Episodic future thinking in 3- to 5-year-old children: The ability to think of what will be needed from a different point of view

Abstract

Assessing children’s episodic future thinking by having them select items for future use may be assessing their functional reasoning about the future rather than their future episodic thinking. In an attempt to circumvent this problem, we capitalised on the fact that episodic cognition necessarily has a spatial format (Clayton and Russell, 2009, Hassabis and Maguire, 2007). Accordingly, we asked children of 3, 4, and 5 to chose items they would need to play a game (blow football) from the opposite side of the table on which they had never before played. The crucial item was the box that was needed by children to reach the table from the other side. Over four experiments, we demonstrated that, while children of 3 perform poorly on future questions and children of 5 generally perform quite well, children of 4 years find a question about what they themselves will need to play in the future harder to answer than a similar question posed about another child. We suggest that this result is due to the ‘growth error’ of over-applying newly-developed Level 2 perspective-taking skills (Flavell et al., 1981), which encourages the selection of non-functional items. The data are discussed in terms of perspective-taking abilities in children and of the neural correlates of episodic cognition, navigation, and theory of mind.

Introduction

Tulving (2005) coined the term ‘chronesthesia’ in order to capture the fact that both when we re-visit the past in episodic memory and envision the future in episodic future thinking (Atance & O’Neill, 2001) we consciously ‘travel through time’ while remaining in the present. If this mental time travel is indeed a unitary capacity, then one would expect re-experiencing and pre-experiencing – we shall borrow the useful term prospection for the latter (Buckner and Carroll, 2007, Gilbert and Wilson, 2007, Suddendorf and Corballis, 2007) – to develop at around the same time. However, in assessing both the development of episodic memory and that of prospection the researcher is dogged by a similar problem. Just as it is one thing successfully to remember a piece of information and quite another thing to recollect, qua re-experience, the picking up of this information, so one must draw a clear distinction between correctly making judgments about, and acting in terms of, future states and needs and actually projecting oneself into a future scenario – prospection. The first kind of capacity is informational and the second is phenomenological. Suddendorf and Corballis (2007) give an explicit analysis of the parallels between episodic memory and prospection in these terms.

Students of episodic memory development have dealt with the issue of how to distinguish between episodic recollection and semantic remembering in a number of ways. Perner, for example, has taken the difference between free recall (episodic) and cued recall (semantic) performance as a measure of episodic memory (Perner & Ruffman, 1995), and has made a similar contrast between the episodic and the semantic in terms of the distinction between memory for directly-experienced events and for ones that were experienced indirectly via a video recording (Perner, Kloo, & Gornik, 2007a). But how does one draw a distinction between successfully knowing about the future and prospecting it? This is the central issue that this paper addresses.

In the first place, we need to distinguish between studies in which the child’s general future-oriented behaviour is assessed and those that at least have the potential for telling us about prospection, given that behaviour can be future-orientated without the agent actually envisioning any future state. In the first category fall studies of planning (Carlson, Moses, & Claxton, 2004) and of prospective memory (Guajardo & Best, 2000). Of course, the capacities just mentioned may indeed involve prospection; but they need not do so prima facie.

As for potential tests of prospection, experimental measures can be divided into three kinds. First, there are those studies that ask children questions about what they would and would not do tomorrow (Busby & Suddendorf, 2005). However, while children may indeed be using prospection to produce their answers, the questions can be answered on the basis of what they know about the future (though see Quon and Atance (2009), for a recent attempt to empirically distinguish between these possibilities). Second, there are tasks that require children to choose appropriately for a future motivational state when this conflicts with their current motivational states. In this domain, Atance and Meltzoff (2006) have shown that children between 3 and 5 years tend to predict future food preferences in terms of their present motivational state, saying for example, that tomorrow they will prefer water to pretzels because they are currently thirsty, having just eaten a lot of pretzels, which they generally prefer to water. That they should do this suggests that they find it difficult to imagine themselves in different motivational state in the future. There was little evidence for development across these ages. Also in this motivational category fall delay-of-gratification tasks in which children are invited to accept a smaller treat in the present in order to gain a bigger one in the future. Children might do this at age 4 (Mischel, Shoda, & Roderiguez, 1989) but individual differences are substantial (Carlson & Moses, 2001). The clear drawback to these two motivational procedures is that, in them, the ability to project mentally into the future is confounded by demands for executive inhibition. Needless to say, many of the tasks we give children in which they have to make a decision in which some options are more attention-grabbing than others (such as the task to be reported here) make executive demands; but in the case of the tasks just mentioned it would appear to be a primary requisite.

