Determining how explicit concepts such as matter, eight, and dog obtain their narrow content, and discriminating content-determining from merely associated conceptual features, is a crucial challenge that remains to be fully articulated in Carey's final chapter (Carey Reference Carey2009). It is important to maintain some division between those mechanisms in the categorization literature that we consider diagnostic of category structure and those that reflect mere belief revision.
I will suggest that Carey's account of content-determination for natural kinds in particular should be amended to allow for revisions in narrow conceptual content that are not exclusively dictated by changes as wide-ranging as scientific progress. Although the suggestion that narrow content is only revised in response to scientific change ensures that Carey's criterion for concepts having shared content across individuals is met, this criterion comes at a cost. It does not allow for bona fide conceptual change as the result of exposure to regularities in smaller-scale causal processes. Such an amendment is therefore critical if The Origin of Concepts (henceforth TOOC) is to allow for natural kind concepts to participate in smaller-scale conceptual change throughout the life course.
Within domain theories, Carey argues, some features are more causally fundamental than others. That is, the category features that are most central within the domain of natural kinds may be those that determine the narrow content of these concepts. I will argue that domain theories, or “intuitive theories,” of natural kind concepts that guide this causal prioritizing are importantly malleable and sensitive to changes in the causal and explanatory role played by instances of these concepts.
Whereas Carey argues that the prototype effects found in the categorization literature do not reflect anything about the deep structure of concepts, she suggests that features that are most causally central in a conceptual structure (Ahn et al. Reference Ahn, Kim, Lassaline and Dennis2000) provide a sound empirical starting point for the determination of narrow content. Carey further suggests that the causal status (Ahn et al. Reference Ahn, Kim, Lassaline and Dennis2000) of a property within its intuitive theory is one model of conceptual structure upon which we might base narrow content. On this account, categories possess a causal structure. Features that are more often causes than effects within this structure have priority in categorization decisions.
One strength of the causal status effect has been its ability to account for previously documented content-based differences in feature centrality between natural kinds and artifacts (Ahn Reference Ahn1998; Keil Reference Keil1989). However, the ability of this phenomenon to assimilate structurally diverse features (e.g., functional features for artifacts and molecular features for natural kinds) may be due in part to the structural guidance provided by the task itself, which often manipulates knowledge about properties by directly stating their statuses as causes and effects. Notably, when this strong manipulation is no longer present, and features are introduced within a functional, rather than a mechanistic, causal–explanatory structure, the causal status effect disappears (Lombrozo Reference Lombrozo2009). Along similar lines, Lombrozo and Carey (Reference Lombrozo and Carey2006) found that preference for teleological explanations in accounting for the functions of natural kind and artifact objects depends on the causal history of the object's function and not on the domain of the object per se. As Lombrozo and Carey argue, with regard to the applicability of teleological explanation, object domains and causal processes do not exhibit a one-to-one mapping. Across natural kinds and artifacts, the causal history of an object influences the way it is explained; conversely, explanatory stance influences a feature's centrality in a category. Given the malleability of their category structure, a similar claim might be made with respect to the structure of natural kind concepts.
Consider how this one-to-multiple mapping operates in Lombrozo's (Reference Lombrozo2009) stimulus set. Participants were presented with a fictional flower (a “holing”) with a certain substance in its stem (“brom compounds”) that caused a second feature (“bending over”). This second feature was also conceived of as a functional feature (“by bending over, the holing's pollen can brush against the fur of field mice, and spread to neighboring areas”). Participants were asked to explain “why holings typically bend over,” and their explanations were classified as appealing to either a mechanistic or a functional structure. They were then asked to categorize new flowers, some of which lacked the mechanistic causal precursor feature (“brom compounds”), and others of which lacked the functional feature (“spreading pollen”). Overall, participants who offered a functional explanation for bending over regarded the functional feature as more causally central in categorization decisions, and those who offered a mechanistic explanation (appealing to the prior cause) regarded the mechanistic feature (“brom compounds”) as more central. A direct manipulation of explanatory stance found similar results. These results indicate that the features of natural kind concepts do not exhibit a clear “default” structure. Participants both entered the experiment with slightly different prior intuitions about the explanatory mode most diagnostic of category structure, and responded to a category structure manipulation.
What do these findings tell us about conceptual change? If conceptual role is central to determining the narrow content of natural kind concepts, and this content only varies in the rare cases in which our scientific knowledge grows, then at best, there is much empirical work left to be done to show how the narrow content of a concept remains consistent across differences in causal history and explanatory stance. Otherwise, these differences suggest that these stances are themselves representational resources highly sensitive to smaller scale causal (and functional) variations in the world.
Of course, learning a fabricated new species of flower is a task limited in scope; the concept token referred to will not likely transform all future instances of “natural kind” encountered. But if some measure of conceptual change remains possible throughout the life course, then concepts themselves, and not just their associated categories, must respond to smaller scale regularities in the causal structure of specific instances of those concepts that are encountered.
