This book "asserts that the origin and spread of languages must be examined primarily through the time-tested techniques of linguistic analysis, rather than those of evolutionary biology" and "defends traditional practices in historical linguistics while remaining open to new techniques, including computational methods" and "will appeal to readers interested in world history and world geography."
Review of The Internal Organization of Phonological Segments
Date: Fri, 9 Sep 2005 15:15:17 -0500 From: Mike Cahill Subject: The Internal Organization of Phonological Segments
EDITORS: van Oostendorp, Marc; van de Weijer, Jeroen TITLE: The Internal Organization of Phonological Segments SERIES: Studies in Generative Grammar 77 PUBLISHER: Mouton de Gruyter YEAR: 2005
Mike Cahill, SIL International
This volume is a collection of eleven papers selected from presentations at the first Old-World Conference on Phonology held in Leiden in January 2003. As the editors note, the rise of Optimality Theory (OT) in the 1990s had the most benefit for subdisciplines of phonology concerned with conflicts. OT excels in matters of alternations and variation, but ignores matters of representation. The papers in this volume are an explicit attempt to complement OT in the area of representations.
After a valuable introductory chapter in which the editors summarize the contributions, the papers are grouped into three areas: five general papers on Feature Geometry, and three papers each on specifics of Nasality and Laryngeal Features in various languages. I will summarize and comment on each paper, than give an evaluation of the book as a whole.
PART 1, on Features and feature geometry, begins with ''Optimal Geometries'' by Christian Uffman. OT does not explicitly deal with the question of representations, but Uffman proposes that representations still have a positive role to play, even within OT. In particular, Feature Geometry (FG) can act as a representational filter on GEN, the generator function of OT. He first points out that there is no generally accepted theory of segment interaction within OT (even the most common phonological processes of assimilation do not have an agreed-upon analysis), and asserts that representations are compatible with OT, unlike some other researchers' proposals, such as strict locality proposals. Far from complicating the phonology, representations can simplify OT by inherently filtering out impossible configurations and thus simplifying the set of constraints that need be considered. He introduces five general constraints that refer to FG structure, relating them to non-structural constraints already in the literature and showing the economy of constraints that results. One result of this approach is that Uffman can do away with constraint proposals that explicitly enforce assimilation (e.g. AGREE) in favor of the interaction of the more basic markedness and faithfulness constraints. For example, for voicing assimilation, he invokes a constraint *[+/- voice], penalizing every separate [voice] autosegment, and incorporates structure by the constraint FAITH (Onset). Uffman also illustrates how the theory would work with long-distance consonant voicing in Kera, vowel epenthesis in loanwords, even in languages with multiple strategies, and in the long- distance voicing dissimilation process known as Dahl's Law. He ends with a proposal that other phenomena such as vowel harmony, traditionally problematic in OT, would be handled more readily by incorporating FG structure into OT.
The next paper is ''Variability in feature affiliations through violable constraints: the case of [lateral]'' by Moira Yip. Her position on the internal organization of speech segments is that they have no internal organization. She argues that [lateral] has been shown to be a dependent of different nodes in FG in different languages, and that this argues against a universal configuration of FG. Yip argues that OT handles these phenomena easily by rankable co-occurrence constraints such as *LateralDorsal. She thereby avoids what she terms the ''excessive rigidity'' of a fixed FG. Yip's constraints generate the inventories found in languages: laterals' preference to be sonorants and the common place being coronal. She uses these constraints to give an an account for [Nl] -> [ll] in Selayarese but [nl] in Chukchi, as well as patterns from Sanskrit, Polish, Flemish, Yanggu Chinese, Tahltan, and Toba Batak. Yip goes beyond Padgett (2000) who replaces FG with feature classes. Yip maintains that in an OT approach, not even feature classes are necessary. Are there then any representational restrictions at the GEN level of OT? The answer is unclear. Two typological predictions that she notes are unattested do weaken the force of the paper: that labial laterals exist, and that assimilation should create velar dorsals. Also, it is surprising that Yip spends so little time and cites such ambiguous evidence for [lateral]'s existence, considering the detailed arguments raised against its existence by Walsh (1997) and others, who deal with some of the same data Yip cites. If [lateral] does not exist, then this might explain why other researchers attempting to find an invariant position for it in FG have come to conflicting conclusions.
