Review Ladefoged, P. (1971 ). Preliminaries to Linguistic Phonetics. Chicago: University of Chicago Press P. Mansell Institute for Phonetics and Speech Communication, University of Munich, West Germany Received 31st August 1972
Introduction
The work under review is a revision of the widely read and influential Ladefoged (1967). As such, the major interest must be in the nature of the revision. Two criteria can be raised. First, the specialist will want to see that recent developments have been to some measure incorporated ; secondly, it must also be recognized that full publication implies a wider readership than simply specialists, a wider readership for whom a relatively self-sufficient account of the field must be provided. On the first count, as will be seen below, only minor reservations can be entered ; indeed, since Ladefoged remains pre-eminent in the field of the relationship between phonetics and phonology, there can be no question of major omissions here. On the second point, however, the success that Ladefoged at first sight achieves, in that he has provided an eminently readable and well-presented text, can be shown to be indeed only superficial. For the major revision that has been undertaken is the removal of the central section of Ladefoged (1967, henceforth LP), Preliminaries to a Theory of Phonetics, from the present text (henceforth Prelims) . This has had the side effect of allowing the combination of the section on the phonetic data and that on the feature system to account for it, thus making once again for greater readability. Yet, it will be argued , the lack of a central statement of a theory of phonetic descriptions renders the intention of the book obscure. The addition of "preliminaries" to the title cannot cover this lack of theoretical orientation, since the book presents at once phonetic data and phonetic feature system. The data is preliminary to the feature system, of course, but so, surely, is the theory. The elaboration of this argument will occupy the first section of this review. More substantive comments, on the phonetic data presented and subsequently on the feature system itself, will follow. Consequences of Lack of Theory
A theory of phonetic representation is required for two reasons. First, since any system of phonetic representation is a limitation on phonetic data, this limitation must be justified. Secondly, we presumably wish to compare systems of phonetic representation, and comparison is only possible given a set of theoretical tenets. Both limitation and evaluation are undertaken in this book [a further revision of LP is the introduction of a final chapter, comparing the feature system set up with that of Chomsky & Halle (1968)]. Hence a theory of representations is required. What should this theory consist of? It seems to the reviewer that the arguments given in LP remain substantially correct. The following principles can be adduced:
94
P. Mansell
(1) The features employed should be sufficient for the statement of lexical contrasts in languages. (2) The features should be adequate to the comparison of languages at a phonetic level. (3) A feature system should not be required to specify sounds which occur only as "coarticulated" (Prelims) or "intrinsic" (LP) allophones. (4) The combination of features should make sense in some physical way. (5) The features and their combinations should be such that appropriate natural classes are provided for the operation of a phonology. We could, of course, add riders to each of these principles. For example, the question of intrinsic and extrinsic allophones, intricately bound up in LP with the notion of articulatory target, remains full of obscurities, as a host of experimental work could show. But despite such riders, the above surely remain the sort of constraints we would wish to impose upon a set of phonetic representations. Indeed, the most telling evidence for the persistence of these constraints is the fact that they can be observed in operation during the practice of this book. Thus, Constraint 2 is in operation when Chomsky & Halle (1968, henceforth SPE) are criticized (p. 98) for failing to provide differentiation between two [b ]s, one in Hindi, one in Margi. Constraint 3 is in operation when labiodental stops are labelled coarticulated allophones (p. 47), or when the "degree of glottal stricture" which distinguishes the [h]s in "that hat" and "my hat" (p. 19) is similarly treated, as being "predictable from the glottal strictures in the adjacent sounds". (One might ask at this point whether, without fuller discussion of the notion of coarticulation here, the reader is provided with the ability to distinguish between this form of predictability and that phonological form of predictability which, say, assimilation rules formulate.) Constraint 4 is in operation when the phonetic unreality of "delayed release" applied to continuants is criticized (p. 106). The use of Constraint 5 will form part of the discussion below. It might be supposed from the above that the theoretical apparatus of this work could be inferred from the practice-itself hardly a recommendation. But matters are made worse by the fact that what theoretical points are made in the introductory chapter serve in the main to create confusion. We may begin with a simple error, since it is pertinent to the discussion. Well on in the text (p. 45) we read that: "In the first chapter we saw that a phonetic theory must be capable of distinguishing not only all the contrasts that