Phillips, Buckwalter, Cushman, Friedman, Martin, Turri, Santos, and Knobe review findings from comparative, developmental, and cognitive psychology on the general theme that knowledge is more basic in our conception than belief. Overall, I find their general argument convincing, particularly as it relates to developmental progressions, evidenced by, for example, the theory of mind scales (Wellman & Liu, Reference Wellman and Liu2004). I would like to see them give a fairer consideration of their alternative “View 1” – the hypothesis that belief attribution is more basic than knowledge attribution (at least regarding developmental progressions). I will present two cases not described in their article to see whether they pose challenges for the framework they propose, and discuss the broader implications of these challenges.
1. Knowledge and belief in pretense
Some have argued that children's ability to pretend demonstrates early representational competence and that children scaffold their representational understanding of pretense to help them make explicit judgments about others' false beliefs (e.g., Leslie, Reference Leslie1987). Others, however, have suggested that although young children engage in pretense, doing so posits only the same representational capacities as moving one's body; pretense is “acting-as-if” (Lillard, Reference Lillard1993a; Nichols & Stich, Reference Nichols and Stich2003; Perner, Reference Perner1991). Support for this perspective comes from variants of the “Moe the troll” paradigm: Children are shown a troll doll (Moe), who is hopping up and down like a kangaroo. Because there are no kangaroos in the land of the trolls, Moe doesn't know what one is, and has never seen one before. Four-year-olds – who pass explicit false-belief measures – will erroneously say that Moe is pretending to be a kangaroo (Lillard, Reference Lillard1993b). Here is a case of judgments about others' (false) beliefs being made developmentally earlier than judgments of others knowledge, particularly as they relate to pretending.
Even if one rejects the acting-as-if hypothesis (Friedman & Leslie, Reference Friedman and Leslie2007) or suggests that the Moe findings reflect children's broader causal reasoning (Sobel, Reference Sobel2009), there is a large body of research that suggests children generally understand false belief prior to their understanding the relation between knowledge and other mental states (reviewed in Lillard, Reference Lillard2001; Stich & Tarzia, Reference Stich and Tarzia2015). Notable for the present argument is Perner, Baker, and Hutton's (Reference Perner, Baker, Hutton, Lewis and Mitchell1994) concept of prelief: On this view, knowledge is not more basic than belief. Rather, pretense and belief are an undifferentiated concept when pretend play emerges. They become differentiated with success on false-belief measures, but prelief itself seems more basic than a concept of knowledge.
2. Knowledge and belief in selective learning
Phillips et al. point out that children selectively learn from others, based on their evaluations of their epistemic competence. They conclude, however, that selective learning relies on “representations of knowledge rather than belief in determining from whom to learn.” It is not clear how they come to this conclusion. Classic measures of selective learning (e.g., Koenig, Clément, & Harris, Reference Koenig, Clément and Harris2004; Koenig and Harris, Reference Koenig and Harris2005) introduced preschoolers to two informants. One labeled familiar objects accurately. The other labeled the same objects inaccurately. Researchers used three different measures: (1) Explicit Judgments questions about the informants – whether one informant was either a good or bad labeler. (2) Endorse questions in which they were shown novel objects that were given different novel labels by each informant (e.g., one labeled it a dax, the other a wug); children were asked whether they thought the object was a dax or a wug. (3) Ask questions in which children were asked from whom they wanted to learn labels of novel objects.
These questions potentially ask different things about the knowledge and belief states of the informants. Ask questions assess what children believe about the two informants' knowledge (i.e., given the demonstrations of epistemic competence you've observed, from whom would you want to learn?). Explicit Judgment questions assess a valence judgment about the informants' knowledge. Endorse questions, in contrast, assess what children believe the label of the object really is (presumably, what children believe the informants believe the label to be). Success on these questions – children's ability to use information from informants selectively – has distinct developmental trajectories. Meta-analyses now suggest that at the youngest ages tested, children perform well on Endorse questions, whereas performance on Ask questions and Explicit Judgment questions develops significantly during the preschool years (Sobel & Finiasz, Reference Sobel and Finiasz2020; Tong, Wang, & Danovitch, Reference Tong, Wang and Danovitch2020). Children's selective learning about the belief states of others seems to be present quite early. Selective inferences about facets of others' knowledge seem to have prolonged developmental trajectories.
3. Do I believe everything I know?
I've focused on instances of belief being more basic than knowledge. There are potentially others. Older preschoolers fail on certain measures of true belief, even when they pass measures of false belief (see Hedger & Fabricius, Reference Hedger and Fabricius2011). False-belief contrastive utterance (“I thought it was an X, but it was a Y”) emerge before children pass false-belief measures (see Bartsch & Wellman, Reference Bartsch and Wellman1995), and lead to the possibility that children's understanding of knowledge itself changes (the “connectionist” construal described on pp. 54–55). Linguistic analysis of adults' usage of the words know and think to children suggest that the data necessary to recognize that know is factive is sparse (e.g., Dudley, Rowe, Hacquard, & Lidz, Reference Dudley, Rowe, Hacquard and Lidz2017).
Therefore, although I suspect the view that Phillips et al. advocate has merit, I'd like to see them more carefully consider the alternative account, particularly from a developmental perspective. The findings I've mentioned here require integration into the framework they have set up, as they shed doubt on the hypothesis that all aspects of knowledge are understood by children earlier than belief, and in some cases, suggest that children's conceptualization of knowledge and belief changes over development. Further integrating their arguments with other developmental findings would make their “call to action” to study the role of knowledge in theory of mind development more compelling.