The next paper is ''The geometry of harmony'' by Don Salting. Salting supports FG within OT, rather than an unrestrained system of featural co- occurrence constraints. He proposes an entirely non-SPE system of height features, more abstract rather than purely phonetically based, which are connected to geometry (his ''Nested Subregister Theory'') to account for vowel harmony. Vowels are assigned features according to their systemic properties, so /e, o/, for example, do not always have the same featural specifications from one language to another. In this proposal, an Aperture node has two daughters, and each of these also has two, making four height categories (degrees of openness), somewhat reminiscent of tone feature proposals by Yip and others. He demonstrates five phonological strategies languages use when there is a gap in the inventory - transparency, opacity, new segment, ''clean-up'', and epenthesis. He glosses over some details and differences between languages in his OT analysis by using a general constraint HARMONY in judging whether the harmony has occurred, and a longer version of this article would profit from an unpacking of the HARMONY constraint. Salting extends his Nested Subregister theory to vowel place as well, with four degrees of backness, phonologically defined for each language. One consequence of this approach is that the opacity of /a/ comes as a result not of a [low] feature, but its marked Place.
The fourth paper of this section is ''Piro affricates: Phonological edge effects and phonetic anti-edge effects'' by Yen-Hwei Lin. In this well- documented contribution examining Piro (Aawakan, Peru), the basic question is whether affricates have a contour representation of [cont] or not, though the focus is not on the specific representation within FG. The author cites past analyses of affricates as phonological contour segments and as phonological stops. Following other researchers, he treats alveolar and palatoalveolar affricates [ts, tʃ] as strident, and the palatal affricate [tç] as non-strident. Using a derivational OT approach (e.g. Kiparsky 2000), he proposes that affricates are strident stops at the lexical level, but at the postlexical level, they are contour segments with ordered [-cont], [+cont] features. The relevant co-occurrence constraints are ranked differently at the lexical and post-lexical levels. The most important conclusion is the basic nature of affricates as stops, though not the strongest version of the Stop hypothesis. The importance of a derivational model of OT is demonstrated, as is the importance of the feature [strident] in the analysis of affricates.
The fifth and final paper of this section is ''On the internal and external organization of sign language segments: some modality-specific properties'' by Els van der Kooij and Harry van der Hulst. The question ''What is a segment in a sign language (SL)?'' is the central point examined here, and the authors, using data from Nederlandse Gebarentaal (a Dutch SL), propose that a sign is a single segment. The feature tree of a sign has daughters of Articulator and Location, and under Articulator is Handshape and Orientation. Under Handshape, Orientation, and Location, there will be one or two features. (The exact identity of these features is not generally agreed upon. They omit an explicit Movement feature, seeing this as redundant once the others are specified.) But since a sign (=segment) has internal temporal structure, this segment can be viewed as dominating a syllabic level. If this is so, then syllables in SLs are not suprasegmentals, but are segment-internal, a reversal of the dominance relationship in spoken languages. An alternative view, that of a sign being a complex segment like an affricate or prenasalized stop, is discussed as a reasonable alternative, especially given the ''intuitive oddness'' of the intrasegmental syllable proposal. Their conclusion is that these alternatives may turn out to be mere terminological variants: the real point is the specific structure they propose. In a survey of signs they see that linear order of features is largely predictable and so needs no lexical representation, with a few exceptions.
