Ontogeny Recapitulates Phylogeny is an often-repeated (albeit also much-disputed) slogan expressing the hypothesis that stages in the development of an animal from embryo through infancy to adulthood mirror stages in the evolution of the animal's own ancestors. Among others, this hypothesis has also been applied in attempts to shed light on the evolution of language, with claims that early child language may provide a model for the evolution of early human or pre-human language.
This paper presents a companion hypothesis encapsulated in the slogan Typology Reflects Phylogeny, arguing that cross-linguistic variation may provide an alternative and in some respects better model for the evolution of language. The substitution of "reflects" for "recapitulates" acknowledges that typology differs in one crucial respect from both ontogeny and phylogeny, namely, it is non-directional. Whereas ontogeny and phylogeny are inherently bound to the arrow of time, linguistic variation has no intrinsic "earlier" and "later". For example, if Czech and Polish differ with respect to the relative order of noun and adjective, then there is no a priori reason to suppose that either of the two orders is more representative of a prior stage in the evolution of language.
However, one basic measure provides a means for characterizing some feature values as likely to be evolutionarily prior to others: complexity. If one particular feature value is simpler than its more complex alternative, then the simpler feature is likely to represent an earlier stage in the evolution of language with respect to the feature in question. Underpinning this inference is the fact that evolution typically proceeds from simpler to more complex. In other sciences, similar inferences are commonplace. For example, life today consists of simpler prokaryotes, more complex unicellular eukaryotes and even more complex multicellular eukaryotes; such progression in contemporary complexity is then taken to represent the evolutionary path of life. By the same token, numerous aspects of cross-linguistic variation involve features with respect to which one particular value may be said to be more complex than another. Most straightforwardly, languages differ with respect to inventory sizes of various items such as phonemes, affixes, parts of speech, speech registers, words, and so forth; for each of these, fewer is clearly simpler than more, and hence fewer may in principle be hypothesized to represent an evolutionarily prior stage in the evolution of that particular linguistic feature.
This paper provides support for the methodology of using cross-linguistic variation as a window into phylogeny, focusing on the domain of compositional semantics. When two or more expressions are brought together, the meaning of the juxtaposition must be associated in some way with the meaning of its constituent parts; this is the most basic principle underlying compositionality. However, languages vary considerably with respect to the extent to which the meaning of such a juxtaposition may be further delimited by various grammatical rules making reference to morphosyntactic strategies such as agreement, flagging, linear order and others. Whereas in some languages, of high articulation, there is an abundance of such rules, substantially limiting the range of possible interpretations, in other languages, of lower articulation, there are fewer such rules, and the range of possible interpretations is thus broader. In terms of their compositional semantics, low articulation languages are thus less complex than their high articulation counterparts, and may accordingly be hypothesized to represent an earlier stage in the evolution of compositional semantics.
This paper presents the results of an ongoing cross-linguistic experiment measuring the degree of articulation of compositional semantics in the domain of thematic role assignment; to date, approximately 50 languages have been examined. The main factor governing the degree of articulation is shown to be the polity type of a language, ranked in accordance with the following four-valued scale of increasing socio-political complexity: (1) regional language until recently not part of a larger polity (e.g. Ju|'hoan, Mursi, Mentawai, Marind, Tikuna); (2) regional language that has been part of a larger polity for a long period (e.g. Galician, Yoruba, Shan, Minangkabau, Q'anjobal); (3) regional variant of a standardized national language (e.g. Neapolitan, Cantonese, Osaka Japanese, Kuching Malay, Riau Indonesian); and (4) standardized national language (e.g. Italian, Mandarin, Japanese, Indonesian, Palauan). The experimental results show that languages of greater socio-political complexity tend to exhibit higher articulation. Thus, within in the domain of compositional semantics, grammatical complexity correlates positively with complexity of polity type. However, with respect to polity type, there is abundant extra-linguistic evidence, from history, archaeology and anthropology, that polity types evolved from simple to complex. Thus, the correlation between polity type and degree of articulation of compositional semantics provides further extra-linguistic support for the claim that low-articulation languages represent a prior stage in the evolution of compositional semantics. In so doing, the results of the experiment provide independent support for the use of cross-linguistic variation with respect to complexity as a window into the evolution of language.