AUTHOR: B. Elan Dresher TITLE: The Contrastive Hierarchy in Phonology SERIES TITLE: Cambridge Studies in Linguistics 121 PUBLISHER: Cambridge University Press YEAR: 2009
Grover Hudson, Program in Linguistics, Michigan State University
According to the publisher's announcement of this book (Linguist List 20.3272) contrast 'is one of the most central concepts in linguistics'. The book 'presents an original account of the logic and history of contrast in phonology', 'provides empirical evidence from diverse phonological domains that only contrastive features are computed by the phonological component of grammar', and 'argues that the contrastive specifications of phonemes are governed by language-particular feature hierarchies'. It assigns 'a key role to abstract cognitive structures' (the feature hierarchies), and challenges 'contemporary approaches that favor phonetic explanations of phonological phenomena'. There are nine chapters as follows.
Ch. 1, 'Introduction'. Rather than 'phonetic contrast' as in phonemic theory, contrast here is 'pattern alignment' somewhat in the sense of Sapir 1925, what D (2; bare numbers in parentheses are pages of the book) terms 'the arrangement of the phonemes of a language, their place in the phonological system'. D reviews some highlights of the theory of contrast in phonology, with the conclusion that modern phonology should 'reconnect with its roots' and 'establish phonological contrast as a central principle of phonological theory' (8). Finally, chapter 1 presents the plan of the book by chapters.
Ch. 2, 'The logic of contrast'. There is a connection to phonemic analysis in the generative grammar sense: how the minimal feature specifications of phonemes (phonemic contrast by phonological features) are discovered. D's idea is that this should be done by 'the Successive Division Algorithm' (16), not by 'the Pairwise Algorithm' (14). Here it must suffice to say that the former involves sorting the phonemes by features in a succession of top-down splits such that feature content may differ depending on the ordering. For example, given /p, b, m/, sorting these first by [voiced] then [nasal] leaves /p/ unspecified for nasality, because /p/ was fully distinct as [-voiced]. Sorting first by [nasal] leaves /m/ unspecified for voice, because /m/ was fully distinct as [+nasal]. D's idea is that one or the other ordering will be preferred according to whether [voiced] or [nasal] is more 'active' in the phonology. The rest of the chapter is concerned to show that the pairwise method, making every phoneme pair contrast by at least one feature, doesn't work anyway.
Ch. 3, 'Contrast in structuralist phonology'. Work of Sapir, Trubetzkoy, Martinet, Jakobson and Lotz, and Hockett is surveyed to show that 'issues of phonological contrast were central to their thinking', despite 'absence of an explicit or consistent approach to assigning contrastive features' (71). According to Trubetzkoy (1969: 100), for example, Artshi consonant rounding is neutralized next to /u, o/, so these 'are placed in opposition with the remaining vowels' /i, e, a/, and must be specified as [+round] rather than [+back, -low], as seems otherwise possible and as in Latin (56-57). If /u, o/ are specified [+back, -low], then [round] is redundant and non-contrastive in them. D concludes from such examples the 'Contrastivist Hypothesis' (74: of Hall 2007: 20): 'The phonological component of a language L operates only on those features which are necessary to distinguish the phonemes of L from one another', such as [round] in Artshi. Such 'active' features are ordered above others (are higher) in the 'contrastive hierarchy', which renders the others less contrastive / more redundant. 'Active' characteristics of contrastive (non-redundant) features include a role in conditioning or undergoing alternations and neutralizations, co-occurrence relations, and the adaptation of sounds of other languages (72).
Ch. 4, 'The rise and fall of the contrastive hierarchy'. Jakobson's collaborator Halle, (especially Halle 1959) promoted a contrastive hierarchy significantly inspired by Jakobson's theory of ordered phonological-feature acquisition by children (Jakobson 1941), implemented in grammars as a tree diagram in which higher(-ranked) features are acquired earlier, for example [+vocalic] and [+nasal] in Russian, which tend to make lower-ranked [voiced] unspecified (91). Stanley's (1967) influential critique of Halle's approach showed how the unspecified (zero-)values could function as additional, third, values in what had been agreed to be a binary-feature system. D agrees (98), but finds this 'inconsistent with the Contrastivist Hypothesis', in which 'the phonology treats contrastive specifications differently from noncontrastive specifications'; that is, zeroes do function as third values. But while Halle's system assigned zero-voiced to the Russian coronal affricates and /x/, which lack contrastive voiced parallels, in D's system as triggers of voiceless assimilation in obstruent sequences these must be assigned 'active' [-voiced] (95).
