حافظه های اپیزودیک

An account of episodic memories is developed that focuses on the types of knowledge they represent, their properties, and the functions they might serve. It is proposed that episodic memories consist of episodic elements, summary records of experience often in the form of visual images, associated to a conceptual frame that provides a conceptual context. Episodic memories are embedded in a more complex conceptual system in which they can become the basis of autobiographical memories. However, the function of episodic memories is to keep a record of progress with short-term goals and access to most episodic memories is lost soon after their formation. Finally, it is suggested that developmentally episodic memories form the basis of the conceptual system and it is from sets of episodic memories that early non-verbal conceptual knowledge is abstracted.

Tulving's concept of episodic memory is principally about a type of memory system ( Tulving, 1983 and Tulving, 1985a). Understanding episodic memory as a system is unquestionably important ( Schacter & Tulving, 1994). There are, however, other valuable questions that can be asked here too and some of these relate to episodic memories as memories. That is, as mental representations with distinct properties, organization in long-term memory, and which are manifest in particular brain circuits. These features are considered in the sections that follow, in the spirit of Tulving (1983), and they lead to proposals about the ontogeny and phylogeny of episodic memory more generally. 1. Properties of episodic memories Table 1 lists nine properties of episodic memories that collectively distinguish them from other types of memory representations. Note that, it is the combination of these features that are proposed as defining. Other types of long-term memory representations may feature one or several of the properties but only episodic memories have all nine properties. These properties are now considered in further detail, as are some of the issues they raise for the study of episodic memories. Table 1. Nine properties of episodic memories. 1. Contain summary records of sensory-perceptual-conceptual-affective processing. 2. Retain patterns of activation/inhibition over long periods. 3. Often represented in the form of (visual) images. 4. They always have a perspective (field or observer). 5. Represent short time slices of experience. 6. They are represented on a temporal dimension roughly in order of occurrence. 7. They are subject to rapid forgetting. 8. They make autobiographical remembering specific. 9. They are recollectively experienced when accessed. Table options The first property in Table 1 captures the idea that episodic memories are summary records of experience. That is to say that they are experience-near and correspond to experience but they are not literal records of experience. In certain special cases, memory for highly self-relevant self-defining moments for instance, they may contain fragments that are close to being literal, veridical, records of experience. Something like this appears to occur in, for example, memory for trauma, (see Holmes, Grey, & Young, 2005) when highly specific details are retained. These can take various forms, for example a vivid recollection of the texture of the material on a car door that a road traffic accident victim looked down on to avoid witnessing the head on impact of an oncoming vehicle, a blue flash of light when an electric cable snapped as a train derailed, the thought a person had during a traffic accident, and so on (see Conway, Meares, & Standart, 2004). Some caution must be exercised here as it is also the case where such details can be incorrect too – the only point is that some are correct and appear to be very close to actual experience, they can also cause an intense feeling of reliving when brought into consciousness. Vivid details like these are also often reported for less traumatic but nonetheless highly personally important self-defining experiences and for memories of significant public events, so called ‘flashbulb’ memories (see Conway, 2005 and Conway and Pleydell-Pearce, 2000; Conway, Singer, & Tagini, 2004, for reviews). Again, these too may not always derive from external experience and, as with all episodic memories, they are also associated with more conceptual knowledge. In general, however, it is suggested that episodic information is more summarized and generic, more representative of an experience than it is a literal record. This proposal raises interesting questions that research has yet to address. For instance, how is experiential processing transformed into episodic information? I think it fair to note that currently we have very little idea, and even less evidence, about how this transformation takes place. Baddeley's (2001) notion of an episodic buffer in working memory constitutes one promising approach to this question. Another thinking about the role of hippocampal processes in configuring knowledge and generating associations ( Squire, 1992) is also highly relevant. Nevertheless exactly how knowledge is extracted from experience and represented in an episodic memory is unknown (but see Moscovitch, 1995, for an particularly interesting