Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

An effective system must be able to monitor and regulate itself. That is, it must be able to evaluate the current state of the system (and how well the system is progressing towards its goals), and the system must be able to alter its behavior, based on these evaluations, in order to achieve the desired goal more accurately or efficiently. In humans (and perhaps nonhuman species; see Smith, Shields, & Washburn, 2003), monitoring and regulation of cognitive operations are achieved via metacognition (i.e., thinking about thinking; see Flavell, 1979).

When an individual evaluates the current state of his or her basic cognitive operations (e.g., memory, perception, problem solving), metacognitive monitoring is said to occur. When an individual uses the output of his or her monitoring evaluations (in conjunction with the current task demands, e.g., for speed or accuracy), to regulate their basic cognitive operations, metacognitive control is said to occur (Koriat & Goldsmith, 1996; Nelson & Narens, 1990, 1994).

Metacognitive monitoring and control may take many forms, depending on the type of task at hand and the stage of the task. This point may be illustrated by considering the example of a student preparing to take an exam. As the student is preparing for the exam, she will need to evaluate how well she has learned the material (i.e., a judgment of learning; see Schneider, Vise, Lockl, & Nelson, 2000), and direct her time and attention to studying materials that she is struggling with (i.e., allocation of study time; Dunlosky & Connor, 1997). During the exam, she may encounter test items for which she does not immediately know the answer; in this case she will have to assess the likelihood that she will be able to remember the answer (i.e., a feeling-of-knowing judgment; see Butterfield, Nelson, & Peck, 1988; Hart, 1965) to decide if she ought to allot extra time to these test items or if her time would be better spent in working on other test items. Finally (if there is a penalty for providing an incorrect response, as is common on many standardized tests), when answering each question, the student must evaluate how certain she is about the likely accuracy of her response (i.e., a confidence judgment; see Roebers, 2002), and decide whether to provide that response (if the level of certainty is sufficiently high) or to refrain from answering the test question­ (if the level of certainty is low) (Koriat & Goldsmith, 1996). Thus, metacognitive monitoring and control can be thought of as a quality control system, whose function is to ensure that only the most accurate and appropriate output is produced.

For this reason, age-related improvements in children’s ability to monitor and regulate their mental operations are widely considered to be a driving force in cognitive development, underlying age-related improvements in accuracy on a wide variety of tasks (Koriat, Goldsmith, Schneider, & Nakash-Dura, 2001; Kuhn & Pease, 2006; Plude, Nelson, & Scholnick, 1998). Accordingly, the development of these skills in childhood has been a central focus of metacognitive research. This work has revealed that critical developments in monitoring and control are observed during middle and late childhood (Ghetti, 2008; Koriat et al., 2001; Lockl & Schneider, 2002; Roebers & Howie, 2003; Son, 2005).

Nevertheless important questions about the emergence and early development of metacognition remain unanswered, as the majority of research has excluded younger children. In large part, this exclusion is due to prevailing views that young children have extremely limited metacognitive skills. This view may be based on findings indicating striking deficits in older children’s ability to monitor and regulate their mental activity (Flavell, Green, & Flavell, 1995, 2000; Lockl & Schneider, 2004), which have lead to the logical inference that young children’s metacognitive skills may be extremely limited.

However, there is some evidence from a small but compelling body of literature to suggest that young children may be more metacognitively skilled than previously assumed (see also Whitebread et al., this volume). In the following sections we review this evidence. Taken together, these findings raise several intriguing questions about the emergence and early development of metacognition, which are discussed in the latter half of this chapter.

2 Metacognition in the Preschool Years

Although young children are often assumed to have limited metacognitive skills, several studies have provided direct or indirect evidence of metacognitive monitoring and control in preschoolers. In the following sections we review this research. First, we consider the evidence indicative of metacognitive monitoring in early childhood; then we examine the evidence indicative of metacognitive control in early childhood.

2.1 Metacognitive Monitoring in Early Childhood

Findings from several lines of research suggest that young children may be more aware of their ongoing mental activity than previously assumed, including studies of children’s ability to monitor their knowledge states (Marazita & Merriman, 2004; Patterson, Cosgrove, & Obrien, 1980), their mental imagery (Estes, 1998), and their memory processes (Cherney, 2003; Cultice, Somerville, & Wellman, 1983). Although some of this research was not originally conducted with the intention of assessing metacognitive monitoring, each offers evidence that young children are capable of reflecting on at least some aspects of their ongoing mental activity. However, the extant literature is sparse and rather disjointed. Thus, in the following paragraphs we discuss each of these lines of research individually, and then provide an integrative discussion.

2.1.1 Monitoring of Knowledge States

Children begin to produce mental verbs referring to knowledge states (e.g., know, think, I don’t know) at the end of the second year of life. Although the first appearance of these words in conversation may simply reflect knowledge of the pragmatics­ surrounding their use, mental verbs are employed to refer specifically to one’s own or others’ mental states by the third year of life (with know and think being the most prevalent; Moore, Bryant, & Furrow, 1989; Moore, Furrow, Chiasson, & Patriquin, 1994; Shatz, Wellman, & Silber, 1983), suggesting that children as young as 3 years of age may be able to reflect upon their states of knowledge or lack of knowledge.

More direct evidence of this ability (to evaluate when one does or does not know something) comes from research on metalinguistic monitoring (Marazita & Merriman, 2004). In this research, children were asked to report whether or not they knew the meaning of real or fake words (read aloud by an experimenter) or whether or not they knew the word labels for real and fake objects (depicted in drawings). Results indicate that children as young as 2.5-years-old are capable of discriminating between instances when they know something (e.g., the meaning of a real word spoken aloud to them or the word label for a real object depicted in a line drawing) and instances when they do not know something (e.g., the meaning of a fake word spoken aloud to them or the word label of a fake object depicted in a line drawing; Marazita & Merriman, 2004), suggesting that well before the age of 3 years, children­ can reflect upon their current knowledge state, discriminating between states of knowledge and ignorance.

However, monitoring of one’s knowledge state involves much more than being able to evaluate whether or not one has knowledge. Often, individuals have some knowledge, but this knowledge is insufficient for them to be able to respond accurately (either because their knowledge state is incomplete or unclear). In these instances, individuals must not only reflect upon whether or not they have knowledge, but how good their knowledge is, in a more graded fashion. One could imagine that this may prove to be more challenging than simply evaluating whether or not one does or does not have an answer in mind.

To date, this ability, to evaluate the graded nature of one’s knowledge state has typically been assessed using comprehension monitoring tests, which assess children’s ability to evaluate how well they understand another speaker’s message. Typically, in these tasks children are given a task to complete (e.g., to follow tape-recorded instructions for building a tower out of blocks), and are given prompts which are problematic (i.e., ambiguous or incomplete messages, such as “put the red one on top”, when there is more than one red object in the stimuli set), and prompts that are non-problematic (i.e., clear, comprehensive, and unambiguous messages). Comprehension monitoring is assessed by examining whether children’s verbal or non-verbal behaviors differ in response to problematic versus non-problematic messages.

Using these procedures, clear evidence of comprehension monitoring has been observed even in very young preschoolers. For example, Patterson et al. (1980) found that preschoolers delayed their responding, spent more time looking at the experimenter, and physically wavered between response options more frequently when they were presented with uninformative messages as compared to when they were presented with informative messages. Children as young as 2.5-years of age have also been found to evince verbal indicators of comprehension monitoring (Revelle, Wellman, & Karabenick, 1985). In this research, children played a game with an experimenter, who requested that the children retrieve items in the playroom. Requests included non-problematic (i.e., clear, unambiguous requests) as well as problematic requests (i.e., inaudible, ambiguous, or impossible requests). Children’s statements were coded for an indication that they detected the problematic nature of the request (e.g., asking the experimenter to repeat or clarify his/her message, stating that the request was not feasible). While younger preschoolers only appeared to monitor comprehension failures due to experimenters’ requests being inaudible, or impossible (e.g., the item requested was absent or too heavy for the child to lift), older preschoolers additionally detected comprehension failures due to ambiguous messages and memory overloads (i.e., the experimenter asking the children to retrieve a long list of items, which the children could not keep in mind).

The results of this line of research suggest that the ability to monitor comprehension may develop in a step-like process, with children first exhibiting awareness of comprehension failures due to striking problems with the incoming messages (e.g., instructions which are impossible to follow). Later, children develop awareness of more subtle problems in comprehension (e.g., ambiguous messages). Furthermore, the somewhat contrasting results of Patterson et al. (1980) and Revelle et al. (1985) suggest that implicit awareness of these more subtle distinctions may be observed before children become able to note them explicitly. Consistent with this notion are findings from a more recent study indicating that preschoolers evince nonverbal indicators of comprehension monitoring (e.g., quizzical facial expressions) much more frequently than verbal indicators of comprehension monitoring (Skarakis-Doyle, 2002). It should be noted, however, that requests for clarification or repetition of a message involve not only metacognitive monitoring (of one’s comprehension) but also metacognitive control (of one’s knowledge state by seeking additional information). Thus, although comprehension monitoring is often assessed using these indicators, in such paradigms it is not possible to disentangle the contributions of monitoring and control to children’s performance.

2.1.2 Monitoring of Mental Imagery

Another domain in which children’s awareness of their ongoing mental activity has been explored is mental imagery. To do so, Estes (1998) asked children to report how they made their judgments on a mental rotation task (Shepard & Metzler, 1971). Children were presented with two images in different spatial orientations (rotated around a center point), and were asked to judge whether the images were the same or different (i.e., “Are these monkeys holding up the same arm or different arms?”). In order to make the judgment, children had to mentally represent one of the images in their minds, mentally rotate the image, and compare that (rotated) mental image to the target object.

When preschoolers were asked how they made their judgments, 40% of 4-year-olds and 56% of 5-year-olds referred to mental activity in their responses. Critically, the behavioral response pattern of these children corroborated their reports (i.e., children who provided “mental” explanations evinced a pattern of responding consistent with actual use of a mental rotation strategy: judgment reaction times increased with the angle of rotation). Children who did not provide mental explanations for how they performed the task did not exhibit this pattern, suggesting that children who referred to their mental activity seemed to be reliably reporting on their mental imagery – for additional indications of early awareness of mental imagery see, also, Estes, Wellman, and Woolley (1989) and Rieser, Garing, and Young (1994). These results provide a straightforward and compelling indication that children, as young as 4 years of age, are capable of metacognitively monitoring the contents of their mental imagery.

2.1.3 Memory Monitoring

Evidence of metacognitive monitoring in young children also comes from studies of preschoolers’ ability to monitor their memory operations. Although relatively few studies have examined memory monitoring in young children, the studies that have done so have provided compelling evidence that young children are capable of reflecting on several dimensions of the retrieval process. For example, Cultice et al. (1983) investigated preschoolers’ ability to provide feeling-of-knowing judgments (i.e., judging whether they would be able to recognize the names of schoolmates, pictured in photographs, whose names they could not currently recall). Results indicated that participants’ feeling-of-knowing judgments corresponded to their subsequent memory performance. Given that, in adults, feeling-of-knowing judgments can arise as a function of the familiarity of the test cues (Koriat & Levy-Sadot, 2001), these results suggest that children as young as 4-years-old may be capable of introspecting on their sense of familiarity for test materials.

More recently, Cherney (2003) examined the extent to which 3- and 5-year-olds’ spontaneous utterances of mental terms indicating subjective certainty (e.g., know, remember, forget) and subjective uncertainty (e.g., think, guess, bet) corresponded to the accuracy of their memory on a spatial location recall task. Although mental terms were uttered infrequently (uttered by about 20% of 3-year-olds and 30% of 5-year-olds), children’s utterances were moderately likely to be consistent with their memory performance (e.g., saying “I know” and then providing a correct response or saying “I forget” and then providing an incorrect response). Correspondence was better for older compared to younger children and was better for terms expressing certainty compared to uncertainty. This latter finding is consistent with research in older children suggesting that children may gain mastery of certainty monitoring earlier than they master uncertainty monitoring (Roebers, von der Linden, & Howie, 2007), suggesting relative continuity in development of metacognitive functioning over the course of childhood.

Typically, monitoring of subjective certainty is indexed using confidence judgments (i.e., a scale-based rating of how confident one is about the likely accuracy of one’s response). These judgments are among the most widely used indices of metacognitive monitoring in research with older children and adults (e.g., Roebers, 2002) and nonhuman animals (Shields, Smith, Guttmannova, & Washburn, 2005; see also, Beran, Couchman, Coutinho, Boomer, & Smith, this volume). This approach to the study of metacognitive monitoring has at least two major advantages. First, confidence judgments can be collected for a wide variety of decisions (e.g., perceptual decisions, memory decisions), and second, collecting confidence ratings in conjunction with decisions affords the opportunity to examine the degree to which individuals’ reports of their subjective mental experiences (i.e., their confidence judgments) correspond to their actual mental experiences (i.e., their memory decisions), providing an objective indicator of monitoring ability (see also, Allwood, this volume).

If confidence judgments associated with accurate responses are higher than confidence judgments associated with inaccurate responses, metacognitive monitoring is inferred. The logic of this approach is that, if an individual is reflecting on their subjective sense of certainty about the likely accuracy of their decisions, they should, on average, report higher confidence for correct compared to incorrect responses. This pattern of results has been observed on memory tasks in children as young as 5 years of age (Ghetti, Qin, & Goodman, 2002). Five-year-olds also report higher confidence ratings on memory tests when recognizing words that have been studied with a picture compared to words that have been studied without a picture (Ghetti et al., 2002) and report that they remember more details about highly memorable life events compared to low memorability life events (Ghetti & Castelli, 2006), offering evidence that young children are capable of evaluating the strength of their memory representations at retrieval.