Part 2 has papers on Nasality, and begins with ''On the ambiguous segmental status of nasals in homorganic NC sequences'' by Laura Downing. NC sequences labeled as prenasalized stops have no consistent phonetic differences from those analyzed as sequences, and no language has an unambiguous contrast between prenasalized segments and NC clusters, so phonological evidence must be the determiner. In Bantu languages, a long vowel preceding NC is generally taken as evidence that the input mora of the nasal has been re-associated to the preceding vowel. The author proposes a different solution, that the N and preceding vowel share the mora, and N is ambisyllabic, linked to both syllables. She supports this by showing that in a variety of languages, pre-NC lengthening can occur without resyllabification of the N. (Reasons for the lengthening could well include enhancement of the prominence of the vowel relative to the nasal without invoking syllables at all.) She offers cross-Bantu evidence from reduplication and tone assignment that N is syllabified as a coda. A fairly detailed OT analysis is offered with constraints to account for the association of moras and segments. She then evaluates evidence that has been previously cited for NC as single segments (Bantu has only open syllables, word-initial NC, extragrammatical tests such as language games and pauses) and shows that these have less force and value than sometimes supposed. The conclusion is that Bantu languages, at least, have no unit prenasalized stops, but that cases of NC are clusters, with a particular geometry.
The next paper is ''Areal and phonotactic distribution of N'' by Gregory D.S. Anderson (''N'' representing the velar nasal here). The author has a database of 512 genetically balanced languages, and shows that while /N/ is common in Australian languages, Southeast Asia, and a band across Africa, it is rare in European ones and not very common in American ones. Word-initial /N/ is more restricted, and other areal phonotactics are presented. This is an interesting paper, but it seems unrelated to the theme of the book.
The last paper in this section is ''Cryptosonorant phonology in Galice Athabaskan'' by Siri G. Tuttle. Galice, an extinct language documented in the 1950s and earlier, has a phoneme that sometimes appears as [d] but acts as a sonorant; this is the ''cryptosonorant'' of the title. Two morphemes, the second person singular subject and the perfective morpheme, evidently consist of an autosegmental [nasal] feature, which nasalizes both the appropriate prefix vowel and changes /d/ to [n]. Sometimes the nasal vowel is separated from the initial nasal consonant by an oral vowel, and this is the challenging pattern to account for. The author builds on Rice's (1993) geometrically represented proposal for Sonorant Voice so that /d/, as a Sonorant, can bear nasality. She proposes that only unlinked [nasal] can spread (nasality from stems does not), and proposes an OT analysis aligning the phonological feature [nasal] to a left edge of a SYLLABLE, and the morphological feature 2sg Subj to the left edge of the STEM. The gapped configuration of the [nasal] feature falls out of the interaction of these constraints. The author seems unclear on whether autosegmental representations are necessary or not: she uses them for illustrative purposes, but for the formal tableaus, they are nowhere in sight, and the constraints make no mention of them. The contribution to ''internal structure of segments'' thus seems ambiguous. It is more of a reinforcement of the notion that in at least some languages, obstruents may have features more commonly associated with sonorants.
Part 3, on Laryngeal Features, begins with ''On the phonological interpretation of aspirated nasals'' by Bert Botma. He observes that besides the normal unmarked voiced nasals, some languages have voiceless, laryngealized, or breathy-voiced nasals. Interesting, while a language may have both laryngealized and another marked nasal, no language (at least of the 12 he presents) has both voiceless and breathy-voiced nasals, and he proposes that these comprise a single phonological category, which he terms ''aspirated nasals.'' He gives a useful review of Element-based Dependency Theory, and then proposes a structure for ''aspirated nasals.'' This Dependency structure includes the elements |L|, showing intrinsic voicing, and |H|, showing voicelessness or aspiration or breathy voice in the theory. The structure does not express linear order, so several phonetic realizations are possible. However, no language appears to have a contrast in these realizations, lending support to the assertion that these are manifestations of a single phonological representation. He amplifies on this with data from a number of languages. The author discusses diachronic evidence showing that aspirated nasals come from a historical *sN sequence, and that his structure accounts for this nicely. He then concludes with a discussion of the interaction of tone with the |H| and |L| elements, concluding that there |H| is involved in both High tone and aspiration, but the relation is not straightforward.
