The attention/likelihood theory (ALT; M. Glanzer & J. K. Adams, 1990) and the retrieving effectively from memory (REM) theory (R. M. Shiffrin & M. Steyvers, 1997) make different predictions concerning the effect of list composition on word recognition. The predictions were empirically tested for two-alternative forced-choice, yes-no, and ratings recognition tasks. In the current article, the authors found that discrimination of low-frequency words increased as the proportion of high-frequency words studied increased. The results disconfirm the ALT prediction that recognition is insensitive to list composition, and they disconfirm the predictions of the REM model described by R. M. Shiffrin and M. Steyvers (1997). The current authors discuss a slightly modified version of REM that can better predict our findings, and we discuss the challenges the present findings pose for ALT and REM.

A solution to the problem of context-dependent recognition memory is presented in terms of the item, associated context, and ensemble (ICE) theory. It is argued that different types of context effects depend on how context information is encoded at both learning and retrieval. Matching associated context in memory and a retrieval cue produces increases in both hit and false alarm rates and may not be accompanied by a change in discrimination. Integrating item and context information in an ensemble and matching ensemble information in memory and a retrieval cue produces context-dependent discrimination. Empirical support for these predictions is presented.

Researchers must consider limitations and assumptions inherent in their measure of source monitoring when drawing conclusions, an important point raised by Murnane and Bayen (1998). However, the issues they raise do not invalidate the conclusions we draw from the findings reported in Henkel and Franklin (1998). Issues regarding conclusions about source monitoring performance, the relation between recognition and source accuracy, and the use of empirical and multinomial analyses are discussed.

A number of studies have reported age differences in memory for the source of information. S.A. Ferguson, S. Hashtroudi, and M.K. Johnson (1992) suggested that older adults do not efficiently use multiple distinctive characteristics of sources to distinguish between sources in source memory tasks. In the study reported here, participants heard information from 2 sources and later decided whether test items had been presented by Source A, by Source B, or were new. The distinctiveness of both perceptual and temporal characteristics of sources were independently manipulated. Older adults benefited more than young adults from multiple distinctive characteristics of sources. These results question the generality of S.A. Ferguson et al.'s hypothesis.

Source identification refers to memory for the origin of information. A consistent nomenclature is introduced for empirical measures of source identification which are then mathematically analyzed and evaluated. The ability of the measures to assess source identification independently of identification of an item as old or new depends on assumptions made about how inconsistencies between the item and source components of a source-monitoring task may be resolved. In most circumstances, the empirical measure that is used most often when source identification is measured by collapsing across pairs of sources (sometimes called “the identification-of-origin score”) confounds item identification with source identification. Alternative empirical measures are identified that do not confound item and source identification in specified circumstances. None of the empirical measures examined provides a valid measure of source identification in all circumstances.

Source monitoring refers to the discrimination of the origin of information. Multinomial models of source monitoring (W. H. Batchelder & D. M. Riefer, 1990) are theories of the decision processes involved in source monitoring that provide separate parameters for source discrimination, item detection, and response biases. Three multinomial models of source monitoring based on different models of decision in a simple detection paradigm (one-high-threshold, low-threshold, and two-high-threshold models) were subjected to empirical tests. With a 3 (distracter similarity) x 3 (source similarity) factorial design, the effect of difficulty of item detection and source discrimination on corresponding model parameters was examined. Only the source-monitoring model that is based on a two-high-threshold model of item recognition provides an accurate analysis of the data. Consequences for the use of multinomial models in the study of source monitoring are discussed.

Theoretical analyses and empirical studies address the issue of how context-dependent recognition is affected by changes in the relative strength of retrieval cues. Analyses of global memory models based on K. Murnane and M. P. Phelps' (1994) general context model showed that if context strength is held constant, context effects are predicted to either increase or remain unchanged when item strength increases. In contrast, the outshining hypothesis (S. M. Smith, 1988, 1994) predicts that context effects will decrease as item strength increases. Three studies are reported in which item strength was manipulated with spaced repetitions, study time, or a levels-of-processing manipulation. The results support the general context model. Implications for the outshining hypothesis and for global memory models are discussed.

