Encoding Specificity predicts an increase in memory performance accuracy when features present during the encoding of information are also present during testing (Tulving & Thomson, 1973). These features include contextual cues found in the environment (i.e., context dependent memory; Smith, 1979). To the extent that environmental elements present during encoding overlap those present at test, memory performance is predicted to improve; for example, Mead and Ball (2007) used background music in major or minor keys to produce context dependent memory effects for lists of words. In addition to features in the surrounding environment, context effects have been found for cues that participants encounter while processing the targets themselves, such as words presented on landscape photographs (Hockley, 2008, Experiment 2) or on colored backgrounds (Isarida & Isarida, 2007).
It is conceivable that for sets of items, inter-item relationships may also qualify as contextual cues, as studying the same target content repeatedly in the same order may cause particular items to be encoded as part of the context for adjacent items. As is found with other types of contextual cues, this inter-item context would be expected to assist retrieval if the studied order of the target items remained the same on the test. Under these conditions, test scores would reflect retrieval of item information boosted by context-cue assistance. Tests on which the target content is ordered differently from that in which the items were studied would fail to provide these inter-item cues and scores would be expected to decline.
Studying a set of items under conditions where the order constantly changes would be expected to prevent the formation of inter-item contextual cues, as the same items are unlikely to precede or follow each other more than once. Scores on the test would now be more likely to represent retrieval of the target information itself, unaffected by matching or mismatching inter-item context. (1)
Several researchers have used college students to compare the effects of matching or mis-matching the order in which information is presented during lecture (or in a textbook) and the order of the corresponding items on the test of that material. In general, test item sequence is either consistent with that in which the information was initially presented, or test items are presented in a randomized order. Despite predictions based on both encoding specificity and the memory benefits of contextual overlap, with few exceptions (Balch, 1989), experiments conducted over the last 30 years have failed to find significantly better scores on matching-presentation ordered tests (Goss Lucas & Bernstein, 2005; Neely, Springston, & McCann, 1994; Perlini, Lind, & Zumbo, 1998; Schmitt & Scheirer, 1977; Spiers & Pihl, 1976; Tal, Akers, & Hodge, 2008).
Two potential explanations for this discrepancy between predictions made by well-established theory and experimental results can be made, depending on the logic behind comparing the lecture order of information to the test order. The first is that the lecture presentation order was intended to be a proxy for the order in which students study that information. In this case, measurement error will be introduced to the extent that students deviate from that presented order. This kind of error often results in "mixed" findings across experiments (as it has on this topic). A second possibility is that existence of inter-item context effects between lecture content and test items themselves was being investigated. Context effects are unlikely to be seen in this situation as both encoding specificity and context dependent memory theories describe the process of episodic memory formation during active learning. Mere exposure to information (e.g., listening to a lecture) may not produce the target-context relationship necessary for an overlapping benefit at test. …