2.1. Introduction

As discussed in the general introduction, the concept of cognitive flexibility has a variety of meanings. In this chapter, we will deal with cognitive flexibility as one of the manifestations of a controlled and goal-directed cognitive functioning. From this perspective, flexibility refers to the ability to adapt one’s representations and processing according to the goal pursued, in response to changes in cues in the environment (e.g. new instructions, failure, disapproving look from an adult, and so on). The development of flexibility is long and progressive throughout childhood, up to and including adolescence (Best et al. 2009; Diamond 2013). This development is particularly dependent on brain maturation. The role of the development of the prefrontal cortical region and the many connections it has with posterior regions is now well established¹ (see, for example, Bunge and Zelazo 2006; Johnson et al. 2013; Chevalier and Blaye forthcoming).

There is a broad consensus on the importance of cognitive flexibility in many fundamental areas of cognitive development, whether in the acquisition of theories of mind (see, for example, Müller et al. 2005) or language (see, for example, Deák 2003). Similarly, links between cognitive flexibility and academic achievement – more specifically, in mathematics and reading – have been regularly highlighted despite some discordant results (e.g. van der Ven et al. 2012; Colé et al. 2014; Cantin et al. 2016; Johann et al. 2020; for a meta-analysis, see Yeniad et al. 2013 and for a failure to demonstrate such a link, see Monette et al. 2011). Beyond correlational studies, several studies suggest that training cognitive flexibility can produce transfer effects on performance in academic tasks (for a recent example, see Johann and Karbach 2020). Understanding the development of cognitive flexibility is, therefore, an important issue.

Since the pioneering work of Miyake et al. (2000), flexibility has been considered one of the three main executive functions allowing the exercise of cognitive control, along with inhibition and updating in working memory. In adults, these three functions are partially independent. In the course of childhood, this structural organization evolves into a progressive differentiation of the three functions: at 3 years of age, the organization is unitary (see, for example, Willoughby et al. 2010, 2011). From the age of 4, a two-factor structuring begins to take place (flexibility may be associated more with the inhibition factor or the working memory updating factor depending on the study (Lee et al. 2013; van der Ven et al. 2013; Monette et al. 2015). A three-factor structure is clearly evidenced only during adolescence (see, for example, van der Ven et al. 2013).

As we will see in this chapter, the question of the structural organization of these functions is partly moot given the significant amount of variance they shared (see Miyake and Friedman (2012) for a revision of Miyake’s original model) and the impurity of the tests to assess them. This impurity is related to at least two reasons: on the one hand, the effectiveness of control processes depends on the mastery of the controlled processes, and, on the other hand, most evaluation tests require the use of several control processes. Thus, demonstrating flexibility to adapt to changes in explicit instructions or to new environmental cues requires updating and then maintaining the new goals in working memory as well as inhibiting the initial representations and processing. More generally, flexibility cannot be studied as an isolated function of cognition, of which it is one of the expressions. It is considered here as an emergent property of cognitive functioning adapted to variations in context (Ionescu 2017; see also Chapter 8). As such, it gives rise to both very early manifestations and late difficulties, as the following points of reference across development will show.

This chapter focuses on the diversity of the cognitive processes underlying cognitive flexibility in different situations.

2.2. Main study paradigms and some developmental points of reference

Cognitive flexibility is mainly studied, from the age of 2, through paradigms of categorization rule change. However, flexibility of behavior is evident in infants from their first year of life. It is revealed in detour behaviors that correspond to the abandonment of an ineffective mode of processing in order to adopt another, more adapted one (see, for example, Diamond 2002) or in the success of infants in modifying the place where they search for a hidden object after a conspicuous change of the hiding place by the experimenter (A-Not-B paradigm: Piaget 1937; Diamond 1990).

The reverse categorization task (Carlson et al. 2004; Carlson 2005), which is proposed to children as young as 2 years old, consists of presenting objects that can be categorized on the basis of a common dimension (for example, size: large versus small). In a first series of trials, children are asked to place each large object in the large container and each small object in the small container (there are two copies of each object offered for sorting, differing only in size). In a second stage, the rule is reversed. The children are told that they will now play a “silly” game by placing the large objects in the small container and the small objects in the large container. This task requires an update of the action rules and a change in response between the two phases. However, there is one dimension that remains relevant in both stages of the task and does not imply a modification of the representation of the pictures: an intra-dimensional change. While 2-year-olds are mostly unsuccessful, 3-year-olds are able to change rules flexibly in this context (Perner and Lang 2002; Brooks et al. 2003; Kloo et al. 2008).

