Chapter I

On the Phonetic Bases of Phonology

1.1.1 Many people reading the word “phonology” for the first time would probably guess that it was the name for the systematic study of sound in language. An appeal to a dictionary would quell any doubts: “The science of vocal sounds (= PHONETICS), esp. the sounds of a particular language; the study of pronunciation; transf. the system of sounds in a language.” The professional linguist is not nearly as secure in defining this term, nor in delimiting the field known as phonology. Mattoso Camara attests to this in his Dicionário de Filología e Gramática (151) where he says “Phonology is a term used by different theoreticians in different and even contrary meanings.” He lists three areas of study: the description of the sounds of a language; the value of the sounds in a given language; and the function of sounds in a given language.

1.1.2.1 Generative linguistics has not narrowed the definition of phonology. In recent years, linguists dealing with phenomena such as suballophonic variation, the articulatory peculiarities of individuals, segmental allophony, phonotactics, segmentation, neutralization of phonemic contrasts, morphophonology, supra-segmentals, and distinctive features all say that they work in phonology, all call themselves phonologists. Broadly speaking, phonology is that area of linguistic endeavor that deals with sound. This requires some clarification. An accepted way of looking at languages appears in the following model:

This means that languages entail both sound and concepts. Sound is the principal medium through which human beings and cultures transmit concepts. The grammar of a language is the code that enables its speakers to translate systematic sound into concepts or concepts into sound. Thus, an extreme position would maintain that phonology and grammar are the same thing, that all grammar is involved in the translation of sound into concept and concept into sound.

1.1.2.2 The generative model of language (Extended Standard Theory) compartmentalizes grammatical phenomena as follows:

This model seems to suggest that phonology is a relatively superficial element of the grammars of languages; yet a quote from Chomsky (1965: 81) suggests deeper origins for phenomena which today we recognize to be phonological. He says “... let us review briefly the operation of the phonological component.... Each lexical formative is represented by a distinctive-feature matrix...” That is, simply, that the phonetic parameters of phonology are located in the lexicon of universalist models of language. Morphology inevitably involves the use of phonological parameters; morphemes have phonetic substance. The model implies an interaction of sonorous and grammatical entities in the production of language. With generative grammar’s claim of universality and its claim to be a model of the linguistic competence of the human race, it reflects the truism that human beings have minds that interpret the sound patterns of their languages, and that they have articulatory organs that produce sonorous representations of whatever they seek to express in language.

1.1.3 The insecurity of linguists in the 1970’s concerning the definition of phonology can be witnessed in recent assertions of practicing phonologists. Foley, in Foundations of Theoretical Phonology, attacks the point of view I shall advocate. He says (52): “Only when phonology frees itself from phonetic reductionism will it attain scientific status.” And (24) “As . . . the elements of psychological theory must be established without reduction to neurology or physiology, so too the elements of a phonological theory must be established . . . without reduction to . . . phonetic characteristics...” He derides all previous work in “phonology”, maintaining (1) that his book “... presents ... perhaps the only genuine theory of phonology in existence...” The bases of these assertions are dubious. Foley’s abstract elements are nothing more than traditional articulatory phenomena that have been renamed with non-articulatory labels (cf. Brakel, 1980). In many cases articulatory data explain the phenomena more coherently than his abstract elements do. Kisseberth and Kenstowicz dedicate a major part (63-130) of their 1977 book to “The NonPhonetic Basis of Phonology.” They assert (63) that while “Phonological alternations generally have their ultimate source (historically speaking) in sounds being effected by the phonological context in which they occur . . . such contexts often become obscured through subsequent historical evolution. The phonetic basis of the original change may thus be lost, resulting in an alternation that . . . from a synchronic point of view lacks phonetic motivation.” The authors examine phonemena such as rule inversions, telescoping, grammaticalization, and regularization (generalization) as linguistic phenomena that tend to obscure the articulatory motivation of certain morphophonemic sound changes in various languages. They also show that non-sonorous entities such as morpheme boundaries, grammatical categories, syntactic patterns, and the lexicality of certain items play an active part in the grammar of sound in a particular language.

1.1.4 No phonologist will deny any of Kenstowicz and Kisseberth’s assertions. Many will maintain, however, that their particular brand of phonology does not concern itself with one or another phenomenon touched on by the former authors. Notwithstanding all that they say in Topics in Phonological Theory, the reason why their field of endeavor is still called phonology is that sound, in this case sound production or articulation, is inevitably involved in whatever phenomena they describe, in whatever rule they may write. It is wrong to consider nonsonorous entities and concepts as the base of phonology; they are theoretical constructs which linguists use to help systematize the sound that they know occurs in the production of all languages. Such constructs are ancillary, the base of phonology is sound produced in the human articulatory apparatus to be interpreted by the human brain.

