Meaningful sequences of words, such as sentences and passages of prose, are easier to remember than sequences that lack syntactic and semantic structure. In this chapter we explore the interactions between short-term memory and stored language knowledge in long-term memory that give rise to this effect. Previous research on sentence processing in the areas of working memory and language comprehension is discussed. We then report two experiments demonstrating a sentence effect in serial recall, indicating that stored language knowledge is applied to the temporary retention of single sentences in a relatively automatic manner, without requiring additional working memory resources. This ``within-sentence binding'' is contrasted with memory for complex, multi-sentence sequences. The work is interpreted within the context of working memory models, with particular reference to the multi-component framework proposed by Baddeley (2000).
The notion that the capacity of short-term memory can be increased through the binding of information into larger chunks has remained popular since initially proposed by Miller (1956). In this seminal work, Miller suggested capacity to be chunk-rather than item-dependent, and that the more ef®ciently larger chunks of information are created, the better subsequent memory performance will be. Processes of chunking have been shown to utilize features of the to-be-remembered stimuli, such as rhythm and timing (Ryan, 1969), as well as word meaning. One important variant of information binding, which presumably involves interaction between temporary processing and stored knowledge, concerns memory for redundant material such as text.