives, glottal stop, and creaky-voiced sono-rants have the featu re [constricted glottis].

    1., ., 4 ""'Jjor place distinctions Our list of distinctive features must also account for contrasts and alternations in place of articulation (abbreviated POA). All languages make place of articulation contrasts among

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 263

    consonants, and many alternations either target a class of sounds based on place (such as vowel lowering adjacent to uvulars) or manipulate place (as in place assimilations). However, developing a set of features for place of articulation has not proved straightfo rward.

    Trubetskoy conceived of POA as a multi -valued feature. The distinctive feature was PLACE, and its values were bilabial, labiodental, dental, alveolar, etc. This corresponds to an accepted understanding of the set o f phonetic distinctions, but does not allow for any natural classes that group places together. For example, given the places of articulation bi la-bial, labiodental, and dental, we might assign bilabial to I PLACE, labiodental to 2PLACE, and dental to 3 PLACE. Under this system, labiodental is equidistant from both bilabial and dental; and involving as it does both the lips and the teeth we would expect labiodental to fo rm a natural class with labia l sounds and with dental sounds equally often. This is no t the case, however. Rather, labiodentals co nsistently form a class with the b ilabials, but not with the dentals.

    In English, fo r exa mple, [ w] cannot follow a labial consonant, where " labia l" includes both bilabia ls and lab iodenta ls: we have "twill," "swill," and "quill," but no t ~ pwill or *fwiiJ. ([9w] is uncom mon, but acceptable, as in "thwart." Clusters with voiced conso nan ts are ma rginal at all places, but "Gwen" and "dwell" are better tha n *bwell. ) In Korean and other languages that do not have (f], the labiodental fricative is co nsistently bo rrowed as [p], no t [t]: " fish" is bo rrowed into Ko rean as [piJi], not [tiJi]. In Setswana, which does no t have [ v], "u niversity" is [u nibesiti], no t [unidesiti ]. To ci te just o ne more example, Tu rkish exhib its a class of words that violate the principle of rounding harmony discussed in Chapter II (example 25). In general, [ u] canno t occur in a non-initial syllable in Turkish unless the initial vowel is also round: [dalw], branch, and [kolu) arm, are fine, but no t *[datu]. The only exceptions are words where a labial consonant, either bilabial or labiodental, intervenes: [karpuz) watermelon, and (havlu] towel are both acceptable.

    This is the common pattern cross-linguistically: labiodentals and bilabials form a natural class, and labiodentals and dentals do not. Dentals, however, form a class with other conso-nants made with the tongue front: the coronals. Evidence for the natural class of coronals comes fro m alternation s such as Arabic [l) assimilation (Chapter II, example 24). Some examples are repeated here:

    ( I) Arabic /I! assimilation:

    ?aJ-Jams ?at-tid3a:ra ?a9-9aqa:fa ?a(Hahab ?az-zuhu:r ?33--:)il: Co mpa re:

    ?al-qamr ?al-badw ?al- filfil ?al-xardal

    the sun the commerce the cultu re the gold the flowers the shadow

    the moon the Bedouin the pepper the mustard

    In Arabic, [ I) assimilates completely to a following dental, alveolar, or post-alveolar, all coronal consonants (the "sun letters.") AU other consonants (the "moon letters," including [f)) induce no change. A similar example is found in Catalan (exam ple 2): [I) assimilates in place of articulation to a following dental, post-alveolar, or retroflex, but not to a labial or labio-dental.

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    264 WHAT IS A POSSIBLE LANGUAGE

    (2) Catalan IV assimilation:

    e[ I pI a e[ I floc e[ ld]ia ""

    e[ l d ic e[.< 3]erma

    the bread the fire the day the rich the brother

    Finally, consider the alternations between stop and fricative in Sudanese Arabic shown in example (3).