1.2. To say that some phonological rules contain non-phonological abstract entities such as morpheme boundaries does not deny the articulatory base of phonology. Such a denial would necessarily conflict with the most obvious fact of language production—all human beings use virtually the same articulatory apparatus to produce the sounds of all languages. This articulatory tract, the organs of speech, is undeniably a linguistic universal. The specific interaction that articulation makes with the grammatical elements of a language is peculiar to whatever language the describer happens to be studying and depends upon the theoretical precepts and primes used. For example, a peculiarity of German and a few other central European languages is that all word-final stops are voiceless. We can express this fact discursively or we can write a rule for it. If we accepted the precepts of American structuralists: A. that the phoneme is the smallest unit of phonological analysis and B. the invariance principle—once a phoneme, always a phoneme; then we would have to consider all words which had a stem-final stop consonant which was voiced when an inflecting suffix was added, e.g. bunde, bergen but voiceless when no suffix was added bund [bunt] berg [berk] to have two allomorphs containing different phonemes. If we accepted the Prague school precepts concerning neutralization of contrasts, then we could represent the stems as /bunD/ and /berG/ and write rules governing the realizations of /D/ and /G/ according to their environments. Or finally, using distinctive features and the generative approach, a rule such as [+occlusion]→[+surd] # could be written and interpreted as either a phonotactic constraint of certain Germanic and Slavic languages or as a rule converting voiced stops to voiceless ones in word-final position; the language and data are the same, their interpretation differs according to linguistic schools.

1.3.1 In recent phonological work, linguists have represented articulation with distinctive features. Distinctive feature (hereafter, D.F.) inventories form an important cornerstone of the generative approach to phonology. If segments were the smallest phonological unit, unification of trans-segmental phenomena in a single rule would be cumbersome and, sometimes, impossible. The sub-segmental nature of the D.F. enables the rule writer to express phonological generalizations for an entire class of segments. For example, many languages have three nasal consonant phonemes /m, n, ñ /which contrast word initially, intervocalicaly, and utterance-finally but assimilate to the point of articulation of any consonant that follows a nasal. In a descriptive approach in which segments were the smallest unit, a rule to express such a generalization would look like this:

Using distinctive features the same rule appears as

1.3.2 Beyond the notational advantages of using distinctive features in rules, a system, an inventory, of D.F.’s is a linguistic hypothesis concerning the phonetic parameters that are necessary to perform the task of describing the sonorous elements that human beings use to distinguish utterances. Any purportedly universal distinctive feature inventory says: 1) These and no more than these parameters are necessary to identify the distinctive articulatory possibilities of all human beings. 2) These parameters are adequate to write all the phonological rules of all human languages.¹ A distinctive feature inventory is the most abstract sonorous element in any grammar of sound, and at the same time it is the most concrete hypothesis about linguistic behavior. It is the focal point of the grammar of any language because it is essential for converting abstract concepts into concrete sounds and concrete sounds back into abstract concepts.

1.3.3 Phonology is the sector of language study in which a hypothesis such as a D.F. inventory has a real chance of becoming a fact. Again, the universality of the articulatory tract is undeniable. The moving parts that produce all linguistic sounds are few, i.e., 4: the vocal bands, the velum, the tongue, and the lower lip; thence careful study of their operations in producing the sounds that linguists identify as segments should reveal precisely which gestures produce all the segmental contrasts of human languages. Some phonologists are embarrassed, however, about the status of the primes used to describe the distinctive sounds of speech. This embarrassment comes from the recognition that theories are explanations of phenomena wherein some of the facts are unknown. In phonology the facts of articulation are known, yet we have, to date, no widely accepted distillation of them into a system of primes (D.F.’s) which represents the distinctive articulations in the vocal tract.

1.3.4.1 Such a distillation is necessary because a satisfactory set of theoretical primes is essential to any investigation. The grammars that linguists create and the conclusions which they draw are shaped by the precepts and primes with which they operate. As examples let us examine two phonological descriptions made with different systems of primes. The contoids that are capable of closing a syllable in Portuguese are/r, 1, n, z/. To describe them using the inventory of D.F.’s in Chomsky and Halle’s Sound Pattern of English one needs the following rule:

This is unsatisfactory because a disjunction is needed to include /n/ and /z/ in the same rule, thus no natural class emerges for this set of segments. The positive specifications suggest, to many linguists, that these consonants are highly marked. A description of them can be made using the system of D.F.’s to be proposed in the present study:

While the number of features is the same, there is no need for a disjunction; thus the consonants pertain to a phonetic natural class. The negative specifications suggest, as well, that these contoids are much less contoidal than others, that they are the unmarked members of the fricative, nasal, lateral, and “vibrant” series of Portuguese phonemes.