occur within a language, but also all the phonetic events that characterize one language as different from another." In fact, there is no such discussion in the first chapter of what is in effect Constraint 2 above. Instead the reader is first told that (p. 1) : "This book is about some of the phonetic events that occur in the languages of the world. The data described consist mainly of contrasts observable at the systematic phonetic level in a wide variety of languages. In many cases I have little or no knowledge of the underlying relations among the sounds in the languages being used to illustrate a particular point. Consequently I in no way wish to imply that the features needed for describing the sounds used in these contrasts are necessarily the same as those needed for specifying contrasts among underlying forms." One thing is clear here: Ladefoged is not concerned with the linguistic status of the contrasts to be described, whether they be "phonemic" or "phonetic" in the terminology of
"Preliminaries to Linguistic Phonetics"
95
Schane (1971). What is not clear, however, and what remains unclear, is how Ladefoged would characterize "systematic phonetic level". This term is, after all, in current usage and in the main associated with SPE. Ladefoged at one point (p. 95) quotes from SPE with a view to showing the inconsistencies exhibited there in the definition of such a level, but appears at no point to offer an alternative definition. Instead, he tries to show (p. 4) that precise characterization of this level (and also, incidentally, and paradoxically, the use of Constraint 4) is irrelevant: "For the purposes of this book it does not matter whether we follow Chomsky and Halle (1968) and try to characterize what a speaker knows about the phonetic facts of his language, or whether ... there should be more constraints on a phonetic description, so that a greater part of it is experimentally observable." Now, it is not clear why this distinction is irrelevant; since a particular level of description has been invoked, why should it not therefore be defined? What is more important than this general point is that Ladefoged's treatment of stress levels in English shows that the distinction mentioned above is far from irrelevant for his practice in the book. If, as is stated (p. 83), the issue with SPE is not the stress rules, but "simply the number of stress levels which occur in English at the systematic phonetic level", and if, as it turns out (p. 84) the number of stress levels recognized in the two systems is not the same, then Ladefoged's systematic phonetics and Chomsky and Halle's are not the same, nor is the distinction between them irrelevant. There is no question, of course, but that using the constraints enumerated above, the notion "systematic phonetics" in Ladefoged's sense could be defined. The constraints remain unexpressed, however, and the level consequently undefined. Proceeding further in the introduction, we may note that although the reader has been prepared for a study of contrasts at the phonetic level, and despite the fa_ct that he is warned that(p.1): · " ... there is no clear evidence showing that the set of features required for specifying phonetic contrasts is the same as that required for specifying the natural classes of sounds required in phonological rules." yet he later reads that: (p. 4): "Accounting for systematic phonetic contrasts is in itself a trivial and uninteresting task that can be done in many ways. It becomes interesting only when we try to constrain our account so that it fits in with the division of sounds into the natural classes required in phonological rules." In other words, Constraint 5. The careful reader might well ask how notions (phonological rules), which are by the author's own admission (p. 1) largely unexplored, can be utilized to establish and validate a phonetic representation, as they are indeed in the text which follows, when it is not clear that the phonetic representation is the same as that employed in phonological processes. Clearly, of course, Ladefoged uses only well-known phonological processes to aid in the establishment of his feature system (lenition, vowel shift, the interaction of labial consonants and vowels, alternations between labial and velar consonants) so that the informed reader finds no contradiction. The question is, however, why, when so much else is left unexplained, the caveat against the identification of phonological and phonetic features should be included. There may well be important reasons for keeping this distinctio~ [see (i) and (iv)below]; but in the text itfts not employed, and when
96
P. Mansell
so much is employed that is not explicitly introduced, its inclusion can only create confusion. To conclude this section, it must' be clearly seen that the complaint here is against the presentation of ideas and aims in this book. For all of the theoretical notions which have been employed in this section are in fact readily accessible in LP or in other works of Ladefoged's. The reader who has not been able to infer the theoretical framework of the book from its practice, or who has been confused by the theoretical statements which do appear will find difficulty in appreciating the major grounds for disagreement with the SPE feature system in Chapter 10, and may well be forced to conclude that the differences are only minor. What is important and should be made important to the reader explicitly is the approach taken in the two feature systems; it is the statement of the principles on which Ladefoged's approach is based that suffers in the revision. Phonetic Description