The next paper is ''The representation of the three-way laryngeal contrast in Korean consonants'' by Hyunsoon Kim. In this, the author examines the well-known lenis-fortis-aspirated contrast in Korean, first giving examples of one class changing to another in appropriate environments. She reviews the extensive literature on proposals on how to represent these contrasts in terms of features. She cites a recent MRI study that measured several dimensions of articulatory movement with these contrasts, and found two independent patterns for the coronal stops and affricates. First, four measures (including closure duration and glottal height) varied from short to long/high in the order lenis, aspirated, and fortis. In contrast, glottal width varied from narrow to wide in a different order: fortis, lenis, and aspirated. As a result of these, Kim proposes that all Korean stops are singletons (no contrastive length) and use features [spread glottis] (s.g.) and [tense]. Stops are specified as follows: lenis are [-s.g., -tense], fortis are [-s.g., +tense], and aspirated are [+s.g., +tense]. (The Jacobsonian feature [tense] is redefined somewhat.) These specifications account for the fact that fortis and aspirated form a natural class around the feature [+tense], and lenis and fortis form a natural class around [-s.g.]. She illustrates this by data from ''intensified expression'' and French and Japanese loans as an illustration of the first class, and English loans illustrate the second class. Specifically, [constricted glottis] is not motivated in Korean.
The final paper is ''Diachronic evidence in segmental phonology: the case of obstruent laryngeal specifications'' by Patrick Honeybone. The author, after an overly long introduction, notes two traditions. The first, perhaps ''standard tradition,'' describes the difference between /p,t,k/ and /b,d,g/ with the feature [voice]. The second tradition asserts that there are two types of languages: Type A (e.g. English, German) has aspirated voiceless stops, little if any voicing in ''voiced stops,'' and assimilation to voicelessness, while Type B languages (e.g. Dutch, Spanish) have unaspirated voiceless stops, fully voiced voiced stops, and assimilation to voicing. The assertion is that in Type A languages the transcription above is not correct, and thus the featural difference in stops is not [voice] in these languages. The diachronic evidence comes from an apparent merger of ''voiced'' and ''voiceless'' stops in some non- standard German varieties to voiced stops, and ''voicing of fricatives'' from Old English to some varieties of Middle English. These seem counterexamples to the claimed markedness universal that voiced obstruents exist only if there also exist voiceless ones. However, this apparent problem disappears if the second tradition is adopted, that is, if the difference is not [voice] at all, but [spread]. In this case, the starting point is aspirated vs. unaspirated obstruents, and one set loses its aspiration. This ''delaryngealization'' is predicted to occur, and is now seen as a natural development. This is a fascinating article, but again seems to have more to do with specific features than how they are organized.
This book is a welcome contribution to phonology, raising the question of representations and featural organization that many phonologists assume is irrelevant. A majority of the authors allow that in this era dominated by OT, there is still a place for representations to make a fruitful contribution (though Yip argues against this). As such, it deserves consideration and further research to respond to the claims advanced here.
The volume is attractively bound, but some editing errors such as misspellings, miscapitalization, misnumbering of sections, and unalphabetical ordering of references (each of which happened multiple times) mar the volume.
It includes a helpful language index, containing over 300 languages referred to in the papers. The author index is welcome but has the curious characteristic of listing co-authors not as individuals, but as a unit, e.g. Kisseberth does not get a line of his own, but is included in ''Kenstowicz, M. & C. Kisseberth.''
Kiparsky, Paul. 2000. Opacity and Cyclicity. The Linguistic Review 17:351- 365.
Padgett, Jaye. 2000. Feature classes in phonology. Language 78.1:81-110.
Rice, Keren. 1993. A reexamination of the feature [sonorant]: sonorant obstruents. Language 69.2: 308-344.
Walsh, Laura. 1997. The Phonology of Liquids. Ph.D. dissertation, University of Massachusetts, Amherst.ʃʃ
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Mike Cahill has done on-site linguistic investigation in the Konni language of northern Ghana for several years, including application to literacy and translation work. He received his Ph.D. from Ohio State University in 1999, and is primarily interested in African phonology, cross-linguistic patterns in tone, and labial-velar stops and nasals. He currently serves as SIL's International Linguistics Coordinator.