1.3.4.2 Another phenomenon which can be described more efficiently with different tools of analysis is what is generally agreed to be contoidal weakening in Spanish and Iberian Portuguese. Intervocalic /b, d, g/² are realized as approximants. To represent this in SPE features the following rule is necessary:

The addition of a feature, [+continuant], to many linguists suggests the addition of a mark which is the opposite of a weakening. In the Chomsky-Halle school of phonology, rules such as this necessitated an elaborate marking mechanism which maintains that certain features represent markedness in certain environments and non-markedness in others. In the system to be advocated here, the following rule describes the intervocalic weakening of voiced stops:

Besides needing fewer features to describe the phenomenon, contoidal weakening is symbolized by the loss of a phonological prime.

1.3.5 Two rules are hardly sufficient for the justification of a new theory of phonology. I shall endeavor to show in Chapter III, where I justify my revision of the inventory, that the features I propose have greater observational, descriptive, and explanatory adequacy than those suggested by Chomsky and Halle (1968) (hereafter SPE). I shall also contrast my D.F.’s with the acoustic primes of Jakobson, Fant, and Halle (1952) in Chapter IV.

1.3.6 Another advantage of theoretical work in phonology and D.F.’s lies in the comparability of hypotheses. Because of the finite articulatory base that human beings have and use to create sound systems, the phonological primes that linguists propose are finite and, as theories, the best D.F. inventories are the simplest ones. The Chomsky-Halle inventory of D.F.’s contained 25 segmental primes of interest here, which, if they could combine freely, would specify a total of 33,554,432 segments (2²⁵)—certainly many more than linguists have any need to identify. Obviously, the total number they can specify, because of the physical limitations of the vocal tract, is somewhat less than the above figure, but 1 doubt that even .01% of the power of this theory is necessary to identify all the segments that linguists have needed to describe in the entire history of linguistics. The Jakobson, Fant, and Halle (1952) (hereafter, JFH) D.F. inventory (defended again in Jakobson and Waugh, 1979 hereafter J&W) contains 12 primes, but, to function, will need two more (see Chapter IV) which will give it a total of 14 primes and a generative capacity of 2¹⁴ or 16,384 segments. These D.F.’s do not include prosodic considerations such as length nor air stream features. Including these in the inventory I propose, gives a total of 19 features and a generative capacity of 2¹⁹ or 524,288 segments which is 1.56% as strong as the SPE inventory.

1.3.7.1 While this work and much of the linguistic research of the 1970’s would be meaningless or impossible without Chomsky and Halle’s SPE, their book has been criticized on many grounds: its D.F. inventory, its approach to markedness, and even, the observational descriptive, and explanatory adequacy of the whole system as a theory of phonology. Ladefoged (1971) argues that some of the SPE D.F.’s are descriptively inadequate and that their strict binarity does not prevail at the level of surface phonetics. Lisker and Abrahamson (1971) criticize Chomsky and Halle’s features which purport to account for laryngeal control during the production of sound as well as their claim that the features proposed in SPE represent articulatory reality. Campbell (1974) shows the inadequacies of SPE D.F.’s in their specification of natural classes of segments. Lass (1976: 168-212) maintains that the SPE D.F.’s are often unable to make generalizations and unite natural classes in individual languages. Kuipers (1975) attacks SPE on the grounds that the semiotic foundation on which the phonetic framework and the marking theory were built was less than solid. Foley (1977) claims that SPE and generative phonologists in general have not achieved any satisfactory explanations of phonological phenomena.

1.3.7.2 Certainly these attacks come as no surprise to Chomsky and Halle—they anticipated them before elaborating their primes (298): “We are well aware of the many gaps in our knowledge that make the success of this undertaking somewhat problematical ...” In view of this, the unquestioning support that their original postulates have been given (e.g. Mira Mateus, 1975; Anderson, 1976) is surprising. Chapter 9 of SPE begins with the caveat (400):

The entire discussion of phonology in this book suffers from a fundamental theoretical inadequacy ... There is nothing in our account of linguistic theory to indicate that the result would be the description of a system that violates certain principles governing human languages ... we have failed to formulate the principles of linguistic theory, of universal grammar, ... we have not made any use of the fact that the features have intrinsic content.

1.4 1 shall show that several of the features really have no intrinsic content (vocalic, syllabic, sonorant), and that the implied content of some is actually counter-productive to efficient description of sound systems. 1 shall attempt to reduce the generative power of the D.F. inventory by reducing the number of primes, yet I shall increase its observational, descriptive, and explanatory adequacy.

Footnotes

¹Phonological rules represent actual articulatory gestures of speech sounds according to their systematic, potentially distinctive, properties. Any detailed, phonetically accurate descriptions or denotations of actual articulations in particular languages usually involves minutia that are phonologically redundant or unimportant in distinguishing utterances’ meanings.

²In Portuguese, the best analysis of these sounds is that they are occlusives. They are obligatorily so only after a homorganic nasal or lateral. It is, however, never wrong to pronounce them as occlusives. The point of the discussion on this matter is, simply, that the SPE features do not portray their allophony in a satisfactory manner.