In sum, the results of several investigations of memory monitoring in young children strongly suggest that children as young as 4 years of age can reflect on critical aspects of their memory operations at retrieval. These dimensions include the feeling of knowing that one will be able to remember something that cannot currently be retrieved, and the strength and likely accuracy of one’s memory representations.

2.1.4 Conclusion

Taken together, these findings suggest that young children may be much more adept at monitoring their mental activity than is often assumed. The extant literature indicates that children as young as 30 months of age can evaluate whether they do or do not know something (Marazita & Merriman, 2004; Moore et al., 1989; Moore et al., 1994). While this ability may be rudimentary, it arguably provides a foundation for subsequent developments in metacognitive monitoring, which are observed during the preschool years.

Between the ages of 3 and 5 years, children demonstrate age-related improvements in their ability to monitor more fine-grained distinctions among their knowledge states. Children’s ability to monitor their comprehension improves during this period (Cosgrove & Patterson, 1977; Pratt & Bates, 1982), as does their ability to introspect on their mental imagery (Estes, 1998) and monitor the vividness and detail of their memory representations (Ghetti & Castelli, 2006; Ghetti et al., 2002). Furthermore, there is initial evidence that children’s ability to reflect upon their sense of certainty about the accuracy of their memory states improves during this period (Cherney, 2003) as well, suggesting that during the preschool years children may develop the ability to reflect upon cues to the subjective experience of high and low certainty (e.g., how fluently one processes information or retrieves a response, how much time is required to process information or retrieve a response). Consistent with this notion are recent findings indicating that between the ages of 3 and 5 years, preschoolers’ confidence judgments increasingly differentiate between accurate and inaccurate perceptual decisions (although even 3-year-olds report higher confidence for accurate vs. inaccurate responses), and that during this period uncertainty monitoring ability begins to extend to additional types of decision making–for example, linguistic decisions (Lyons & Ghetti, in press).

There are, of course, limitations in young children’s monitoring abilities, perhaps most strikingly demonstrated by Flavell et al. (1995, 2000) who documented that introspecting on the contents of one’s thoughts poses great challenges for 5-year-olds. Nevertheless, the results of several studies provide compelling evidence that young children are in many ways quite capable of monitoring their ongoing mental activity. This ability is critical as it provides the foundation for future action aimed at improving one’s knowledge or the accuracy of one’s performance. Of course, this benefit can only occur to the extent that children are able to act on the basis of monitoring evaluations to metacognitively control their cognitive activity. Several lines of research suggest that this ability may be observed even in very young children.

3 Metacognitive Control in Early Childhood

Relatively little research has directly examined early self-regulation from a metacognitive perspective. However, findings from several lines of research suggest that young children readily engage in specific behaviors aimed at improving the precision of their understanding or the accuracy their performance, including studies of children’s regulation of their knowledge states, children’s use of rudimentary memory strategies, and children’s control of goal-directed activity. This work has been conducted in a rather fragmented fashion; therefore, in the following sections we review each of these lines of research individually and then provide an integrative discussion.

3.1 Control of Knowledge States

To improve their knowledge states, children as young as 12 months of age seek information from adults (through expressions and gestures), and preschoolers ask an average of more than 70 information-seeking questions per hour; it has been argued that children ask these questions when they are faced with an experience of uncertainty (Chouinard, 2007), although this proposal has yet to be tested empirically. Furthermore, recent research indicates that even 3-year-olds seek information selectively and are more likely to trust individuals who were previously correct compared to individuals who declared their ignorance (Koenig & Harris, 2005), demonstrating that preschoolers actively seek information from reliable sources in order to improve their knowledge.

The results of comprehension monitoring studies offer more direct evidence that young children act deliberately to overcome limitations in their knowledge states. For example, even young preschoolers can be trained to request clarification in response to ambiguous messages to improve the accuracy of their performance (Pratt & Bates, 1982). However, older preschoolers respond to comprehension failures in a more problem-focused way (e.g., requesting clarification or elaboration that would be helpful in resolving the failure) to a greater extent than younger children (Revelle et al., 1985; see also Walters & Chapman, 2000), suggesting that the ability to overcome limitations in one’s current knowledge state by seeking information does appear to improve with age. Taken together, these findings suggest that young children actively seek additional information from reliable sources in order to improve the accuracy and clarity of their states of knowledge and understanding.

3.2 Control of Memory Performance

The majority of research on metacognitive regulation of performance in early childhood has been conducted in the context of memory tasks (Acredolo, Pick, & Olsen, 1975; Deloache & Brown, 1983). Several studies have found that very young children selectively engage in “studying” behaviors in the context of a memory task (but not on parallel tasks in which the memory demands are removed). For example, Wellman, Ritter, and Flavell (1975) found that 3-year-old children spent more time looking at and pointing at the location of a hidden object when they were told to “remember” the location of a toy hidden under one of several opaque cups than when children were simply told to “wait” during the delay period; the amount of time spent engaged in these behaviors positively correlated with memory performance.

