Different developmental researchers view ‘cognitive flexibility’ differently.
The definition of cognitive flexibility is elusive since different experiments focus on different aspects of the cognitive operations required for that specific task.
That said, cognitive flexibility is likely to be greater than the sum of experimental findings! One example of the inconsistent views of cognitive flexibility is that cognitive flexibility is often defined conceptually in the introduction of a journal article, and then defined operationally in the discussion sections. At the conceptual level, cognitive flexibility is often defined as the ability to simultaneously hold goal representations online, resolve conflicting information, and shift attention and respond appropriately to environmental changes (e.g.
Davidson, Amso, Anderson, & Diamond, 2006; Diamond, Carlson, & Beck, 2005; Zelazo et al., 2003), thus encompassing multiple cognitive operations.
However, the specificity of these core cognitive operations remain largely elusive (Cragg & Chevalier, 2012) .
In comparison, operationally, cognitive flexibility in developmental research is often defined with a specific theoretical account relevant for a specific experimental result only. The tendency to propose a restricted theoretical account for a complex behaviour reflects the methodological constraints of behavioural studies, since only a limited number of independent variables can be tested at one time in an experimental setting. Thus, it is not always clear how these precise theoretical accounts relate to the multi-faceted definition of cognitive flexibility at the conceptual level. Below I will give an overview of what some of the theoretical accounts are, and how it may be possible to approach cognitive flexibility more liberally, by moving beyond adopting a specific theoretical account of flexible behaviour.
(1) Cognitive flexibility may be about the ability to construct complex task rules. Zelazo’s Cognitive Complexity Control Theory (Zelazo et al., 2003) posits that young children fail on the DCCS task because they are unable to form a complex, hierarchical ‘if-if-then’-like rule structure.
However, other studies that required a similar task structure have shown that 3-year-olds are able to switch their responses on tasks that require a similar hierarchical task rule, if the need to redescribe the same object is removed (Kloo & Perner, 2005; Perner & Lang, 2002). Kloo and Perner (2005) argued that young children may simply be unaware that one object can be two distinct things at the same time (e.g. a red object and a truck), and that this lack of awareness
prevents them from forming an appropriate task set. Crucially, whether an appropriate task representation is formed or not may not necessarily implicate the development of domain-general cognitive control, but may have more to do with learning about different possible concepts. If anything, DCCS tasks highlight the possibility that young children may form task representations differently from older participants under some situations, due to their limited experience with the task attributes and task structures.
(2) Cognitive flexibility may be about the ability to actively maintain information. Some authors have argued that a major source of performance error in children comes from their difficulty in actively maintaining the relevant goal state (Blaye & Chevalier, 2011;
Chevalier & Blaye, 2008; Freier et al., 2017; Marcovitch, Boseovski, &
Knapp, 2007). According to this view, external factors that assist children in maintaining goal states should facilitate performance. One way to facilitate goal representation is through transparent task cues, which has been shown to be particularly effective in improving performance in 4-year-olds (Blaye & Chevalier, 2011). Furthermore, it was found that young children sometimes fail to switch responses not necessarily because they perseverate with the previous task rule, but because they fail to activate the current task rule (Chevalier & Blaye, 2008; Marcovitch et al., 2007). Similarly focusing on representations, Munakata (Morton & Munakata, 2002; Munakata, 2001) argued that representations at different levels of processing are graded rather than all-or-none. When young children pass in one task but fail on another task that taps into the same kind of knowledge (e.g.
colour/shape), the underlying cause may be the difference in the requirement of how strongly these representations need to be activated to succeed on these two tasks. To support the graded representation account, Blackwell, Cepeda and Munakata (2009) found that 4- to 5-year-olds who were quicker in responding to simple no conflict task-rule queries were also better at switching between tasks with multivalent stimuli than those who were slow at answering simple questions. This was true even after controlling for simple processing speed, indicating that the ability to actively represent rule may underlie the differences between switcher and non-switchers.
(3) Cognitive flexibility may be about the ability to resolve information interference. Most developmental studies of cognitive flexibility involve task sets that have overlapping task attributes, either in the stimulus set, the response set, or both. Thus, one defining feature of cognitive flexibility, at least in the laboratory tests, is the ability to resolve interference.
Trial-to-trial and task-to-task interferences can exist at different levels of processing, and different ages may be more or less affected by the types of interference. Several studies have suggested that children aged between 5 and 7 years experienced a greater level of stimulus-based interference than older children and adults, but not necessarily response-based interferences (Cragg, 2016; Cragg &
Nation, 2009). The general procedure in those studies involved independently varying the degree of dimensional overlaps among the stimuli (e.g. univalent colour-only stimulus vs. bivalent integrated coloured shape) for stimulus-based interference and varying how
many stimuli are mapped onto the same response for response-based interference. The interference in this context should be classed as representational incongruency, since the interference arises from how task sets were defined the first place. Thus, children’s inflexible behaviours may be due to the lack of ability in overcoming representational interferences between stimulus attributes associated with the task goals.
