3 January 1995
Episodic Memory and Autonoetic Consciousness:
Developmental Evidence and a Theory of Childhood Amnesia.
Josef Perner
Laboratory of Experimental Psychology, University of Sussex,
and Max-Planck Institute for Psychological Research, Munich.
and
Ted Ruffman
Laboratory of Experimental Psychology, University of Sussex,
Running Head: EPISODIC MEMORY & AUTONOETIC CONSCIOUSNESS
published as:
Perner, J. & Ruffman, T. (1995). Episodic memory and autonoetic consciousness: Developmental evidence and a theory of childhood amnesia. Journal of Experimental Child Psychology,59, 516-548.
Abstract
This research draws together Tulving's (1985) view on episodic memory and research on children's developing "theory of mind". Episodic memory, in its technical meaning given by Tulving, requires the autonoetic consciousness of having experienced remembered events, but developmental findings suggest that children cannot encode events as experienced before the age of about 4 or 5 years. Before this age they have insufficient understanding of what constitutes experience, specifically they do not reflect on the perceptual origin of their own knowledge. To demonstrate such a link children between 3 and 6 years were assessed for their understanding of the perceptual origin of their own knowledge on different "see-know tests," in particular a test assessing understanding that our senses inform only about certain aspects of the perceived objects. A significant association was found between passing see-know tests and free recall, which persisted even when cued recall and verbal intelligence are partialled out. These results are used to argue that between 3 to 6 years children develop the ability to remember events as experienced and that this development can explain adults' inability to have recollective experiences of childhood events before that age (childhood amnesia).
Episodic Memory and Autonoetic Consciousness:
Developmental Evidence and a Theory of Childhood Amnesia.
Tulving (1985) proposed--a view which he traced back to Ebbinghaus (1885)--that different types of memory are bound up with different levels of consciousness. In particular, a distinction is drawn between knowing about past events (which forms part of semantic memory in Tulving's terminology) and remembering them. Only remembered facts constitute episodic memory. Facts are said to be 'remembered' if, in Ebbinghaus' words, they are "recognized as something formerly experienced." (Tulving 1985, p. 3). This autonoetic (self-knowing) component of having experienced an event was also emphasized by William James (1892) as the distinguishing feature of 'memory' (in the strict sense): "Memory, says James, 'is the knowledge of an event, or fact ... with the additional consciousness that we have thought or experienced it before.'" (as cited by Lockhart, 1984, p. 133).
For instance, consider two ways in which one might retain the fact that the word "butter" was on a list. For merely knowing that there was the word "butter" on the list it is sufficient to represent the proposition there was the word "butter" on the list. In contrast, for an act of proper remembering, in this specific narrow sense, it is also necessary to represent the additional fact that one has seen the word on the list and that, therefore, one knows that the word has appeared on the list. In other words, one has to represent one's own act of experiencing. These additional representations are autonoetic, because they represent one's own knowledge and experiences and they make higher order experiences possible, namely the experience of having experienced the events that one knows about. This higher order experience is a kind of recollective experience (Tulving, 1985, p. 8) which confers the special phenomenal flavor to the remembering of past events (p. 3).
There is relevant developmental work suggesting that children may lack the prerequisite representations for recollective experiences until they are 3 to 5 years old. This is not to say that children before that age may not, in Tulving's words, know something about the past as they certainly do. For instance, Fivush and Hamond (1990) have shown that even at 2½ children can retrieve much detail about a trip to the zoo. Such demonstrations of early competence, however, do not show that children have recollective experiences of remembering these past events. Such recollective experiences may not be had before the age of 3 or 5 years, because children up to that age do not understand the connection between informational access (e.g., seeing) and knowledge (Wimmer, Hogrefe and Perner, 1988; Wimmer, Hogrefe and Sodian, 1988). Without such understanding children could have no or at least only very impoverished recognition of what they know as formerly experienced, because part of what it means to experience an event is forming knowledge of this event by observing it.
Wimmer, Hogrefe, and Perner tested 3- and 4-year-old children. They either let children observe which object was put inside a box or told them what the object was or gave them no information at all. Although children at all ages were highly accurate specifying whether they knew what was in the box or not, there was a sharp distinction between 3- and 4-year-olds when asked to provide reasons for their knowledge or ignorance. Hardly any 3- but practically all 4-year olds answered correctly that they knew because they had been shown, or told, or didn't know because they had not been shown or told.
