The Ocp in Functional Phonology

Though seemingly a good candidate for a universal output-oriented constraint, the OCP does not occur as a constraint in the production grammar. Instead, it handles, in interaction with the NoCrossing Constraint, the correspondence between acoustic cues and perceptual feature values in the perception grammar. Because faithfulness constraints use the perception grammar to evaluate the similarity between the perceptual specification and the perceptual output in the production grammar, the OCP does influence the evaluation of candidates in the production grammar. As a result, adjacent identical elements are avoided because they constitute PARSE violations. Dissimilation at a distance, by contrast, is due to a constraint against the repetition of articulatory gestures. In this paper and in Boersma 1998a, I point out the advantages of distinguishing between articulatory and perceptual features in autosegmental phonology. According to McCarthy (1988), the only phonological processes that can be accepted as primitives in autosegmental phonology, are spreading, deletion, and the obligatory contour principle (OCP). While Boersma 1998a centres on spreading, the current paper will tackle the OCP. McCarthy (1986) expresses the Obligatory Contour Principle (OCP) in its naked form as follows: “adjacent identical elements are forbidden” As we will see, many phenomena have been ascribed to this principle. 1 Functional interpretation of the OCP From a functional standpoint, the OCP is not a single primitive principle, but branches into two fundamentally different ones. Furthermore, one of these two principles is naturally embedded in a set of constraints on simultaneous and sequential combinations of gestures and perceptual features. 1.1 Articulatory and perceptual motivations If we distinguish between articulatory and perceptual phenomena, the OCP branches into two principles. The first is a general principle of human perception, not confined to phonology. In designing a country map of Europe, the cartographer can choose to fill in the countries with the minimal number of four colours that are needed to give every pair of adjacent countries different colours. If she decided to paint both the Netherlands and Belgium red, 1 This paper is chapter 18 of Boersma (1998b), which is available from the author (http://fonsg3.hum.uva.nl/paul/) or from the publisher (http://www.hagpub.com). Paul Boersma The OCP in functional phonology 2 the reader of the map would not be able to identify them as separate countries; thus, in cartography, adjacent identically coloured countries are avoided. Likewise, if a morph ending in /-m/ is concatenated with a morph starting with /m-/, the usual timing of syllable-crossing clusters will result in the long consonant [-m ̆-]. The perceptual identity of one of its constituents is therefore lost, violating featural faithfulness. Some of the information about the existence of two morphemes is kept in the timing, but if the language is adverse to geminates, it may just end up with [-m-], violating some more faithfulness. The problem of the long perceptually homogeneous sound can be levied by inserting a pause between the two consonants (i.e., drawing a black border between the Netherlands and Belgium): giving [[-m_m-]]. This violates a FILL (pause) constraint: a pause can be perceived as a phrase boundary. Another strategy would be to insert a segment (declaring the independence of the southern provinces of the Netherlands, and painting them blue), which will give [-m ́m-] or so: another FILL violation. Languagespecific rankings of all the faithfulness constraints involved will determine the result. The perceptual nature of this first functional correlate of the OCP is shown by the rules of vowel insertion in English, which are hard to capture with generalizations over single tiers in feature geometry. Thus, the insertion of /I/ before the morpheme /-z/ occurs in bridges but not in tents, exactly because [dZz] would contain a perceptually unclear boundary (The Netherlands in red, Belgium in purple), and [nts] would not; likewise, the insertion of /I/ before the morpheme /-d/ occurs in melted but not in canned, because the boundary would be lost in [lt ̆] but not (or less so) in [nd]. The second functional correlate of the OCP is simply the tendency not to repeat the same articulatory gesture: an articulatory *REPEAT constraint. The features involved in this constraint are arguably of an articulatory nature: the Japanese constraint against two separate voiced obstruents within a morpheme obviously targets the articulatory gesture needed for the voicing of obstruents, not the perceptual voicing feature, which would also apply to