Cross-linguistic Patterns and their Significance for the Performance-Grammar Interface
John A. Hawkins, USC

Joe Greenberg was an incredible pattern-seeker across and within languages. He sought and found patterns that would establish new genetic groupings of languages, e.g. in Africa and the Americas, as well as larger "macro"-groupings such as Eurasiatic. He sought and found many universal regularities of different types (e.g. implicational and distributional) across all languages. He was inspirational in his proposals regarding the significance of these regularities for our deeper understanding of why languages are the way they are. He stressed the inseparability of diachrony and synchrony and he formulated diachronic regularities that made sense of many synchronic variation patterns and of intermediate synchronic types. He also stressed the inseparability of performance and grammars. In his (1966) discussion of feature hierarchies, such as Singular > Plural > Dual, he drew our attention to correlating patterns of grammars and performance. Morphological systems, declining allomorphy and increased formal marking provided grammatical evidence for these hierarchies, while declining frequencies of use in languages such as Sanskrit suggested not only a correlation with performance but a possibly causal role for it in the evolution of the grammatical regularities themselves.

In this paper I pursue this relationship between performance and grammars and I argue that there is a profound performance basis to the syntactic universals that Joe proposed as well as the morphological ones. I outline a general principle of efficiency and complexity that emerges from clear patterns of preference (in corpora and psycholinguistic experiments) in languages with variation. This principle defines a general minimization preference for syntactic and morphological processing. I then show how the preferences of performance have been conventionalized in grammars in languages and structures with less variation, resulting in the implicational patterns of the "Greenbergian correlations" and in other regularities that Joe discovered. Joe always encouraged interdisciplinary synthesis and the present paper will draw on insights from parsing and psycholinguistics, as well as from formal grammar and language typology, which I will try to integrate. The paper will conclude that psycholinguists need to view grammars as "frozen" processing preferences, that typologists should not shun formal grammars but should make use of them when explaining typological variation, and that formalists should try to motivate and ground their fundamental grammatical principles in performance, in accordance with Joe Greenberg's insight about their inseparability.