Carey's illuminating discussion of the empirical motivation for a dual factor theory of content is remarkably broad (Carey Reference Carey2009), but there are some lingering worries one might have regarding narrow content. Carey wants to avoid radical semantic holism because of its troubles with explaining how interpersonal communication is possible. Yet she also wants narrow content to have a role in determining the semantics of concepts. In order to do so, she has to specify which computational roles are content-determining and which are merely extraneous beliefs or part of the sustaining mechanisms. Theories that cannot distinguish the content-determining inferences from the non–content-determining inferences must say that all inferences are content-determining, thereby embracing a radical semantic holism. Carey sees this problem and puts forth some forceful suggestions concerning the two types of concepts that her book focuses on: concepts stemming from core cognition, and bootstrapped concepts. She contends that for core cognitive concepts (e.g., agent, object), the inferences that are content-determining are the innately specified and unrevisable ones; for concepts that are bootstrapped (e.g., force, division) the inferences that form the original hermeneutic circle are the content-determining inferences (e.g., the narrow content of force is specified by its connection to mass and acceleration). These appear to be quite different criteria for the two different types of concepts. Presumably what is underwriting the criteria is that the narrow content in question is apt to stay constant across people, therefore allowing for interpersonal communication.
So far, so good. However, a worry arises once we look at concepts that are neither derived from core cognition nor bootstrapped in a Quinean way. Take, for example, the definitionally and interpretationally primitive concept fence (if it turns out fence isn't definitionally and interpretationally primitive, then substitute your favorite concept that is both definitionally and interpretationally primitive and not the output of core cognition, e.g., fan). Presumably, such a concept is neither the output of core cognition nor the result of some bootstrapping process (at the very least it does not fit the canonical pattern of placeholder structures that Carey describes for other bootstrapped concepts). Imagine three people introduced to fences in the following ways: one is introduced to fences by reading Robert Frost and comes to believe that what is most important about fences is that they make good neighbors. Another is introduced to fences in New York, where they are mostly chain-link and topped with barbed wire. The third has grown up next to a field that has a few feet of picket fence placed right in the middle of it, separating nothing of consequence. What content is the shared narrow content among these people?
The current worry is that for concepts that aren't the outputs of core cognition or bootstrapped via placeholders there will be no set of inferences that are interpersonally shared. In the above example, the central computational roles of fence don't appear to be shared. The concepts of these three people needn't even share beliefs about the function of fences (the third person may have the concept even without having the belief that fences are used to separate areas) nor need their concepts share any perceptual features. But it seems plausible that such a group could still communicate about fences. If this is so, how do we distinguish the inferences that are content-constitutive from the ones that are not? What types of narrow content must be shared among such people in order to facilitate communication? In other words, what is the narrow content of fence?
Carey says that this type of problem, the problem of specifying the gritty details of content, is endemic to all competitor theories. Certainly this is true, but there is more reason to be skeptical about her particular proposal than she lets on; what a theory of narrow content would need in order to ground shared content (and thus shared concepts) is nothing short of an analytic/synthetic distinction. Carey does an admirable job of motivating something like an analytic/synthetic distinction for core cognitive and bootstrapped concepts, which is no small feat. However, she accomplishes this via appealing to data that point to shared psychological structures that, as a matter of nomological fact, underwrite the acquisition of certain concepts. The question remains: for the multitudinous class of concepts that are simple yet not the result of innate machinery, what are the shared psychological structures underwriting the shared narrow content? They can't be prototypes, as Carey herself claims that “[prototype] theories also make a mystery of shared concepts, failing to address the problem of disagreement: my prototype of a dog (or the dog exemplars I represent) must be different from yours, yet we both have concepts of dogs” (Carey Reference Carey2009, p. 497).
As suggested, Carey might respond by saying that all the competitor theories face similar problems of specifying the details. However, the burden of proof does not appear to be identical across theories. For example, Carey criticizes Fodor's theory for not detailing the sustaining mechanisms for concepts. But Fodor can't characterize the “mere engineering” (see, e.g., Carey Reference Carey2009, p. 535) because he supposes that they are a heterogeneous array of mechanisms, and consequently research will have to proceed piece by piece (see, e.g., Fodor Reference Fodor1998). There is no in-principle argument against this piecemeal endeavor, even though surely we are quite a long way from the completion of any such task. In contrast, for Carey's dual content view to work out, we need something akin to an analytic/synthetic distinction which, even without the unnecessary epistemological baggage, is an elusive prize. Nevertheless, we may hope that such a distinction can be found: after all, Carey has been able to impressively describe something akin to it for certain kinds of concepts. That being said, those concepts seem like they may be the exception, not the rule. So I close by asking: what could the shared narrow content for simple, non-core, non-bootstrapped concepts be?
