Source monitoring and memory distortion MARCIA K. JOHNSON Department of Psychology, Princeton University, Princeton, NJ 08544-1010, USA ([email protected]) SUMMARY Memory distortion re£ects failures to identify the sources of mental experience (reality monitoring failures or source misattributions). For example, people sometimes confuse what they inferred or imagined and what actually happened, what they saw and what was suggested to them, one person's actions and another's, what they heard and what they previously knew, and ¢ction and fact. Source confusions arise because activated information is incomplete or ambiguous and the evaluative processes responsible for attributing information to sources are imperfect. Both accurate and inaccurate source attributions result from heuristic processes and more re£ectively complex processes that evaluate a mental experience for various qualities such as amount and type of perceptual, contextual, a¡ective, semantic and cognitive detail, that retrieve additional supporting or discon¢rming evidence, and that evaluate plausibility and consistency given general knowledge, schemes, biases and goals. Experimental and clinical evidence regarding cognitive mechanisms and underlying brain structures of source monitoring are discussed. 1. INTRODUCTION (Johnson & Raye l981; Johnson l985, 1988, 1991, 1997). Reality monitoring failures occur when people misat- The problem of verifying memories and separating true tribute something that was re£ectively generated to from false memories has received national attention in perception or vice versa. According to this view, recent years, for example in highly publicized cases of reality is not directly given in remembering, but is an children's testimony and apparently long-forgotten attribution that is the outcome of judgement processes. memories of childhood abuse that are remembered by Reality monitoring is just a particularly interesting adults, sometimes in the course of therapy (Ceci & subset of the more general cognitive activity of source Bruck 1993; Loftus 1993). These socially signi¢cant monitoring, which includes distinguishing between instances point to fundamental issues about human external sources or between internal sources, as well as memory in general. Psychologists have known for between external and internal sources. For example, some time that memory does not record experience source monitoring includes identifying who told you like a video camera (Bartlett 1932). When we read something, whether you saw an event in real life or on novels or listen to friends or see an accident, we are television, whether an event happened a week ago or a comprehending the information in light of our expecta- month ago, and whether you told a secret to your friend tions, making assumptions and inferences and ¢lling in or only thought about telling it. We have described based on prior knowledge, and perhaps failing to notice source monitoring as an attribution which is the or incorporate information that is incongruent with our outcome of judgement processes that take into account currently active cognitive schemas. Likewise, when we certain phenomenal characteristics of memories, as well think back on events and remember them, the same as consistency checks, plausibility judgements, and interpretive, inferential and motivational processes are other extended retrieval and reasoning strategies (e.g. operating. The resulting memories include not only Johnson 1988; Johnson et al. 1993; Johnson & Raye perceptual information but our inferences, imagina- 1981). Memory distortions can happen in many tions and thoughts (e.g. Bransford & Johnson 1973; see contexts and forms, but most memory distortion also Alba & Hasher 1983; Johnson & Sherman 1990, involves failures in monitoring the source of informa- for reviews). tion. To manage as well as we do, given such a construc- tive memory system, we proposed that the system must have some mechanisms or processes designed to help us 2. A SOURCE-MONITORING sort out information that was primarily derived from FRAMEWORK perception from information that was derived from our My colleagues and I proposed an integrative frame- thoughts; we called these processes reality monitoring. work for characterizing and investigating processes That is, reality monitoring refers to the processes involved in identifying the origin of information: the involved in discriminating memories and beliefs gener- source-monitoring framework (SMF) (Johnson et al. ated by re£ection from those derived from perception 1993; Johnson & Raye 1981). This approach explains Phil. Trans. R. Soc. Lond. B (1997) 352, 1733^1745 1733 & 1997 The Royal Society Printed in Great Britain 1734 M. K. Johnson Source monitoring and memory distortion both veridical and distorted memory with a common set A third fundamental idea is that the judgement of principles. Fundamental are several ideas. First, an processes used in any given situation and the criteria event memory consists of sets of phenomenal character- adopted will be a¡ected by such factors as the nature istics, which are the result of multiple distributed of the task, cost of mistakes, our preconceptions, cognitive processes during the event (and after if the amount of distraction and the social context. For memory is reactivated or retrieved) (e.g. Johnson 1983, example, criteria are likely to be more lax in a casual 1992). Second, memories from di¡erent sources conversation with a friend (e.g. recalling the source of di¡er in characteristic ways that can, in principle (but gossip) than in a professional meeting (e.g. recalling the only on average), be used to identify the origin of source of scienti¢c ideas), or more lax if a memory ¢ts information. with what one already believes or wants to believe than For example, perceived and imagined classes of events if it does not. di¡er in average value along a number of dimensions or attributes. Memories originating in perception typically have more perceptual detail or features (e.g. colour, (a) Sources of error and distortion in memory sound), more contextual detail such as time and place, Using this source-monitoring framework, there are more semantic detail and more a¡ective information. various ways in which source memory errors could In contrast, memories originating in re£ection typically occur (for recent reviews see Ceci & Bruck 1993; have more accessible information about cognitive opera- Johnson et al. 1993; Roediger 1996; Schacter 1995; tions, that is, about those perceptual and re£ective Wilson & Brekke 1994): The schema-based inferences processes that took place when the memory was estab- we make in simply comprehending events initially can lished. Judgement processes capitalize on these average later be misattributed to perception. For example, if di¡erences. Thus, di¡erences in average value between someone tells you that John pounded the nail and you perceptually and re£ectively derived memories along infer he used a hammer, you may later be likely to these dimensions form one basis for deciding the origin claim that the speaker told you John had a hammer of a memory. For example, one would be likely to decide (Johnson et al. 1973). Anything that disrupts the that a memory with a great deal of perceptual informa- binding of features (e.g. semantic content, spatial loca- tion and very little cognitive operations information was tion, colour) into a complex event memory, such as externally derived (Johnson et al. 1979, 1981). Similarly, stress, distraction or brain damage, will reduce source one might decide that a statement was said by person A memory (Chalfonte et al. 1997; Jacoby et al. 1989; rather than person B because the auditory information Schacter et al. 1984; Shimamura et al. 1990). Certain in the memory matches A's voice or the visual informa- types of emotional focus can disrupt feature binding as tion matches A's appearance (e.g. Ferguson et al. 1992; well (Johnson et al. 1996b). Memories that are similar Johnson et al.1996b). in perceptual features or semantic content can easily Other processes involved in source monitoring tend be confused (Johnson et al. 1988a; Lindsay et al. 1991). to be more complex. You may decide when an event Misinformation introduced through suggestion can be occurred because you can relate it to another event misattributed to a target event as shown by studies of memory that contains more de¢nitive time informa- eyewitness testimony (Belli & Loftus 1994; Zaragoza tion; or you may decide you read something in the & Lane 1994). Reactivating and retrieving memories newspaper rather than saw it on TV news because you strengthens them, but also potentially embellishes remember thinking it was consistent with the newspa- them (Suengas & Johnson 1988; Ceci et al. 1994). Any per's position, or because you cannot retrieve disruption of the ability to revive or evaluate additional perceptual information about the broadcast that you supporting or discon¢rming information may create feel you should be able to if you had seen it on TV. source monitoring errors (e.g. Craik 1982; Dodson & That is, more extended source monitoring may involve Johnson 1996; Jacoby 1991). Inappropriately lax criteria beliefs about memory and cognition as well as can produce source monitoring failures (Lindsay & retrieving additional information from memory and Johnson 1989; Dodson & Johnson 1993; Multhaup evaluating the source of the target memory given these 1995). And source memory is a¡ected by the social beliefs, other speci¢c