sonorants. A clear difference with the first principle is exhibited by a morphemestructure constraint in Arabic, which does not allow two labial consonants within a root; apart from disallowing two appearances of /b/, it does not even allow /m/ and /b/ to appear together. This generalization over plosives and nasals is typical of the articulatory labial gesture, which does not care whether the nasopharyngeal port is open or not, whereas the divergent behaviour of plosives and nasals in melted versus canned is exactly what is expected from a perceptually conditioned phenomenon. The predicted correlations between near OCP effects and faithfulness constraints, and between distant OCP effects and articulatory constraints, are verified in this paper. 1.2 Simultaneous and sequential combinations of features I will identify the “perceptual” OCP as one of the four constraint clans that handle combinations of articulatory gestures or perceptual features. In a functional theory of phonology (Boersma 1998b), we express articulatory implementations in articulatory features (gestures) or their combinations, and we express perceivable sounds in perceptual features or their combinations. Paul Boersma The OCP in functional phonology 3 Articulatory constraints on combinations of gestures. From general properties of the acquisition of human motor behaviour (namely, the ability to group simultaneous or sequential gestures into a more abstract coordination or motor program), we can posit the unity of often-used coordinated gestures (like the lip and tongue body gestures in [u], in a language where this sound is common), and the unity of common sequences of gestures (like the lip closing and opening gestures in [apa], in a language where [p] often occurs intervocalically), which leads to assuming separate constraints for these more abstract articulations: *COORD (a1: g1; a2: g2 / ...): “do not combine the gesture g1 on the articulator a1 with the gesture g2 on the articulator a2.” *SEQ (a1: g1; a2: g2 / ...): “do not follow the gesture g1 on the articulator a1 with a gesture g2 on the articulator a2.” Faithfulness constraints on combinations of perceptual features. From general properties of the acquisition of human perception (namely, the ability to group simultaneous or sequential percepts into a more abstract representation), we can posit the unity of often-heard simultaneous features (like labiality and nasality in [m], in a language where this sound is common), and the unity of common sequences of features (like the nasal murmur, the silence, and the explosive burst in [ampa], in a language where [mp] often occur in sequence), which leads to assuming separate faithfulness and correspondence constraints for these more abstract percepts: *REPLACEPATH (f × g: x × z, y × w): “do not replace the values x and z on the combined perceptual tiers f and g with the different combination y and w.” For instance, depending on the language, a surfacing of /n/ as /m/ may involve a violation of *REPLACE (place: cor, lab / nas) or a violation of *REPLACEPATH (place × nas: cor × +nas, lab × +nas). OCP (f: x; q1 ñ m ñ q2): “A sequence of acoustic cues q1, q2 with intervening material m is heard as a single value x on the perceptual tier f.” The OCP is just one of the four combination constraints. It belongs in the perception grammar (Boersma 1998b: ch. 6, 8, 15) since it handles the categorization of acoustic input into perceptual features. As such, it also plays a crucial role in featural correspondence in the production grammar (Boersma 1998b: §12.3). 2 History of the OCP In order to be able to defend the descriptive adequacy of a functional account of the OCP, we have to investigate first the various interpretations it has suffered throughout the years, and the types of phenomena it has been invoked to explain. 2.1 The original OCP The first expression of the OCP is commonly attributed to Leben (1973). In his defence of suprasegmental phonology, he demonstrated that tone features and nasality show suprasegmental behaviour in several languages. For example, in Mende, monoand Paul Boersma The OCP in functional phonology 4 bisyllabic monomorphemic nouns have the following possible tone sequences (H = high, L = low, HL = falling, etc.): 1. Nouns of one syllable have H (/kç@/ ‘war’), L (/kpa$ / ‘debt’), HL (/mbu^/ ‘owl’), LH (/mba&/ ‘rice’), or LHL (/mba& › / ‘companion’), but not HLH. 2. Nouns with two syllables have H-H (/pE@lE@/ ‘house’), L-L (/bE$lE$/ ‘trousers’), H-L (/ke@ ̄a$/ ‘uncle’), L-H (/n"$ka@/ ‘cow’, or L-HL (/ ̄a$ha^/ ‘woman’). The five tone sequences for the bisyllabic nouns can be seen to be equal to those of the monosyllabic nouns, if we represent Mende tone in a suprasegmental way: