MNEMONIC ADVANTAGESOF THE SELF
word in a semantic condition describes themselves. This kind of interference should occur less frequently in a between-subjects design, thus the SRE could be larger. The question is an empiri- cal one, however, because virtually all studies in the literature that involvedintimate other targets used within-subjects designs.
Finally, two moderators--participant population and mode of presentation--were found to exert opposite effects for the two manipulation classes. For the SR-OR class, larger SREs were found when undergraduate students were not the participants; in the SR-semantic class, when undergraduates were the partici- pants, significantly larger SREs were obtained than for other populations. This is a difficult finding to explain, and one that has received little attention because it has not been considered important theoretically (except to the degree that, e.g., the popula- tion possessed some characteristic that affects SR, such as depres- sion; Kuiper & Rogers, 1979). An examination of the 17 nonun- dergraduate studies shows that the samples were from a variety of other populations (e.g., participants with depression, healthy adults recruited from local communities, etc.). We present the results of this model as a design consideration that may be of importance to some researchers who choose to test other popula- tions, whether out of convenience or necessity.
The second moderator, mode of presentation, is similarly in- triguing in that it results in opposite patterns for the two manipu- lation classes. Specifically, in the SR-OR class, SREs were not obtained on average when a projector was used (monitors or orally presented stimuli resulted in significantly larger SREs). In contrast, projectors resulted in the largest SREs for the SR- semantic class. Different modes of presentation have largely been taken for granted in SRE study designs; presumably, the choice has been made based on convenience and history. The differences that we found may point to subtle differences in processing that may occur as a consequence of the choice of presentation mode. In particular, it should be noted that, in many cases in which projectors are used, the choice is often made, because the experimenter wishes to test more than one partici- pant at a time. It may be that, when an SR-OR manipulation is used, the presence of other people may provide cues for participants (even though fellow participants are not the refer- enced target) that may help to facilitate subsequent retrieval. Although admittedly speculative, it is an interesting question that merits consideration in future primary research.
used highly intimate target others did obtain a significant mean SRE, this effect was significantly smaller than that obtained for studies that used low-intimacy target others. It should be noted that familiarity of the target other had no effect on the magnitude of observed SREs for studies in this class? We specifically included ratings of both familiarity and intimacy in the meta- analysis because it was obvious that researchers did not discrim- inate between these variables. For example, Johnny Carson may be referred to as a highly familiar target in some studies, but he is not likely to be a highly intimate target to the average participant. Thus, it does not come as a complete surprise that familiarity does not result in a significant model test when both one's mother and Carson are classified as highly familiar targets. Our hypothesis that intimacy is a significant predictor of the magnitude of the SRE was supported. Results show a weak, but still significant, SRE for studies in the high-intimacy class, and study findings are homogeneous. Even though the referenced target is highly intimate, there is still a slight advantage of SR. Moreover, there is a significant difference between the class of studies that used highly intimate targets and low-intimacy tar- gets, with significantly larger SREs in the low-intimacy class. Thus, reference to highly intimate others promotes memory al- most as well as SR.
Thus, consistent with our hypotheses and contentions in the literature, results suggest that referencing a very well-known other, information about whom is presumably well organized and elaborated on in memory, has a facilitating effect on recall. Because the self is, presumably, even better known and informa- tion about the self has been even more frequently elaborated on and organized, SR facilitates memory even better. As Aron et al. (1991) have suggested, to the degree that a relationship is intimate, it is also likely to overlap with one's self-representa- tions. Memory representations of one's mother are very likely to be fraught with SR associations, for example. It is possible that autobiographical retrieval may be more sensitive to this possibility than are trait descriptiveness tasks. However, only two studies in the literature used an episodic retrieval task in the SR-OR class .(Bower & Gilligan, 1979), and both used high-intimacy targets, thus no comparison to test this possibility was possible. Future research should address this possibility.
Type of task was shown to be a significant moderator of SRE size in the SR-OR class. Results show that studies that used
Variation Within the Two Manipulation Classes
At this point, we turn to a number of task parameters in SRE studies that explain variation within the two manipulation classes (SR-OR and SR-semantic). Several of these model tests suggest pragmatic issues in the design of SRE studies that researchers may have overlooked that directly affect the magnitude of the SRE; others have theoretical significance and may suggest new avenues of exploration. We begin with the set of moderators, unique to a particular manipulation class, that suggest either important task parameters that significantlyaffect the magnitude of the SRE or important theoretical variables.
Models for the SR-OR class. only to the SR-OR class were
Three model tests that pertain significant: (a) intimacy of the
target other, (b) type of SR task, and (c) results of the first model test show that,
type of OR task. The although studies that
imagery tasks did not observe SREs on average, confirming Lord's (1980) hypothesis that the SRE will not occur with imagery instructions. Moreover, whereas studies that used imag- ery tended not to obtain SREs, all other types of tasks (SR or OR) in the SR-OR class did tend to produce SREs. Model tests for both type of SR task and type of OR task reveal that studies that used descriptiveness tasks found significantly larger SREs, on average, than any other type of task. Lord offered an explana- tion for this finding: He theorized that imagery involves the same kind of visual processing a person uses when he or she
3We performed continuous model tests as well for both familiarity and intimacy.The continuous models were based on ratings made by undergraduates who had judged the intimacy and familiarity of each referencedtargetused in the SRE literature.The findingsfor the continu- ous models parallel the categorical models we presented based on our codings. We thus omitted them to avoidredundancy.