The unique value of this book must lie in the variety of phonetic facts presented and the way in which observations are in the main backed by reference to experimental findings. Further, practically every chapter here includes additions toLP. Notably, there is a wholesale re-organization of the chapter on secondary articulations, including (p. 65) a very useful summary of "all the actions of the lips that we have to take into account within a feature system". There is also a large addition to the chapter on vowels, as well as a great improvement in the coverage of prosodic features. The reviewer is incompetent to comment in detail on the mass of phonetic observations, and can only record that when Ladefoged's themes are those of current experimental controversies (in the main when dealing with SPE, e.g. pharynx enlargement in voiced stops, p. 96/7) the text is informative and reasonable. Further, Ladefoged's predilection (p. 2) for basing his text on personal observation leads to a refreshing absence of the conventional survey of previous opinions at every turn. Finally, also refreshing is the willingness to concede that "further research is needed" as in the case of fricatives (p. 49) and the relative merits of rival tongue behaviour features (p. I 04) The above remarks hold for the majority of the text. There are certain exceptions, however. Thus, when describing the action of the velum (p. 34) Ladefoged ceases to be a sure guide to current research. He refers approvingly to Moll & Shriner's (1967) model of the velum, which suggests a single velic "target" modified by coarticulation effects. What Ladefoged does not report is the extremely reliable research of Lubker (1968) and Fritzen (1969) which purports to show that this model is incorrect. The reader is thus given a false impression of the current state of research. Further, although, as stated above, references to the literature are not numerous, there are some occasions (e.g. p. 55 on the experimental justification of the feature "strength") on which untraceable references are apparently gratuitously introduced with no communicative function at all. Phonetic Representation
The major interest of the present version of the system of phonetic representation lies in two points: (i) that the phonological facts appear to call for the use of acoustic or auditory features in addition to articulatory features. In this way bilabial and velar stops can be linked by the feature of "gravity", and necessary distinctions among fricatives can be made with the feature "sibilance".
"Preliminaries to Linguistic Phonetics"
97
It would appear that certain, presumably acoustic, features of sounds have achieved a universal significance in phonological terms. The question which should surely interest Ladefoged, given his initial warning against the identification of phonological and phonetic features mentioned earlier in this review, is whether such features have a place at the output of a phonology, in other words at the systematic phonetic level. The answer depends in part upon our interpretation of the relationship between a systematic phonetic level and articulation itself. Ladefoged evidently does not view features as representing necessarily completely independent components of the articulation process [seep. 31 and (iv) below]; yet surely the features of "gravity" and "sibilance" are at the other extreme. At the output of a phonology they are nothing but completely dependent: there is nothing one can do beyond perform the articulation specified by other features to make [b] and [g] "grave" and [s] "sibilant". Further, at the output of a phonology, if acoustic features are to be included, what mechanism blocks the introduction of other possible acoustic generalizations, ones which might not have phonological significance? (ii) Features both at the phonetic level and in the phonology itself can be multi-valued. This can also hold true of lexical representation, but (p. 90):
"Note that, without including the tone features, twenty features are used as binary oppositions at the systematic phonemic level. Of the remaining six features, two (glottalicness , and rate) are either binary or not used at all in all but a handful of languages; and two others (glottal stricture and voice onset) are used as binary oppositions in the majority of languages. There seems to be no doubt that . .. the binary principle is a major factor in human communication." Within multi valued features there are two types, one, called "scalar", has values symbolizing "more or less of the same", while the other type is called a "multivalued independent feature". In the latter case, for example, places of articulation might form an unordered set within a multivalued independent feature, with each of the members of the set being regarded as equal distances apart. The prime phonological example of the need for a scalar feature is shift of vowel height (see, e.g., p. 103). Place of articulation, on the other hand, offers, as Ladefoged notes (pp. 43/44) an interesting problem. There are phonological processes involving adjacent places of articulation which favour a scalar interpretation of this feature; while there are also processes involving alternations between non-adjacent places of articulation which could be better expressed if the feature were seen as multi valued independent. Of the latter type of process Ladefoged discusses only alternations between labials and velars, and proceeds to introduce (p. 44) the feature of gravity to account for them. In this way he leaves the impression that the balance of the evidence is in favour of a scalar interpretation. One could note, however, that there are cases such as that of the consonant in the English prefix feed-/ where a segment is prone to assimilate to the place of articulation of most following consonants. Under a multivalued independent interpretation of the place feature the behaviour of the segment could be expressed in a single generalization; this would not be possible under a scalar interpretation. Further, it would seem that as yet there has been no debate as to whether a rule such as: [o: place] ->-[o:- 1 place]/-- [o:- 1 place]
