In lieu of an abstract, here is a brief excerpt of the content:

  • Letters to Language
  • Gregory R. Guy

Language accepts letters from readers that briefly and succinctly respond to or comment upon either material published previously in the journal or issues deemed of importance to the field. The editor reserves the right to edit letters as needed. Brief replies from relevant parties are included as warranted.

Grammar and usage: A variationist response

July 8, 2005

To the Editor:

The recent discussions of grammar and usage in the pages of Language—by Frederick Newmeyer (N) (‘Grammar is grammar and usage is usage’, 79.4.682–707, 2003; ‘A reply to the critiques of “Grammar is grammar and usage is usage”’, 81.1.229–36, 2005), Brady Clark (C) (‘On stochastic grammar’, 81.1.207–17, 2005), Ritva Laury and Tsuyoshi Ono (LO) (‘Data is data and model is model: You don’t discard the data that doesn’t fit your model!’, 81.1.218–25, 2005), Charles F. Meyer and Hongyin Tao (‘Response to Newmeyer’s “Grammar is grammar and usage is usage”’, 81.1.226–28, 2005)—make little reference to the extensive work in variationist sociolinguistics addressing these issues. Worse, one overt reference (N 2005:229) clearly misrepresents a foundational work in this field—Henrietta I. Cedergren and David Sankoff’s ‘Variable rules: Performance as a statistical reflection of competence’ (Language 50.2.333–55, 1974; CS).

Work in this tradition explicitly problematizes N’s grammar/usage distinction, not by rejecting grammar, but by improving its adequacy. Variationist research seeks not merely to generate all and only the grammatical utterances, but also to adequately characterize their contextual appropriateness and likelihood. N apparently postulates that grammar must be limited to categorical principles; hence he rejects out of hand any stochastic or probabilistic elements in the grammar. Variationist work begins with the contrary postulates enunciated by Uriel Weinreich, William Labov, and Marvin Herzog (‘Empirical foundations for a theory of language change’, Directions for historical linguistics: A symposium, ed. by Winfred P. Lehmann and Yakov Malkiel, 95–188, Austin: University of Texas Press, 1968; WLH): inherent variability and orderly heterogeneity. Inherent variability, as C observes, depicts variation, rather than perfect invariant consistency, as the normal state of linguistic systems; orderly heterogeneity is WLH’s term for the fact that much of this variability in language is highly structured, showing probabilistic regularities, if not categorical ones.

Based on these principles, variationist work offers integrative resolutions to the grammar/usage differences that N sees as irreconcilable. Limited as to space, I restrict my comments to three of these here: probabilistic grammar, the use of corpora, and the multiplicity of styles and genres.

The quantification of grammar follows from inherent variability. Speakers manifestly are sensitive to frequencies and manipulate them systematically, and a probabilistic grammar more adequately models such competence than a categorical one, while maintaining the capacity to incorporate categorical statements (these are the limiting cases of probability values at zero or one). N’s position, that there are no weights in the grammar other than zero and one, is simply an assumption, one that is massively falsified by the data.

N’s principal argument against quantification is the claim that statistical regularities are ‘overwhelmingly epiphenomenal’ (2003:696). To variationists, this is an extreme case of special pleading. In fact, the multivariate analytical techniques like VARBRUL used in this research specifically control for the uneven data distributions that N invokes, and the results of studies of probabilistic linguistic constraints on variation typically find explanations in terms of recognizable grammatical properties. Two examples will illustrate this point. First, coronal stop deletion in English is probabilistically favored by phonological similarity with the preceding segment (there is more deletion in tokens like wrist than rift, slapped than slammed), a probabilistic instantiation of the obligatory contour principle (OCP), which disfavors adjacent identical elements in phonological structures. The OCP is a well-established principle of phonology with many categorical effects. The relationship between probabilistic and categorical instantiations of the OCP is that some particular sequences are absolutely prohibited, while others are probabilistically avoided; but in the latter cases, the disfavoring is significant, not epiphenomenally derived from frequencies of occurrence of the segments and features involved. This is the issue that C...

pdf

Share