You have
Access
Phillips, Buckwalter, Cushman, Friedman, Martin, Turri, Santos, and Knobe review findings from comparative, developmental, and cognitive psychology on the general theme that knowledge is more basic in our conception than belief. Overall, I find their general argument convincing, particularly as it relates to developmental progressions, evidenced by, for example, the theory of mind scales (Wellman & Liu, Reference Wellman and Liu2004). I would like to see them give a fairer consideration of their alternative “View 1” – the hypothesis that belief attribution is more basic than knowledge attribution (at least regarding developmental progressions). I will present two cases not described in their article to see whether they pose challenges for the framework they propose, and discuss the broader implications of these challenges.
1. Knowledge and belief in pretense
Some have argued that children's ability to pretend demonstrates early representational competence and that children scaffold their representational understanding of pretense to help them make explicit judgments about others' false beliefs (e.g., Leslie, Reference Leslie1987). Others, however, have suggested that although young children engage in pretense, doing so posits only the same representational capacities as moving one's body; pretense is “acting-as-if” (Lillard, Reference Lillard1993a; Nichols & Stich, Reference Nichols and Stich2003; Perner, Reference Perner1991). Support for this perspective comes from variants of the “Moe the troll” paradigm: Children are shown a troll doll (Moe), who is hopping up and down like a kangaroo. Because there are no kangaroos in the land of the trolls, Moe doesn't know what one is, and has never seen one before. Four-year-olds – who pass explicit false-belief measures – will erroneously say that Moe is pretending to be a kangaroo (Lillard, Reference Lillard1993b). Here is a case of judgments about others' (false) beliefs being made developmentally earlier than judgments of others knowledge, particularly as they relate to pretending.
Even if one rejects the acting-as-if hypothesis (Friedman & Leslie, Reference Friedman and Leslie2007) or suggests that the Moe findings reflect children's broader causal reasoning (Sobel, Reference Sobel2009), there is a large body of research that suggests children generally understand false belief prior to their understanding the relation between knowledge and other mental states (reviewed in Lillard, Reference Lillard2001; Stich & Tarzia, Reference Stich and Tarzia2015). Notable for the present argument is Perner, Baker, and Hutton's (Reference Perner, Baker, Hutton, Lewis and Mitchell1994) concept of prelief: On this view, knowledge is not more basic than belief. Rather, pretense and belief are an undifferentiated concept when pretend play emerges. They become differentiated with success on false-belief measures, but prelief itself seems more basic than a concept of knowledge.
2. Knowledge and belief in selective learning
Phillips et al. point out that children selectively learn from others, based on their evaluations of their epistemic competence. They conclude, however, that selective learning relies on “representations of knowledge rather than belief in determining from whom to learn.” It is not clear how they come to this conclusion. Classic measures of selective learning (e.g., Koenig, Clément, & Harris, Reference Koenig, Clément and Harris2004; Koenig and Harris, Reference Koenig and Harris2005) introduced preschoolers to two informants. One labeled familiar objects accurately. The other labeled the same objects inaccurately. Researchers used three different measures: (1) Explicit Judgments questions about the informants – whether one informant was either a good or bad labeler. (2) Endorse questions in which they were shown novel objects that were given different novel labels by each informant (e.g., one labeled it a dax, the other a wug); children were asked whether they thought the object was a dax or a wug. (3) Ask questions in which children were asked from whom they wanted to learn labels of novel objects.
These questions potentially ask different things about the knowledge and belief states of the informants. Ask questions assess what children believe about the two informants' knowledge (i.e., given the demonstrations of epistemic competence you've observed, from whom would you want to learn?). Explicit Judgment questions assess a valence judgment about the informants' knowledge. Endorse questions, in contrast, assess what children believe the label of the object really is (presumably, what children believe the informants believe the label to be). Success on these questions – children's ability to use information from informants selectively – has distinct developmental trajectories. Meta-analyses now suggest that at the youngest ages tested, children perform well on Endorse questions, whereas performance on Ask questions and Explicit Judgment questions develops significantly during the preschool years (Sobel & Finiasz, Reference Sobel and Finiasz2020; Tong, Wang, & Danovitch, Reference Tong, Wang and Danovitch2020). Children's selective learning about the belief states of others seems to be present quite early. Selective inferences about facets of others' knowledge seem to have prolonged developmental trajectories.
3. Do I believe everything I know?
I've focused on instances of belief being more basic than knowledge. There are potentially others. Older preschoolers fail on certain measures of true belief, even when they pass measures of false belief (see Hedger & Fabricius, Reference Hedger and Fabricius2011). False-belief contrastive utterance (“I thought it was an X, but it was a Y”) emerge before children pass false-belief measures (see Bartsch & Wellman, Reference Bartsch and Wellman1995), and lead to the possibility that children's understanding of knowledge itself changes (the “connectionist” construal described on pp. 54–55). Linguistic analysis of adults' usage of the words know and think to children suggest that the data necessary to recognize that know is factive is sparse (e.g., Dudley, Rowe, Hacquard, & Lidz, Reference Dudley, Rowe, Hacquard and Lidz2017).
Therefore, although I suspect the view that Phillips et al. advocate has merit, I'd like to see them more carefully consider the alternative account, particularly from a developmental perspective. The findings I've mentioned here require integration into the framework they have set up, as they shed doubt on the hypothesis that all aspects of knowledge are understood by children earlier than belief, and in some cases, suggest that children's conceptualization of knowledge and belief changes over development. Further integrating their arguments with other developmental findings would make their “call to action” to study the role of knowledge in theory of mind development more compelling.
Financial support
The author was funded by NSF grants 1661068, 1917639, and 2033368 during writing of this commentary.
Conflict of interest
None.