This article reports a meta-analysis comparing the size of repetition priming in full and divided-attention (DA) conditions. The main analysis included 38 effect sizes (ES) extracted from 21 empirical studies, for a total of 2074 (full-attention) and 2148 (divided-attention) participants. The mean weighted ES was 0.357 (95% CI = 0.278–0.435), indicating that divided attention produced a small, but significant, negative effect on implicit memory. Overall, the distinction between identification and production priming provided the best fit to empirical data (with the effect of DA being greater for production tests), whereas there was no significant difference between perceptual and conceptual priming. A series of focused contrasts suggested that word-stem completion might be influenced by lexical–conceptual processes, and that perceptual identification might involve a productive component. Implications for current theories of implicit memory are discussed.
A great research effort has been devoted to the study of the relationship between attention and implicit memory (see Mulligan and Brown, 2003 and Rajaram, 2007, for reviews). Indirect memory tasks differ from explicit ones because they do not require intentional retrieval of the encoded information. Learning is typically demonstrated by an increase of the accuracy and/or speed of elaboration, identification and generation of studied vs. unstudied stimuli (Stone, Ladd, Vaidya & Gabrieli, 1998). Early studies suggested that divided attention (DA) at encoding had differential effects on implicit and explicit memory. This manipulation reduced performance in tasks of free recall and recognition (Craik, Govoni, Naveh-Benjamin, & Anderson, 1996), whereas it had no effects on implicit tests (Kellogg et al., 1996, Parkin et al., 1990, Parkin and Russo, 1990, Russo and Parkin, 1993, Schmitter-Edgecombe, 1996 and Szymanski and MacLeod, 1996). However, subsequent experiments found a number of exceptions (Gabrieli et al., 1999, Light and Prull, 1995, Mulligan and Hartman, 1996, Mulligan, 1997, Mulligan, 1998 and Wolters and Prinsen, 1997). The present meta-analysis was specifically aimed at testing a number of different explanations about the effects of DA manipulations on implicit memory. In particular, the primary aim was to ascertain whether incongruent results could be reconciled by taking into account the nature of implicit tests (based on perceptual vs. conceptual or on identification vs. production processes) and the difficulty of interference tasks. These issues are briefly illustrated in the following paragraphs. In addition, two separate sections will be dedicated to the discussion of the problems concerning task classification and the use of different types of dependent variables.1
Summing up, the present meta-analysis showed that DA at encoding produced a negative, albeit mild, effect on indirect tasks, contradicting the idea that implicit memory is exclusively based on automatic encoding processes (Parkin et al., 1990). More importantly, it turned out that the model representing the distinction between identification and production priming provided the best fit with empirical data, whereas the numerical difference between perceptual and conceptual priming was unreliable.
In addition, our data support the hypothesis that, as far as the difficulty of the secondary task is concerned, two critical factors are: a) the frequency of response to interference items and b) the synchronicity between the presentations of target and distractor stimuli (Mulligan et al., 2007). Finally, the present results are also in agreement with the proposals that performance in word-stem completion might be affected by conceptual processes (Gabrieli, 1991 and Keane et al., 1991), and that perceptual identification might include a production component (Mulligan & Peterson, 2008).
