Why are some languages more regular than others?

Many years ago, I did anthropological fieldwork among the Dorze of Southern Ethiopia. Since no grammar of the Dorze language was available, I had to find out what were its basic morphological and syntactic rules. The good news was that once I had identified a rule, I could apply it across the board: there were hardly any exceptions. From this point of view, Dorze stood in sharp contrast with Amharic, the dominant language of then imperial Ethiopia. Amharic (like English) is a language with many irregularities. Dorze regularity was found not only at the morphological level, but also at the phonological level. The many words that had been borrowed from Amharic into Dorze had all, except for the most recent ones, acquired fully-regular dorze phonology.

Why are some languages quite regular and others not? I remember posing the question to the historical linguist Robert Hertzron, whom I met at the time in Addis Ababa. It is, he suggested, because, in the process of language acquisition, children tend spontaneously to over-regularize. They apply any rule they have acquired to all possible instances (in English, for instance, they may over-generalize the ordinary rule for past tense and say “he goed” instead of “he went”). In societies where adults correct children, these mistaken regularization are suppressed and irregularities are maintained; in societies where adults leave children alone in this respect, irregularities are less stable, and the language tends to be more regular. Gary Marcus et al. in their monograph on “Overregularization in language acquisition” (1992) quote Jill de Villiers half-joking: "Leave children alone and they'd tidy up the English language."

The view that children are the main source of language regularization is an old one. According to Max Müller, for instance, “It is likely… that the gradual disappearance of irregular declensions and conjugations is due, in literary as well as in illiterate language, to the dialect of children. The language of children is more regular than our own” [Müller 1890: 75].

This view however is far from being generally accepted. Marcus et al. (1992) for instance, , hold that over-regularization errors are rare. Language learners, they argue, tend to assume that only one form expresses a grammatical function (either “goed” or “went”, but not both) and that the form they hear from adults (e.g., “went”) is the right one (but see Maratsos 2000 [1]). Moreover, apparently inconsistent with Hertzron suggestion, it is generally agreed that the role of explicit correction in language learning is at best quite marginal.

Still, the question I am raising is not the historical question: why is there a tendency towards regularization in the evolution of a given language (in the absence of typically language-contact-related sources of irregularity)? It is the comparative question: why are some languages more regular than others (even limiting the comparison to related languages with similar morphologies)? It could be that both children and adults contribute to historical regularization, or even that adult contribution is more important, and that, nevertheless, children are more conformist; if they are given evidence that a certain form is the socially approved one, they may inhibit over-regularization to a higher degree. This evidence, by the way, can take many form and be more subtle than explicit correction. If on the other hand, adults don’t seem to care and there is no such evidence, then children may over-regularize more freely, with the cumulative result that irregularities would disappear in fewer generations. (It would be useful, incidentally to have evidence on the frequency of over-generalization in children drawn not just from English, where there is social pressure for conformity, as is the case now, but also for languages where there is no such pressure, such as Fijian.)

So, I don’t know what is the right answer to the question I was puzzling about in Ethiopia. I would welcome 1) any theoretical suggestion, and 2) any relevant (positive or negative) evidence of a correlation between, on the one hand, regularity/irregularity in morphological and phonological forms in the language and, one the other hand, the prevalence in the linguistic communities of views regarding ‘correct’ and ‘incorrect’ usage, views expressed in particular in the form of feedback, however subtle, to language learners.

[1] Maratsos, M. (2000). More overregularizations after all: new data and discussion on Marcus, Pinker, Ullman, Hollander, Rosen & Xu. Journal of Child Language, 27(1), 183-212.