Memory and Theory of Mind

XXX

Draft of 10 June 1999

Memory and Theory of Mind

Josef Perner

University of Salzburg, Austria

Draft of Chapter 21 in E. Tulving and F.I.M. Craik (Eds.). The Oxford Handbook of Memory. New York, NY: University of Oxford Press.

"Theory of mind" is primarily a label for the research area that investigates the conceptual system that underlies our ability to impute mental states (what we know, think, want, feel, etc.) to others and ourselves. The study of these concepts is essential for our understanding of memory insofar as memory is not just storage of information, but is also dependent on knowledge of our own information storage processes. In Tulving and Madigan's (1970, p. 477) words we should "start looking for ways of experimentally studying and incorporating into theories ... of memory one of the truly unique characteristics of human memory: its knowledge of its own knowledge." To have such higher order knowledge one needs a concept of knowledge and other mental states. Memory development should, therefore, be seen in the light of the acquisition of mental concepts, i.e., the child's growing theory of mind.

1. What is Our Theory of Mind?

Our theory of mind is constituted by our everyday mental concepts. The analysis of these concepts is influenced by traditions in analytical philosophy of mind. They are characterized as propositional attitudes (Russell, 1919), i.e., they describe a relation between an organism (person) and a proposition, e.g., the sentence, "Max knows that the chocolate is in the cupboard" describes knowing as a relation between Max and the proposition 'the chocolate is in the cupboard'.

Organism ------Attitude------> Proposition

e.g.: Max knows the chocolate is in the cupboard

Since cognitive science has taken on board this common sense view of the mind, an important question is, how such a relationship to a proposition can be implemented. The representational theory of mind (RTM, Field. 1978; Fodor, 1978) assumes that a propositional attitude consists in holding a representation of the proposition and that this representation plays a certain functional role in the economy of mental states. This can be best illustrated with the two core concepts: belief and desire. These are core concepts since knowing what someone believes (thinks) to be the case (e.g., Max thinking the chocolate is in the cupboard and thinking that going there will get the chocolate into his possession) and what that person desires (wants) (e.g., Max wanting the chocolate to be in his possession), allows us to make a behavioral prediction that Max will approach the cupboard. This kind of inference is known since Aristotle as the Practical Syllogism.

Searle (1983, after Anscombe, 1957) points out that these two states are mirror images in terms of causal direction and direction of fit. The function of a belief is to be caused by reality and the believed proposition should match reality. For instance, The chocolate being in the cupboard should be responsible for Max believing that the chocolate is in the cupboard (world to mind causation) and the proposition 'the chocolate is in the cupboard' should thus match the relevant state of affairs in the world (mind should fit world). The function of desire (want) is to cause a change in the world (mind to world causation) so that the world conforms to the desired proposition (world should fit mind), e.g., if Max wants the chocolate to be in the cupboard, then this desire should cause action leading to a change of the chocolate's location such that it conforms to what Max desires.

This trivially sounding example does highlight the important point that our mental states are not represented but are constituted by the functions that the represented propositions are to fulfil. It also serves to highlight some important distinctions.

1.2 Three Important Distinctions

First- vs. third person. One important distinction is between first-person and third-person attribution of mental states. A third person attribution is an attribution to another person and a first person attribution is one to myself. For instance, if Max erroneously believes that the chocolate is still in the cupboard (because he didn't see that it was unexpectedly put into the drawer) then a third person observer will attribute a false belief to Max. In contrast, Max himself will make a first-person attribution of knowledge to himself. The observer can capture this difference between her own and Max's subjective view by the second-order attribution that Max thinks he knows where the chocolate is. This is useful to keep in mind when it comes to false memories. Since a memory can only be a recollection of something that actually occurred, a false memory is not a memory by third-person attribution, although it is by first-person attribution.

Sense and reference. A related second point has to do with Frege's (1892/1960) distinction between sense and reference. Since mental states involve representations, they connect us to objects and events in the real (or a possible) world. Famously, Oedipus knew and married Iocaste (referent: a particular person) but he did not know or marry her as his mother but as an unrelated queen (sense: how Iocaste was presented to Oedipus' mind). Thus, in third-person parlance we can say that Oedipus married his mother, if we use the expression "his mother" to pick out (refer to) the individual that he married without implying that he knew Iocaste under that description. In first-person description of the event Oedipus would not have used the descriptor "my mother". These distinctions are useful to keep in mind when discussing infants’ ability to remember particular events: Whenever a memory trace of a unique event can be demonstrated then one can conclude (in third person parlance) that the infant remembers that particular event. However, one cannot conclude that the event is remembered by the infant (in first-person parlance) as a particular event, i.e., that the infant makes cognitive distinctions that represent that event as a particular event.

Having vs. representing a mental state. The third important distinction is that between BEING in a mental state (or HAVING an attitude) and representing that mental state. For understanding or knowing that a person is in a mental state, or to reflect on one's own mental states, one has to be able to represent that state. In order to be able to represent a state one needs a concept of that state, i.e., a rich enough theory of mind. The study of how children acquire the requisite theory of mind is therefore essential for our understanding of how children come to understand memory. Furthermore, since some memorial states are reflective or self-referential, children need a theory of mind for being in such states or having such memories.

