Phonological explanation in the theory of phonetics: the ...

University of Calgary

PRISM: University of Calgary's Digital Repository

Calgary (Working) Papers in Linguistics Volume 04, Spring 1978

1978-05

Phonological explanation in the theory of phonetics:

the distributional frequency of half-nasal consonants

Herbert, Robert K

University of Calgary

Herbert, R. K. (1978). Phonological explanation in the theory of phonetics: the distributional

frequency of half-nasal consonants. Calgary Working Papers in Linguistics, 4(Spring), 27-44.

http://hdl.handle.net/1880/51271

journal article

Downloaded from PRISM: https://prism.ucalgary.ca

Phonological Explanation in a Theory of Phonetics:

The Distributional Frequency of Half-Nasal Consonants

Robert K. Herbert

1. Introduction

The past several years have witnessed a significant increase in the role accorded explanation within linguistic theory. There have been numerous attempts to explain linguistic phenomena on all levels of analy­sis by reference to linguistic and extralinguistic factors. This concern is not novel. The direction it has taken presently, however, can be traced to a dissatisfaction with the very formal and abstract conceptuali­zation of explanation within early generative grammar. In this paper, I would like to consider briefly the present status of explanation in the area of intersection between phonology and phonetics and to suggest that just as we have come to recognize the perils of a phonetics-free theori of phonology, there can be no such phonology-free theory of phonetics.

2. Explanation

One of the more important notions central to the evaluation of any theory is explanation. The explanatory goals of modern linguistics have been explicitly formulated by a large number of scholars. However, the term explanation is used, not always appropriately, at many levels of analysis. Scheffler (1963), for example, cites the following four dis­tinct levels of explanation:

i. classification of terms or statements, e.g. definition of umlaut, definition of syllable coda, definition of a Bantu language, etc.

ii. provision of reasons in support of a judgement, e.g. why [h] is not a syllable coda in English, why Kikuyu is a Bantu language, etc.

iii. weaving together a theoretical fabric within which credible generalizations may occupy determinate places, e.g. construc­tion of a theory of syllable structure, a theory of what Proto-Austronesian may have been like, etc.

iv. causal diagnosis of particular events, occurrences, and facts, e.g. the development of open syllables in the history of Malagasy, the acquisition of click sounds in some Bantu languages, etc.

Two important points need to be noted in this schema. First, the so­called different levels of explanation are by no means independent entities. For example, an explanation of the first level, a definition of syllable coda, necessarily incorporates large theoretical claims such as the existence of syllables and the possibility of decomposing

- 27 -

- 28 -

linguistic form into syllable units. Second, it is highly questionable that the fourth level of explanation, the level of causal diagnosis, can be achieved in any science, even when the explanation has predictive power. On the one hand, it is always possible to 'push' the explanation one step further back. For example, the fact that the plural of Modern English mouse is miae is explained by the fronting and later unrounding of the vowel of Pre-Old English mUs which is in turn explained by the generalization that all u's became y in certain environments. This fact is explained by the generalization that all back vowels were fronted under certain conditions which is in turn explained by umlauting. Umlaut is explained by a theory of vowel harmony which is explained by a general theory of assimilation which is then explained by various phonetic factors such as coarticulation. However, even when this line of reasoning is exhausted, there exists a wide range of psychological and sociolinguistic factors which need to be cited to explain the form miae in Modern English. Further, a theory of explanation is incomplete unless it can explain what does not happen as well, 2 e.g. explain why Primitive Germanic and Modern English tolerate sequences which Old English did not. Viewed in this light, it is apparent that a theory of explanation in linguistics, viz. a theory of causality, is impossible. Although the concept disappears, the term ea:pZanation is retained and the notion of functional association is substituted for causality. Similarly, it is necessary to distinguish ea:planation in its causal sense from e:x:pZanatopy sketah which is incom­plete, a partial explanation. This is, at best, the level of explanation which can be achieved in linguistics.

2.1 Explanation in Phonology

In phonology, the search for a more satisfying level of explanation, seen in the light of the so-called explanations of orthodox generative phonology, has taken a particular direction. Specifically, there have been numerous attempts to explain various synchronic and diachronic phenomena by reference to physiological and acoustic-auditory phonetics. Ohala (1975:289-91) states explicitly that it is "difficult to give serious consideration to phonological works which purport to explain the naturalness, expectedness, or unmarked character of sound patterns while ignoring or explicitly denying the relevance of anatomical, physiological, or acoustic-auditory aspects of speech to their task." Among the sound changes which have been examined in this light are:

1. the lowering of tone after voiced consonants 2. the fronting and affrication of velars before front vowels 3. the frequent changes of fricative to fricative 4. the nasalization of vowels 5. the tendency for nasalized vowels to lower.

