Abstract
The primacy effect in memory for repetitions describes the phenomenon that participants, when presented with repeated items, are more likely to remember attributes of the first presentation than attributes of later repetitions (DiGirolamo & Hintzman, 1997). Interestingly, this effect is assumed to be due to less thorough encoding of the second and later repetitions. This account resembles the explanation typically invoked to explain spacing effects in item memory. Based on the similarity between the accounts of the spacing effect and the primacy effect in memory for repetitions, we suggest that the primacy effect of repetitions should also be moderated by the lag between repetitions. This prediction is tested in two experiments using a source memory paradigm that allows for studying the primacy effect of repetitions, spacing effects in item memory, and source memory in a unified paradigm. The results show that the primacy effect for memory of repetitions occurs at short lags between repetitions but not under long lags.
Acknowledgements
We thank Simone Malejka for her help with the data collection and André Aßfalg, Jan Rummel, Thorsten Meiser, and Edgar Erdfelder for helpful comments on an earlier draft.
Notes
1Hintzman and Curran (1995), were the first to note the similarity of the accounts of the spacing effect and the registration without learning effect. But see also Malmberg and Shiffrin (2005), who proposed a similar explanation to account for moderator effects of massed versus spaced strengthening operations on the list strength effect.
2 This measure is analogous to the measure used by DiGirolamo and Hintzman (1997) to investigate the primacy effect. However, the current measure is conditionalised on item memory so it reflects the proportion of items attributed to one source only (e.g., the first source of presentation) given the item was recognised as old. Because DiGirolamo and Hintzman did not include distractors in their memory test, no such conditionalising was necessary in their study.
3 Hit rates and false alarm rates were computed according to the correction suggested by Snodgrass and Corwin (Citation1988), that is, adding a constant of 0.5 to the frequency of hits and false alarms and adding a constant of 1 to the number of old items and the number of new items in order to avoid hit rates of 1 and false alarm rate s of 0. Analyses with d’ as dependent variable yielded the same conclusions in all analyses and are thus not reported.
4 We thank an anonymous reviewer for bringing this point to our attention.
5 We thank an anonymous reviewer for raising this important issue.