Although younger children (2-year-olds) failed to demonstrate such an effect in Wellman et al. (1975), a similar pattern of selective “studying” behavior was observed in toddlers as young as 18 months of age, when the task involved searching for a stuffed toy hidden in one of several naturalistic hiding places (e.g., under a pillow, behind a chair) (Deloache, Cassidy, & Brown, 1985). Again, these behaviors (e.g., talking about the toy and/or its location) were associated with subsequent memory performance and were rarely observed when the memory demands were removed (e.g., when the child was told that the experimenter would search for the toy at the end of the delay, or the child was told that their task would be to “wake up” the toy, which was placed in plain sight, after the delay). Furthermore, these behaviors were more readily engaged in when the task was performed in a novel (laboratory) setting than a familiar (home) setting, offering a preliminary indication that these behaviors are under strategic control, and engaged in more readily when children are less certain about their performance ability (as would be the case in a novel setting). Experimental research corroborates this interpretation, as similar increases in studying behaviors have been documented as a result of incentive manipulations (i.e., preschoolers engage in more studying behaviors when the incentives for accurate performance are increased; see O’Sullivan, 1993).

Further evidence of control has been obtained by examining children’s behaviors at retrieval. For example, Deloache and Brown (1984, 1987) observed that at retrieval, upon failure to locate a hidden toy (either because children forgot the location or because the toy was surreptitiously moved), children as young as 2 years of age directed their subsequent search to related locations (i.e., adjacent locations or analogous locations such as behind the pillow on the opposite side of a couch). Moreover, children were more persistent in searching when the toy had been surreptitiously moved (i.e., when they should be highly confident in their memory for the location), compared to when they forgot the location (i.e., when they should be less confident in their memory for the location). Taken together, these results suggest that, when faced with retrieval failure, children continue to act in a strategic fashion in order to produce optimal performance.

More recent findings indicate that preschoolers also strategically regulate the accuracy of their performance by withholding inaccurate information, when they are given the option to refrain from responding. For example, in an eyewitness memory interview study, Mulder and Vrij (1996) found that 4- to 5-year-old children were more accurate (i.e., provided fewer incorrect responses) when they were instructed to say “I don’t know” when they were uncertain about their answers. Impressively, the advantage observed in this age group was similar to that of older children (8- to 10-year-olds). In another study, 3-year-olds evinced higher accuracy rates on a paired-associates memory test when they were given the option to skip trials, compared to when they were required to provide an answer on all trials (Balcomb & Gerken, 2008). However, clear limitations in metacognitive control were observed, as more than one-third of the children never used the opt-out option or used the opt-out response on all trials (with the former being significantly more common, exhibited by 89% of the children who consistently used or failed to use the opt out response).

This finding is consistent with the well-documented challenges that young children evince with inhibitory control (Gerstdat, Hong, & Diamond, 1994; Reed, Pien, & Rothbart, 1984). From this perspective, young children’s challenges with metacognitively regulating their accuracy may be due to limitations in executive function, rather than an inability to act on the basis of monitoring evaluations to guide strategic responding. Extant procedures have thus far been unable distinguish between the contributions of metacognitive control and inhibitory control to preschoolers’ performance on these tasks. Although research on the early development of metacognitive control has largely been conducted in parallel to research on the development of cognitive control and executive function during the preschool years, these domains are conceptually, theoretically, and developmentally intertwined (Fernandez-Duque, Baird, & Posner, 2000; Shimamura, 2000; see also Whitebread et al., this volume). Thus, the early development of metacognitive control needs to be examined while accounting for global developments in executive function and self-regulation skills in preschoolers (Carlson, 2005).

Following this approach, a recent study explored the relation between individual differences in inhibitory control and preschoolers’ ability to take advantage of a withholding strategy to improve the accuracy of their performance on a perceptual-identification task (Lyons & Ghetti, 2010). Replicating previous findings, preschoolers across age groups evinced higher accuracy rates when the option to withhold was present compared to when it was not. Critically, participants reported higher confidence in association with responses that were provided rather than withheld, offering clear evidence that metacognitive control was influenced by the results of metacognitive monitoring evaluations. Further evidence consistent with this notion was the finding that individual differences in uncertainty monitoring ability positively predicted regulation of accuracy (via withholding of incorrect responses).

However, an additional path to metacognitive control was observed. Specifically, inhibitory control and uncertainty monitoring ability were independent positive predictors of accuracy improvements (due to strategic withholding). These findings are notable because they suggest that control of accuracy may be achieved if children approach tasks cautiously, treading carefully when there are any cues to indicate that the task at hand may be difficult to perform or that the risk of making an error is high (independently of uncertainty monitoring ability).