The (re-)phoneticization of phonology is not, of course, the only reaction to the abstractness of generative phonology which has arisen and gained some prominence. There are several broader theories of "natural phonology" whose genesis can be seen as a reaction to The Sound PattePn of English (Chomsky and Halle 1968) and similar studies. Although all

•

•

- 29 -

versions incorporate the notion of phonetic plausibility to which Chomsky and Halle pay lip service, there is wide variation in the amount of im­portance attached to this criterion and the role it plays in the theory. Natural Generative Phonology allows underlying representations which correspond to the level of systematic phonetics and only phonological rules which are phonetically motivated. Some work has been done in de­termining directions of change which are context-free, especially by Stampe (1972, 1973) and his students (e.g. Miller 1972). On the other hand, various context-sensitive changes have been very effectively treated by Ohala (1974, 1975) and other members of the Berkeley phonology labora­tory. These last two approaches attempt to explain in rather different ways, why some phonological processes occur again and again in diverse languages whereas others less often occur and some never. 3

2.2 Explanation in Phonetics

The intersection of phonetics and phonology is much deeper and more fundamental than reference to phonetic facts in explaining phonological patterns. There exists no general phonetic theory which is phonology-free since there necessarily exist certain phonological presuppositions in any phonetic theory. 4 For example, the division of the speech continuum into discrete phonetic segments for classification is basically a phonological task based on the recurrence of certain units. The classification of sounds in the alphabet of the International Phonetic Association is basic­ally a phonological grouping which extracts certain parameters as dis­tinctive such as plosive, nasal, bilabial, etc.

Phonetic theories fall basically into two major groups which re­flect two very different views of what we should expect of a theory of phonetics. These approaches are appropriately termed anthropophonic and linguistic phonetics. The anthropophonic approach, which term was popular­ized by Baudouin de Courtenay, is concerned with the so-called "sound pro­ducing capabilities of man" and attempts to delimit all the possible para­meters which can be manipulated and all possible combinations of parameters which produce sound. There are no real applications for a general theory of language which are forthcoming from such innnense concerns. For example, the theory does not ask why the same sound types occur again and again in the world's language. Two of the prime proponents of such a theory of phonetics are Pike (1943) and Catford (1968, 1977).

The theory of linguistic phonetics, on the other hand, deals only with sounds found to be used in language. In its narrow form, the theory is concerned only with those sounds used distinctively whereas in its broad form there is no question of distinctiveness. The theory poses such questions as those relating to the distributional frequency of speech sound types, the determination of what features are to be used in analysis, etc. Due to its nature, this theory needs to be constantly revised and expanded as new sounds are discovered in language.

It is now commonly agreed that just as we expect of a theory of phonology more than that it be taxonomic, so too our theory of phonetics must pass beyond simple classification. Just as we expect a theory of

phonology to explain why certain sound changes commonly occur whereas others rarely or never do, our theory of phonetics must explain why some speech sound types commonly occur, some rarely, and some never although they are all within "the sound producing capabilities of man." This concern is, of course, not novel; the principles of least effort and maximal perceptual contrast have long been cited, with varying degrees of success, as functional explanations in linguistics. In recent years, there has been a great deal of very sophisticated phonetic work which addresses itself to these concerns and others such as the origin of the phonological feature system. Lieberman (1970) has proposed that phono­logical features correspond to signalling ~nits which take advantage of various innate mechanisms of the human vocal apparatus and the auditory and perceptual systems such as acoustic feature detectors, the "ease" of certain articulatory maneuvers, and the stability of acoustic outputs produced by certain articulatory maneuvers. Stevens (1972) has demon­strated that there are certain acoustically "natural" points of articu­lation in which relatively large variation in articulation has small acoustic effect and that these ranges are bounded by others which are acoustically sensitive to slight changes in articulation. This explains, for example, why the sama set of places of articulation for stop conson­ants is utilized again and again throughout the world's languages whereas others occur much more rarely. Lieberman (1974) argues that determinatness, i.e. a one-to-one relationship between formant patterns and particular vocal tract configurations, along with other principles such as Steven's acoustic stabiZity, structure universal features. Finally, Lindblom (1972) and Liljencrants and Lindblom (1972) provide a numerical model of the acoustic space of the vocal tract and an algorithmic procedure for finding a set of n vowel qualities based on a certain definition of per­ceptual contrast. This ~rocedure is in good agreement with attested vowel systems of 3 - 8 vowels. There are some problems with systems of more than 8 vowels, namely the model generates too many central vowels and no [~]. Similarly, Hombert (1976) presents a mathematical model for pre­dicting the most likely tone shapes of a tone system based on a trade-off betweEm maximal perceptual difference and minimum articulatory difficulty. These attempts to construct quantitative models with predictive power seek to demonstrate the unmarked character of certain contrasts and thus the origins of phonological structure. Although these avenues of research seem very fruitful, it is clear that they will not be able to account for all the data on the frequency of occurrence of sound types.

3. Half-nasal Consonants

We will now examine the relationship between two "exotic" speech sound types: the so-called 'half-nasal consonants i.e. prenasalized and postnasalized consonants. Both types are consonantal segments which are characterized by changing values of nasality during the articulation of the segment. Prenasalized consonants are geographically widespread, occurring in Africa, Amerindia, and large parts of the Pacific; post­nasalized consonants occur distinctively in New Guinea, Australia, New Caledonia, and the surrounding areas. In some languages, there is a close affinity between the two types of sounds. For example, both types of half-nasal consonants may be produced contextually by shielding processes with perceptual motivation. In Kaing~ng, a language of Brazil:

..

..

