Processing Difficulty and Principles of Grammar

Abstract

Some words are more difficult to pronounce than others, and, similarly, some sentence types are harder to process than others. Such processing differences are due to properties of the human parser, and these may be responsible for certain laws of natural language syntax is not a new idea: it is a key concept in linguistic typology. This chapter investigates the merits of the proposal that two core principles of current generative syntax, namely the principle that syntactic movement is costly, and the principle that the costs of movement are proportional to the distance covered (Chomsky 1995), can be explained in terms of processing theory. The chapter is not the first attempt of relating abstract syntactic laws to processing facts (cf., e.g., Marcus 1980, Staudacher 1993), but for such an approach to be successful, several requirements must be met: • assumptions concerning processing difficulty must be justified independently • the full range of empirical facts of syntax must be captured, and, • the approach must be explicit about the link between processing difficulty and syntactic laws. After introducing the facts to be explained, this chapter makes a few principled but obvious remarks on the last aspect, to which we return in the final section. The third section sketches the general line of the argument. The fourth section introduces evidence for the claim that movement is cognitively costly, and discusses processing models that predict such costs. Particular attention is given to approaches involving memory load. The fifth section tries to assess to what extent the two key laws of syntax introduced above can be derived from such processing considerations. The paper reports work in progress, so some of the points presently depend more on plausibility arguments than on hard core data. The conclusion we will arrive at is somewhat different from the one we originally found plausible. We still believe that the processing facts are compatible with the view that processing shapes grammar, but there are too many gaps in the grammaticalization account for it to be likely to be true.