When two-dimensional stimuli are used and the imposed rule change involves changing the categorization dimension (extra-dimensional change), flexible adaptation becomes difficult for 3-year-olds.

This is demonstrated by the Dimensional Change Card Sorting (DCCS) task (Frye et al. 1995; Zelazo 2006), which has become paradigmatic for assessing flexibility in the preschool period (see Figure 2.1).

NOTE.– Gray arrows correspond to correct answers in the shape game; colored arrows correspond to correct answers in the color game.

Here, children have to match two-dimensional test pictures to two equally two-dimensional target pictures, in turn, in two series of trials, according to one and then the other dimension. Each test picture is identical to one of the target pictures on one dimension (e.g. color) and to the other target picture on the other dimension (e.g. shape), thus creating a conflict of responses that can be overcome by taking into account the instruction associated with each series of trials. The game to be played – shape game versus color game, respectively – in the series of trials before and after the instruction change (pre-switch and post-switch) is recalled before each trial, along with the associated pair of action rules (e.g. for the color game, the child is told, “remember, we’re playing the color game; in the color game, reds go with reds and blues go with blues”; see Chapter 1 for a detailed description of the tasks). While children as young as 3 years of age succeed perfectly in the pre-switch series of trials, regardless of the dimension proposed first, they massively fail at the post-switch stage, continuing to categorize following the pre-switch sorting criterion. Such perseverative behavior only disappears about a year later, as the majority of 4-year-olds show flexibility in this new context. Given the characteristics of the task, the lack of flexibility is expressed here by the perseveration of responses that were correct in the pre-switch stage.

The assessment of cognitive flexibility beyond the age of 4–5 years is generally based on the task-switching paradigm (see, for example, Monsell 2003). Like the DCCS task, this involves alternating between two ways of processing bidimensional stimuli. However, in addition to two blocks of single-task trials (single-task blocks), participants are confronted with mixed blocks with possible task switches from one trial to the next. In the cued version of this paradigm, goal cues displayed concurrently with – or immediately before – each picture indicate which processing should be implemented on each trial. Two crucial differences characterize this paradigm compared to the DCCS task (Cragg and Chevalier 2012). In addition to the absence of a verbal instruction at the beginning of each trial, this paradigm involves a large number of changes (task switches) in an unpredictable sequence; only the processing of the goal cues, which are mostly visual, provides knowledge of the goal to be pursued for each trial.

Research in adults has clearly established a decrease in performance in mixed blocks compared to single-task blocks (performance measured in rate of correct responses and/or time of correct responses). An indicator of the gradual development of flexibility is specifically the reduction of this cognitive cost with age during childhood and adolescence (see, for example, Cepeda et al. 2001). A critical feature of this paradigm is that it allows one to differentiate between two components of flexibility. Indeed, mixed blocks contain repeated trials (where the task to be performed is the same as in the previous trial) and switch trials requiring a change of task compared to the previous trial. Comparing performance on these two types of trials sheds light on the specific cost of the switching process (local switch cost). This cost evaluates the difficulty of abandoning one task-set² and activating another, once the new task is identified. The comparison of performance on repeated trials in mixed blocks with that on single-task trials further estimates a second component, that is, the cost of managing two goals in the mixed block compared to one in the single-task blocks (mixing cost). This cost is observed even though no task change is required in any of the considered trials. Although both types of costs decrease over the course of childhood (see, for example, Cepeda et al. 2001; Chevalier and Blaye 2009; Cragg and Nation 2009), the cost of managing two goals has been found to be the most sensitive to development (see, for example, Reimers and Maylor 2005; Chevalier et al. 2010).

