Defining Features and Constraints in Terms of Complex Systems: Is UG Too Complex?

Douglas Pulleyblank
University of British Columbia

Segmental inventories range from simple to complex, as do phonological patterns such as featural harmony. Under the assumption that Universal Grammar is ‘universal’, it follows that the theory must be sufficiently rich to account for the most complex of inventories and patterns. For example, if inventories containing 3, 5, 7, and 9 vowels are encountered, the theory of features must be up to the task of describing the 7 and 9 vowel systems, not just the 3 and 5 vowel inventories. Similarly, the constraints required to account for a simple harmony system – e.g. agreement within the word for feature F – must be up to the task of accounting for a complex pattern – e.g. one exhibiting domain restrictions, patterns of transparency and/or opacity, directionality differences.

Establishing theoretical machinery sufficient for the most complex patterns raises problems of indeterminacy when simple systems are considered. Imagine a vowel system with a small number of contrasts, a system of, say, the following five vowels: {i, e, a, o, u}. Assuming that more complex inventories motivate a feature system including [high], [low], [rtr], [back], [round], this 5-vowel vowel system could be analysed in a number of different ways: (i) [high], [low], [back]; (ii) [high], [back], [round]; (iii) [low], [back], [round]; (iv) [high], [low], [round]; (v) [high], [rtr], [back]; etc. Somewhat troubling analytically is the fact that the systems exhibiting such indeterminacy are the most common patterns attested in natural language phonologies. That is, languages exhibiting relatively small numbers of vowels vastly outnumber the more differentiated systems that require the full set of features. By postulating a universal set of features, redundancy and indeterminacy are introduced into the phonological systems of the more commonly attested languages.

Focussing on the complex vowel harmony patterns of Kinande (Bantu), Setswana (Bantu) and Lango (Nilotic), this paper addresses two ways in which complex systems force indeterminacy into the analysis of simple systems. It is shown that a consideration of complex systems raises questions about the postulation of universal feature and constraint sets. It is suggested that the postulation of language-particular feature sets may reduce indeterminacy in phonological systems, while providing analyses of comparable or greater restrictiveness.

download handout (pdf)