3.3 Self-Directed Metacognitive Control

Historically, the topic of self-directed control has received substantial attention in the developmental literature, at least in part stemming from the Vygotskian view that young children talk to themselves as a form of early self-regulation (Manning, White, & Daugherty, 1994). Consistent with this view is research from naturalistic settings (namely, preschools) indicating that older preschoolers engage in self-talk in a systematic way, for example, more frequently doing so while they are working alone than while in groups or working with adults, and more frequently when they are engaged in focused and prolonged goal-directed activities. While self-talk is also observed in younger preschoolers, it is not systematically associated with the same situations as it is for older preschoolers, suggesting improvement in the ability to self-regulate one’s thoughts and actions during the preschool years (Winsler, Carlton, & Barry, 2000).

More recent evidence indicates that in addition to verbal self-regulation, preschoolers also evince nonverbal metacognitive regulation, including error detection and checking (e.g., that one is doing a task appropriately), redirection of activities (e.g., directing attention back to the main task), help-seeking (e.g., from peers or adults), and transferring previously successful strategies to new situations (Dermitzaki, Leondari, & Goudas, 2009; Whitebread, Bingham, Grau, Pino Pasternak, & Sangster, 2007). These behaviors are also more readily engaged in when children work more independently and are less frequently observed when adults are involved in interactions (Whitebread et al., 2007), suggesting that these behaviors may be under conscious control and more likely to be engaged in when the need for self-regulation is greater.

3.4 Conclusion

In sum, several studies have documented metacognitive control (or metacognitive control-like) behaviors in preschoolers. Children begin to seek information from adults around their first birthday (through gestures and expressions; Chouinard, 2007). During the preschool years children’s ability to metacognitively control their knowledge states through information-seeking develops substantially, as they become increasingly likely to request information from reliable sources (Koenig & Harris, 2005), and their questions become more problem-focused, and directed towards gaining information that will be helpful in improving their knowledge and understanding (Revelle et al., 1985).

In addition to these strategic actions aimed at improving their knowledge states, preschoolers also act to improve the accuracy of their performance, engaging in deliberate strategies to improve memory retention (Deloache et al., 1985; Wellman et al., 1975), and retrieval accuracy (Deloache & Brown, 1984, 1987). These rudimentary strategies are observed very early in life (well before the preschool years, under some conditions; Deloache et al., 1985); however, the ability to engage in more advanced forms of strategic control of performance (e.g., through the strategic withholding of incorrect responses) improves with age during the preschool years (Balcomb & Gerken, 2008; Lyons & Ghetti, 2010; Mulder & Vrij, 1996), although there is evidence that metacognitive monitoring evaluations guide strategic responding even in young preschoolers. These age improvements in strategic control of performance accuracy develop in parallel with age improvements in children’s ability to use verbal and nonverbal strategies for self-regulation when engaged in goal-directed behaviors (Whitebread et al., 2007; Winsler et al., 2000), consistent with the well-documented age-increase executive function observed during the preschool years (Carlson, 2005).

4 Towards a Comprehensive Model of Early Metacognitive Development

Overall, the findings reviewed above provide compelling evidence that critical milestones in metacognitive monitoring and control are achieved in early childhood. These early abilities likely provide the foundation for learning in a host of domains as well as subsequent metacognitive development. Nevertheless young children have exhibited some clear limitations in their ability to monitor their mental activity (e.g., being greatly challenged at reporting about what they were thinking about only a moment earlier; Flavell et al., 1995, 2000), as well as in their ability to strategically regulate their mental activity and behavior (e.g., in directing their attention during study only to items which they will be tested about later; Miller, 1990), underscoring the complex nature of early metacognitive development.

Thus, although important (but often unrecognized) steps have been made towards developing a comprehensive understanding of early metacognitive development, several fundamental questions remain unanswered. In the following sections we raise some of these questions about the early development of metacognition and offer some thoughts on how these issues may be resolved. Answering these questions will elucidate the mechanisms underlying early metacognitive development, and will help lay the foundation for a comprehensive model of the emergence of metacognition in early childhood.

