Although we find his proposal thought-provoking, we disagree with Carruthers' conclusion that “developmental evidence is of no use” (target article, sect. 4, para. 3) in determining the relationship between mindreading and metacognition, because it is based primarily on an incorrect assumption that false belief performance is the only relevant measure of mindreading and metacognition in children. False beliefs tasks are just one way of measuring mindreading, and not necessarily a good way (Bloom & German Reference Bloom and German2000). Indeed, most developmental psychologists believe that mindreading and metacognition involve a number of subcomponent skills. Some of these skills, such as the ability to understand that others have intentions, may be mastered by infants and toddlers (e.g., Meltzoff Reference Meltzoff1995), whereas others, such as the ability to appreciate false beliefs, are mastered in the early preschool years (Wellman et al. Reference Wellman, Cross and Watson2001). Other skills, such as the ability to understand that a person can have conflicting desires (e.g., Choe et al. Reference Choe, Keil and Bloom2005) or the ability to accept that two people can interpret the same information in different ways (e.g., Carpendale & Chandler Reference Carpendale and Chandler1996), are not grasped until middle childhood. Indeed, many of the component skills that enable mindreading and metacognition are unaccounted for in Carruthers' account of why developmental evidence is not useful. Thus, by examining self/other comparisons in the development of these skills, developmental data can help us better understand potential asymmetries in the emergence of mindreading and metacognition.

Carruthers identifies one possible area for further examination. In his introduction, he points out that mindreading involves not only intentions and beliefs (which are measured in the false belief task), but also components such as knowledge. However, his review of the developmental literature does not take into account the growing body of research examining young children's assessments of what other people know versus what they themselves know.

Determining what another person is likely to know can be considered mindreading because it requires an understanding of the contents and limitations of that person's thoughts. In order to make a judgment about what another person knows, children cannot always rely on associations or broad generalizations. Rather, the evidence suggests that young children have a sophisticated understanding of what other people know or do not know. For instance, they understand that one person can be knowledgeable about some things but not knowledgeable about other things (e.g., Harris Reference Harris2007; VanderBorght & Jaswal, in press). Even 3-year-olds, who might be considered poor at mindreading based on the false belief task, are often adept at identifying knowledgeable individuals when contrasted with ignorant or inaccurate sources (e.g., Birch et al. Reference Birch, Vauthier and Bloom2008; Jaswal & Neely Reference Jaswal and Neely2006; Koenig & Harris Reference Koenig and Harris2005; Sabbagh & Baldwin Reference Sabbagh and Baldwin2001). By age 4, children are also capable of drawing inferences about another person's knowledge that go beyond familiar associations to reflect an understanding of the underlying principles that make up a person's expertise (e.g., Lutz & Keil Reference Lutz and Keil2002). This suggests that young children understand both the contents and limits of another person's knowledge.

Conversely, young children's ability to assess their own knowledge accurately – a form of metacognition – is quite weak (Flavell et al. Reference Flavell, Friedrichs and Hoyt1970; Mills & Keil Reference Mills and Keil2004). For instance, 4- to 5-year-old children are notoriously poor at realizing how much they have learned of a new piece of information (Esbensen et al. Reference Esbensen, Taylor and Stoess1997; Taylor et al. Reference Taylor, Esbensen and Bennett1994). The younger they are, the more children overestimate their understanding (and underestimate their ignorance) of familiar objects and procedures, often not appropriately assessing their own level of knowledge until age 9 or later (Mills & Keil Reference Mills and Keil2004). Thus, despite the evidence that children accurately judge what other people know, there seems to be a real developmental gap in applying these principles to their own knowledge.

One obstacle in applying this research to determining whether metacognition or mindreading is mastered first is that the kinds of questions used to ask children to reflect on their own knowledge are very different from the questions used to measure children's understanding of others' knowledge. Research examining how children think about their own knowledge often requires them to evaluate their own knowledge level using a scale or to estimate the number of problems they can answer correctly. In contrast, studies examining how children think about the knowledge of others typically require them to choose the source that would be able to give the most helpful or accurate information. Given that little research has been conducted directly comparing children's accuracy estimating the knowledge of others to their own knowledge, many open questions remain.

Therefore, in order to provide direct evidence in favor of or against Carruthers' hypothesis, we propose that additional research is necessary. Comparing the development of an understanding of one's own knowledge with an understanding of others' knowledge, or making self/other comparisons based on the other subcomponent skills required for mindreading and metacognition besides false belief, can provide information regarding the developmental trajectory for these skills. In designing this research, it is essential to identify analogous subcomponent skills important for mindreading and metacognition and test them using parallel measures. For instance, one could compare children's ability to identify what they know about a novel object versus another person's knowledge, given the same exact experience with the object. We believe such developmental evidence could go a long way in resolving this debate, and that, based on the research so far, it is likely to support Carruthers' hypothesis.

To conclude, there may be times when developmental evidence is of little use in resolving philosophical debates, such as when the debate is over the existence of free will or whether God exists. We do not believe this is one of those times; rather, Carruthers' proposal addresses a debate that is perhaps best understood in terms of child development. Certainly his hypothesis that mindreading is essential for metacognition poses specific empirical questions that can be tested using developmental methods. We feel strongly that careful research identifying analogous subcomponent skills necessary for mindreading and metacognition and examining their developmental trajectory will provide valuable evidence of whether mindreading really, truly is prior.