Ch. 5, 'Generative phonology: contrast goes underground'. Chomsky and Halle (1968) discontinued Halle's feature hierarchies in favor of a universal but incomplete theory of context-sensitive feature markedness. As D (108) says: 'no well-formed branching tree can model these markedness conventions'. In the 1980s, theories of phonological 'activity' based on contrastiveness reappeared: in Kiparsky's (1985) Lexical Phonology, and the opposed 'underspecification' theories of Archangeli (1984) and Steriade (1995). Non-contrastive features are inert in Kiparsky's lexical stratum (e.g. typically voice in sonorants), but become active in the post-lexical (allophonic) stratum (where sonorants can assimilate voice). In 'Radical Underspecification' (Archangeli), only the marked value of a feature is lexical, and not even this if it is derivable, like [+voiced] typically in sonorants, but also all -F if +F is lexical. Feature markedness depends on language-specific feature activity, prominently evidenced by the minimally specified, inactive, vowel. In 'Contrastive Underspecification' (Steriade), both values of a contrastive feature are lexical, e.g. typically [voiced] in obstruents. Non-contrastive (redundant) features are zeroed, e.g. [-high] in /a/ given contrastive [+low]. Unspecified for [high], /a/ may be neutral toward height harmony, but is zeroed as logically redundant (127). In 'Feature Geometry' (cf. Clements and Hume 1995), feature hierarchies are implemented in feature trees in which, in assimilations, specified features spread to nodes unspecified for the features.
Ch. 6, 'Contrast in Optimality Theory'. D surveys and exemplifies Optimality Theory (OT) constraints which optimize language-specific feature hierarchies (in which noncontrastive specifications are absent). These are faithfulness constraints preserving particular features e.g. Max [sonorant], and markedness constraints which disprefer feature combinations e.g. *[round, +low]). Lacking in OT is a constraint type or theory of constraints which might optimize D's theory (161) that the Contrastivist Hypothesis is 'implemented by the Successive Division Algorithm operating on a contrastive feature hierarchy', and 'the phonological component of a language L operates only on' the contrastive and specified features established by the hierarchy.
Ch. 7, 'Evidence for the contrastive hierarchy in phonology'. Five sorts of data for D's theory are examined, concerning vowel harmony, metaphony (regressive vowel-to-vowel assimilation), consonant co-occurrence restrictions, loanword adaptation, and child acquisition. An example of each is mentioned here. Classical Manchu vowel harmony is said to reflect its contrastive vowel hierarchy according to which vowels participate unexpectedly in assimilations: for example open-/o/ triggers labial harmony but /u/, being contrastively back and unspecified for labial, doesn't; schwa triggers [+ATR] harmony (against [-ATR] /a/) but ATR-noncontrastive /i/ doesn't (177). Pasiego Montañes metaphony is an argument for different contrastive hierarchies in different 'domains' within a language -- here, stem vs. affix vowels (192). Bumo Izon consonant co-occurrence restricts words to either plosive or implosive, except for the plosive and implosive velars, which co-occur with both sets, a result D explains by the implosive velar being labio-velar so arguably unspecified for the implosion (glottalic) feature. Hawaiian and Maori adapt English coronal obstruents differently (Hawaiian as /k/ and Maori as /h/), explained by absence of glottal stop in the latter, with the result that /k/ is the default obstruent in Hawaiian and /h/ in Maori (201). Since Jakobson (1941), phoneme acquisition order has been explained by a feature hierarchy, famously unmarked [a,i] before [u]. Concerning acquisition, against Pinker's (1994: 265) claim that learners have phonemic behavior 'before they produce or understand words', D (205) finds it 'hard to see how infants can acquire phonemes without knowing if two utterances are same or different' (contrast).
Ch. 8, 'Other approaches to contrast in phonology'. Several 'other approaches', all 'advanced in the recent phonological literature', are found wanting to the extent they diverge from D's theory. Included are Dispersion Theory (Padgett 2003), Structured Specification (Frisch, Pierrehumbert and Broe 2004), Clements' (2009) argument for universally determined feature markedness, and appeals to 'minimal contrast ... in the work of a number of contemporary phonologists', which 'arises from the concept of pairwise comparison, which was shown to have fatal flaws in chapter 2'.
Ch. 9, 'Conclusion'. This 2-page chapter reviews main points to conclude that 'the contrastive hierarchy' is a 'pivotal principle of linguistic structure', and 'the Contrastivist Hypothesis is a fruitful approach to phonological theory'.
Especially given D's thorough demonstration of the long history in phonology of deferring the matter, it does seem important, now, to seek principles of contrast in phonological feature analysis. Agreement on the evidence, however, and even on the data, seems sure to prove difficult in the absence of agreement on the feature set, and with D's acceptance of 'abstract' analyses employing unphonetic lexical form. D accepts and employs privative and binary features, some of both, and actually ternary features since he admits zero as a third value in addition to plus and minus; he uses both ATR and RTR; front, coronal, and back; peripheral, vibrant, etc. Schwa is termed non-low in Inuit (165) and low in Classical Manchu (176). Concerning Inupiaq (165), Nez Perce (185), and Yawlumne (=Yawelmani, 208), critical data involve (underlying) contrasts never heard in the languages ('absolute neutralization').