Finally, another task, from which the DCCS task is inspired, is often used in neuropsychology, particularly in cases of suspected dysexecutive syndrome: the Wisconsin Card Sorting Task (WCST) (Grant and Berg 1948). This latter task involves flexibility that can be described as inductive insofar as the picture sorting rule (varying here on three dimensions) to be implemented on a series of trials must be inferred from the psychologist’s feedback on the adequacy of the rule used. Flexibility is estimated here in terms of the number of perseveration errors on the previous rule after negative feedback and the number of rules discovered with a finite number of pictures knowing that the participant is forced, by negative feedback, to abandon the current rule after 10 consecutive trials using the same rule. The level of flexibility corresponding to the mature level of young adults is reached in adolescence in this test as well³ (Chelune and Baer 1986; Huizinga et al. 2006).

These few points of reference based on different flexibility tasks highlight the need to consider flexibility as being closely dependent on the contexts in which it is assessed, and hence on the processes that interact to allow its expression. These processes are partly different according to the nature of the tasks proposed to children of different ages. In this chapter, we shall consider them precisely by examining in turn:

• – research that considers the development of flexibility as reflecting the ability to overcome the tendency to persevere on an initial representation;
• – research that considers progress in flexibility as evidence of increasingly effective goal management.

2.3. How can we account for perseveration behaviors in the preschool years?

The preponderance of the use of the DCCS task in the study of flexibility has led researchers to debate the origins of perseveration behaviors on the initial criterion observed in the post-switch stage of this task. The theoretical proposals put forward are not to be conceived as incompatible, but rather as shedding light on the diversity of the processes at play in the expression of flexibility of behaviors at pre-school age (see also Clément forthcoming).

Zelazo et al. (2003), authors of the DCCS task, proposed an explanatory model of perseveration behaviors based on the analysis of the complexity of the rule structure to be used (Frye et al. 1995; Zelazo et al. 2003). Thus, failure to follow a new sorting instruction in the post-switch stage would be due to difficulty in co-ordinating the pairs of rules of the post-switch stage (e.g. reds with reds and blues with blues) with the incompatible pair of rules in the pre-switch stage (e.g. rabbits with rabbits and boats with boats) within an overarching rule that takes into account the task conditions (“if shape game then” versus “if color game then”). It would thus be the development of the capacity to reflect on this system of rules that would allow both the disengagement from one pair of action rules and the re-engagement in another. However, this type of explanation does not account for why, at the same age, children are able to demonstrate flexibility in versions of the task involving the same rules embedded within an overarching rule when, for example, the target pictures are replaced by two puppets whose preferences change between the pre- and post-switching stage (Perner and Lang 2002; Kloo and Perner 2003). Numerous interpretations of perseverative behaviors that emphasize other aspects of young children’s functioning have been proposed.

The DCCS task requires flexibility in representations due to the change of rules (e.g. the picture that was “a red one” becomes “a rabbit”), but also flexibility in the production of responses (e.g. the picture placed in the “red boat” box in the color game must be placed in the “blue rabbit” box in the shape game). Therefore, it was legitimate to wonder about the level of young children’s difficulties: that of the control of responses execution or control of representations. By transforming the DCCS task into an error-detection task, while observing a doll performing the pre- and post-switch sortings, Jacques et al. (1999) showed that the difficulty was not due to a lack of control over the motor execution of the responses that had become preponderant in the pre-switch stage, but rather to a form of “inflexibility” of representations. Thus, 3-year-olds do not detect errors in the doll’s perseverative responses in the post-switch stage, even though they do not have to exercise any control over their actions.

Reasons for this representational inflexibility can be illustrated by the (approximate) 1-year lag between the expression of flexibility in situations of intra-dimensional rule change (as in the reverse categorization task) and that observed in the DCCS task involving an extra-dimensional change. It would seem that there is a specific difficulty related to the need to consider two representations of the same object in turn. For Brooks et al. (2003), this difficulty is due to the need to pay selective attention to a single dimension of two-dimensional stimuli in the DCCS task (Brooks et al. 2003; Hanania and Smith 2010). However, Kloo et al. (2008) found that children aged 2–4 years performed equally well on two reverse categorization tests using one-dimensional and two-dimensional stimuli, respectively. Moreover, both tests were consistently more successful than the standard DCCS test. The two-dimensional reversal version requires precisely the same changes of responses as between the pre- and post-switch stages of the standard DCCS task in responses as the standard version, but does not require a change in representation of the pictures. Such results support the hypothesis proposed by Perner and Lang (2002) of a conceptual difficulty present in young children who do not understand that the same stimulus can give rise to several representations: after having considered an object as “a red one” in the pre-switch stage, they would fail in the same situation to consider it as “a rabbit” in the post-switch stage (for a more detailed presentation of this hypothesis, see Chevalier and Blaye 2006). Such a hypothesis is obviously consistent with what Piaget defined as the egocentric thinking of the pre-operational child.

