Analogy Causes Distorted Memory by Blending Memory Episodes

Analogy Causes Distorted Memory by Blending Memory Episodes Margarita Pavlova (margarita.velinova@gmail.com) Boicho Kokinov Department of Cognitive Science and Psychology, New Bulgarian University, 21 Montevideo Street Sofia 1618, Bulgaria Abstract The present study compares two cases of blending between episodes: blending of episodes that share a number of elements (superficially similar episodes) and blending of episodes that share the same structure of relations but do not share the same elements. According to classic theories and models of blending, superficially similar episodes are more likely to be blended because there is a larger overlap of the feature vectors representing them. In contrast, according to the AMBR model of analogy-making and memory, analogous episodes are more likely to be blended. The results obtained in the present study support the prediction of the AMBR model: people blend structurally similar episodes much more often than superficially similar ones. These results are consistent with previous experiments on the influence of analogy- making on constructive memory. Keywords: analogy; memory; memory distortions; blending of episodes; psychological experimentation. Constructive Memory Two popular metaphors for memory are known to cognitive psychologists. The classic one suggests that it is a physical place where our memories for events, concepts, and objects are stored and later on retrieved. The more recent metaphor for memory views it as a dynamic and constructive process (Moscovitch, 1995; Loftus, 1997; Neisser & Harsch, 1992; Roediger & McDermott, 1995; Schacter, 1999). This relatively modern concept of memory suggests that we reconstruct episodes by taking a small piece of information about the episode and then with the help of general knowledge about the category to which it belongs, we reconstruct it altogether (Kokinov & Petrov, 2001). According to this constructivist point of view on memory, recollections can not only be partial or lost, but they can also be distorted. There are two known mechanisms of memory distortion – episode blending and schematization (use of general schematic knowledge). There is much evidence now that people do distort real episodes. Sir Frederick Bartlett (1932) first demonstrated a case of memory distortions as a result of intrusions from general schematic knowledge in the reconstruction of the target event. Loftus and Palmer (1974) enriched these findings and demonstrated that depending on the schema activated, the reconstructed episodes varied. Furthermore, various other experiments by Loftus (1977; 1997; 2003) have indicated various instances of blending of episodes – a slideshow and a piece of consecutive information concerning the target object; a real event and an imagined one; a personal experience and an advertisement; an actual autobiographical recollection and an implanted one. All these findings suggest that misinformation alters one’s recollections about an event regardless of it being actual or not. One important conclusion is drawn from these studies: people pick elements from one event and implant them into another, thus mixing up the two events and blending the recollections of both. Simarly, the case study of John Dean’s memory (Neisser, 1981) provided an analysis of his recollections and revealed a lot of inaccuracies. Discrepancies and fault memories are common even for highly emotional events. Neisser & Harsch (1992) focused on the so called flashbulb memories and explored the participants’ immediate and delayed recollections of the Challenger incident, showing that the participants wrongly recalled the event and reported various details most probably taken from different events, hence they strongly believed that their memories were accurate and vivid. Nystrom and McClelland (1992) also obtained similar effects: after studying a number of sentences, their participants blended them. Roediger and McDermott (1995) demonstrated that a word not present in a list could be wrongfully recalled if it was associated with some of the words presented in the list. All of the above mentioned data have been extended and supported in a number of psychological experiments, as well as in cognitive neuroscience, brain imaging, and brain lesions studies (Moscovitch, 1995; Schacter, 1999; Kokinov & Hirst, 2003). Despite the extensive experimental work demonstrating memory distortion, the mechanisms of memory construction are still a topical issue. Various researchers suggest different points of view on the mechanisms and explanation of memory construction. This lack of theory ground in the field makes the research move in various directions with no clear predictions and explanations. Among the few existing models that do explain and reproduce episode blending are Murdock’s TODAM model (1993; 1995), McClelland’s Trace Synthesis Model (1995), and Metcalfe’s CHARM model (1990). Although they are quite different in a number of aspects, these models share a