The theory seems to require every language to offer learners decisive evidence for its contrastive hierarchy: phonological 'activity' evidenced consistently if variously (72), contra Hockett (1955: 174, quoted p. 57): 'a given feature or difference turns up in some contexts as of primary relevance, in other contexts as subsidiary'. Regarding the problem of 'abstractness', the Inuit/Inupiaq data (164-7) may be instructive. There are dialects having 'strong' and 'weak' /i/. Strong /i/ 'palatalizes' alveolars and weak /i/ doesn't; the latter, said to be underlying historical schwa, also 'undergoes a variety of assimilation and deletion processes'. Most dialects, however, no longer distinguish 'the two kinds of /i/', and in these /i/ does not palatalize. D's interpretation is that in the latter dialects /u/ is labial and /a/ is low, so /i/ necessarily lacks 'some contrastive feature that triggers palatalization' (165, footnote).
There is another interpretation: in the palatalizing dialects instead of underlying schwa there is in addition to ordinary /i/ an exceptional, palatalizing, and lexically encoded /i/ (encoded in the manner of /f/s in English which alternate with /v/s as in life/lives). This is indeed the marked vowel, so the alternation has naturally been levelled in the majority dialects in favor of unmarked (probably more frequent) ordinary /i/. Necessarily the result is no palatalization. D's analysis works, but cannot explain why the palatalization-alternation was levelled, except by invoking some principle different from that (preference for one-form, one-meaning) which elsewhere and ordinarily explains levelling (as of the English f~v alternation). (Then there is the problem, mentioned above, of the palatalizing feature -- presumably not [coronal], which is said to be contrastive for 'active' /i/.)
D argues persuasively from a variety of data and with insightful reference to a long history and a breadth of theoretical perspectives. Helpful (if in a longer book) would be data from a single language across the range of phenomena thought to provide consistent evidence of phonological 'activity' supporting a contrastive hierarchy. The most exemplified language of the book, Classical Manchu, unfortunately lacks speakers and acquirers who could provide such data. (For analyses of this language differing from that presented by D, see Ard 1984.)
REFERENCES Archangeli, Diana (1984). Underspecification in Yawelmani phonology and morphology. PhD dissertation, Cambridge: MIT (published 1988, New York: Garland Press).
Ard, Josh (1984). Vowel harmony in Manchu: a critical overview. Journal of Linguistics 20.57-80.
Chomsky, Noam, and Morris Halle (1968). The sound pattern of English. New York: Harper & Row.
Clements, G. N. (2009). The role of features in phonological inventories, Contemporary views on architecture and representations in phonological theory, Eric Raimy and Charles Cairns, eds., 19-68. Cambridge: MIT Press.
Clements, G. N, and Elizabeth Hume (1995). The internal organization of speech sounds, Handbook of phonological theory, John Goldsmith, ed., 245-306. Oxford: Blackwell.
Frisch, Stephen, Janet Pierrehumbert, and Michael Broe (2004). Similarity avoidance and the OCP. Natural Language and Linguistic Theory 22.179-228.
Hall, Daniel C. (2007). The role and representation of contrast in phonological theory, PhD dissertation, Toronto: University of Toronto.
Halle, Morris (1959). The sound pattern of Russian. The sound pattern of Russian: a statistical and acoustic investigation. The Hague: Mouton.
Hockett, Charles (1955). A manual of phonology (Indiana University Publications in Linguistics 11). Baltimore: Waverly Press.
Jakobson, Roman (1941). Kindersprache, Aphasie, und allgemeine Lautgesetze. Uppsala: Uppsala Universitets Arsskrift (translated by A. R. Keilor, 1968. Child Language, Aphasia, and Linguistic Universals. The Hague: Mouton).
Kiparsky, Paul. 1985. Some consequences of Lexical Phonology, Phonology Yearbook 2: 85-138.
Padgett, Jaye (2003). Contrast and post-velar fronting in Russian. Natural Language and Linguistic Theory 21.39-87.
Pinker, Steven (1994). The language instinct. New York: William Morrow.
Sapir, Edward (1925). Sound patterns in language. Language 1.37-51 (also 1957, Readings in Linguistics I, Martin Joos, ed., 19-25. Washington: Council of Learned Societies).
Stanley, Richard (1967). Redundancy rules in phonology, Language 43.393-436.
Steriade, Donca (1995). Underspecification and markedness, Handbook of phonological theory, John Goldsmith, ed., 114-174. Oxford: Blackwell.
Trubetzkoy, N. S. (1939). Grundzüge der Phonologie. Göttingen: Vandenhoeck & Ruprecht (translated by Christiane Baltaxe, 1969. Principles of phonology. Berkeley: University of California Press).
Underhill, Robert (1976). The case for an abstract segment in Greenlandic. International Journal of American Linguistics 42.349-358.
ABOUT THE REVIEWER
ABOUT THE REVIEWER:
Grover Hudson taught phonology, historical linguistics, and Ethiopian
linguistics including Amharic language at Michigan State University. He is
author of a comparative dictionary of Highland East Cushitic languages, an
introductory linguistics textbook, with Anbessa Teferra a recent book on
Amharic, and articles on phonology and Ethiopian descriptive and historical