Retrieval of autobiographical memories: The mechanisms and consequences of truncated search

Abstract

Five studies examined the extent to which autobiographical memory retrieval is hierarchical, whether a hierarchical search depends on central executive resources, and whether retrieving memories that are “higher” in the hierarchy impairs problem‐solving ability. The first study found that random generation (assessed using a button‐pressing task) was sensitive to changes in memory load (digit span). The second study showed that when participants fail to retrieve a target event, they respond with a memory that is higher up the hierarchy. The third study showed that memory is more generic only when participants use low imageable cues under cognitive load. The final two experiments showed that experimental manipulation of memory specificity affects problem solving (MEPS performance). The data are consistent with Conway and Pleydell‐Pearce's hierarchical retrieval model of autobiographical memory, and suggest that overgeneral memory in nonclinical participants is associated with reduced executive capacity only when retrieval is “top‐down” (generative).

Notes

This research was supported by the Wellcome Trust (GR06779 (University of Oxford) and by the Medical Research Council (University of Bangor; Grant No G9621313). Special thanks to Paul Wallace who carried out some of the testing in Bangor.

There are a number of ways to assess randomness. We piloted tested 6 parameters: chi square, mean difference, auto correlation index, Evans RNG, triplets, and a phase measure. We found that Evans RNG was the most sensitive to differences between human participants and computer‐generated or tabled random numbers (Healy, 1997, pp. 161–163). This replicated previous findings (Spatt & Goldenberg, 1993) so we used the RNG measure.

Note that this time limit is less than that usually required for the AMT. For this study, we had been concerned that a longer time interval might allow too much time when the participant would not need to be retrieving. The disadvantage of this procedure is that mental effort of retrieval is confounded with the effort of reporting. However, there is no way of avoiding this unless the randomisation task stops when the participant begin to respond. Given the need to collect 100 responses to generate a reliable index of randomness, we decided to give participants cue words in relatively quick succession.

The following procedure was used to establish the reliability of the AMT for this series of studies. The experimenter's (H.H.) ratings for a random sample of 10 participants for a prior experiment involving 18 cues varying in imageability and valence from the same database (180 responses in all) were given to an independent rater who was blind to the experimental hypotheses. This yielded a reliability coefficient of .82, comparable with previous studies in our laboratory. This experimenter's ratings were then used for Studies 2 and 3. If there were any responses for which H.H. had doubts, she consulted with J.M.G.W. who gave a decision remaining blind to the treatment condition of the participant.

We are grateful to an anonymous reviewer for this suggestion.

For a reason that remained unclear, this effect was only found in male participants by Goddard et al. In the current study we were unable to recruit sufficient male participants to be able to verify whether our results interacted with gender.

Note that this design does not deal with the possible confounding of retrieval and reporting noted for Study 2, though the proportion of time allocated to retrieval overreporting should be relatively greater in this present study.

Interestingly, although this was not known when these studies were designed, Watkins and Moulds (in press) have now found a direct link between type of self‐focus and problem solving. In clinically depressed patients, a self‐focus manipulation (derived from Nolen‐Hoeksema & Morrow's 1991 rumination manipulation), previously found to increase specific autobiographical memory recall, improved the effectiveness of problem solving, relative to a self‐focus manipulation known to reduce specific autobiographical memory recall (Watkins & Teasdale, 2001, 2004). Importantly, in this study, the specificity of mental representations for the problems produced by the different manipulations mediated their effect on problem solving, indicating a causal link between memory specificity and problem solving.

Note that this scale is different to that used in Study 3. However, note that in this case we are not interested in these data as a measure of outcome, but only to check that the induction had worked.

Ideally, a formal test of mediation would be conducted to examine the impact of memory induction condition on problem‐solving performance. However, in the current study, because the memory induction was so successful, induction condition and memory specificity were completely confounded, with very little nonoverlapping variance to be explained.