Conjunction recall errors for compound words but not names

Abstract

The misidentification of conjunction words has been investigated with a recognition paradigm (e.g., an "old" judgement for moonlight after studying moonshine and limelight; Underwood et al., 1976). However, there is an absence of research examining these errors in recall. Three experiments investigated the extent to which memory conjunction errors occurred in free recall for two-syllable family names (e.g., recalling Milnar or Conroy after studying Milroy and Connar; Experiment 1) and compound words (e.g., moonshine, limelight; Experiments 2 and 3). For the sake of comparison, a recognition test was included in Experiments 1 and 2. In the recall procedure subjects recalled short study lists containing pairs of parent items that could be recombined to create conjunction items. In Experiment 3, the proximity between the presentations of parent items was manipulated and confidence ratings were obtained. Conjunction errors have presented themselves quite readily in recognition, however, there seems to be little agreement as to the underlying mechanisms that lead to their occurrences. Reinitz and colleagues have argued that they are the result of memory processes used during binding (Reinitz & Demb, 1994; Reinitz & Hannigan, 2001, 2004; Reinitz et al., 1992; Reinitz et al., 1994; Reinitz, et al., 1996), while Jones and colleagues have argued that they are based on an automatic process, familiarity, in the absence of a controlled process, recollection (Jones & Atchley, 2002, in press; Jones & Jacoby, 2001; Jones, Jacoby, & Gellis, 2001). The results showed that conjunction errors occur in free recall for compound words where study words were presented in separate trials. However, there was no evidence of conjunction recall for names. Compared to recognition, conjunction errors in free recall were extremely rare both in terms of the overall level and the proportion of participants producing them. The effect of parent proximity was small but suggesting that features presented in closer proximity were more likely to be conjointly recalled. A general feature-configuration theory can account for many of the findings, however, the results are more consistent with the idea that automatic and controlled processes contribute to memory performance (e.g., Jacoby, 1991).