- 31 -

/m/ ---> [mb] I V [bmJ I v

[bmbJ I v v [m] I v---V

These contextually-derived half-nasals have been the subject of much recent discussion by advocates of autosegmental phonology and internally struc­tured features. Both approaches are critically treated in Herbert (1977a). Additionally, certain prenasalized consonants arise by a "series gener­ating component" of prenasality whereby underlying voiced stops are real­ized with an initial period of nasality, e.g. Fijian, some dialects of Malay, some New Guinea languages, some dialects of Modern Greek, etc. In Fijian, the period of nasality is optionally absent in absolute initial position. In this paper, we are concerned neither with the processually produced half-nasals nor with those which are produced by a series gener­ating component. Our concern is only with those half-nasal consonants which can be attributed to underlying nasal and oral components, i.e. those which might be specified at the underlying level by some ad hoc feature such as [prenasal] or ~ostnasal]. It is the exact nature of the relationship between the oral and nasal components which is the subject of this paper.

In the majority of languages which exhibit half-nasal consonants, only one type occurs, but the two types do contrast in a few languages. S. Hyengh~ne, a language of New Caledonia, is reported to have the follow­ing consonant inventory (Haudricourt 1971:368):

unaspirated p pW t c k

aspirated Ph pwh th h kh c

postnasal pm pwm tn en k?J

prenasal m m w nt n ?Jk p p c

vd. nasal m mW n fi ?J

vl. nasal, h h w h h_ h

asp. m m n n ~ 0 0 0 0

Both prenasalized and postnasalized consonants are analyzed as secondary nasal consonants, i.e. nasal phonemes whose most characteristic allophone is not a voiced nasal stop (Ferguson 1966). 6 Intuitively, both types seem equally complex phonetically, involving in the one case a raising and in the other a lowering of the velum within a single consonantal segment. Yet, on a distributional level, prenasalized consonants occur far more frequently than postnaslized consonants. This fact requires explanation.

References to the ease of articulation of one type of sound over another are typically circular: it is asserted that there exists a general preference for "easier" sounds and therefore these sounds occur more fre­quently in human languages. The status of "easier" sounds can be deter­mined only by reference to those sounds which occur most frequently, how­ever. A formal attempt to incorporate these concerns into phonological theory is the basis of markedness theory as exemplified in Chomsky and Halle (1968) and Postal (1968). However, as many critics have pointed

- 32 -

out (e.g. Lass 1972), the theory is still largely devoid of content.

In the case of complex phonetic units, it might be possible to provide a physiological basis for the ease of articulation of one type over another. 7 Bjork (1961) and other researchers since have demonstrated that the velum can be lowered more quickly and with greater precision than it can be raised. 8 Thus, this would seem to imply a preferred status _for sequences of oral-nasal compounds over nasal-oral compounds. The only problem with such an analysis is that it gives the wrong result, namely it predicts that postnasalized consonants should be easier to articulate than prenasalized consonants and therefore more common among the world's languages.9 The facts of frequency of occurrence are true not only for the two types of half-nasal consonants but for consonant clusters composed of oral and nasal segments as well. This is by no means a fortuity of occurrence as is demonstrated below.

A prenasaZized ~onsonant is defined formally as a necessarily homorganic sequence of nasal plus oral consonantal segments which to­gether exhibit the surface duration approximate to 'simple consonants' in those language systems within which they function. The same criteria hold for postnasaZized consonants except that the order of elements is reversed. Ignoring for the moment the general question of how complex surf ace units are represented at the underlying level, the crucial characteristic which distinguishes half-nasal consonants from clusters is duration, i.e., a cluster manifests the surf ace duration equivalent to the combined duration of its individual members.

One of the phonological arguments usually cited in favor of analyz­ing prenasalized and postnasalized consonants as underlying units is that they function within single surface syllables in many languages. For example, in Runyankore, a Bantu language of Uganda with a fairly extensive inventory of prenasalized consonants, surface syllabification is as below:

(a) kutui;iga ku.tu.i;iga 'to get' kugyenda ku.gye.nda 'to go' ekintu e.ki.ntu 'a thing' ekicoiico e.ki. co.nee •a gift' kumbi ku.mbi 'badly'

(b) nimbara ni.mba. ra 'I am counting' /ni+N+bara/ (cf. nitubara 'we are counting')

empapura e.mpa.pu.ra 'paper' /e+N+papura/ (cf. oru2a2ura 'piece of paper')

empene e.mpe.ne 'a goat' /e+N+hene/ (cf. akahene 'a small goat')

andeeba a.ndee.ba 'he sees me' /a+N+reeba/ (cf. areeba 'he sees')

The obvious generalization here is that all surf ace syllables are open, i.e. end in a vowel. Half-nasal consonants are analyzed as units in many languages precisely because of such syllabification and the fact that no other clusters occur. Although I agree with the surface syllabification given above, I claim that at a more remote level of organization the

•

- 33 -

syllabification is:

(c) kutuJJga kugyenda ekintu ekicofico kurnbi nirnbara empapura empene andeeba

ku.tuJJ.ga10

ku.gyen.da e.kin.tu e. ki. con.co kum.bi nim.ba.ra em.pa.pu.ra em.pe.ne an.dee.ha

Underlying syllables are of the general form: (C)V1 (N). 11 There is in­strumental evidence to support the assignment of preconsonantal nasals to the syllables to their left as well as evidence relating to compensatory lengthening and other phonological processes (Herbert 1975, 1977c).

