Recall of compound words in simple and complex span tasks

Abstract

There has been little research exploring whether different memory processes (i.e. related to short term memory (STM), working memory (WM) and long term memory (LTM) may be differentially sensitive to representation and processing aspects of compound words. This study investigated how compound words are represented in STM in immediate serial recall tasks and in WM in complex span tasks that combine processing and storage demands. The simple span STM task was comprised of solely a list of memory words, whereas the two complex span WM tasks interleaved sentence processing between presentation of memory words. They varied in the presence of a pause after presentation of each memory word and before onset of the following distractor sentence for processing. The absence of a pause was intended to minimize opportunity for subvocal rehearsal, whereas the presence of a pause encouraged rehearsal. To increase chances of recombination errors for error analyses, lists of memoranda were manipulated so that each set (list) of four compound words contained one “lure” pair (e.g. pinstripe + warhead = pinhead) in which the modifier and head constituents from separate compound words could recombine to form a new, legal word. Recall performance was better in the simple span and complex span pause tasks compared to the complex span no pause task. Whole compound and left constituent frequencies played opposite roles, helping and harming, respectively. Error types reflecting decomposition of the compound words to their constituents were more common in simple span than in complex span. Omissions were more common in complex span. We discuss how different memory processes may be differentially sensitive to representation and processing aspects, and how recall of compound words is affected by various lexical variables.