The mirage of morphological complexity
Fermin Moscoso del Prado · Cognitive Science · 2011
The Mirage of Morphological Complexity Ferm´in Moscoso del Prado Mart´in ([email protected]) Laboratoire Dynamique du Langage, CNRS & Universit´e de Lyon, Lyon, France & Institut Rhˆ one-Alpin des Syst`emes Complexes (IXXI), Lyon, France Abstract 1. Inflectional morphology expresses semantic content that is required for successful communication. I present a study on the morphological complexity of six European languages. A theory-free measure of the complexity of a language’s inflectional morphology, is derived from Gell-Mann’s concept of Effective Complex- ity. Using a parallel corpus, I show that disconsidering word order information results in the classical gradation of inflectional complexity: Languages in which words have more inflected variants seem to be more complex than languages with fewer variants. It also appears that the presence of the inflectional system increases the com- plexity of languages. However, when word order infor- mation is explicitly considered, the apparent gradation in complexity across languages vanishes. Furthermore, it becomes clear that the presence of inflections reduces the overall complexity of languages. In sum, inflection is used whenever its presence simplifies a language’s de- scription. Inflectional morphology is not a capricious feature, as some argue, but rather an effective tool for complexity reduction. Keywords: Effective Complexity, Entropy, Morphol- ogy 2. The presence of inflectional categories and markers re- duces the overall complexity of a language. What is morphological complexity? Introduction Words are usually accepted as being the fundamental units of syntax. However, in many (or most) languages, words can be related to each other by (quasi-)regular re- lations between their orthographical forms, their gram- matical functions, and their meanings. These intra- lexical regularities are referred to as morphological struc- ture, and constitute one of the main areas in the study of human languages, both from a linguistic and a psy- chological perspective. Morphology: ‘By itself ’ ? The classical structuralist position on inflectional sys- tems is that they are largely random classifications that need not serve any useful function within the overall lin- guistic system. This is still a common view both among linguists (e.g., Aronoff, 1994; Bloomfield, 1933) and psy- chologists (e.g., Maratsos, 1979). For instance, Aronoff (1994) argues that morphology must be considered ‘by itself’, not merely from the perspective of its interface with syntax, phonology, or semantics, and that linguis- tic theory should therefore consider a separate and au- tonomous morphological component. In contrast, sup- porters of ‘functionalist’ theories of grammar would ar- gue that the presence of inflectional systems must be useful from a communicative perspective. Otherwise, in line with Zipf (1949)’s ‘Principle of Least Effort’, the cognitive system would have discarded them for requir- ing unnecessary efforts. From this perspective, the pres- ence of inflectional morphology is only justified if one or both of the following conditions apply: Notice that the discussion above assumes that some ob- jective quantification of the complexity of a morpho- logical system is available. However, no definition or quantification of the complexity of a linguistic system is widely accepted. The concept of complexity has cen- tral importance for theories of language structure and processing. For instance, McWhorter (2001) raised an interesting controversy over his claim that pidgin and creole languages are intrinsically less complex than other languages, reflecting the very early stage in their devel- opment. However –despite its frequent invocation– lin- guistic complexity remains a rather ill-defined concept. Indeed, much previous research has been aimed at pro- viding a concise and workable definition of linguistic – and morphological– complexity (e.g., Goldsmith, 2001; Greenberg, 1954; Juola, 1998, 2007; Malvern, Richards, Chipere, & Dur´an, 2004; McWhorter, 2001; Nichols, 2010; Shosted, 2006; Siegel, 2004; Vulanovi´c, 2007; Xan- thos & Gillis, 2010). Many of these studies define com- plexity in terms of either counting forms, or detecting the presence or absence of certain morphological phe- nomena or structures that are chosen as representative of complexity based on some theoretical premises, an ap- proach which was also used by Lupyan and Dale (2010). However, such theory-bound measures can result in con- siderable vagueness and subjectivity with regard to what is more or less complex (see, e.g., McWhorter, 2001; Nichols, 2010; Shosted, 2006, for debate on this issue). To obtain more objective estimates, some have measured the complexity of inflectional paradigms as the num- ber of inflected variants that can be formed for a sin- gle word (e.g. Malvern et al., 2004; Xanthos & Gillis, 2010). This simple approach presents the problem that it ignores that there are sometimes very regular relations between different inflected forms (e.g., the plural of an English word is in most cases the same as the singular plus an ‘s’), and not much information is required to encode this (what some term the Paradigm Cell Filling Problem; cf., Ackerman, Blevins, & Malouf, 2009). To avoid this, some have taken a theory-free approach in- vestigating to what extent linguistic utterances or struc- tures can be compressed (Goldsmith, 2001; Juola, 1998,