4.1 How Do Metacognitive Monitoring and Control Develop in Early Childhood?

An important first step in developing a comprehensive model of early metacognitive development will be to characterize the exact mechanisms through which ­metacognitive monitoring and control develop. It has been proposed that awareness and regulation of mental activity may progress through a series of increasingly complex levels of self-reflection. For example, Zelazo and colleagues (Zelazo, 2004) have proposed an information processing account of the development of conscious awareness. This model contends that at birth children are minimally conscious of their world; they are aware only of the stimulus which they are currently encountering and whether this stimulus gives rise to a pleasurable or a negative feeling (i.e., “minimal consciousness”). Around the first birthday, infants achieve the ability to bring back to mind stimuli which are no longer in the environment and consider them in relation to one another (i.e., “recursive consciousness”); that is to say, they are able to bring to mind and reflect upon the contents of “minimal consciousness”. Around the second year, children achieve a third level of conscious awareness (i.e., “self-consciousness”), at which they are able to bring to mind the contents of “recursive consciousness” and are able to explicitly reflect on them. Additional levels of conscious awareness are achieved as children progress through further iterative recursions, bringing to mind and reflecting upon the contents of their mental activity from lower levels of consciousness. It is proposed that children’s ability to control their behavior increases as a function of achievements in the highest level of reflection that children can engage in (Zelazo, Gao, & Todd, 2007). Support for this proposal comes from a series of experiments demonstrating that in the first several years of life, children are increasingly able to follow complex embedded rules (e.g., on the dimensional change card sort task, “If we are playing the color game the red truck goes with the red cards, but if we are playing the shape game, the red truck goes with the truck cards”) (Zelazo, 2004). Although a number of experiments have provided evidence in support of components of the model, to date, the model has not been tested in its entirety within a single study.

Similarly, Flavell (2003) speculated that children may develop awareness of uncertainty in a four-stage process. At the first stage of development (i.e., at birth), infants may not have any experience of uncertainty. During this period, children would not exhibit any behavioral differences in the face of certain or uncertain situations. At the second stage of development, children may have a subjective experience of uncertainty but fail to be consciously aware of it. During this period, children may have slower reaction times when asked to predict the outcome of uncertain compared to certain situations, but they would not report feeling any differently about responding to the two kinds of prompts.

At the third stage of development, children may be consciously aware of the subjective experience of uncertainty, but may not attribute it as such. Although for adults and older children, the subjective experience of uncertainty is readily identified as such, one could imagine that when awareness of these feelings first emerges, young children may not identify them as indicating uncertainty. Even adults sometimes misattribute the sources of their subjective experiences; for instance, they may misattribute their arousal (actually resulting from experimental manipulations) to feelings of romantic attraction (Dutton & Aron, 1974; White, Fishbein, & Rutstein, 1981). Moreover, there is evidence for young children’s wishful thinking and biases towards overconfidence (Schneider, 1998). Therefore, it seems probable that young children may progress through a phase in which they may experience the feelings associated with subjective uncertainty but they may not associate these feelings with uncertainty. However, as children gain more experience (perhaps as they learn to associate their subjective feelings of uncertainty with instances in which they produced the wrong answer or response), they would progress to the final stage of uncertainty monitoring, at which they are consciously aware of their subjective experience of uncertainty and recognize it as uncertainty. This proposal has yet to be empirically tested.

To fully characterize the emergence and early development (and indeed development throughout the lifespan) of metacognition, future research should examine how the ability to monitor and regulate mental activity develops over time and across domains, using procedures that can be used with individuals from a wide age-span and across tasks. Developing such procedures may be challenging, but it would prove invaluable in helping to elucidate the reasons why young children are able to reflect on some types of mental activity (e.g., images, word knowledge; Estes, 1998; Marazita & Merriman, 2004) but not others (e.g., thoughts; Flavell et al., 1995, 2000), as well as why young children are capable of engaging in some aspects of metacognitive control (e.g., strategic withholding; Balcomb & Gerken, 2008), but not others (e.g., selectively attending to relevant information during encoding; Miller, 1990).

4.2 What Is the Relation Between Monitoring and Control in Early Childhood?

Research with young children to date has established a set of conditions under which young children can monitor their thinking and control their performance. However, it remains unknown to what extent young children’s metacognitive control operations are guided by the results of monitoring evaluations. What level of awareness is required for children to be able to control their behavior? Do children act on the basis of implicit metacognition (e.g., asking questions to resolve ambiguity, despite not being explicitly aware that they are uncertain)? If so, how should this level of “awareness” be characterized, and how does acting on the basis of such monitoring influence awareness?

The question of whether monitoring evaluations guide control processes, and whether control can occur in the absence of monitoring has been debated in recent years in the adult metacognitive literature. Although traditional models of metacognition hold that control occurs as a consequence of monitoring, recent findings (from the adult behavioral and patient literature) suggest that control may occur in the absence of conscious monitoring (Moulin, Perfect, & Fitch, 2002), and that monitoring may even occur under some circumstances as a consequence of control operations (Koriat, Ma’ayan, & Nussinson, 2006).