2. Why We Need a Theory of Mind for Memory: Consciousness

We probably do not need a theory of mind for implicit memory but for explicit memory we do, since "explicit memory is revealed when performance on a task requires conscious recollection of previous experiences.” (Schacter, 1987). To be conscious of a fact one requires to be also aware of the state with which one beholds that fact. The higher-order-thought theories of consciousness make this their core claim (Armstrong, 1980; Rosenthal, 1986). For instance if one SEES a state of affairs X (e.g., that the chocolate is in the cupboard) then this seeing is a first-order mental state (attitude). To be conscious of this state of affairs means, according to theory, that one entertains a second-order thought about the seeing, i.e., the second-order thought represents the first-order seeing. A weaker version does not require that one has to entertain the second-order thought but only that one has to have the potential for having the second-order thought (Carruthers, 1996). That some such condition must be true can be seen from the following consideration: Could it ever be that I can genuinely claim that I am consciously aware of the chocolate being in the cupboard, but claim ignorance of the first-order mental state by which I behold this state of affairs, i.e., by claiming that I have no clue as to whether I see, or just think of, or want the chocolate being in the cupboard.

The important point of these conceptual analyses is that to be conscious of some fact requires some minimal concept of knowledge or of some perceptual state like seeing. Unfortunately, there is no clear evidence when children understand a minimal state of this sort. There is some evidence of understanding (mother's) emotional reactions and seeing (direction of gaze) in the first year of life (see Perner, 1991, ch. 6; Baldwin & Moses, 1996; Gopnik & Meltzoff, 1997) for summaries and discussion of problems of interpretation). There is also some recent evidence that between 8 and 12 months children might be inferring people's intentions to grasp an object from where that person looks (Spelke, Phillips & Woodward, 1995) and even between 5 to 9 months from how a person touches an object (seemingly intentional or accidentally, Woodward 1998),[1] and by 18 months (where children's understanding of mental phenomena seems to flourish in general) children imitate people's intended actions even when they observe a failed attempt (Meltzoff, 1995a) and they understand differences in preferences (e.g., that someone else can prefer cauliflower over biscuits, Repacholi & Gopnik, 1997).

Evidence that children distinguish their knowledge from ignorance is available at a relatively late age. Povinelli, Perilloux, and Bierschwale (1993) asked children to look for a sticker under one of three cups. Children were first trained to look under the cup at which the experimenter had pointed. After some training even the youngest were able to do this. When asked to look without the experimenter pointing, an interesting developmental difference emerged. Children older than 2 years and 4 months acted without hesitation when they knew which cup the sticker was under but hesitated noticeably when--in the absence of the experimenter's pointing--they had to guess where it was. Interestingly this age is also the age at which children start using the phrase "I don't know" (Shatz, Wellman, & Silber, 1983). In contrast, children younger than that showed no comparable difference in reaction time. This may indicate that young 2-year olds do not yet reflect on what they do and do not know.

So, theory of mind research is not yet able to give a good guideline for when infants might develop explicit, conscious memories. Memory development may help out on this point. Meltzoff (1985, 1995b) demonstrated that 14 month old infants can re-enact a past event (e.g., they imitate the experimenter leaning forward to touch a panel with forehead so that panel lights up) after several months. Recently this has been demonstrated in 11 month olds with a delay of 3 months. Since this is achieved from a brief observational period and does not require prolonged learning and since patients with amnesia cannot do this (McDonough, Mandler, KcKee, & Squire, 1995) it is tempting to conclude that such enactment demonstrates explicit, conscious memory. One should, though, keep in mind that delayed imitation that is based on a single event (third person view) is not to be equated with a memory (knowledge) of that event as a single, past event (first person view) (see chapter 20 by Nelson & Fivush, this volume, for a related point on infant memory).

More impressively, Rovee-Collier (1997) reviews research on memory in infants as young as 2 or 3 months for mobiles that they had learned to activate with a foot kick. Investigation of over ten different independent variables that produce an implicit-explicit memory dissociation in experiments with adults showed that infants’ delayed recognition of the mobile behaves like explicit memory, while reactivation of foot kicking (child can activate the mobile a day after a single re-exposure of the mobile) behaved like implicit memory. However tempting it is to conclude that 2 or 3 month old infants have conscious memory, one needs to keep in mind that explicit and conscious memory may not coincide in the infant as they seem to do in the adult. Hence, the memory parameters indicating explicit, conscious memory in adults may depend on features of explicit memory that might exist without consciousness in infants (see also chapter 19 by Rovee-Collier & Hayne).

In any case, conscious knowledge of past events is not all there is to memory proper. Theory of mind research does have highly relevant findings for that further development.

3. Episodic Memory as a Self-Referential Mental State

For the possession of some mental states--and 'memory proper' (James, 1890) is one of them--more than just a higher order mental state is required: specific features need to be represented, in particular their causal connection to the world. Searle (1983) argued that such causal self-referentiality is required for intentions (underlying intentional action) as well as remembering. That is, intentions to act must specify (represent as their content) that the intended action be caused by them. Similarly, proper memories of a specific event must not only specify the event but also that they have been caused by that event. This analysis fits well Tulving's (1985; see chapter 38 by Wheeler this volume) analysis of autonoetic episodic memory as the remembering of personally experienced events that need—in Ebbinghaus's (1885) words—"be recognized as something formerly experienced." William James (1890, p.648) gave a similar definition as the distinguishing feature of memory (in the strict sense): "memory proper (or secondary memory) is the knowledge of an event, or fact .... with the additional consciousness that we have... experienced it before." Ebbinghaus's addendum (in light of the sense-reference distinction discussed above) is critical: A personally remembered event must be remembered as personally experienced. Perner (1990) spoke of experiential awareness. Applied to a typical word list memory experiment in which, e.g., the word "pear" appeared, this would mean that one has to encode (or later reconstruct) a metarepresentational comment of the following kind (Perner, 1991, p.163):