account). Although it seems that in everyday experience this is a largely non-conscious process and not within intentional control. Interestingly, much the same might be said about property two in Table 1 which focuses on the fact that once an episodic memory is formed, episodic information (episodic detail) within the memory is differentially accessible. Racsmany and Conway (2006) developed this idea in a series of experiments that showed that items that were inhibited and therefore difficult to access were only inhibited in an episodic memory of their acquisition. The same items if accessed in conceptual, non-episodic, representations were either not inhibited or were primed. Thus, somewhat paradoxically the same item can be simultaneously inhibited (in it's representation in an episodic memory) and primed (in it's representation in conceptual knowledge structures). This pattern was termed episodic inhibition. Episodic inhibition captures the idea that in any episodic memory there is a pattern of activation/inhibition over the contents of the memory and it is this pattern that determines the accessibility of episodic details: inhibited details are difficult to access and activated details have their access facilitated. The activation/inhibition levels of details in an episodic memory are probably determined by many factors, although we have suggested that the goal structure of an experience may be critical in that it drives attention, action, and affect, and must thereby influence encoding processes, cf. Conway (2005). Indeed, one of the main functions of episodic memories might be to keep a highly specific record of aspects of experience relevant to recent goal processing ( Brewer and Dupree, 1983 and Lichtenstein and Brewer, 1980; Williams, Conway, & Baddeley, 2008). Being able to remember in a relevant way, rather than literally, what has been recently said and/or done is critical to focussed and fluent everyday cognition and action: exactly the type of relatively routine daily behaviour that patients with, for example, anterograde amnesia can find so difficult. Details in an episodic memory are then at different levels of accessibility. Some of these levels will be preset by encoding others will emerge over repeated episodes of accessing a memory and yet others will be determined the nature of the search and cues that feature in it. A question of considerable interest here is: What in memory represents the boundaries of an episodic memory? Indeed, boundaries must be especially important as they separate memories into discrete entities and because of that are, presumably, highly available. Unfortunately relatively little is known about this (see Williams et al., 2008, for some recent findings). Properties 3, 4, and 5 in Table 1 concern the nature of episodic memories. It has long been known that episodic memories are dominated by imagery and particularly by visual imagery (Brewer, 1988 and Galton, 1883). Indeed, brain damage that leads to the loss of the ability to generate visual images may as a secondary consequence give rise to retrograde amnesia (Conway, 1996, Conway, 2005 and Rubin and Greenberg, 1998). This type of retrograde amnesia is marked by a lack of specificity in memory although more general knowledge of the patient's life is often retained (Conway, 1996 and Conway, 2005). Visual images contain information that is configural. That is to say that the objects represented in a visual image are represented in relation to each other and because of this visual images may maximize the amount of information they contain (Conway, 1988). The contents of a visual episodic image may be highly sensitive to visual cues that can activate the image, either by some direct mapping from cue to image content or, perhaps, by some mapping onto relations configured in the image. As visual processing is so central to human cognition it follows that visual episodic images are probably being accessed a great deal of the time (although they do not necessarily enter conscious awareness) and are generally highly responsive to visual cues. Interestingly, it seems that visual episodic memories always have a perspective. This was first noticed by Henri and Henri (1898), further developed by Freud (1915), and reintroduced into modern memory research by Nigro and Neisser (1983); see too (McIsaac and Eich, 2002 and Robinson and Swanson, 1993). In the modern parlance memories are said to have a ‘field’ or an ‘observer’ perspective (Nigro & Neisser, 1983). A visual episodic memory with a field perspective is considered to preserve a person's original perspective or something approximating to that perspective. In contrast in a visual episodic memory with an observer perspective the rememberer looks into the memory and sees themselves in the memory. Although systematic properties of field and observer memories have been investigated, for example field memories have been found to be more strongly associated with recollective experience than observer memories (Libby and Eibach, 2002 and McIsaac and Eich, 2002), and recent memories are more likely than older memories to be retrieved with a field perspective (Robinson & Swanson, 1993), there remain interesting aspects of episodic perspective yet to