Although there is some evidence (Hogrefe, Wimmer & Perner, 1986; Pillow, 1989; Pratt & Bryant, 1990) that many 3-year-old children make some link between seeing and knowing, it seems clear that they do not understand the real importance of seeing (informational access) for knowing because when asked explicitly about their reason for knowing, as in the original paradigm by Wimmer, et al. (1988a), few of them succeed. Their failure cannot be due to the form of question asked, because same age children are quite proficient in answering a similar question about the reason for why a person is hungry (Perner & Ogden, 1988). Furthermore, their difficulties persist even when the question is posed as a forced choice, i.e., "Do you know because I showed you or because I told you?" (Gopnik & Graf, 1988; Poole, 1988).
Following the theoretical analysis above, the implications of these findings for the development of episodic memory are striking. Before the age of about 4 years children cannot have episodic memory in Tulving's sense. That is, they are not able to remember in the narrow sense of having recollective experiences of having experienced known events. The question that faces us now is how this theoretical claim can be empirically substantiated?
Existing evidence for Tulving's distinction between knowing (semantic memory) and remembering past events (episodic memory) as a function of a difference in consciousness is based on adults' ability to make this distinction phenomenally. For instance, Gardiner (1988) instructed subjects to write an "R" for remember, if their recognition of a word was accompanied by a conscious recollection of its prior occurrence in the list or a "K" for know if the word was recognized on some other basis. It was found that subjects are willing to use these labels and that frequency of attributions can be systematically manipulated. Using similar instructions Tulving (1985) reported that items recalled under free recall are more often deemed "remembered" than items which could only be retrieved with the help of semantic cues. Further, Tulving (1985) showed that over the first day the proportion of correctly recognized items which are judged to be "remembered" decreases with retention time faster than the proportion judged "known". Then both proportions decline at an equal rate for the next 6 months (Gardiner & Java, 1991). Similarly, the number of "remember" judgments declines as a proportion of the total number of words correctly recognized when encoding conditions change from semantically more demanding conditions to less demanding conditions, i.e., finding a meaningful associate to the target word versus just a word that rhymed, having to generate the target word versus just reading it (Gardiner, 1988), being presented rare (low frequency) words versus high frequency words, or words versus non-words (Gardiner & Java, 1990). In contrast, when processing resources are diverted by a distractor task the opposite effect occurs: the proportion of "known" words increases relative to "remembered" words (Gardiner & Parkin, 1990).
A particularly compelling finding in all these recognition studies is that "remembered" items were less prone to include false positives than "known" items. Yet, Gardiner & Java (1990) found that "remember" and "know" judgments are not just judgments of confidence, since confidence judgments ("sure", "unsure") changed in a different direction in response to the word versus nonword manipulation and in response to masked repetition (Rajaram, 1993). Similarly, Parkin & Walter (1992) found that in old age the proportion of "remember" judgments declines while "know" judgments increase and that no such change is observed in the proportion of "sure" and "unsure" judgments.
Unfortunately, this direct method of asking subjects to judge whether they remembered or just knew on the basis of their phenomenal experience is not useful for testing young children. It is difficult enough to get the intended phenomenal distinction across to adult subjects, not to mention very young children. So we need to find a less direct but more child adequate method of assessment. To this end we go back to Tulving's (1985) first demonstration experiment.
Tulving argued that to retrieve items from memory one has two types of information available: semantic cue information and episodic trace information. In cued recall semantic cue information is quite rich whereas in free recall such information is relatively poor since no explicit retrieval cues are given. Hence, free recall depends much more heavily than cued recall on the availability of episodic trace information, in particular information about the experience of having seen the item on the list (experiential awareness). It is for that reason, Tulving argues, that almost every item retrieved under free recall was judged as "remembered", whereas under cued recall relatively more items were judged as merely "known" because they were retrieved with the help of cues only, without adequate episodic trace information for a genuine recollective experience to justify a judgment of "remembered".
So, younger children, who fail tests assessing their understanding of how seeing (informational access) relates to knowing, should do much worse on free recall tasks than children who pass such tests. This is so, because children who fail these tests should lack the autonoetic awareness that they know something because they had experienced it, and because without this awareness they lack the essential episodic trace information which is particularly helpful for free recall. In contrast, no, or at least a much smaller difference should exist between these two groups of children in terms of cued recall since these two groups are not expected to differ in their ability to use semantic cue information.
EXPERIMENT 1
This experiment tests the hypothesis that there is a developmental relationship between children's growing conceptualization of their own experiences and episodic memory in Tulving's sense. As an indirect measure of episodic memory children's relative success on free over cued recall was used, since free recall depends on episodic trace information more heavily than cued recall. Progress in conceptualizing experience was assessed by measuring children's understanding of the origin of their own knowledge in two different "see-know tests".