(4) Cognitive flexibility may be about inhibitory control on prepotent responses. Children’s difficulty with flexible behaviours may arise because they are unable to stop themselves from making a response, even when they know what the relevant task attribute is. In a series of spatial-compatibility studies using speeded tasks with children aged 4 to 13 years old and adults, Davidson et al. (2006) found that young children’s performance was particularly affected by spatial incompatibility between the location of the stimulus and the response (this is also known as the Simon Effect), indicating difficulty in suppressing prepotent responses caused by spatial compatibility. In another study, Ling, Wong, & Diamond (2016) examined whether young children’s difficulty in the Day/Night task (i.e. saying Day when seeing a picture of a moon) was due to response prepotency or weak task representation. They found that if there was an imposed delay between the stimulus and the response stages, such as the use of task-unassociated ‘ditty song’, children’s performance on Day/Night task improved. This result suggests that young children may have a particular difficulty in inhibiting a prepotent response, particularly if the response can be strongly triggered by external cues, and that the
ability to inhibit a response appears to be independent to the control component of task representation (e.g. Diamond & Kirkham, 2005).
Although inhibition of a prepotent response may be important in some situations, not all tasks necessarily elicit strong habitual responses.
Thus, it is unclear what role inhibitory control plays in tasks where responses are less habitual in nature.
(5) Cognitive flexibility may be about the ability to update task approaches and monitor performance adaptively. This ability cannot be measured at a single specific time point in the task, such as the point of switch, but rather, should be examined through the continuous changes in behaviours. In this view, flexibility is defined as the ability to take in new information and dynamically revise task approaches (i.e. plans and actions). This means that even when the behaviour initially appears inflexible, through continuous revision of internal representations and planned actions, adaptive flexible behaviours can emerge later on. There are reasons to believe that young children’s ability to change their behaviour within a task context may be better than what the standard laboratory tests suggest. For example, Kloo and Perner (2003) found that false-belief training improved performance on the DCCS task, and that DCCS training also improved performance on false-belief tasks in 3- and 4-year-olds.
Both false belief and DCCS tasks required the participants to view the same object from an alternative perspective. Their result shows that once children are encouraged to view the same object differently in the first task, they are able to apply a similar approach to a different task. Importantly, this training effect is not to do with the development
of control itself, but to do with the understanding and exploration of different approaches in a given task context.
In all likelihood cognitive flexibility is closely associated with most if not all of the above abilities, and modulated by task-specific demands and structural variables such as cueing, feedback, probabilities of switches, task attributes, task conflicts, and so on. Although the picture outlined here may seem overwhelmingly complex, when broken down, each ability may only matter to different stages or aspects of information processing. For example, to undertake a task, one must form a task representation within a specific context, and exhibit the ability to form an appropriate rule structure (Point 1). The person then needs to employ some form of control to maintain this task representation throughout the duration of the trial (Point 2). To be able to carry out actions purposefully, one must be able to select what information is relevant to the current goal, and filter out the irrelevant and interfering information (Point 3). Once a change of goal is recognised, the person must be able to adjust their response plan according to the new task rule. This new response plan may only be weakly represented in the face of strongly triggered habitual response. Thus, the person needs to overcome the urge to respond too quickly through mechanisms such as inhibitory control (Point 4). And finally, the person may monitor their current state in the overall task context and form an internal representation of the task events, and consciously or unconsciously update task strategies that alter the performance in future tasks (Point 5).
When multiple information processes are in operation, and when each process is dependent on different abilities and experimental factors, any observed behaviour may not be easily attributable to a single cause. For
example, any developmental differences in active task representation may be masked by the response prepotency if the context strongly triggers a habitual response. Similarly, any developmental difference in conflict resolution may be masked by the differences in structural representations of the task set (i.e. when the participant failed to know what the current task goal is). Thus, our understanding of cognitive flexibility may need to move beyond a single factor, and importantly, beyond the sole measure of accuracy. Thus, in this thesis, I will argue that cognitive flexibility may be better investigated by placing greater emphasis on understanding the information processes that take place within the task context, within an individual, and within the developmental constraints (for a similar perspective, see Deák & Wiseheart, 2015).
In the next section I will be describing task-switching studies of adults that focus on the information processing underlying flexible rule switches, and look for both conceptual and procedural overlaps between adult and developmental studies of cognitive flexibility. These overlaps will provide the framework for how cognitive flexibility is viewed and investigated in the current thesis.