be investigated. In our laboratory for instance we have often asked our participants when they reported observer memories: what do you look like? The answers are surprisingly varied. For memories from childhood a representation of themselves from a photograph is often reported or a sort of stereotyped image is described. For memories some years old but not from childhood the image of the self in observer memories is frequently a generic image rather than a specific image from a unique moment. For more recent memories the images of the self in a memory are more frequently episodic than generic. In contrast we found field perspective to be less varied. It seems, as Freud (1915) originally noted, that the observer perspective in an episodic memory indicates more memory construction and the incorporation of other knowledge into an episodic memory. That this appears to occur for older compared to recent memories perhaps indicates a slow integration of episodic memories with more generic and conceptual autobiographical knowledge. Episodic memories, especially visual episodic memories, represent short time slices of experience (Anderson and Conway, 1993, Neisser and Harsch, 1992 and Williams et al., 2008). It is, however, neither known how this occurs nor what the nature of these time slices is. We have proposed that the boundaries of episodic memories are marked at the opening boundary by information about actions and at the closing boundary by facts that are often details about the outcomes of actions (Williams et al., 2008). This is consistent with our more general view that episodic memory and autobiographical knowledge are about goals. That is to say they preserve information that is highly relevant to goal processing including goal generation, plan execution, outcomes and evaluations (for related views see Brewer and Dupree, 1983 and Lichtenstein and Brewer, 1980). We have suggested that one function of episodic memories is to provide a means to accurately check on recent progress with current goals, i.e. in the preceding few minutes, hours, days, or last few days. In providing this detailed record of progress with very specific goals episodic memory also forms the basis for future goals and plan-implementation. Tulving (2002) has pointed to the important role of episodic memory in underpinning goals and plans for the future and how these fail in anterograde amnesics who cannot retain episodic memories. It is an aspect of episodic memory generally and specific episodic memories in particular that remains under-investigated. Finally, and briefly, consider the last four properties in Table 1. The notion of temporal order is hardly a new one in the study of episodic memory and indeed an early and simple model of random access memory was able to show how temporal order can be retrieved even from a relatively unsophisticated system (Landauer, 1975; see Kahana, Howard, & Polyn, 2008 for a recent review). The ability to recall the day's events, in backward and/or forward order, may well be a key cognitive ability that underlies coherent planning and goal pursuit. The preservation of temporal order in episodic memory may then be especially important. One aspect of this that has not been investigated is the feelings of beforeness and afterness that preservation of temporal order may support. Feelings ultimately about the direction of time that must also include at least an anticipatory sense of the immediate future. It is interesting that patients with, for instance, closed head injuries and wide spread brain damage often complain of losing feelings like these (we refer to them as cognitive feelings see Conway, 2005) and because of that are disoriented in time. The temporal dimension in episodic memory extends then both backward and forward in time and we have recently termed this the remembering-imaging window ( Conway, in press). The idea partly derives from the observation that in terms of brain activations remembering and imaging barely differ ( Conway, Pleydell-Pearce, Whitecross, & Sharpe, 2003; Schacter & Addis, 2007). Thus, remembering the past and imaging the future take place in the same system, a system we might call the remembering-imaging system. The notion of a remembering-imaging ‘window’ relates to work currently ongoing in our laboratory in which participants list as many specific memories as they can for yesterday, 2 days ago, 3 days ago, and so on for several more days. There is a steady decrease in the number of memories listed with increasing retention interval and, interestingly, at about the 3-day retention interval and further back memories appear that are not specific but instead are more general, more generic, and much more concerned with routines and schema than with specific episodic memories although, of course, some of these persist. The same participants then also list, day-by-day, specific events they plausibly expect to occur in their lives over the next 5-day period. A similar pattern is observed and the number of specific events listed decreases day-by-day and after about the 2/3 day mark there is a sharp rise in listings of routine schematic events. Some specific anticipated events do of course appear at later points but these are not numerous. This then