Although this revised syllabification complicates the statement of underlying syllable structure, it is itself not too marked a structure in the sense that many languages which are otherwise characterised by only open syllables allow closed syllables which end in a nasal, but no other consonant codas. This is the case in Tikar, a Bantoid language of Came­roon. Additionally, in many of the Bantu languages of West Africa, the range of syllable codas is very restricted, but nasals are not excluded. Among the languages discussed by Richardson (1957), Basosi allows closed surface syllables which end in a nasal or glottal stop. In Elong, closed syllables end in /1, N, k, ?, p/; in Nyokou and Dyanti, they end in /p, N, s, ?, r/ and /r, k7

, y, n, Q/ respectively. These facts are not limited to Bantu. In Kanakanava (Sung 1966), a language of Formosa, only nasals occur as syllable finals. Maryott (1965:118) gives the segmental invent­tory of Tabukang Sangir, an Austronesian language as:

p t k ?

b d g i 4: u w s y h m n lJ e 0

1 i r a

Most words conform to one of the following patterns:

in which:

c1 includes all consonants c2 includes all consonants except /w, y, r, !, !J/ c3 includes only nasals and glottal stop

- 34 -

c4 includes only /ij, ?/ v1 includes all vowels v2 includes all vowels except /i/

In the development of some of the Miao languages of South China and Burma, all s~llable-final consonants were lost except nasals, which all became [ij]. 1 Bell (1971) surveys 39 languages with restricted patterns of syl­lable codas. Most of these exhibit nasal codas, but among the 39, some languages do not permit final nasals. For example, the only offset allowed in eight languages is glottal stop, which is, in some sense, the most ob­struent-like consonant and therefore the most expected coda. In some languages, final glottal stop is the relic of a previous system of final stops. Among these languages, Bell lists Burmese. However, Okell (1969) argues that both /?/ and /n/ need to be recognized as syllable codas al­though the latter is often realized only as vowel nasalization. Thus, there are processes which may account for the absence of nasals in restric­ted syllable coda inventories.

It is in a real sense natural that nasals should function as syl­lable offsets in languages where other consonants cannot. Very of ten liquids also function within the class of restricted codas. Dixon (1976: 263) notes that the most common type of cluster in Australian languages involves an apico-alveolar first member (/n, 1, r/) followed by a peri­pheral stop or nasal (/b, g, m, n/). That /n, 1, r/ are all sonorants seems to be the more significant generalization. This behavior is a natural relfex of the fact that nasals and liquids, as sonorants, are permitted vowel-like behavior. The ultimate generalization may be that nasals and liquids function as part of complex syllable nuclei. 1 3 This generalization accounts for the fact that sequences of nasal and oral consonants are often possible whereas the reverse sequences more rarely occur and are often separate morphemically although they may arise via vowel loss as in Nemi, a language of New Caledonia (Haudricourt 1964):

*tama > tna *tina > tne

4. Syllables

'father' 'mother'

The explanatory sketch which is proposed to account for the dis­tributional frequency of half-nasal consonants draws a fundamental dis­tinction between the segmental and syllabic levels of organization. These two levels of organization reflect directly the distinction between under­lying clusters and derived surface units. In fact, the unification of independent segments into unitary sounds may be motivated b? the transition from segmental to syllabic, i.e. larger-than-segment unit,l organization.IS The various evidence in support of the syllable unit will not be reviewed here. The evidence is largely fragmentary and refers to such diverse facts as the "syllable size" domain of prosodic features, constraints on syllable structure defined as "morpheme structure conditions", co-articulation and malfunctioning of production which seem to imply that the unit of articu­latory programming is larger in size than the segment, the syllable as a

"

- 35 -

domain of phonological processes, etc. The crucial problem involved in the incorporation of the syllable as a formal unit into generative phono­logy is a definition for the term syllable. This problem has, of course, a much longer history, and a good deal of confusion seems to arise from a failure to distinguish fully between phonetic and phonological levels of organization. To be sure, Grammont (1950) and many others have dis­tinguished between two types of syllables. However, Grammont looks for a physiological correlate of phonetic syllables which, he claims, are formed by "the principle of muscular tension". He treats phonological syllables as purely theoretical constructs which represent the rhythmic element of language. In this paper, syllable is defined as a highly struc­tured larger-than-segment unit of organization which plays an especially important role in timing organization in spoken language. Organization of structured units occurs at both the underlying and surface levels of linguistic organization. Therefore, the syllable, by definition, has both phonological and phonetic existence. It is assumed that in most cases underlying units of organization will be identical with surface syllables. However, there is no a priori reason why this is necessarily the case. The treatment of half-nasal consonants which is proposed here rests upon the claim that segmental restructuring and reorganization are possible and, in fact, fairly common occurrences on a universal basis.

4 .1 Unification

The input to the segmental level of organization, insofar as half­nasal consonants are concerned, is a sequence of oral and nasal conson­antal segments. 16 This is obviously correct for half-nasals which are morphologically complex such as those listed in (b) above. There is further a great deal of evidence, both phonetic and phonological, to support this claim for other half-nasals, which cannot be reviewed here (cf. Herbert 1975, 1977a). It should be noted that various phonetic · processes such as position assimilation of the nasal, voicing assimilation, post-nasal hardening, etc. occur at the segmental level of organization and th·eir effects are evident not only in half-nasal consonants in some languages but in clusters in many languages which do not fuse the two segments. The actual process of unification is a syllable level process in which sequences function primarily as components of the syllable and are subordinated to constraints determining syllable structure, e.g. temporal adjustment and reduction.