The third kind of study that at least has the potential for being a test of prospection are studies in which children have to select items for future use. The present study falls within this category, in fact. Proposals of this kind were simultaneously published by Tulving (2005) and by Suddendorf and Busby (2005), the latter authors backing up the idea with an experimental study. In fact, the general proposal was first made by Suddendorf (1994) in an unpublished Masters thesis.

Tulving’s inspiration sprang from an Estonian folk tale about a small girl who dreamed one night about attending a party where a delicious-looking chocolate pudding was being served, which she was unable to taste as she had no spoon. The next night, on retiring to bed, she placed a spoon under her pillow. We shall refer to tasks of this kind as employing a ‘spoon-test’ methodology. Similarly, Suddendorf and Busby (2005) argued that if children can select items for a future need then they should, compared to a control group, be more likely to take items with them into a room in order to alleviate future boredom. In this ‘rooms task’, there were two rooms, one containing a puzzle board minus the puzzle pieces (‘empty’ room) and the other containing a number of items including puzzle pieces (‘active’ room). At issue was the question of whether children would take puzzle pieces from the active to the empty room to alleviate anticipated boredom. In the control group there was no puzzle board in the empty room. At around 4–5 years children did indeed select puzzle pieces as an insurance against future boredom more often than did control children. Loosely comparable procedures have been employed with apes (Mulcahy and Call, 2006, Osvath and Osvath, 2008); though note that in this case the animals received multiple trials. Note also that, while the present study employs something comparable to the spoon test procedure, we use verbal methods, asking the children what they or another child will need.

A further study by Atance and Meltzoff (2005) has addressed similar issues, although in this case the 3-, 4-, and 5 year olds were presented with stories and pictorial scenes designed to evoke thoughts about thirst, cold, and hunger. The children were then asked to imagine themselves in these scenarios (e.g., a snowy mountain) and point to one item on a picture card that they would need in that scenario. The items could be functional (e.g., a winter coat for cold) or non-functional but semantically associated (ice cubes for cold). Across scenarios, 5 year olds selected the functional items more than did the 3 and 4 year olds. Moreover, in a number of scenarios, 3- and 4 year olds chose the non-functional but semantically-associated items just as frequently as they chose the functional items. Atance and Meltzoff (2005) proposed that children develop the ability to anticipate future needs before age 5, but that this ability is not fully mature and is prone to error in certain situations, particularly those where the available options are semantically associated with the scanario.

There is, however, an interpretive problem with the spoon-test methodology which, as adumbrated earlier, derives from the fact that it does not necessarily test for children’s ability to imagine themselves in a future scenario. That is to say, their choices may be based on impersonal reasoning and functional knowledge. For example, the child’s decision to take puzzle pieces with her into the room could have been based on no more than reasoning of the kind: “Anybody left alone in that room will need puzzle pieces, so I’ll pick some up.” All that is required for this kind of reasoning is the general semantic or ‘script-like’ knowledge that board games cannot be played without the relevant pieces. This is to be distinguished from imaging oneself playing the game in the prospective ‘theatre’ (Suddendorf & Corballis, 2007) of the mind’s eye. Similarly, in the Atance and Meltzoff (2005) study the child may simply be reasoning “What is needed in cold weather is a warm coat, so I shall pick that.” Imaginative projection is not required in these cases.

How might one distinguish empirically between selecting an item for a future need on the basis of functional reasoning and doing so by prospection? We suggest that this can be done by exploiting the fact that episodic memory – and indeed episodic cognition more generally – essentially involves egocentric spatial representation. To explain this claim we shall briefly précis the case for a ‘perspectival’ interpretation of episodic memory made by Clayton and Russell (2009), generalising this to episodic cognition, so as to include prospection.

• If episodic cognition amounts to re-experiencing (Tulving, 1999, p. 13) and pre-experiencing (Atance & O’Neill, 2001), then these two will inherit the necessary features of experience itself.

• Following Kant (1781/1998) in the Transcendental Aesthetic section of the Critique of Pure Reason, all objective experience (where ‘objective’ can include bodily feelings) necessarily has spatial and temporal formats. That is, space and time are ‘a priori’ in the sense of being necessary for experience, rather than products of it. The spatial content refers to the perspectival spatial relation between subject and objects (egocentric relations such as left/right, up/down). For Kant, the temporal element refers to succession versus simultaneity (of elements within an event); but as these must be bound to the spatial element, the temporal element inherits this perspectival quality.

• Accordingly, if we wish to assess whether or not subjects, such as a children, are engaging in episodic cognition then we must gauge the extent to which they are representing the spatial perspective of the past or future self in an episode.

