cient for memory of input to be facilitated in comparison to a semantic orientation task" (p. 392). Moreover, after 2 decades, researchers seem divided between those who are willing to attri- bute the SRE to special mnemonic properties of the self (e.g., Maki & McCaul, 1985; Rogers et al., 1977) and those who are not (e.g., Brown, Keenan, & Potts, 1986; Klein & Kihlstrom, 1986).
These inconsistencies were the impetus for our meta-analytic investigation. The first important question we address is whether SR tends to facilitate memory more effectively than other encod- ing tasks. That is, how consistently robust is the SRE? By ad- dressing this question, we provide an empirical test of conclu- sions by narrative reviewers (e.g., Higgins & Bargh, 1987) that variation in the results of SRE studies imply that it is not SR per se that facilitates memory. Our second and more important question concerns the conditions under which the SRE is most likely to occur. That is, assuming that SRE reviewers were cor- rect to conclude that studies' findings were inconsistent, we feel that it is most important to detail methodological features of. SRE studies that can be used to account for these inconsisten- cies. Obviously, such explanations represent the most theoreti- cally interesting aspect of this investigation.
In this article, we adopt the perspective that the SRE results primarily because the self is a well-developed and often-used construct in memory that promotes both elaboration and organi- zation of encoded information--a perspective first advanced by Klein and Loftus (1988). The respective roles of elaboration and organization in the SRE have been discussed by several researchers (e.g., A. G. Greenwald & Banaji, 1989; Klein & Kihlstrom, 1986; Klein & Loftus, 1988). In general, these re- searchers emphasized the role of ordinary memory processes, which leads to the conclusion that there is nothing special or unique about SR that renders it more effective as a mnemonic device than other encoding tasks. According to Klein and Loftus, although SR may be ordinary in the sense that it can be ex- plained by these properties, it is distinguished from many other comparison tasks (e.g., synonym judgments) in the sense that SR promotes both elaboration and organization simultaneously, resulting in a mnemonic advantage. However, only Klein and Loftus's study provides empirical support for the joint elabora- tion-organization model. Their findings imply that, across all the studies in the literature, SR tends to result in superior mem- ory compared with tasks that promote either organization or elaboration separately (e.g., a synonym judgment task) but not combined.
We further submit that, to the extent that SR is spontaneous or habitual (e.g., Markus, 1977), the major benefit of SR lies not in its ability to invoke organizational or elaborative pro- cessing per se but rather its likelihood to spontaneously create matching between encoding and retrieval conditions (see Wells, Hoffman, & Enzle, 1984). This effect distinguishes the SR task from other tasks and may be the primary reason why it promotes memory more than other kinds of processing in the typical incidental learning situation.
This study provides a meta-analytic integration of the SRE literature that allows an examination of the consistency and generality of the SRE and the conditions under which it is likely to occur. Furthermore, it provides a meta-analytic test of the joint roles of elaboration and organization. Specifically, we present
evidence that the SRE occurs as a result of two features of the experimental task in a typical SRE study: (a) the nature of the comparison task (person-reference vs. semantic processing) and (b) the likelihood that the semantic or other-reference (OR) comparison task promotes both organization and elaboration. First, we show that comparisons that involve person reference (e.g., SR vs. OR) have smaller SREs relative to SR- versus semantic-encoding comparisons. Even more interesting, we show that comparisons of SR versus OR appear to be sensitive to certain task parameters (e.g., expectation of a test) that do not affect comparisons of SR with semantic processing. In addi- tion, we show that the mnemonic advantage following SR tends to diminish when SR is compared with encoding tasks that invoke memory structures, which resemble the self in terms of the amount of development and use (e.g., OR tasks in which the target is very well known). Second, we show that, when the task compared with SR is judged to promote both elaboration and organization of stimulus words, the SRE is smaller than when the comparison task promotes only organization (rela- tional processing) or only elaboration (item-specific processing; Klein & Loftus, 1988).
Historical Overview of SRE Research
In a seminal pair of studies, Rogers et al. (1977) extended the depth of processing (DOP) paradigm (Craik & Tulving, 1975) to the realm of the self. The basic strategy of the DOP paradigm is to compare the responses produced by encoding tasks that are presumed to differ in depth, or extensiveness, of processing (Eysenck & Eysenck, 1979). Essentially, DOP the- ory assumes that recall is a function of trace elaboration at the time of encoding. Differences in responses to different tasks, therefore, reflect underlying differences in the processes used to encode stimulus materials, such as word lists. DOP research- ers had already demonstrated that semantic-encoding tasks ( "Does the word mean the same as xxx?" ) resulted in superior recall compared with phonemic ("Does the word rhyme with xxx?") or structural encoding ("Does the word have capital letters?") tasks (e.g., Craik & Tulving, 1975). Rogers et al. used these standard DOP encoding tasks and added a new one, an SR task (i.e., "Does the word describe you?"). Showing that memory was even better for the SR condition than the semantic condition, Rogers et al. concluded that the self acts as a "superordinate schema" (p. 686) to facilitate encoding and retrieval of the information.
Researchers in subsequent studies generally confirmed that SR produces superior memory relative to semantic encoding. These researchers obtained an SRE with different (a) SR-encod- ing tasks (e.g., self-descriptiveness, autobiographical retrieval, imagery; Bower & Gilligan, 1979; Brown et al., 1986); (b) to- be-remembered materials (e.g., traits, nouns, prose; Bellezza, 1984; Klein & Loftus, 1988; Maki & McCaul, 1985; Reeder, McCormick, & Esselman, 1987); and (c) populations (e.g., children and adults, Pullyblank, Bisanz, Scott, & Champion, 1985; participants with and without depression, Derry & Kuiper, 1981). The SRE appeared to be a robust phenomenon and quickly surfaced in introductory social and cognitive psychology textbooks.
However, along with those who found SREs, researchers also