is the remembering-imaging window in which specific memories for recent experiences and anticipated future experiences constitute a window of episodic consciousness that functions to keep us tightly connected to our current goals and plans. Despite the importance of episodic memories in preserving temporal order and, possibly, supporting our feelings of beforeness, afterness, nowness, and anticipation of the future, episodic memories do not endure for lengthy periods of time in long-term memory. It is clear that many episodic memories from a single day can be recalled at the end of the day. But as the retention interval increases access to many of these memories is rapidly lost. Few episodic memories can be recalled, even of a distinctive day, at a retention interval of one week and of a mundane routine day virtually none (Williams et al., 2008). This process of rapid forgetting, although first reported over a century ago by Ebbinghaus (1885/1964) is one of the under-investigated properties for episodic memories of everyday events. Recent work in our laboratory using a portable camera worn around the neck that automatically takes photographs in response to sensory changes, e.g. in movement, luminance, etc., suggests that the many episodic memories formed in a typical day are not themselves lost and can often be accessed using sequences of photographs from the camera (Berry et al., 2007 and Loveday and Conway, 2008; St.Jacques, Conway, Lowder, & Cabeza, 2009). This raises the fascinating prospect of large numbers of long-lasting but inaccessible episodic memories enduring in long-term memory, perhaps across the full lifespan. The potential of such large sets of inaccessible, but available, episodic memories to have what might be quite powerful implicit effects, is intriguing. It perhaps suggests too, that give sufficiently effective cues it might be possible reinstate, at least partly, a previous window of remembering-imaging. Finally with respect to Table 1 consider the role of episodic memories in providing memory specificity and in triggering recollective experience (properties 8 and 9). Memory specificity is interesting. Why should we have specific memories when surely a general knowledge of the world and some generic recollection of the past would almost certainly be sufficient to ensure survival? If one function of episodic memories is to keep an adaptive record of recent goal processing so that progress with goals can be readily assessed then clearly episodic memories need to be specific enough to provide the appropriate information. Hence, perhaps, the high degree of specificity that they have in the short-term at least. I suggested that only a relatively small proportion of episodic memories remains accessible in long-term memory and perhaps this is because of their relevance to longer-term goals and consequently to the future. It is particularly noteworthy that clinically depressed patients who have over-general memories (lack of normal access to specific episodic memories) also have over-general futures that lack specificity in plans and goals (see Williams et al., 2007, for a review). It may be that for plans and goals to be specific they have to be grounded in specific episodic memories of both the past and the future. A further factor driving the long-term retention of specific episodic memories relates to their role in learning and the acquisition of knowledge. In the main anterograde amnesics who apparently no longer have accessible episodic memories also show major impairments in learning and especially in the acquisition of new knowledge. Although it appears that some greatly reduced acquisition of new knowledge, for example new words does take place it is, nonetheless, not even remotely close to the levels and extent of acquisition of new knowledge present when episodic memory is intact. To-be-sure, there are certain individuals with extensively reduced hippocampal volume and abnormal episodic memory whose learning can approach normal levels (Vargha-Khadem et al., 1997). In all these cases it seems likely that there is residual functioning hippocampal circuitry that supports some, usually very little, knowledge acquisition (Squire, 1992). Or possibly these low levels of learning are mediated by other areas of brain that can support in a minor way some hippocampal functions. There is, however, evidence that episodic memories are crucial in the acquisition of new knowledge and learners may pass through a phase during which knowledge is gradually abstracted from episodic memories in the process of becoming part of more general long-term conceptual knowledge (Conway, Gardiner, Perfect, Anderson, & Cohen, 1997). Indeed, episodic memories may remain closely associated with some long-term conceptual knowledge (Conway, 1990 and Westmacott and Moscovitch, 2003), although the more usual case is that as conceptual knowledge is acquired access to memories of moments when the knowledge was first processed is rapidly lost. The suggestion is then that episodic memories are the basis of concept acquisition and may be especially critical when new knowledge is to be acquired. This proposal raises interesting questions about the development of memory and these are returned to later.