last changed 31 January 2006
This is a draft from 1995 which was never revised and submitted for publication - I no longer remember why not. Maybe it was because I felt I ought to find out more about theories of memory first. Ten years later I still think the general ideas are right, especially for readability based on dependency distance, though I now have a better developed theory of parsing based on spreading activation, binding and default inheritance. As for the second problem, centre-embedding, maybe this theory provides a better explanation; but the details remain to be worked out. I never got round to completing the appendix of textual frequencies but there are now other sources of WG data on Japanese (So Hiranuma's 2001 PhD: The syntactic difficulty of Japanese sentences) and on German and English (Eva Eppler's 2004 PhD: The syntax of German-English code-switching).
The paper distinguishes two kinds of processing difficulty caused by syntactic structure: memory overload, which can be predicted from measures of dependency distance and (perhaps) dependency density, and self-embedded structures, which are difficult because two words are too similar to keep conceptually separate. The ideas about self-embedded structures are developed more fully in The difficulty of (so-called) self-embedded structures.