Correlating complexity: A typological approach

UC Berkeley Phonology Lab Annual Report (2005) Correlating complexity: a typological approach Ryan K. Shosted University of California, Berkeley “In sum, linguists don’t even think of trying to rate languages as good or bad, simple or complex.” (O’Grady et al. 2005: 7) Abstract Proceeding from the assumption that all languages are equally complex, there exists a corollary, widely held but poorly documented, herein referred to as the negative correlation hypothesis. It states: if one component of language is simplified then another must be elaborated. Here, this assumption is reformulated in terms of a scientific hypothesis and subjected to statistical analysis. Thirty-two geographically diverse languages representing 30 language families and two isolates are tested for syllable count and inflectional synthesis on the verb as a means of rating their phonological and morphological complexity, respectively. The correlation between these measures is found to be slightly positive (r=0.0704), but statistically insignificant (p>0.05), indicating that the negative correlation hypothesis, if it is to be retained, still awaits scientific confirmation. Introduction The complexity problem and holistic typology Among some linguists, it seems there is a tacit assumption that the various components of human language (phonology, morphology, the lexicon, syntax and semantics) correlate negatively within a single language based on complexity. From such an assumption, it could be argued that if a language is simple in terms of its phonology (e.g. if it has a low number of contrastive elements, distinctive features, allophonic alternations, etc.) the language as a system will somehow force elaboration in the areas of syntax or morphology, etc. Throughout this paper, I will refer to this notion as the “negative correlation hypothesis.” Introductory linguistics textbooks often note in their first few pages that all grammars are, in a sense, “created equal.” For example, Akmajian et al. (1997: 8) declare, “Although it is obvious that specific languages differ from each other on the surface, if we look closer we find that human languages are surprisingly similar. For instance, all known languages are at a similar level of complexity and detail— there is no such thing as a primitive language.” In their text, the first stated goal of Cipollone et al. (1998: 2) is to convince the reader that “[e]very language is enormously complex.” On the matter of what they call grammatical “equality,” O’Grady et al. (1997: 6) reason that “[i]t is futile to attempt to ‘rank’ languages in terms of relative sophistication.” Nonetheless, it seems clear to even the casual observer that some languages do in fact have smaller/simpler/less sophisticated linguistic components than corresponding components in other languages. For example, Ju|’hoan (Northern Khoisan) has 93 phonemic consonants and Yimas (Lower Sepik) has only twelve (see Table 2). How, then, can these two languages end up with “a similar level of complexity and detail” when all other linguistic components are taken into account? Based on the foundational assumption of “grammatical equality,” there are two possible answers: (1) Yimas compensates for what it lacks in terms of phonological complexity through elaboration (beyond that of Ju|’hoan) in terms of its syntax, morphology, semantics, etc.; or (2) There is some aspect of phonological complexity that is not effectively captured by simply counting the consonants in these two languages (i.e. the phonologies are equally complex when other variables are considered). Both are valid responses to the perplexing position we inherit from the dictum that all languages are equally complex (and the corollary, in the case of (1), that their components may not be). However, neither response, to my knowledge, has ever been validated using quantitative methods that produce falsifiable results—an astonishing lacuna, considering the widespread acceptance of grammatical equality. Several enlightening discussions of differential linguistic complexity have recently come forth. Dahl (2004: 42) observes: “Given that a language as a system can be seen as involving both resources [the inventory of structured linguistic units] and regulations [roughly, ‘rules’], it follows that a language could