MNEMONIC ADVANTAGESOF THE SELF
(e.g., Kuiper & Derry, 1982; Symons, 1990). As researchers continue to examine specific processing strategies that may pro- duce the SRE, these issues should be illuminated.
Theoretical Differences Between Person-Reference and Semantic Processing
The results of several model tests are interesting in that they allow us to speculate about the differences between person- reference and semantic processing. Our conclusions are based on a complex of model tests that were significant (or not sig- nificant) for one manipulation class versus the other. Specifi- cally, some model tests were significant for the SR-OR class but not for the SR-semantic class; some, such as stimulus type, show different patterns for the two classes. The results of these model tests allow us to theorize about differences between the SR-OR and SR-semantic manipulation classes and, in general perhaps, between person-reference and semantic encoding.
The first important theoretical model that we tested that re- flects differences between person-reference per se (SR-OR comparisons) and SR-semantic comparisons is the model test for stimulus type. Results show that, when nouns were used as stimuli, there was no SRE for the SR-OR class. However, when traits were used as stimuli, the studies obtained a significant mean SRE that was significantly larger than the mean SRE
two manipulation classes. In both manipulation classes, there is a significant difference between studies that used recognition versus those that used free recall. In the SR-semantic class, however, SREs were significantly larger for studies that used recall rather than those that used recognition; in fact, studies that used recognition tended not to observe SREs at all. In contrast, SR-OR studies that used recognition observed sig- nificantlylarger SREs on average than those that used free recall. It seems then that the presence of retrieval cues following an SR-semantic manipulationprovides an advantage following se- mantic encoding. A recognition task may provide cues that sup- ply an advantage equivalent to that inherent in an OR or SR task, which may have the benefit of a category label to facilitate retrieval. In contrast, the presence of retrieval cues following OR seems to disrupt retrieval, resulting in larger SREs than those obtained with free-recall tasks. It may be that experi- menter-imposed retrieval cues may actually interfere with sub- jective categorical structures imposed by the participant when encoding words about another person. Although this explanation is purely speculative, this pattern may be related to our findings for the next set of model tests, which indicate that OR may be sensitive to interference with short-term memory stores. As a final precaution, we note that we did not compare semantic encoding with OR directly; therefore, care should be observed with regard to comparisons of the two types of processing.
found for nouns. In contrast, in the SR-semantic class, a sig- nificant mean SRE appeared when either traits or nouns were used as stimuli, but the SRE was significantly larger with traits than nouns.
OR Tasks: Sensitive to Disruptions in Short-Term Memory?
Many researchers have speculated about the ways in which traits are represented in memory (e.g., Breckler, Pratkanis, & McCann, 1991). It is interesting that nouns seem to be equally well remembered in both SR and OR (i.e., person-referent) conditions across studies in the SR-OR manipulation class. This finding may have implications for the assertion that nouns are sometimes part of the self-representation (e.g., Klein & Kihlstrom, 1986). If we make the assumption that nouns are not part of the self-representation and the SRE is indeed a "self- based" phenomenon, then SR encoding of nouns should not be expected to pose an advantage for later memory (Maki & McCaul, 1985). However, if we hold to the assumption that nouns are not an integral part of the self-concept (and that, therefore, SR should not facilitate memory for nouns), we must also be able to explain why it is that we find an SRE for SR- semantic manipulations that used nouns as stimuli. The most likely and parsimonious explanation for the finding of no SRE when nouns are used in SR-OR studies is that, of the 26 studies that used nouns, 13 also used imagery tasks, which tend not to find SREs on average, as we discussed earlier. The rest of the studies used nonstandard tasks, either judgments about a non- specific other (for the OR task; k = 8) or tasks that were judged to be unusual and could not be included with other classes. Again, this is an important line of questioning by which pro- cessing assumptions about SR versus other kinds of processes could be tested by careful attention to the kinds of stimuli that are presented to participants.
The model test for dependent variable is consistent with our earlier discussion of the effects of organization and elaboration. Results show opposite patterns for recall and recognition in the
The final set of models that we examined are specific to the SR-OR manipulation class and suggest that OR may be particularly sensitive to task restrictions that interfere with short- term memory stores. First, the model for expectation of test shows that there was no SRE when participants expected to have their recall tested following the encoding task but that studies using incidental learning paradigms observed SREs on average. Second, the use of distractor tasks resulted in a larger mean SRE than when distractor tasks were not used in SR-OR studies. Because (a) the expectation of a memory test should increase the use of rehearsal strategies and (b) distractor tasks disrupt any rehearsal strategies that participants may decide to undertake, both of these model tests suggest that rehearsal is disrupted and results in a disadvantage for retrieval following OR that does not affect SR-semantic comparisons. It is interest- ing that both of these models are consistent with Kuiper and Rogers's (1979) theory that OR may require some sort of re- hearsal strategy to be mnemonically effective.
Two interesting continuous models suggest that OR may be sensitive to task restrictions that affect time or short-term mem- ory immediately following encoding tasks: SREs increased (a) as time between encoding and retrieval increased and (b) as length of stimulus presentation decreased. These findings sug- gest that, the longer the time between encoding and retrieval, the more likely it is that SR will promote more memory than OR. (These patterns also hold true for semantic encoding, al- though the magnitude of the effect is somewhat smaller.) For OR manipulations, this finding is further support for the idea that OR may benefit from a rehearsal strategy: The longer parti- cipants are required to retain the information in short-term mem-