There are, of course, languages in which half-nasal consonants occur word-initially. In these cases, it would seem difficult to maintain an underlying heterosyllabic representation. However, in some cases medial prenasalized consonants become sequences of syllabic nasal plus oral con­sonant in initial position, e.g., Ganda:

embogo eJJkuba

ipbogo ~kuba

'buffalo' 'rain'

Conversely, in Anyula, an Australian language of the Northern Territory, prenasalized consonants occur only initially and syllable-initially after

- 36 -

another consonant; intervocalically, they are heterosyllabic surface clusters. There are languages in which prenasalized consonants occur morpheme-initially in which there is no alternation with sequences of nasal plus consonant. This is the case, for example, in Fula, a West Atlantic language. In fact, Fula contrasts prenasalized consonants, nasal plus stop clusters, and nasal plus prenasalized clusters. A contrast between half-nasal consonants and the corresponding surface clusters is extremely uncommon; if such contrasts cannot be explained in sollle non-ad hoc fashion, our claim regarding the non-unit status of half-nasal consonants is falsi­fied. Although the initial prenasalized consonants of Fula have generally been used to argue for an underlying unit analysis, there are other limi­tations on their distribution which suggest that they are not underlying units. For example, all initial consonants may occur as the first member of CC clusters. There exist (nonhomorganic) nasal-stop clusters, glide­stop clusters, .liquid-stop clusters, fricative-stop clusters, and stop-stop clusters. However, prenasalized consonants never occur in this position.

The only other case cited in the literature in which a language contrasts prenasalized consonants and nasal-oral clusters is that of Sinhalese, an Indo-European language of Sri Lanka. Feinstein (1977) has demonstrated that in both of the above cases, Fula and Sinhalese, the surf ace distinction can be attributed to language specific rules of syllable boundary placement which are independently motivated. Feinstein treats initial prenasalized consonants simply as marked initial clusters which are, in our terms, unified into single complex segments. The treatment proposed in Herbert (1977a) is somewhat different in that initial nasal-oral sequen­ces are treated as heterosyllabic as well as medial sequences. This follows from the vowel-like status of nasals, especially preconsonantal nasals. If nasals may occur as independent syllables initially, it may be that they also do so in non-initial position. Thus, at a remote level of organization, CVNCV strings may be syllabified CV.N.CV. There is some evidence to support such an analysis, including facts of timing and various tonal perturbations which suggest that the nasal bears an underlying tone in some languages. This would parallel a treatment of long vowels in which they are analyzed as sequences of two identical vowels with an intervening syllable boundary. This extension of the model we propose does not crucially affect the point at hand and is not pursued further here.

4.2 Motivation for Unification

The very important question which needs to be asked is simply the following: why does unification occur in some languages and not in others so that identical underlying sequences have different realization in, for example, Runyankore and English? Granting the correctness of the analysis which has been proposed here, it is still desirable to provide some moti­vation for the absence of the process in many languages and, more import­antly, for the discrepancy between underlying and surface organization in those languages where it is present.

Both synchronically and diachronically, changes in syllable struc­ture, i.e. changes in the number of segments and/or the syllabic organi­zation of a segment string, are not rare. Bell (1971:52-53) lists various

•

- 37 -

phonetic and morphological processes which shape syllable structure: vowel loss, nuclear fusion, cluster simplification, consonant loss, consonant vocalization, vowel epenthesis, consonant epenthesis, meta­thesis, word combination, analogic extension of inflectional and de­rivational elements, and metanalysis. It is proposed to add consonant unification ~r fusion) to that list.

A great deal of work has been done attempting to delimit the con­ditions under which, for example, vowel loss is most likely to occur. It appears that such factors as the quality of the vowel, the position of accent within the word, position of the vowel with respect to various boundaries, the nature of contiguous segments, and constituent membership are potentially important. However, the above factors do not explain why vowel loss occurs; they simply delimit the most probable context for such loss. In contrast, we understand the motivation for vowel addition some­what better. For example, vowels may be added to break up impermissible clusters, to make initial sequences conform to the MSC's of a language, to make all words end in a vowel, etc. These processes are often evident in loan phonology, e.g. Runyankore:

En8lish Runyankore

report ripoota blanket burangiti glass egiraasi cement simenti scout omusikautu petrol peteroro

Thus, vowel addition may be directly motivated by constraints governing syllable structure, and in this respect is similar to unification involved in the derivation of half-nasal consonants.