    (3) Assimilation in Sudanese Arabic:

    book daughter fish kita:b bit samak X kita:f fat 11i b it fathi samak fathi Fathi 's X kita:p sa mi:r b is sami:r samak sami:r Sameer's X kita:p Jati:f biJ Jari:f samak Jari:f Sharif's X kita:p xa: lid b it xa:lid samax xa:lid Halid's X kita:p hasan bit hasan samak hasan Hasan's X

    T he pattern is tha t the word-final stop assimilates to a follow ing fricative just in case the two share an act ive articulator. Bilabial [b] assimilates to labiodental [fl (shared labial ar ticula-tor), alveolar [ d] assimilates to both alveolar [s] and postalveolar [JJ (shared tongue front articulator), and [k] assimilates to [x] (shared tongue body articulator). No stop assimilates to pharyngeal [h], because no stop shares the tongue root articulator with [h].

    Alternations such as those listed above have led phonologists to propose a set of unary place features based on active articulator, as listed below. These features depart from the )akobsonian pattern in two ways: they are unary, and they are defined in purely articulatory, not acoustic, terms. They do however, seem to work. That is, they make accurate predictions about the possible natural classes found in the languages of the world.

    [ labial] lltnn constriction at the lips

    1\!Jlur l

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 265

    Table 12.4 Features based on active articulator. The fricative symbol stands for all sounds made at the corresponding place of articulation.

    tP f 8 s I X X fj II

    Labial L L

    Coronal c c c c c

    Do rsal (D) D D Pharyng.,al Ph

    La ryng.,al Lar

    [laryngeal) I ~~h11 r: co nst riction at the glottis N.ltUJ".ll ,J"' [hi a nd(?].

    Additional fen tures ~ re necessa ry to make distinctions with each of the classes defi ned by the fea tu res in Tab le I 2.4. These subsidiary place features are lis ted in Table 12.5 a nd d iscussed below.

    12.2.5 subr;-i;1ry place distinctions No language has all of the fr icatives shown in Table 12.5, or uses all these places of articula-tion, but for any pair, there is some language that creates a contrast between those two sounds. Therefore, we need a set of features sufficient to uniquely identify each column.

    We begin with the coronal consonants, since there are five different possible places of articulation that can be contrasted within the coronal articulator. For a first distinction, the feature [anterior) divides sounds made at or in front of the alveolar ridge ([+anterior!) from those made behind the alveolar ridge ((- anterior)) . Many languages have alternations similar to the palatnlization rule of English: (t, d, s, z] become (tf,

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    266 WHAT IS A POSSIBLE LANGUAGE

    in use is [distr ibuted], going back to a Jakobsonian feature that was proposed to apply across all places of articulation.

    [ +/- distributed) [+distributed] sounds have a relatively long constnct10n. In practice,

    this means that !aminal consonants are (+distributed] and apical consonants are [-distributed].

    " [s] and[~] are [-distributed); (9 ), [J) and [10] are [+distributed].

    Thus among the coronal fricatives, [9] is [+anterior, +distributed], [s] is [+anterior, - dis-tributed], [SJ is [- an terior,+distributed] and[~] is [-anterior,-distributed]. Stops and nasals at the same places of articulation would be distinguished in the same ways.

    Amo ng the coronals, that leaves the distinction between post-alveolar [J] and palatal [10] . One observa tio n to make about this contrast is that no language is known to contrast post-alveolar and palatal stops or nasals: th is d isti nctio n seems to apply only to fricat ives and affricates. T hus we can appeal to the fact that these frica tives sound very diffe rent: lJ] and [s] (known as "sibilan ts") have a no ise compo nent that is both high-amplitude and high-frequency, whi le the other coronal fricatives are softer. A fea ture that references this di fference is I+/- striden t]. This is o ne of the few features that has a primarily acoustic defi nition. The feature [s trident] has the added advantage of being able to distinguish higher-pitched labi-odental [ f] from softer b i lab ial[~], another place distinction that applies only to fricatives. It can also dist inguish [9) from [s] for those who make [9] with the tongue tip.

    [ +/- strident) [+strident] sounds have high-amplitude, high-pitched frication .

    ~ sibilants (alveolar and post-alveolar fricatives and affricates), as well as labiodental and uvular fricatives and affricates are [+st rident]. All other fricatives and affricates are [- strident] . This feature is only defined for fricatives and affricates.