Studying the emergence and early development of these processes may provide critical insight into the nature of the relation between metacognitive monitoring and metacognitive control. Recent findings suggest that children as young as 3-years-old engage in control operations on the basis of monitoring evaluations of certainty (refraining from responding when their confidence is low and providing responses when their confidence is high; Lyons & Ghetti, 2010). However, there may be a period earlier in development in which a dissociation is observed between monitoring and control (such that young children are capable of metacognitive control before they are capable of consciously reflecting on their ongoing mental activity), which would raise questions about how control is achieved in the absence of monitoring. One possibility, certainly worth investigating, is that children may proceed more cautiously when environmental cues suggest that the risk of making an error is high (e.g., if a task is unfamiliar).

4.3 Do the Monitoring and Control Processes Differ Across Domains in Early Childhood?

A comprehensive understanding of the early development of metacognition, must include a description of (and account for) the domain specificity or domain generality of metacognitive operations across cognitive functions. That is to say, are monitoring and control skills similar across domains in young children? Do young children develop the ability to monitor and regulate their mental states similarly for different kinds of cognitive functions (e.g., memory, perception, language) and/or for different aspects of cognitive functioning (e.g., judgments of learning, confidence ratings, feelings of knowing)?

The complex patterns of research findings to date suggest that the development of metacognitive monitoring and control are unlikely to develop uniformly across domains. Instead, it seems that children may develop the ability to monitor certain kinds of mental content (e.g., monitoring of one’s mental imagery; see Estes, 1998; see also Lyons & Ghetti, in press) before they develop the ability to monitor other forms of mental content (e.g., monitoring­ of thoughts in one’s stream of consciousness; Flavell et al., 1995, 2000). Similarly, children may develop the ability to engage in some forms of metacognitive regulation earlier than others–for example, strategies aimed at preventing forgetting (Wellman et al., 1975) may precede strategies aimed at remembering specific items (Miller, 1990). By characterizing the development of monitoring and control processes across domains, and identifying the characteristics of monitoring and control processes which emerge earlier and later in development, future research could provide critical insight into the mechanisms underlying age-related improvements in metacognition. Such information would be invaluable in helping to develop a comprehensive model of metacognitive development, and may provide insight into the nature of metacognitive processes generally.

4.4 What Are the Neural Bases of Monitoring and Control and How Do They Develop in Early Childhood?

A comprehensive model of early metacognitive development must take into account the neural underpinnings of monitoring and control, and how these change with age and experience. To date, very few studies have examined the neural correlates of procedural metacognition in young children. However, there is a growing body of cognitive neuroscience research (from patient and neuroimaging studies) which provides some insight into the neural bases of metacognitive operations in adults and older children.

There is good evidence, from several lines of research, that monitoring and regulation are supported by the prefrontal cortex (Bunge & Zelazo, 2006; Shimamura, 2000) and the anterior cingulate cortex (Fernandez-Duque et al., 2000). With regard to the former, studies with confabulating patients indicate that damage to ventromedial cortex causes impairments in individuals’ ability to monitor the output of their retrieval (for accuracy), resulting in confabulation, that is, the reporting of false information (Schnider, 2003). Patients with prefrontal lobe lesions also exhibit deficits in the feeling-of-knowing evaluations (Shimamura & Squire, 1986), source monitoring (Janowsky, Shimamura, & Squire, 1989), and feelings of subjective recollection (Ciaramelli & Ghetti, 2007). Research with healthy adults also indicates that the prefrontal cortex is involved in monitoring of retrieved information on memory tasks (Ranganath, Johnson, & D’Esposito, 2000; Ranganath & Paller, 2000) and assessments of feeling of knowing (Paynter, Reder, & Kieffaber, 2009). Neuroimaging studies of executive functioning in typically developing adults and older children have also identified midfrontal circuits (involving the frontal cortex and the anterior cingulate cortex) as supporting error monitoring and cognitive control (Fernandez-Duque et al., 2000).

Taken together, these findings suggest that age-related changes in the structural and functional maturation of the frontal lobes and anterior cingulate cortex may underlie age-related improvements in monitoring and control. Integrating neuroimaging methods with behavioral assessments of metacognitive development in early childhood in future research will provide an important foundation for developing a model of early metacognitive development.

5 General Conclusions

An effective system must be able to monitor and regulate itself. Although young children are often assumed to have extremely limited metacognitive skills, there is good evidence to indicate that young children may be more metacognitively skilled than previously assumed. Nevertheless, the extant research is fraught with apparently discrepant findings, with young children demonstrating competence in some domains of metacognitive monitoring and control, while exhibiting striking limitations in others. These seemingly contradictory findings may be resolved by future studies addressing several remaining fundamental questions about the origins and early development of metacognition. However, the extant literature provides good evidence that critical achievements in metacognition are achieved in early childhood, providing the foundation for learning in a variety of domains.