You have
Access
Carey's illuminating discussion of the empirical motivation for a dual factor theory of content is remarkably broad (Carey Reference Carey2009), but there are some lingering worries one might have regarding narrow content. Carey wants to avoid radical semantic holism because of its troubles with explaining how interpersonal communication is possible. Yet she also wants narrow content to have a role in determining the semantics of concepts. In order to do so, she has to specify which computational roles are content-determining and which are merely extraneous beliefs or part of the sustaining mechanisms. Theories that cannot distinguish the content-determining inferences from the non–content-determining inferences must say that all inferences are content-determining, thereby embracing a radical semantic holism. Carey sees this problem and puts forth some forceful suggestions concerning the two types of concepts that her book focuses on: concepts stemming from core cognition, and bootstrapped concepts. She contends that for core cognitive concepts (e.g., agent, object), the inferences that are content-determining are the innately specified and unrevisable ones; for concepts that are bootstrapped (e.g., force, division) the inferences that form the original hermeneutic circle are the content-determining inferences (e.g., the narrow content of force is specified by its connection to mass and acceleration). These appear to be quite different criteria for the two different types of concepts. Presumably what is underwriting the criteria is that the narrow content in question is apt to stay constant across people, therefore allowing for interpersonal communication.
So far, so good. However, a worry arises once we look at concepts that are neither derived from core cognition nor bootstrapped in a Quinean way. Take, for example, the definitionally and interpretationally primitive concept fence (if it turns out fence isn't definitionally and interpretationally primitive, then substitute your favorite concept that is both definitionally and interpretationally primitive and not the output of core cognition, e.g., fan). Presumably, such a concept is neither the output of core cognition nor the result of some bootstrapping process (at the very least it does not fit the canonical pattern of placeholder structures that Carey describes for other bootstrapped concepts). Imagine three people introduced to fences in the following ways: one is introduced to fences by reading Robert Frost and comes to believe that what is most important about fences is that they make good neighbors. Another is introduced to fences in New York, where they are mostly chain-link and topped with barbed wire. The third has grown up next to a field that has a few feet of picket fence placed right in the middle of it, separating nothing of consequence. What content is the shared narrow content among these people?
The current worry is that for concepts that aren't the outputs of core cognition or bootstrapped via placeholders there will be no set of inferences that are interpersonally shared. In the above example, the central computational roles of fence don't appear to be shared. The concepts of these three people needn't even share beliefs about the function of fences (the third person may have the concept even without having the belief that fences are used to separate areas) nor need their concepts share any perceptual features. But it seems plausible that such a group could still communicate about fences. If this is so, how do we distinguish the inferences that are content-constitutive from the ones that are not? What types of narrow content must be shared among such people in order to facilitate communication? In other words, what is the narrow content of fence?
Carey says that this type of problem, the problem of specifying the gritty details of content, is endemic to all competitor theories. Certainly this is true, but there is more reason to be skeptical about her particular proposal than she lets on; what a theory of narrow content would need in order to ground shared content (and thus shared concepts) is nothing short of an analytic/synthetic distinction. Carey does an admirable job of motivating something like an analytic/synthetic distinction for core cognitive and bootstrapped concepts, which is no small feat. However, she accomplishes this via appealing to data that point to shared psychological structures that, as a matter of nomological fact, underwrite the acquisition of certain concepts. The question remains: for the multitudinous class of concepts that are simple yet not the result of innate machinery, what are the shared psychological structures underwriting the shared narrow content? They can't be prototypes, as Carey herself claims that “[prototype] theories also make a mystery of shared concepts, failing to address the problem of disagreement: my prototype of a dog (or the dog exemplars I represent) must be different from yours, yet we both have concepts of dogs” (Carey Reference Carey2009, p. 497).
As suggested, Carey might respond by saying that all the competitor theories face similar problems of specifying the details. However, the burden of proof does not appear to be identical across theories. For example, Carey criticizes Fodor's theory for not detailing the sustaining mechanisms for concepts. But Fodor can't characterize the “mere engineering” (see, e.g., Carey Reference Carey2009, p. 535) because he supposes that they are a heterogeneous array of mechanisms, and consequently research will have to proceed piece by piece (see, e.g., Fodor Reference Fodor1998). There is no in-principle argument against this piecemeal endeavor, even though surely we are quite a long way from the completion of any such task. In contrast, for Carey's dual content view to work out, we need something akin to an analytic/synthetic distinction which, even without the unnecessary epistemological baggage, is an elusive prize. Nevertheless, we may hope that such a distinction can be found: after all, Carey has been able to impressively describe something akin to it for certain kinds of concepts. That being said, those concepts seem like they may be the exception, not the rule. So I close by asking: what could the shared narrow content for simple, non-core, non-bootstrapped concepts be?