is an illegitimate generalization for use with numbered independent features, to express, say, a fronting or backing process involving a number of adjacent places of articulation. The following minor points can also be made:
98
P. Mansell
(iii) There may be dependency relations between features. There are, for example, "constraints on the co-occurrence of some of the values of the glottal stricture and the voice onset features" (p. 31). In other cases features may be dependent in one language and independent in another. It may be suggested that a valuable addition to this work would have been a discussion of how the concepts "independence and dependence of features" illuminated or were illuminated by the concept of "markedness". It would seem, for example, that where dependence can be shown to be the result of physiological factors, then phonetics illuminates marking conventions. But where, despite evident phonetic independence, features show dependence in a large number oflanguages (for example, in the case ofbackness and rounding for vowels, p. 71) then it would seem that phonological conventions, and whatever explanatory concept lies behind them in such cases, should illuminate the phonetic facts. (iv) This comment concerns the feature "rate" first introduced on p. 56. This is a scalar feature. The basic question is, whether the various uses to which it is put constitute anything like a scale. It is true that Ladefoged notes on p. 31 that: " ... we will find that we cannot always enforce the constraint that sounds differing in terms of a feature should differ only in degree, and that the differences should be quantifiable in terms of a single measurable parameter .. ." but the conditions under which the constraint can be legitimately relaxed are not given. Here the feature is generalized to cover not only the process "stop becomes flap" and the process "trill becomes tap", but also contrasts between long and short consonants and in vowel length. Perhaps the difficulty here is only nominal; I cannot conceive of a definition of "articulatory rate" which covers both flapped articulations and long vowels. On the other hand, were the feature labelled "length" perhaps the generalization would be easier to accept. There is a further complication, one which again perhaps favours a "length" interpretation. For the feature must in some way be limited to the time domain, since otherwise it offers itself as a mediator of place assimilation both extrinsic and intrinsic. And as a final point here, any feature operating in the ,time domain which does not limit itself to simply stating before and after relations, as in Prelims "prenasality" and SPE "delayed release", has the problem of how duration should be measured. One answer is Ladefoged's "some arbitrary percentage longer than normal"; but normality, as Ladefoged implies (p. 82) is relative to overall rate of articulation. (v) Finally, Ladefoged (p. 48) notes that a palatographic examination showed [s] and [J] to be in some people's speech articulated in essentially the same place, but with a different tongue shape. It is not of concern here whether this is evidence of the need for a feature differentiating grooved and slit fricatives (p. 49). What is of interest is that in the phonology of the place feature in English [J] is clearly not alveolar. This may be further evidence for distinguishing between features within a phonology and features at the output of a phonology. Concluding remarks In that the present book covers a broad spectrum, from experimental phonetics through to phonological processes, it is to be recommended over the more conventional handbooks on which it is in part (p. 1) modelled, and as such welcomed. Where direct criticism has been offered, as in the first section of this review, the purpose has been to show the
"Preliminaries to Linguistic Phonetics"
99
theory in which Ladefoged's practice is based is incompletely or confusingly presented. Some caution is therefore required in the book's use. References Chomsky, N . & Halle, M. (1968). The Sound Pattern of English. New York : Harper & Row. Fritzell, B. (1969). The velopharyngeal muscles in speech: an electromyographic and cinefluorographic study. Acta Oto-laryngologica Suppl. 250. Ladefoged, P . (1967). Linguistic phonetics. UCLA Working Papers in Phonetics 7. Lubker, J. (1968) . An electromyographic-cinefluorographic investigation of velar function during normal speech production. Cleft Palate JournalS, 1-18. Moll, K. L. & Shriner, T . H . (1967). Preliminary investigation of a new concept of velar activity during speech . Cleft Palate Journal4, 58-69. Schane, S. A. (1971) . The phoneme revisited . Language 47, 503-521.