This same difficulty of dual representation of an object has been invoked in many other cognitive acquisitions during the same developmental period. This is the case, for example, in the understanding of false beliefs (Kloo and Perner 2003), or in the use of symbolic representations (Deloache 2004). Indeed, in these domains, like in the DCCS task, experimental manipulations that either eliminate the need for dual representations or clarify the need for a second representation improve performance.

For Kirkham et al. (2003), perseverative behavior is less the result of a conceptual difficulty than of an executive difficulty. According to the authors, preservative behavior is due to attentional inertia, that is to say, the difficulty to inhibit the initial representation. Therefore, conditions likely to favor the process of inhibiting the first representation which has become irrelevant should favor flexibility. These authors validate this prediction by comparing the standard version of the DCCS task with a condition in which children have to name the card by its newly relevant dimension (e.g. “rabbit” for an picture considered red or blue in the pre-switch stage), before producing each response in the post-switch stage.

The difficulty of inhibiting the initial representation is undoubtedly a central aspect of the explanation of perseveration behaviors, but the low working memory capacities of young children could also contribute to these behaviors. Morton and Munakata (2002) proposed an interpretation of perseveration in terms of competition in working memory between the representation of the pre-switch rule (latent representation) and that of the post-switch rule (active representation). The central idea of their model is the assumption of graded representations (Munakata 2001). Developmental studies too often tend to consider an erroneous response as something that reflects the absence of a correct representation. However, a representation may already exist in the repertoire of knowledge, but may not be sufficiently active, given the characteristics of the situation (in this case, the interference of previously activated rules), to guide the control of responses. This interpretation resolved the apparent paradox observed in 3-year-olds between knowledge and action. Indeed, after solving the DCCS task, most 3-year-olds are able to answer a question about the post-switch rules correctly (e.g. if “shape” is the post-switch criterion, the question is “where do the rabbits go in the shape game?”), whereas they have produced perseverative responses by sorting the pictures according to the initial criterion of color. Munakata and Yerys (2001) have shown that the reduction of conflict between latent and active representations, and thus the lesser activation of the latent representation during verbal questioning, is at the origin of the better performances to the knowledge question. Indeed, the formulation of the question makes the dimension to be ignored (color of the picture in our example) disappear, whereas this dimension strongly interferes during the effective sorting of the pictures. As soon as this dimension is reintroduced in the knowledge question (for example, “where do the red rabbits go in the shape game?”), the dissociation between knowledge and action is no longer systematic.

Older works on functional fixity phenomena provide empirical arguments supporting the role of activation competition between representations to be ignored and those to be mobilized. Duncker (1945) had already shown in adults, with the candle task, that the effect of functional fixidness was reduced as soon as the “container” function of the box of tacks was not made salient. The latter was then more easily considered in a “support” function for the candle. Thus, the reduction of the activation level of the initial representation favors the consideration of an alternative representation of the object (for similar results in a developmental perspective; German and Defeyter 2000).

The above-mentioned research remains focused on the search for an interpretation of the perseveration behaviors considered as the hallmark of inflexibility. However, this constitutes a fragmented view of the processes at play and leads to ignoring other expressions of inflexibility likely to reveal other underlying mechanisms that will now be considered.

2.4. Beyond perseveration

The paradigmatic task of DCCS has undoubtedly contributed greatly to an almost exclusive focus on the phenomenon of perseveration. Indeed, it only allows perseveration errors in the post-switch stage. Moreover, insofar as the expected sorting criterion is formulated verbally before each trial, the difficulty of maintaining the goal seems to be minimized. Yet, in 3-year-olds, distraction errors in the post-switch stage are at least as common as perseveration errors when using flexibility tasks that can detect them. This type of error corresponds to a response that differs from both the expected response and the relevant response in the pre-switch stage.