It is hypothesized that the motivation for consonant unification, at least insofar as the fusion of oral and nasal segments is concerned, is provided by constraints or general tendencies for preferred syllable structures. Specifically, unification is initiated by the same general preference for surface open syllables as vowel addition. Although I claimed above that CVN syllables are not too marked a structure, it appears that there are nevertheless processes which attempt to convert all such forms into syllables of the optimal CV shape in many languages. Jakobson and Halle (1956) note that this is the only universal type and that in some languages every syllable is of this shape. 17 Martinet (1955) in­vestigated various diachronic tendencies which conspire to produce open syllables. For example, he notes that the development of French up to the 16th century exhibits simplification of geminates, loss of preconsonantal ~ internally, loss of final nasals after nasalization of preceding vowels, simplification of diphthongs or restructuring of the sort [o!] > [wE]. Thus,

1 , . . , d . l' on est eg1t1mement tente e voir a de faits connexes pour lesquelles on tendance ~ l'ouverture des syllabes.

toute une serie peut parler d'une (1955:326)

- 38-

Bell (1971) criticizes the typological approach to the study of syllable structures where syllables are treated as "disembodied units" and proposes instead such a processual approach. He states:

This convenient simplification which treats syllables as abstracted from segment strings, fails to capture the role of the syllable as a unit of organization of strings of segments. The view ofi syllable structure as consisting of the ways that segment strings can be organized into syllables is a prerequisite to our deeper understanding of it. (1971 :27)

Accepting that the notion of unification motivated by syllable structure constraints accounts for the derivation of half-nasal consonants which are underlying sequences has two important advantages. On the one hand, it explains why prenasalized consonants are more conunon than post­nasalized consonants. Since nasal codas are more conunon, particularly in languages with restricted coda inventories, it follows naturally that nasal plus oral sequences are unmarked medial clusters. 18 In many languages, this is the only type of consonant cluster which is permitted. This analysis also explains why half-nasal consonants most commonly do not occur initially or finally since their origin is in the juxtaposition of two underlying syllables.19 Additionally, this analysis accounts in ·a prin­cipled manner, for the high co-occurrence of half-nasal consonants and open syllable languages. Of course, we ultimately want to explain the open syllable nature of these languages. A possibly fruitful avenue of research here might be the actual timing systems of these languages. There appears to be a hierarchical relationship between syllable timing and stress timing in language so that, for example, children's speech is often first syllable timed as in much language disorder speech. Syllable timed languages typic­ally have simple syllable structures, i.e. CV is an optimal unit of timing whereas in stress timed languages other considerations exist. The question with regard to the co-occurrence of syllable timing and simple syllable struct~re is which, if either, is causal. The best we can do at present is to note this assoi;:.iation.

5. Conclusion

By way of summary, I have discussed the goals of a theory of phono­logy and phonetics and attempted to delimit constraints which must ulti­mately be placed on these explanatory goals, We examined in brief detail a phonetic problem involving the distributional frequency of occurrence of various speech sound types. Unlike the more purely phonetic explanations which have been proposed for similar cases, I attempted to demonstrate how phonological considerations such as the naturalness of various underlying units or organization must be taken into account in an explanation of phonetic fact. The title of this paper mirrors deliberately earlier papers by Wang (1972), Ohala (1974b), and ·others which stress the need for phonetic explanation in phonology. I argue that not only does phonetics explain phonology, but that phonological explanation has a role in a theory of phonetics as well. 20 The correct generalization is probably simply that phonetics and phonology are ultimately a single science.

•

•

Footnotes

1The research for this paper was in part supported by a Fulbright­Hays grant to the author at the University of Edinburgh, which support is gratefully acknowledged. I am also grateful to Gillian Brown, Mark Fein­stein, Robert Jeffers, Roger Lass, John Laver, Ilse Lehiste, Elizabeth T. Uldall, William S.-Y. Wang, and Arnold Zwicky for discussions and comments on various topics treated in this paper. None of the above should be held responsible for my weaving together of topics and facts, and, of course, all oversights and analytical flaws are my own responsibility. A shorter version of this present work was presented at the Annual Meeting of the Linguistic Society of America, December 1976.

2The explanation of non-events actually incorporates two types of phenomena: i) explaining isolated changes which are unattested and un­expected based on articulatory, acoustic, and auditory non-similarity, e.g. ~ > tsU, and ii) more importantly, explaining non-changes of the sort p, t, k, > p, t, k. One explanation of this latter type which comes to mind is the frequent stability of nasals in ludo-European, Semitic, Finno­Ugric, Bantu, etc. based on the distinct perceptual characteristics of nasals as a class although they are much confused among themselves (Ohala 1975:294). This stability is not universal, however; for example, see the Formosan reflexes of Proto-Austronesian *n (Dahl 1973:70-75)

3However, neither of these approaches attempts to deal explicitly with non-events as discussed in Footnote 2.

4The one exception to this generalization which comes to mind is cardinal vowel theory, the reference points of which are arbitrarily selec­ted, exactly determined, and of invariable quality, i.e., they are defined independently of linguistic form.

5Lindblom (1972:85) notes that perceptual contrast is not the only determinant involved in shaping phonological systems. Other facts of phonetic (cf. Stevens 1972), social, and historical nature must also be considered.

6Ferguson (1966) mentions only prenasalized voiced stops, but the classification of the other prenasalized consonants and postnasalized consonants is implicit in his treatment.

7Postal (1968:170;71) notes that physiological and perceptual in­vestigations will hopefully provide evidence for the assignment of marked and unmarked values in phonology.

8This fact should account for the greater tendency for vowels to nasalize after a simple nasal than before one, though tautosyllabicity is also an important factor.

9o. Fujimura (personal communication) suggested that the fact that there are muscles whose active function is to raise the velum whereas the

- 39 -

- 40 -

lowering of the velum is a muscularly passive activity may account for the greater frequency of nasal-oral sequences over oral-nasal sequences. How­ever, as I argue below, the ultimate explanation lies elsewhere in phonetic fact.