How nested contexts and shared contexts affect episodic memory is an important and often overlooked problem for general theories of memory. A simple modification of Humphreys et al.'s (in press) computation-level theory of memory of proposed that solves both problems.

The effects on recognition of changes in environmental context between learning and test are examined. A context effect occurs when memory tests that take place in an environmental context that is different from the learning context produce consistent differences in performance. A formal model of context-dependent recognition within a global activation framework is presented. The model generates the predictions that (1) context effects will be present when items are tested in a new context that was not seen during learning and (2) context effects will be absent or very small when items are tested in a context that was experienced during learning but that differs from the context in which the test item was learned. Both predictions were verified in an experiment that varied the nature of the different-context test within subjects. Implications for research concerned with context-dependent recognition are discussed.

B. B. Murdock and M. J. Kahana (1993a) presented a continuous memory version of the theory of distributed associative memory (TODAM) model; they claimed that this model predicts list-strength and list-length findings, including those reported by R. Ratcliff, S. E. Clark, and R. M. Shiffrin (1990) and K. Murnane and R. M. Shiffrin (1991a). This model is quite similar to one discussed by R. M. Shiffrin, R. Ratcliff, and S. Clark (1990), who rejected the model on the basis of its inability to predict both an absent or negative list-strength effect (when strength is varied by repetitions) and a present list-length effect. In this comment we elaborate the earlier discussion and demonstrate that the version of TODAM proposed by B. B. Murdock and M. J. Kahana (1993a) indeed fails for this reason. We show this first for a somewhat simplified version of the model for which derivations are obvious and then in a simulation of the complete version using the parameter values suggested by B. B. Murdock and M. J. Kahana (1993a).

A number of prior studies have not found declines in recognition performance when testing occurs in an environmental context that is different from the learning context. These findings raise serious problems for global activation theories of recognition which predict that hit and false alarm rates will decline when the test context does not match the learning context. Environmental context was manipulated as a unique combination of foreground color, background color, and location on a computer screen in three experiments using intact-rearranged recognition testing and two experiments using single-item testing. Changes in context resulted in reduced hit and false alarm rates as predicted by global activation theories in all five experiments. Mental reinstatement of the learning context was also examined

Most current models of memory predict that the presence of increasingly well-learned, or strong, items in memory will cause increasing interference. This phenomenon, the list-strength effect, occurs as predicted when memory is tested by free recall but not when a recognition test is used. Four experiments use end-of-session testing to demonstrate that redistribution of storage time or effort from strong to weak items on mixed lists does not occur and therefore cannot be masking interference by strong items. Delay between study and test is found to cause memory loss independent of the basic list-strength findings. It is concluded that the presence of strong items in memory does not interfere with recognition performance and that interference is due to failures of retrieval rather than to composition or other forms of destructive interaction during storage.

When some items on a list are strengthened by extra study time or repetitions, recognition of other, unrelated, list items is not harmed (Ratcliff, Clark, & Shiffrin, 1990). Shiffrin, Ratdiff, and Clark (1990) accounted for this list-strength finding with a model assuming that different items are stored separately in memory, but that repetitions are accumulated together into a single stronger memory trace. Repeating words in the context of different sentences might cause separate storage of the repetitions of a given word, because either word or sentence traces are stored separately. Separate storage would, in effect, convert a list-strength manipulation into a list-length manipulation and thereby induce a positive list-strength effect. In Experiment 1, this result was produced for single-word recognition and for two types of sentence recognition. In Experiment 2, both words and sentences were repeated together, which should have caused repetitions to be stored in a single, stronger, trace. As expected, the list-strength effect was eliminated. A sentence trace model was fit to the data, supporting the account of Shiffrin et al. (1990) and supporting an account of word and sentence recognition in which activation is summed for representations of all list items. The results from the two studies are inconsistent with most current models of memory (as shown by the theoretical analyses of Shiffrin et al., 1990) and pose an additional challenge for theory.