Chevalier and Blaye (2008) proposed a new flexibility task (PAST: Preschool Attentional Switching Task) in which young children are asked to indicate the shape of one of three color-designated pictures on on-screen test cards (see Figure 2.2 for an illustration in a two-picture version). The three pictures on the same card differ from each other in both shape and color. Participants must point to the response option corresponding to the shape of the C1 color picture in a first set of trials (stage 1), then to the C2 color picture in a post-switch stage (stage 2). Thus, in stage 2, the errors may reflect perseveration in the designation of the picture C1 or correspond to the designation of the shape of the picture C3, thus evidencing distraction and insufficient maintenance of the stage 2 goal. At age 3, distraction errors are as frequent as perseveration errors and concerned a majority of children. At age 4, only perseveration errors are still observed in about one third of children. A similar developmental hierarchy between perseveration and distraction errors has been found in recent works (Carroll et al. 2016; Blakey and Carroll 2018), thus suggesting that a first step in the development of flexibility between action rules would be to successfully apply the same rule over a series of consecutive trials when the task does not require changing it (Ionescu 2012, 2017).

Thus, perseveration behaviors are not the only manifestation of a lack of flexibility and probably not the earliest. Moreover, as we will now illustrate, these perseveration errors do not necessarily result from a perseveration process on an initial representation. The DCCS task, given its characteristics, did not allow for testing this hypothesis. Instead, this was made possible by a version of the PAST task with two pictures per test card and two different colors (Chevalier and Blaye 2008; see Figure 2.2). More precisely, the authors contrasted the hypothesis of a perseveration process on an initial representation that had become irrelevant with that of a difficulty in reactivating a representation that had to be ignored during the initial stage. Three versions of the task were proposed to three groups of 3-year-olds. These three versions of the task respectively allowed for either both perseveration and reactivation difficulty (Control condition: the pre- and post-switch pictures containing the same two colors), or only perseveration (Preservation condition: the focus color of phase 2 not being one of the colors of phase 1), or only reactivation difficulty (Reactivation condition: focus color in phase 2 corresponds to the color to be ignored in stage 1, the color to be ignored in stage 2 being new in comparison to stage 1). The lack of flexibility in young preschoolers is due more to difficulty in reactivating the processing of the picture of the previously ignored color (64% failure in the reactivation condition) than perseveration on the processing of the picture of the initially relevant color (11% failure in the perseveration condition). Zelazo et al. (2003) and Müller et al. (2006) reached similar conclusions with variants of the DCCS task, and explicitly invoke a negative priming effect.

NOTE.– The red circle indicates the color to be considered at each stage: (a) control condition: both types of process, perseveration and reactivation failure, can be at work in case of error; (b) perseveration condition: errors can only result from a perseveration process; (c) errors can only result from a reactivation failure of the previously ignored color.

Thus, a flexible behavior allowing the updating of representations and action rules to adapt to a change of goal depends both on the progress in inhibition to succeed in ignoring information that has become irrelevant (proactive interference), but also on the capacity to counteract the effects of the, probably automatic⁴, inhibition, of information that were irrelevant in the stage that immediately preceded the change of goal.

Work with young children has emphasized the cost of disengaging attention from one property of the object and reengaging it on another previously ignored property, despite the explicit formulation by the adult of both the need to change processing and the nature of the new processing to be implemented. However, the need for behavioral flexibility is not always signaled by verbal instructions as explicitly as in the DCCS task. Demonstrating flexibility sometimes requires detecting and processing contextual cues that may signal a need for change (Chevalier 2015a). As we shall see, monitoring these cues to identify the goal to be pursued and processing them in an optimal temporal sequence is the key to the development of flexibility beyond the preschool period. Important qualitative changes related both to a global change in the modes of control engaged and to the development of metacognitive skills take place between the end of the preschool period and school age.

2.5. Flexibility: a question of goal management

We can define the goal with Altman and Trafton as: “[...] intention to accomplish a task, to achieve some specific state of the world or take some mental or physical action” (2002, p. 39).