    The features [anterior] and [strident] sometimes seem to duplicate effort: the distinction between [s] and [9], for example, can be defined in terms of either [+/- strident] or[- /+ distributed]. But the features defin e different natural classes, both of which seem to be useful. [+strident) is crucial in defining the class of sibilant consonants. For example, the English plural allomorph [tz] occurs only after the coronal sibi lants: [coronal, +strident). On the other hand, the fea ture [strident] is not available for stops and nasals, and there are languages (such as Tiwi, see Chapter 10 exercises) that distinguish lam ina! denta l vs. apica l alveolar, or lam ina! postalveo lar vs. apical retroflex, stops and nasals.

    As wns noted in the defi n itio n, the feature [strident) ca n be used to d isti ngu ish [- strident] velar frica tives fro m ha rsher [+strident] uvular fricat ives. But another fea tu re is needed to distinguish velar and uvular sto ps and nasals. The featu re [distributed] will no t do it, as both vela rs and uvulars are made wi th the same art iculato r, the to ngue body. The most salient dist inction between these two is positio n of the tongue body: the feature [h igh] can b e rec ruited to distinguish these sounds. This fea ture, which is defined o nly for sounds that use the tongue body as an active articulator, is mostly ac tive for vowels, and is de fi ned (s traight-forwa rd ly) in Section 12.2.6.

    The aforementioned set of features is hypothesized to be sufficient for accurately and succinctly describing aU the consonantal contrasts and natural classes that are found in the languages of the world.

    There is one consonantal natural class that has not yet been covered, however. Languages that utilize the uvular, pharyngeal, and laryngeal places of articulation (including the Semitic languages, and a number of Caucasian and Native American languages) often

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 267

    group these consonants into a natural class (which does not include the velars). For example, we saw in Chapter I I that, in Syrian Arabic, [e) lowers to [a] when it follows a uvular, pha-ryngeal, or laryngeal consonant. The data are repeated in example (4):

    (4) Syrian Arabic: dara3 -e step Jerk-e society mad ras-e school

    wa:3h-a display mni:n-a good dagga:R-a tanning

    This class is often referred to as the class of guttural consonants, and [guttu ral) is sometimes pro posed as a una ry articulato r feature, para llel to [labial], [coronal] and [do rsal). Un li ke those three featu res, however, [guttura l] does not refer to a si ngle active ar ticulator: uvulars use the tongue body, pha ryngea ls the tongue root, and laryngeals the larynx. (One intriguing possib ili ty, sti ll speculative however, is that the guttu ral consonan ts may compr ise the class of consonants that do not involve ra ising of the jaw.) In part, the issue is di ffi cul t to resolve because pharyngeal consona nts, which a re crucial to determining the correct genera liza tio ns, are not among the most commo n. Here, while we reco gnize that the issue is far from reso lved, we will no te the affinity of guttural consonants and low vowels, and specify uvulars, pha-ryngeals and laryngeals as [+low]. Like [+/- high], [+/ - low) is most often a featu re of vowels, and it is to features of vowels that we now turn.

    Table 12 Expanded table ol place dtstinctions

    "' f 8 s I X X II II

    Labial L L

    Coronal c c c c c

    Dorsal (D) D D Pharyngeal Ph

    Laryngeal La r

    Anterior + + Distributed + + + Strident + + + + High + +

    Low + + +

    12.2 .6 tea turPs for vowels Vowels always involve movement of the tongue body, as thus are always considered to be [dorsal]. ln additio n, since rounding is implemented with tbe lips, round vowels are [ labia l] . The distinctive features needed to distinguish vowels are straightforward: [ +/- high], [ +/-low), [+/-back!, [+/-advanced tongue root) (ATR) and [+/ - round]. A subset of these features can also be used to indicate secondary articulations on consonants (vowel-like constrictions articulated simultaneously with the consonant): !+low) for pharyngealization,

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    268 WHAT IS A POSSIBLE LANGUAGE

    [+back] for velarization, [- back] for palatalization, and [+round) for rounding. The feature [+/-round] must be distinct from [labial), as languages may distinguish r p) ([ labial]) from [p"J ([labia l, +round]); and [t], for example, can become [+round) [t") without becoming labial (pJ.