10The question as to whether some or all of these preconsonantal nasals are unspecified for point of· articulation is a moot question.

11The facts are actually somewhat more complex in some languages. For example, it is also necessary to recognize a syllable type N which occurs only initially in Runyankore: · fpora 'slowly'; Mbarara 'Mbarara', ndeeba 'see me!'. In many languages, alf-nasals also occur morpheme­lnitially, but there is evidence for an underlying cluster analysis of these units as well. This point is treated further below.

That CVN syllables do not occur word-finally in these same languages is perhaps a reflex of the fact that languages of ten allow different codas in syllable-final and absolute final position, though historical consider­ations also play a role since, for. example, all CVN sequences in Runyankore ultimately derive from CVNV with vowel loss which could not be accomplished in absolute final position.

12There is a great deal of evidence to suggest that [~] is the un­marked nasal in final or pre-pausal position.

13In some languages, e.g. Lithuanian, a post-vocalic sonorant con­sonant is functionally the same as the second element of a diphthong.

14Simple reference to the 'larger-than-segment' size of syllables will obviously not suffice, even on a mere descriptive level, as a defin­ition sin:e words, phrases, sentences, etc. are also larger-than-segment units of organization. The problem of definition is a very large one which is beyond the scope of this present paper.

15Diachronically, unification may be motivated by such other phono­logical processes as vowel loss, for example.

16That is, half-nasal consonants which are not derived processually or by a series generating component. For a complete survy of processes which give rise to half-nasal consonants, see Herbert (1977a, Chapter 7)

17Sommer (1968) has argued that Kunjen, an Australian language of North Queensland, has no CV syllables, but only types such as VC, VCC, vccc, vcccc.

18This fact may account for the rapid evolution of oral-nasal se­quences in languages where they develop. For example, postnasalized consonants in many languages of New Caledonia and the Loyalty Islands have become simple aspirates (pm>~), aspirates followed by nasal vowels (~ > ph-), fricatives (.I?!!!>_!) optionally followed by nasal vowels, or simply h followed by nasal vowels (.I?!!!> h-) (Haudricourt 1964, 1971). In many of the Kwa languages of West Africa, there is a general metathesis of oral-nasal sequences which arise from historical vowel deletion:

..

•

- 41 -

*CVNV > CNV > NCV

which accounts for the unexpected presence of vowel nasalization after prenasalized consonants in some dialects (Hyman 1972, Williamson 1973).

19It appears that unification occurs only across, not within, syl­lable boundaries. This is in keeping with the notion that unification is a function of processes which optimize, and therefore change, syllable structures. The evidence for this claim is found in Herbert (1977a).

20v. Fromkin (personal connnunication) has pointed out that the validity of this conclusion, based on the evidence which I have presented, is not conclusive since there is wide disagreement among linguists as to what distinguishes phonetic and phonological facts. While I agree with the utlimate impossibility of delimiting absolutely where phonology ends and phonetics begins, my analysis of underlying units of organization as phonological entities is based on a traditional di.stinction between sur­face and underlying syllables which are labelled phonetic and phonological syllables respectively (cf. Grammont 1950).

Of course there is ultimately a phonetic basis for the phonological patterning which is cited to explain another phonetic fact, viz. the dis­tributional frequency of certain sound types. Thus, the real generali­zation is that phonetic and phonological explanations are very closely interwoven; the question then is whether it will be possible or profitable to separate the two factors.

References

Anderson, Stephen R. 1974. The organization of phonology. New York: Academic Press.

Bell, Alan. 1971. Some patterns of occurrence and formation of syllable structures. Working Papers on Language UniversaZs 6:23-137.

Bjork, L. 1961. Velopharyngeal function in connected speech. A~ta RadioZogiaa Supplement 202.

Catford, J. C. 1968. The articulatory possibilities of man. ManuaZ of Phonetics, ed. by Bertril Malmberg, 309-333. Amsterdam: North­Holland Publishing Co.

Catford, J. c. 1977. FundamentaZ probZems in phonetics. Bloomington: Indiana University Press.

Chomsky, Noam and Morris Halle. 1968. The sound pattern of EngZish. New York: Harper & Row.

Dahl, Otto Chr. 1973. Proto-Austronesian. Lund: Scandinavian Institute of Asian Studies Monograph No. 15.

Dixon, R. M. W. 1976. by S. A. Wurm.

Review of Languages of Australia Language 52:260-66.

and Tasmania,

- 42-

Feinstein, Mark. 1977. The Linguistia natux-e of prenasaZization. Ph.D. dissertation, Graduate Center, City University of New York.

Ferguson, Charles A. 1966. Assumptions about nasals: a sample study in phonological universals. UniversaZs of Language, ed. by Joseph H. Greenberg, 53-60. Cambridge: MIT Press.

Fischer-J~rgensen, Eli. 1975. !'Pends in phonoZogiaaZ theo:roy. Copenhagen: Akademisk Forlag.

Grammont, Maurice• 1950. TPaitJ de phonJtique. Paris: Librairie Dela­grave.

, , Haudricourt, Andre G. 1964. Les consonnes postnasalisees en Nouvelle

Caledonie. PPOaeedings of the ninth intePnationaZ aongress of Zinguists, 460-61. The Hague: Mouton.