Cognitive control models in adults have long established the importance of the current goal representation in guiding behavior by reinforcing the activation of the most appropriate actions to achieve that goal (see, for example, Miller and Cohen 2001). The difficulty in maintaining the goal in situations requiring control has been identified in adults through the phenomenon of goal neglect, defined by Duncan (1996) as ignoring the requirements of an instruction even though the participant knows and understands it.

2.5.3. Toward an optimal sequencing of the information gathering process: from reactive to proactive control

The efficiency of processing also depends on the organization of the sequence of information intake about goal cues, on the one hand, and objects requiring a response, on the other hand. Young children tend to focus on the objects they need to act on (i.e. those that require a response), whereas adults and school-aged children inquire about the goal before considering the object to be processed. A recent eye-tracking study in a task-switching paradigm revealed such a reversal in the processing order of stimuli and visual goal cues during development (Chevalier et al. 2010, 2018). Each stimulus picture was offered along with the transparent goal cue specifying the type of match expected with two target pictures present on the screen (based on shape vs. color). The analysis included eye trajectories of children aged 3–12 years and a group of adults. In single-task blocks of trials (single blocks), the absence of goal uncertainty made cue information processing unnecessary. Indeed, only fixation of the stimulus was found to be the major pattern and no significant developmental changes were observed. In the mixed blocks, however, different fixation patterns were observed. The youngest children produced a majority of fixations on the stimulus to be processed before shifting their gaze to the cue. With age, this pattern becomes less frequent in favor of a fixation of the cue before the stimulus (see Figure 2.3). The distribution of these two patterns at different ages reveals a shift in information prioritization around 8.5 years of age (see Figure 2.3). It is noteworthy that the “cue then stimulus” fixation pattern is associated, in all children, with better performance in blocks requiring flexibility between the two sorting tasks.

This developmental shift in information prioritization illustrates a shift between two modes of cognitive control characterized by Braver (Dual Mechanisms of Control) (see Braver et al. 2007; Braver 2012). Braver, thus, identifies a reactive mode of control consisting of wondering about the goal only at the moment when a response is required and a proactive mode of control involving taking information from the context, before confrontation with the stimulus, to identify the goal, and then maintaining this information in order to prepare to respond as soon as the stimulus appears. Developmental studies have clearly established a developmental shift in terms of preferred mode of control (Chatham et al. 2009; Munakata et al. 2012; Lucenet and Blaye 2014; Gonthier et al. 2019): younger children predominantly opting for a reactive mode while older children more often engage a proactive mode. These two modes have different costs and benefits and are more or less adapted to different contexts (internal: e.g. working memory capacities; external: e.g. predictability of future events). Optimal efficiency, thus, lies in a context-adjusted coordination of these two control modes (Chevalier 2015b).

Recent studies suggest that it is possible to elicit such proactive processing of goal cues for the benefit of flexibility. Lucenet and Blaye’s (2019) study showed the benefits of explicitly prompting proactive goal thinking (through translation of goal cues, prior to stimulus display). Chevalier et al. (2015) tested the effectiveness of an implicit support of proactive control through the manipulation of situational constraints. 5- and 10-year-old children were assigned to three conditions in a cued task-switching paradigm: either the goal cues were displayed in advance of the stimulus and remained present during the stimulus display (making it possible to process the cue before or after the stimulus display), or they were displayed in advance but disappeared at the stimulus onset (prompting processing of the cue in advance), or finally, they were displayed synchronously with the stimulus (making it impossible to process the cue before the stimulus display). The authors took both measures of the time course of pupil dilation, assumed to reflect the dynamics of cognitive effort, and measures of event-related potentials. These two types of indicators converged and suggested that 5-year-olds tended to engage in reactive control (post display of the stimulus) when the cue was maintained, but became capable of proactive control when it was induced by the disappearance of the cue at the onset of the stimulus. 10-year-olds, on the other hand, engaged in proactive control as soon as it was possible.