    Vowel features can be defined in both articulatory and acoustic terms. Each lip and tongue body position results in a typical resonance pattern. (If you haven't studied acoustic phonet-ics (Chapters 6 to 8), and these descriptions of resonance patterns mean nothing to you, ignore them and concentrate on the articulatory definitions.)

    [+/- round] t!H ., definition: [ + round] segments have pursing of the lips. A~ definition: rounding lowers aU resonances .

    f' natural classes and alternations: round vowels, including r u, u, o, :>, y, a:] are [+round], as are rounded consonants such as [t", s", g"J.

    [+/- high] '\ lie td.ll,,ry defi nitio n: fo r [+high] segmen ts, the tongue body IS raised from the neut ral positio n. t\c st defin ition: [+high] segm ents have low Fl. I 1 I''' ' natural classes and alternations: high vowels, including (i, u, u, i y,] are [+h igh], all o ther vowels are [-high]. May also be used to distinguish [+high] palatals and velars from [- high] uvulars.

    [+/- low ] definition: for [+low] segments the tongue body is lowered from the

    neutral position. definition: [+low] segments have high Fl.

    natural classes and alternations: [a ], (a] and [:rJ are [+low]. Guttural conso-nants may also be described as [+low].

    [+/- back) Jl 1 ' definition: for [+back] segments, the tongue body is moved back from

    the neutral position. definition: for [+back] vowels, the first two resonances (FJ and F2) are close

    together, for [- back] vowels, F1 and F2 are farther apart. I !'t natura l classes and alternations: Front vowels (including [i, 1, e, , re) are [-back]. Back and cent ra/vowels (includ ing [u, i, o, A, a) are [+back]. See discussio n below.

    [+/- advan ced tongue root] ( [+/- ATR)) \ lcttl.tlory defin ition: [ +ATR] vowels have the tongue root pu lled forward. \c t' ,, de fi nition: [+ATR] vowels have lowered Fl compared to their [-ATR]

    counterparts. I !'' natural classes and alternat ions: [ i, e, u, o ] are (+ATR), [1, u, e, :>)are [- ATR]. The [ ATRI feature is active in vowel harmony in a number of West African languages, such as lgbo (Chapter 11, example 14). In other languages, [+ATR) vowels may be termed [+tense). See discussion below. Low vowels are sometimes considered to be [-ATR), o r may be undefined for [ATR) .

    Vowel features are in general quite straightforward, but there are a few points worth noting. First, a note on vowel height. The usually-accepted features for vowel height are [ +/- low]

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 269

    and [ +/- high]. There is no d istinctive feature [mid] . T he mid vowels [e, e, o, o, A[ are [- high, - low[. Other apparent or subsid iary differences in tongue body height may be sub-sumed with the feature [ +/-ATR] . There may be a d ifference in tongue body height between [i] and [], but the primary distinction is that [i] is [+ATR] and [l] is [- ATR].

    Concerning the feature [ATR], we wiU use [+ATR] for the English vowels that are usually described as tense, and [-ATR] for the English vowels that are usually described as lax. Introducing a separate feature r +/- tense] would, first, create redundancy, since [+/- ATRI and[+/- tense] target what are essentially the same vowel contrasts. Second, it would make wrong predictions. Generally, we expect featUies to cross-classify: if we have features [ +/-high) and [+/-back]. we expect that all four combinations [+ back, +high], [+back, -high], [-back, +high) and [-back, -high] should occu r. With [ATR) and [tense], the combinations [-ATR, + tense] and [+ATR, -tense] would never occur. O ne possible approach would be to assume that som e languages use [ATR] and some languages use [tense], contra the claim tha t featu res are universa l. Less radica lly, one could propose that [ATR] and [tense] refer to a single universally-availab le di mension of contrast (essentially [i] vs. []) and should there-fore be represented by a single distinctive feature (whichever term is chosen), b ut tha t lan-guages may di ffe r so mewha t in the exact pho netic implementa tion. T his is the solution chosen here. This weakens the requ irement that fea tures have a un iversa l, shared, pho netic defi ni tion, and thus opens the possibi lity that some distinctive features might no t be pho netically-based at a II. The subject o f the precise phonetic defin ition o f fea tu res continues to be one of debate.