You have
Access
Determining how explicit concepts such as matter, eight, and dog obtain their narrow content, and discriminating content-determining from merely associated conceptual features, is a crucial challenge that remains to be fully articulated in Carey's final chapter (Carey Reference Carey2009). It is important to maintain some division between those mechanisms in the categorization literature that we consider diagnostic of category structure and those that reflect mere belief revision.
I will suggest that Carey's account of content-determination for natural kinds in particular should be amended to allow for revisions in narrow conceptual content that are not exclusively dictated by changes as wide-ranging as scientific progress. Although the suggestion that narrow content is only revised in response to scientific change ensures that Carey's criterion for concepts having shared content across individuals is met, this criterion comes at a cost. It does not allow for bona fide conceptual change as the result of exposure to regularities in smaller-scale causal processes. Such an amendment is therefore critical if The Origin of Concepts (henceforth TOOC) is to allow for natural kind concepts to participate in smaller-scale conceptual change throughout the life course.
Within domain theories, Carey argues, some features are more causally fundamental than others. That is, the category features that are most central within the domain of natural kinds may be those that determine the narrow content of these concepts. I will argue that domain theories, or “intuitive theories,” of natural kind concepts that guide this causal prioritizing are importantly malleable and sensitive to changes in the causal and explanatory role played by instances of these concepts.
Whereas Carey argues that the prototype effects found in the categorization literature do not reflect anything about the deep structure of concepts, she suggests that features that are most causally central in a conceptual structure (Ahn et al. Reference Ahn, Kim, Lassaline and Dennis2000) provide a sound empirical starting point for the determination of narrow content. Carey further suggests that the causal status (Ahn et al. Reference Ahn, Kim, Lassaline and Dennis2000) of a property within its intuitive theory is one model of conceptual structure upon which we might base narrow content. On this account, categories possess a causal structure. Features that are more often causes than effects within this structure have priority in categorization decisions.
One strength of the causal status effect has been its ability to account for previously documented content-based differences in feature centrality between natural kinds and artifacts (Ahn Reference Ahn1998; Keil Reference Keil1989). However, the ability of this phenomenon to assimilate structurally diverse features (e.g., functional features for artifacts and molecular features for natural kinds) may be due in part to the structural guidance provided by the task itself, which often manipulates knowledge about properties by directly stating their statuses as causes and effects. Notably, when this strong manipulation is no longer present, and features are introduced within a functional, rather than a mechanistic, causal–explanatory structure, the causal status effect disappears (Lombrozo Reference Lombrozo2009). Along similar lines, Lombrozo and Carey (Reference Lombrozo and Carey2006) found that preference for teleological explanations in accounting for the functions of natural kind and artifact objects depends on the causal history of the object's function and not on the domain of the object per se. As Lombrozo and Carey argue, with regard to the applicability of teleological explanation, object domains and causal processes do not exhibit a one-to-one mapping. Across natural kinds and artifacts, the causal history of an object influences the way it is explained; conversely, explanatory stance influences a feature's centrality in a category. Given the malleability of their category structure, a similar claim might be made with respect to the structure of natural kind concepts.
Consider how this one-to-multiple mapping operates in Lombrozo's (Reference Lombrozo2009) stimulus set. Participants were presented with a fictional flower (a “holing”) with a certain substance in its stem (“brom compounds”) that caused a second feature (“bending over”). This second feature was also conceived of as a functional feature (“by bending over, the holing's pollen can brush against the fur of field mice, and spread to neighboring areas”). Participants were asked to explain “why holings typically bend over,” and their explanations were classified as appealing to either a mechanistic or a functional structure. They were then asked to categorize new flowers, some of which lacked the mechanistic causal precursor feature (“brom compounds”), and others of which lacked the functional feature (“spreading pollen”). Overall, participants who offered a functional explanation for bending over regarded the functional feature as more causally central in categorization decisions, and those who offered a mechanistic explanation (appealing to the prior cause) regarded the mechanistic feature (“brom compounds”) as more central. A direct manipulation of explanatory stance found similar results. These results indicate that the features of natural kind concepts do not exhibit a clear “default” structure. Participants both entered the experiment with slightly different prior intuitions about the explanatory mode most diagnostic of category structure, and responded to a category structure manipulation.
What do these findings tell us about conceptual change? If conceptual role is central to determining the narrow content of natural kind concepts, and this content only varies in the rare cases in which our scientific knowledge grows, then at best, there is much empirical work left to be done to show how the narrow content of a concept remains consistent across differences in causal history and explanatory stance. Otherwise, these differences suggest that these stances are themselves representational resources highly sensitive to smaller scale causal (and functional) variations in the world.
Of course, learning a fabricated new species of flower is a task limited in scope; the concept token referred to will not likely transform all future instances of “natural kind” encountered. But if some measure of conceptual change remains possible throughout the life course, then concepts themselves, and not just their associated categories, must respond to smaller scale regularities in the causal structure of specific instances of those concepts that are encountered.
ACKNOWLEDGMENT
The author thanks Steve Sloman for comments on a previous draft.