2.6. From imposed flexibility to self-regulated flexibility

Such metacognitive monitoring processes play an even more critical role in the implementation of flexible behaviors when the child is no longer guided by explicit instructions or salient contextual cues, but has to decide for themselves the appropriateness of a processing change. Sometimes, the identification of an error is likely to trigger the change in representation or strategy that characterizes flexibility (Clément 2006). This identification is facilitated by the presence of explicit feedback on performance. The Wisconsin Card Sorting Task (see section 2.1) reveals the difficulties of children up to 10–12 years of age in inferring and applying a change in picture matching rules from response feedback. Hence, the development of flexibility may depend, at least in part, on progress in processing feedback (see, for example, Zanolie et al. 2008). Chevalier et al. (2009) observed such links in 4- to 6-year-old children to whom they proposed an inductive version of the PAST task described above (see section 2.4). Children were asked to indicate the shape of a certain color picture from among three different color pictures, with the need to focus attention on a new color to be inferred from feedback on the responses. Simpler than the WCST task, notably because it involved an intra-dimensional rule change (change of target color), the task revealed significant progress in flexibility between children of 4 and 6 years of age, associated with high intra- and inter-individual variability in the processing of feedback. Older children and those with better working memory capacity were the most likely to improve their feedback processing during the task.

In other situations, it is a matter of being able to achieve a task switch in the service of a higher goal, in the absence of immediate feedback. Who has not experienced the difficulty of switching between using their computer as a social networking tool and as a work tool? This form of flexibility, which could be termed endogenous since it requires the individual to set the goal change themselves, is relatively understudied in children. A possible context for assessing this competency is the uncued task-switching paradigm in which the participant is instructed to follow a particular sequence of changes between processes (e.g. AABBAABB). Although the changes of the goal are imposed by an initial instruction (sequence to be repeated), they then remain at the initiative of the participant. Such situations are frequent in kindergarten, where pupils receive instructions on a series of activities to be carried out successively. Working memory capacities are critical in this very demanding situation in terms of updating goals. In the study by Dauvier et al. (2012), the 5- to 6-year-olds with the lowest abilities either persevered on a single task or randomly switched goals, thus making a very high number of errors. A majority of other children of this age, on the other hand, were able to systematically update the sequence to be completed (shape-shape-color-color).

Verbal fluency tasks can be considered more demanding in terms of endogenous flexibility since the participant is not given any indication that goal changes are necessary, but is only given an overall goal: to produce the largest number of items of a category in a given time. A strategy for maximizing the number of copies produced is to retrieve them from memory, subcategory by subcategory. It is, therefore, a question of being flexible by changing sub-categories as soon as production runs out, and this without external injunction. Snyder and Munakata (2010) have shown that 5-year-olds fail to make these types of changes without assistance. In general, progress is observed in verbal fluency tasks until adolescence (Kave et al. 2008). Further research is needed to evaluate the hypothesis that the development of metacognitive processes for monitoring and evaluating performance is influential. In support of this hypothesis, a recent study suggests that supporting these processes with a requirement for systematic performance assessment on each trial allows 5- to 7-year-olds to engage with more proactive control (Hadley et al. 2020). Finally, the work of Munakata et al. (2012) and Snyder and Munakata (2013) suggests that the development of increasingly abstract representations during childhood may contribute to the development of increasingly self-regulated flexibility. Indeed, in addition to the ability to detect when a change is necessary, this form of flexibility requires the success of selecting a new option among multiple possibilities. Representing subcategories rather than individual exemplars in a fluency task, for example, can facilitate both the selection of possible options and reduce the set of competing items by allowing their retrieval in a structured manner.

In general, the expression of flexibility depends on the conceptual contents to be mobilized, as shown by studies on categorical flexibility (see, for example, Bonthoux et al. 2004; Blaye et al. 2006, 2007; Maintenant and Blaye 2008; Blaye and Jacques 2009). Better conceptual control ensures better top-down control over the representations that are best suited to situational demands.

2.7. Conclusion

The study of cognitive flexibility in young children has given rise to an explosion of research over the last three decades. These studies have rapidly shown that flexibility cannot be isolated from other control functions such as inhibition and the updating of information in working memory. In the same way, it cannot be reduced to the sole aspect of switching between two representations of the same reality. Rather, recent research suggests that its development is underpinned by progress in identifying the relevant goal to be pursued, then maintaining it and monitoring contextual cues to change it. Thus, put into perspective as an integral part of a set of interrelated control processes, its development is unsurprisingly subject to more general qualitative changes, such as those relating to the dynamics of control engagement and/or to metacognitive progress. This new descriptive framework opens up new avenues for thinking about possible interventions to support flexible cognitive functioning.