Haudricourt, Andr~ G. 1970. Consonnes nasales et demi-nasales dans !'evolution des systemes phonologiques. ·Aates du xe CongrJs IntePnationaZ des Linguistes, vol. 4, 105-108. Bucharest: l'Acad­emie de la Republique.

Haudricourt, Andre G. 1971. New Caledonia and the Loyalty Islands. Cux>rent trends in Linguistias, ed. by Thomas A. Sebeok, vol, 8, 359-396. The Hague: Mouton.

Herbert, Robert K. 1975. Reanalyzing prenasalized consonants. Studies in AfPiaan Zinguistias 6:105-23.

Herbert, Robert K. 1977a. Language UniversaZs, Markedness Theo:r>y, and Natura.Z Phonetia Proaesses. Unpublished Ph.D. dissertation, Ohio State University.

Herbert, Robert K. 1977b. Phonetic Analysis in Phonological Description: Prenasalized Consonants and Meinhof's Rule. Lingua 43:339-373.

Herbert, Robert K. 1977c. Principles of Segment Organization and Consonant Unification. Symposium on Segment Organization and the Syllable. University of Colorado, Boulder.

Hombert, Jean-Marie. 197-6. Tone Space and Universals of Tone Systems. Linguistic Society of America Annual Meeting, Philadelphia.

Hyman, Larry M. 1972. Nasals and Nasalization in Kwa. Studies in AfPiaan Linguistias 3:167-205.

Jakobson, Roman and Morris Halle. 1956. FundamentaZs of La.nguage. The Hague: Mouton.

Lass, Roger. 1972. How intrinsic is content? Markedness, sound change, and 'family universals'. Edinbux-gh Working Papers in Linguistias 1:42-67.

Lieberman, Philip. 1970. Towards a unified phonetic theory. Linguistia inqui:roy 1:307-22.

Lieberman, Philip. 1974. Physiological structuring principles and phon­etic theory. Linguistic Society of America Annual Meeting, New York.

"

..

•

•

- 43 -

Liljencrants, Johan and Bjor:n Lindblom. 1972. Numerical simulation of vowel quality systems; the role of perceptual contrast. Language. 48:838-62.

Lindblom, Bjorn. 1972. Phonetics and the description of language. Pro­ceedings of the Seventh International Congress of Phonetic Sciences, 63-97. The Hague: Mouton.

Martinet, Andr~. 1955. Economie des changements phon~tiques. Berne: A. Francke.

Maryott, Kenneth R. 1965. The phonology and morphophonemics of Tabukang Sangir. Philippine Social Science Review, 26:111-26.

Meillet, A. and M. Cohen. 1952. Les Zangues du monde. Paris: H. Champion.

Miller, Patricia. 1972. Some context-free processes affecting vowels. Ohio State University Working Papers in Linguistics 11:136-67.

Ohala, John. 1974a. Experimental historical phonology. Historical Linguistics II: theory and description in phonology, ed. by John M. Anderson and Charles Jones, 353-87. Amsterdam: North-Holland Publishing Co.

Ohala, John. 1974b. Phonetic explanation in phonology. Papers from the parasession on natural phonology, 251-274. Chicago: Chicago Linguistic Society.

Ohala, John. 1975. Phonetic explanations for nasal sound patterns. NasaZfest, ed. by c. A. Ferguson et al., 289-316. Stanford: Language Universals Project.

Okell, J. 1969. A reference grammar of coZZoquiaZ BUX1!11ese. Oxford: Oxford University Press.

Pike, Kenneth. 1943. Phonetics. Ann Arbor: University of Michigan Press.

Postal, Paul. 1968. Aspects of phonoZogicaZ theory. New York: Harper & Row.

Richardson, Irvine. 1957. Borderland, vol. 2.

Linguistic Survey of the Northern Bantu London: Oxford University Press.

Scheffler, Israel. 1963. The Anatomy of Inquiry. New York: Alfred A. Knopf.

Sommer, B. A. 1968. Kunjen phonology: synchronic and diachronic • Pacific Linguistics, Series B, No. 11.

Stampe, David. 1972. On the natural history of diphthongs. Papers from the Eighth Regional Meeting, 578-90. Chicago: Chicago Linguistic Society.

Stampe, David. 1973. A dissertation on natural phonology. Ph.D. dis­sertation, University of Chicago.

Stevens, Kenneth N. 1972. The quantal nature of speech: evidence from articulatory-acoustic data. Human aommunication: a unified view, ed. by E. David and P. Denes, 51-66. New York: McGraw-Hill Book Co.

- 44 -

Sung, Margaret M. Y. 1966. Phonetic and phonemic system of the Kana­kanavu language, Formosa. BuZZetin of the Institute of History and PhiZoZogy 36:787-800. Shanghai: Academica Sinica.

Vennemann, Theo. 1974. Words and syllables in natural generative grammar. Papers from the para.session on natuzoaZ phonology, 346-74. Chicago: Chicago Linguistic Society.

Wang, William S.-Y. 1972. Phonetic explanations in phonology. Second International Phonologie-Tagung, Vienna.

Williamson, Kay. 1973. More on nasals and nasalization in Kwa. Studies in African Zinguistias, 4:115-38

..

•