Consider, in illustration, what we would naturally count as a case of memory in the sense of re-experiencing. If somebody says that he can remember the scoring of the winning goal in a soccer match but cannot recall whether he saw the goal from side-on, from behind the goal-mouth or from the far end of the pitch, from the back of the stadium or near the front, then we would be inclined to deny that this was a case of episodic recall. (If recall is ever reported from a third-person point of view, as some adult cases in Nigro and Neissser (1983), then it is either an imaginative reconstruction, or the original experience was of self-as-observed, something one would expect not to find in young children.) Similarly, if somebody says that she can envision “in my mind’s eye” her dinner with John tomorrow night but is agnostic about whether she will be sitting vis-à-vis or side-by-side him then we would tend to say that, despite what the speaker claims, this cannot be a case of episodic future thinking.

This leads to the following empirical claim about children’s performance in spoon-task situations. If children are really envisioning a future state of the self within a scenario, then they will not only be able to select items they need in that scenario but will be able to select items they need to use from a particular spatial location within that scenario. Moreover, in order to ensure that they are not performing this spatial projection task by duplicating their present spatial position, it needs to be the case that they have to project to, and select an item needed at, a position at which the item has never been used. (Compare the case in which somebody prospects seeing a goal scored from their usual seat in the stadium, which can be done from memory, with the case in which this is done from a position in which the individual has never before sat.)

Before we give the specifics of such a task a caveat has to be entered. The claim is not being made here that if children pass a task of this kind then they must necessarily be doing so by prospection. As will be seen, the claim is rather that, when combined with different kinds of question, such as task can yield significant clues about whether a-personal, functional reasoning or prospection is being used.

The task we used is called the Blow Football task. Blow football is a form of table soccer in which each of two players defends a goal on one side of a table and attacks the opponent’s goal on the other side. The ball is a ping-pong ball and each player projects it by blowing through a drinking-straw. The prototypical experimental situation was this. Our blow football table was such that only one side of it was reachable by a preschool child without standing on a yellow box. Children stood on this box before playing. The side on which the box was needed was the blue side and the reachable side was red. The child plays the game with the experimenter, with the child standing at red and experimenter at blue. After this, the child is asked to pick, from 6 items, which two he or she will need when she/he returns to play the game again the next day from the blue (unreachable) side. See Fig. 1 for a schematic illustration. There were two functional items (the yellow box and the straw) and four non-functional ones (such as a doll referee). The ball was always present. We included the straw as a functional item for two reasons: (1) to ensure the child was choosing items to play the game and not merely to reach the table or make the table-platforms symmetrical; (2) to make a child’s being correct by chance less likely.

If a child correctly chooses the straw and box, then, while it is possible that this was done by prospection, that interpretation is by no means mandatory, as the task will inevitably be passable by purely functional reasoning such as ‘anybody who plays on the blue side will need the box’. However, it is possible to discover something about whether prospection is being used here by comparing children’s answers to self-projection future questions with their answers to other-projection questions (e.g., “What would another little girl just like you need…?”). This is because the essence of prospection is projection of the self towards a future state and not merely thinking functionally about that state insofar as it would make functional demands on anybody. If the answers do not differ, then the appropriate conclusion is that there is no evidence for prospection in children of this age. However, if there is a difference in difficulty in one direction or another then it is likely that the achieved (self condition superior) or inadequately attempted (other condition superior) act of prospection is playing a role.

We tested children between 3 and 5 years because, as the previous literature review suggests, it is within this age range that future-orientated thinking and behaviour seems to develop. Moreover, in an extensive analysis using a number of tasks (item selection for the future, delay of gratification, planning, prospective memory) Atance and Jackson (2009) confirmed that it is within these ages that significant development takes place.

Finally, in order to ensure that it is their being asked about future states that is challenging the children, and not the perspective-taking or other demands of the task, our first study included conditions in which children are questioned about the present and the past. In any event, perspective-taking of the kind required here is generally taken to be within the capacity of children within the age-range studied, with 3 year olds being above chance in Newcombe and Huttenlocher’s (1992) task in which children had to project themselves to different locations and report on the positions of landmarks (left–right, etc.) in relation to their body at these new locations. For reviews see Newcome and Huttenlocher, 2000, Newcombe and Huttenlocher, 2006.

We shall report four experiments:

• to confirm that children of 3–5 years can pass the Blow Football task when (a) the self-projection question is about the present and (b) the ‘other’ question is about the past not the future, with these conditions crucially having in common their not being about the future;

• directly to compare performance on self- and other-projection future questions at all three ages;

• given the outcome of Experiment 2, to do the same as (2) for age 4 only;

• to ask whether the self-other differences in 4 year olds that emerged in (3) were specific to questions about the future, or whether the same would be found with self-other questions framed in the present tense.