    The central vowels, such as [i] and [A], can be defined in featural terms as [+back, - round]; and they contrast with [ u] and [ o] tha t are [+back, +round] . That is to say, although there are three degrees of backness phonetically (front, central, back), there are only two dimensions of contrast phonologically: [-back) front vowels and [+back] back and central vowels. In retrospect, it would have been better to have named the feature [+/-front], but Chomsky and Halle happened to choose [back], and it seems to be too late to change it now.

    There are few, if any, cases of languages contrastmg a bac~ vs. central vowel without also ma~ing a contrast in round1ng. There are no complete~ convincing cases of languages that contrast central[t] and back [UJI. or centra liA! and bad [T]. A few possible back vs. central contrasts, in Nimboran, Norwegian, and Bostonian are discussed in Seaion 4.3 on p. 59, but each is subjea to reana~sis. Bostonian, for example, contrasts "shack" [J~k] vs. shock" [Jok] vs. shark" [Jak]. but in that case [a] may be argued to be underlying (ar). so that back /a/ and central /a/ are not real~ contrastive.

    Table 12.6 gives the featu re values of a representative set of vowels. The [ATR] values shown in parentheses are tentative, as these vowels do not commonly participate in [ATR] contrasts or alternations, so definitive data as to thei r value is lacking.

    Finally, note that, secondary articulations aside, the set of"place" features for vowels and consonants are largely mutually exclusive. Only consonants can be [coronal!, for example, and only vowels are specified for [ATR]. For the most part, consonants are not specified as [high], [back], or [low] . This is by design, allowing for simpler and more general statements of the alternations that affect only consonants or only vowels, which constitute the majority.

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    270 WHAT IS A POSSIBLE LANGUAGE

    Table 1 J .6 Place features for a representative set of vowels. .

    I I e E u () 0 II y 0 i A

    Back + + + + + + + H1gh + + + + + + Low + + ATR + + (- ) + + (-) + + (- ) (- ) Round + + + + + +

    T his decision, however, complicates cases where vowels and co nso nants in teract. Palataliza-tio n is a case in point. The alternation "[s) is realized as (J) before [i)" is quite commo n, and its phonetic grou nding is clear: the fronted and raised tongue body position of [ i] is more co mpatible with (J] than with [s), making palatalizat io n a kind of ass imila tio n. This compat-ibili ty is not captured in fea tures, however. The change from fsl to (J] is o ne of [+anterior] to [-a nter ior}, a feature that is not even defined for [ i] . Reacti ng to this problem, some pho nologists have proposed a unified set of fea tures for vowels and consonants, suggesting that fron t vowels should be [coronal] instead o f [-back]. This fa ils to acco unt, however, fo r the many times tha t coronal consonants and [i] fa il to interact. Finding a si ngle representa-tion that elegantly captures both interactions and lack of interactio ns between vowels and co nso nants continues to be a matter of research and debate.

    T here are nvo mo re vowel features that we will find useful at this point to define natural classes and alternations. We need a feature [ +/- long) to distinguish vowels that differ in length but not quality, such as are found in the Japanese words [biru) building vs. [bi:ru) beer, and to capture alternations Like compensatory lengthening and closed syllable shorten-ing. We will also find it useful to refer to a feature [+/-stress! in order to account for alter-nations that are conditioned by stress, such as English tapping and vowel reduction. We will see, however, that a more carefully-elaborated theory of timing, syllable structure, and stress, to be developed in Chapters 15 and 16, will obviate the need for these two features. Chapter 17 will discuss the features n ecessary to account for alternations and contrast in tone.

