Cross-linguistic variation and efficiency by John A. Hawkins (review)

How do performance factors influence the fixed grammatical conventions of languages? How does crosslinguistic variation in grammatical structure inform our understanding of language use? In Cross-linguistic variation and efficiency, John A. Hawkins presents new applications and refinements of the principles proposed in his earlier book Efficiency and complexity in grammars (2004), representing the latest development in a long line of research on the relationships among grammar, typology, and performance. As in his previous major works on this topic (Hawkins 1983, 1994, 2004), H argues that factors related to language use have left their mark on grammars in such a way that one can find evidence of their influence in numerous grammatical rules and crosslinguistic generalizations. In turn, he shows how linguistic evidence from grammatical rules and crosslinguistic generalizations can shed light on aspects of performance that might otherwise remain obscure.

The current work applies the principles proposed in Hawkins 2004 to a wider range of empirical data and addresses a number of recent developments in language processing, syntactic theory, historical linguistics, and linguistic typology. Ch. 1 summarizes the performance grammar correspondence hypothesis (PGCH) and lays out some of its predictions. The PGCH is stated as follows: ‘Grammars have conventionalized syntactic structures in proportion to their degree of preference in performance, as evidenced by patterns of selection in corpora and by ease of processing in psycholinguistic experiments’ (3). The primary evidence for this hypothesis is found in parallels between patterns of variation in language use and patterns of variation among grammars, of which numerous examples are given throughout the book.

Ch. 2 reviews the three major efficiency principles from Hawkins 2004: minimize domains (MiD), minimize forms (MiF), and maximize online processing (MaOP), and gives examples from both performance and grammar illustrating how these principles cooperate and compete with each other. One example is the distribution of gaps versus resumptive pronouns in relative clauses. Whereas gaps are favored by MiF, since there are fewer morphological forms to process, resumptive pronouns are favored by MiD, since they provide a local argument for processing lexical cooccurrences within the relative clause (24). H provides evidence both from systematic crosslinguistic variation in the distribution of gaps versus resumptive pronouns and from variation in usage for languages like Hebrew and Cantonese that allow the option of either gap or resumptive pronoun in certain positions. Both sources of data confirm the same pattern: resumptive pronouns are preferred more often in structurally more complex environments for which overall efficiency is achieved through domain minimization at the expense of form minimization.

Ch. 3 situates the three major efficiency principles in the context of current findings in psycholinguistics. One current line of research develops the idea that linguistic elements are easier to process when they are more predictable within a given context (Levy 2008), and the related idea that speakers maintain, on average, a uniform information density in spontaneous speech by reducing the form of more-predictable information and elaborating the form of less-predictable information (Jaeger 2006). H’s MiF captures the same general idea, but lacks the precise metrics used to quantify predictability (or surprisal) and information density. H acknowledges the usefulness of these measures and sees them as a complement to his own proposals. Another current issue addressed in this chapter is the apparent contradiction between studies that show locality effects (i.e. a processing advantage for local dependencies as compared with nonlocal ones), and those that show antilocality effects (i.e. a processing advantage for nonlocal dependencies). H exemplifies this contrast with the case of relative clause extraposition in German and English (52–57). Konieczny’s (2000) German data showed antilocality effects in word-by-word reading, in the form of shorter reading times on the verb when preceded by an object-modifying relative clause, but longer reading times on the verb when the relative clause was extraposed (55). In contrast, Francis’s (2010) English data showed locality effects in that whole-sentence reading times were faster...