One test (why-know test) is a replication of the method used by Wimmer, et al. (1988a) asking children for the reason for their knowledge or ignorance of a particular fact. However, since Perner (1991, ch. 7) argued that this method might overestimate children's actual understanding of the see-know relationship, the more stringent know-guess test was also used. It tests whether children understand that their knowledge crucially depends on the availability of relevant information independent of other typical indicators of knowledge, i.e., successful action. Such understanding is required in the "lucky-guess" condition of the know-guess test. The child is prevented from seeing where something is put but, by sheer luck, finds it. The child is then asked whether "she really knew where the object was or whether she just guessed". Since successful search is a very strong indicator of knowledge, children who do not yet understand the full significance of informational access answer that they knew (because they found the object). Only children who understand that there can be no knowledge without informational access will understand that they did not know and just guessed, despite their success.
The overall logic of this experiment is to see whether there is a substantial difference in free recall between children who pass a particular see-know test than children who fail on that test. Any difference in free recall between these groups must go beyond any potential difference in cued recall.
Method.
Subjects.
Fifty-eight children (36 girls and 22 boys) from a kindergarten in a rural community near Munich, Bavaria participated in this study. There were 28 "4-year olds" aged 3 years and 4 months (i.e., 3;4) to 4;5 and 30 "5-year olds" aged 4;6 to 5;6.
Material and Procedure.
Children were tested in a quiet room of the kindergarten. They were seated opposite to the experimenter at a low table. Each child was given 4 memory tasks and 2 see-know tasks consisting of 3 subtasks each. Two of the memory tasks involved free the other two cued recall. In one task of each recall condition items were presented in order by category while in the other task they were presented in random order. This difference in presentation was introduced to create tasks of different difficulty in order to avoid potential problems with ceiling or floor effects.
Administration of all tasks was spread out over two sessions a few days apart. Each session started with presentation of items for a memory test, followed by one of the 6 subtasks (which took about 2 minutes each) constituting the 2 see-know tests, followed in turn by recall of memory items, another see-know subtask, presentation of material for the second memory task, a third see-know subtask, and recall of the second set of memory items. The different memory conditions and see-know subtasks were presented in random order, with the exception that the free recall condition with mixed presentation of items was always presented first (to avoid the possibility that preceding tasks may induce children to spontaneously structure the material by categories).
Memory Tasks. For the memory tasks 12 sets of 4 pictures of familiar objects or animals belonging to the same category were used (see Appendix for list of items). Three sets were randomly chosen for each task. Each set was used only once for a particular subject. Pictures were presented one after the other and left in the child's view until the whole set was shown. As each picture was produced the child was asked to name it. Slight misidentifications (e.g., calling the lorry a car, or the pot plant a flower) were accepted. Completely wrong answers were corrected.
Each set of pictures was randomly assigned to the 4 experimental conditions created by the 2 factors of recall (free vs. cued) and presentation (mixed vs. by category). For presentation by category, pictures were arranged in 3 rows one for each category, and children were prompted to name pictures in each row by asking, e.g., "Tell me what kind of FRUIT these are!", "Tell me what kind of LIGHTS these are!", etc. For mixed presentation, pictures were put down in random order and children were simply asked, "Tell me what is in the pictures!". After all pictures were on the table children were instructed to "look at the pictures carefully, because I will take them away and ask you later what was in them!". A few seconds later the pictures were picked up.
After an intervening see-know subtask children were reminded: "I've shown you some pictures earlier on. Can you remember?" In free recall conditions they were then simply asked "What was in these pictures?". In cued recall conditions they were asked for each category: "There were some, e.g., pieces of FURNITURE, what were they?"
See-Know Tasks. There were two tasks assessing children's understanding of the connection between informational access and their resulting knowledge of where an object was put: Their ability to explicitly justify how they knew where the object was (how-do-you-know task), and their ability to differentiate between knowing and guessing (know-guess task). Each of these tasks consisted of 3 subtasks. They all involved the same material: a green and blue box and some candy which was put inside one of the boxes.
In one of the 3 subtasks of the How-Do-You-Know Task children simply observed the candy being put inside one of the boxes. In the other two subtasks the candy was put away behind a large screen that had been erected between the boxes and the child. In one of these tasks, however, the child was told where the candy was while in the other no information was given, so that the child did not know where the candy was. The child was then asked whether he/ she knew where the candy was. Mostly correct answers to this question were anticipated. The real measure of interest was the ensuing justification question asking for the reason for the child's knowledge or ignorance: "How do you know that?" or "Why don't you know that?".