    12.3 how have our hypotheses fared? We began this chapter with a set of hypotheses about natural classes in hu ma n languages. Hypo theses 2 and 3 are repeated here:

    Hypothesis 2: Pho no logical contrasts and natural classes ca n be defined in terms of a closed set of phonetic parameters tha t is the same across all languages: the set of distinctive features.

    Hypothesis 3: Features are defined in terms of binary (+/- ) oppositions (e.g., every segment is either [+voice] o r [- voice)).

    The bulk of this chapter has been spent specifying the set of distinctive features. Is this set able to account for all phonological contrasts and natural classes? As far as contrasts go, the hypothesis is pretty well supported. The features are defined clearly enough that we know

  • Zsiga, Elizabeth C.. LAWZ : Linguistics in the World : Sounds of Language : An Introduction to Phonetics and Phonology. Somerset, NJ, USA: John Wiley & Sons, 2012. ProQuest ebrary. Web. 25 March 2015.Copyright 2012. John Wiley & Sons. All rights reserved.

    WHAT IS A POSSIBLE LANGUAGE 271

    what kind of datn would render them inadequate: a language that required six degrees of vowel height could not be accounted for with the features ( +/- high) and [ +/- low], or a language that contrasted four degrees of rounding could not be accounted for with the feature [ +/- round]. Thus far, languages such as these hypothetical cases have not been described.

    As far as natural classes go, the story is more complicated, and we have seen that the set of distinctive features has undergone significant revisions since the original work of )akob-son, and new revisions continue to be proposed. Vt/e have seen that there are numerous cases for which hypothesis 3 has not been supported. The lack of evidence for targeted classes such as [-nasal] and [-lateral] have led to the suggestion that at least some features are unary rather than binary. In particular, unary features for place of articulation based on active articulator have proven to be useful.

    Once unary features are introduced, it becomes clear that no t all segments are specified for every featu re. A lack o f co mplete fea tural specification of a given segment is termed underspecification. Absence of a unary feature (as when non-nasal consona nts si mply have no [nasa l] specificatio n) is one type of underspecification. Another type is when a b ina ry featu re is no t defined or not ava ilab le fo r a certa in class of so unds: (+/- an terior] is defined o nly fo r coro nal consona nts, for example, and (+/- strident) is defined only for frica tives and affricates. Other types of underspecification are possible. For example, if we know that all nasals are (+voice), do we need to specify nasal consonants as both [+nasal] and [+vo ice] in underlying represen tatio ns? There is reason to think we sho uld not. This and similar cases are discussed in Chapter 13.

    Cases that continue to be the subject of research and debate include the definition of the feature [continuant], the possibility of unary [voice), features of guttural consonants, the relation between "tenseness and ATR, the best representation for vowel height, and the interaction or non-interaction of vowels and consonants. These specific cases feed into larger questions of whether the set of distinctive features is truly universal, how dearly-defined in articulatory or acoustic terms each feature must be, and whether features are learned or innate. Different hypotheses have been put forward, each one in service of a slightly different theory of what human languages can be like. All the problems have not been solved- if they were, phonologists would be out of a job. In that case, we could declare that we had learned all there is to know about human language contrasts and alternations, and retire to our respective beach houses. We're not there yet, but t he set of features described in this chapter does ho ld up very well against the data. We will use this set of features in subsequent chapters as we investigate the next set of hypotheses regarding the relatio nship between underlying and surface representations.

    The set of dist inctive features co nstitutes a hypothesis about the dimensions of con trast and alternations avai lable in human languages. As such, they define the space of possible languages. The distinctive features proposed by )akobson, Fant, and Halle were binary, defined in both articulatory and acoustic terms, and fully specified for all segments.

    Subsequent analyses, based on a range of cross-linguistic contrasts and alternations, have suggested that some features are unary, some are based on the active articulator used, and not all features are specified for all segments. A suggested set of features and feature values are found in Table 12.2 (major class fea-tures), Table 12.3 (laryngeal features), Table I2.5 (place features), and Table 12.6 (vowel

    chapter summary