There are few developmental psychologists who have attempted to elucidate in detail the origin of concepts in infancy. As one of that band, I applaud Carey's book (2009) but have some suggestions as to how to simplify her view of core cognition.
Carey states that it is impossible to derive concepts solely from sensorimotor information. I agree. However, elaborate conceptualizing machinery may not be necessary. Throughout her book, she defends discontinuity in development and the importance of bootstrapping. Therefore, we must consider the possibility that core cognition may be leaner than she suggests, with more of the work being accomplished by associative learning enriching a small number of innate conceptual representations (followed by linguistic and analogical bootstrapping).
Carey offers dedicated innate analyzers for objects, agents, and number plus a rather unspecified central processor that also has innate concepts. The latter can be combined with outputs from the dedicated analyzers to produce, for example, the concept of cause. Not much is said about the interactions between dedicated analyzers and innate central concepts. It is difficult to disentangle them, and this approach may allow too many alternative empirical outcomes. Furthermore, a more parsimonious system can accomplish the same goals. A single innate analyzer (such as Perceptual Meaning Analysis; Mandler Reference Mandler2004; Reference Mandler2008; Reference Mandler2010) that simplifies attended spatiotemporal information into a small number of conceptual primitives, can produce first concepts of objects, agency, and causality. It also allows combinations of them and provides first concepts for relations such as containment, occlusion, and support. The resulting representations, in conjunction with information directly supplied by the perceptual system, are sufficient to account for current infant data, including early language understanding (although this approach has not yet been extended to number).
The uncertainty Carey expresses as to how a concept of cause originates (Carey Reference Carey2009, Ch. 6) illustrates the problem of assuming that some innate concepts are the product of separate analyzers and others part of a central mechanism. Carey argues against Michotte's (Reference Michotte and Miles1946/1963) view that perceiving motion transferred from one object to another is obligatory and foundational for understanding causality, because she states that whether objects are inert or animate affects causal interpretation right from the beginning. However, not only is this conclusion debatable, it does not invalidate the view that motion transfer is obligatory and foundational for causal understanding. For example, the fact that infants are not surprised when animates move without contact does not refute Michotte's claim; there is no evidence that infants conceive of self-starting motion as causally based.
Carey's stronger argument depends on the claim of simultaneous emergence of concepts of contact causality and change of state causality. However, in my reading of the literature, I do not find enough evidence for simultaneous emergence. Adult-like response to both contact causality and entraining has been demonstrated at 3 to 4 months of age (Leslie Reference Leslie1982) whereas change of state causality has not been not shown before 8 months of age, and not even then unless a hand is involved. In infancy research, this is a sizeable gap. Leslie's experiments differ in detail from his definitive work with 6-month-olds, but the outcomes at 3 to 4 months are essentially the same.
Further, there is the dynamic aspect of causality. If forceful causality can be learned, then why cannot change of state causality also be learned, leaving motion transfer sufficient for core cognition? A way to do this (Mandler Reference Mandler2008; Reference Mandler2010) is by an innate conceptual primitive of “make move” based on seeing motion transfer from one object to another, with force added to the concept only when infants begin to move themselves around in the world and experience their own exertions in manipulating objects. Three-to-four-month-olds have little, if any, such experience. Once they do in the second six months, there is an already organized representation of caused motion available to be integrated with feelings of bodily exertion. Change of state causality is apt to be conceptualized even more slowly than adding force, because although infants may notice the relevant correlations, they need a more complex chain of associations to reach the core “make move” concept. Even adults often misconstrue change of state causality when it is not associated with motion transfer; it is not obligatory in the same way.
Carey also rejects my single analyzer approach (p. 195) because she says there is no known way that Perceptual Meaning Analysis could transform spatio-temporal properties into representations of intentional agency. But agency (goal-directed behavior)
can
be defined in spatial–temporal vocabulary, and there is evidence that this is indeed how it begins in infancy, as the observation of repeated paths of motion taking the most direct possible paths to the same end point (e.g., Csibra Reference Csibra2008). Infants learn early on that people are the most likely agents, but they accept inanimate boxes as agents too if they follow contingent paths. Even adults sometimes do, suggesting the obligatory character of this core concept. Understanding agency in terms of mental intentions is not part of core cognition but a late development, requiring infants' own attempts (and failures) to reach goals to become associated with the earlier established representations of agency in terms of paths of motion. Associating eyes (or head turns) with goal-directed paths is easy enough to learn, but mental intentionality is difficult and may even require language to become established.
Another concern is how Carey's core cognition enables the recall of event sequences and mental problem solving that have been demonstrated in the second six months. Such mental activities require explicit concepts, but the latter are not part of her core cognition and in her account explicit concepts appear to require language (Ch. 1). Therefore, although concepts are defined as units of thought, it is not clear how preverbal infants manage such thoughtful processes as recall and problem solving. An advantage of a mechanism such as Perceptual Meaning Analysis is that it creates iconic representations enabling imaginal simulations that even preverbal infants can use for thought.
I agree with much that Carey proposes but suggest that a single innate mechanism may suffice as the origin of core concepts.
There are few developmental psychologists who have attempted to elucidate in detail the origin of concepts in infancy. As one of that band, I applaud Carey's book (2009) but have some suggestions as to how to simplify her view of core cognition.
Carey states that it is impossible to derive concepts solely from sensorimotor information. I agree. However, elaborate conceptualizing machinery may not be necessary. Throughout her book, she defends discontinuity in development and the importance of bootstrapping. Therefore, we must consider the possibility that core cognition may be leaner than she suggests, with more of the work being accomplished by associative learning enriching a small number of innate conceptual representations (followed by linguistic and analogical bootstrapping).
Carey offers dedicated innate analyzers for objects, agents, and number plus a rather unspecified central processor that also has innate concepts. The latter can be combined with outputs from the dedicated analyzers to produce, for example, the concept of cause. Not much is said about the interactions between dedicated analyzers and innate central concepts. It is difficult to disentangle them, and this approach may allow too many alternative empirical outcomes. Furthermore, a more parsimonious system can accomplish the same goals. A single innate analyzer (such as Perceptual Meaning Analysis; Mandler Reference Mandler2004; Reference Mandler2008; Reference Mandler2010) that simplifies attended spatiotemporal information into a small number of conceptual primitives, can produce first concepts of objects, agency, and causality. It also allows combinations of them and provides first concepts for relations such as containment, occlusion, and support. The resulting representations, in conjunction with information directly supplied by the perceptual system, are sufficient to account for current infant data, including early language understanding (although this approach has not yet been extended to number).
The uncertainty Carey expresses as to how a concept of cause originates (Carey Reference Carey2009, Ch. 6) illustrates the problem of assuming that some innate concepts are the product of separate analyzers and others part of a central mechanism. Carey argues against Michotte's (Reference Michotte and Miles1946/1963) view that perceiving motion transferred from one object to another is obligatory and foundational for understanding causality, because she states that whether objects are inert or animate affects causal interpretation right from the beginning. However, not only is this conclusion debatable, it does not invalidate the view that motion transfer is obligatory and foundational for causal understanding. For example, the fact that infants are not surprised when animates move without contact does not refute Michotte's claim; there is no evidence that infants conceive of self-starting motion as causally based.
Carey's stronger argument depends on the claim of simultaneous emergence of concepts of contact causality and change of state causality. However, in my reading of the literature, I do not find enough evidence for simultaneous emergence. Adult-like response to both contact causality and entraining has been demonstrated at 3 to 4 months of age (Leslie Reference Leslie1982) whereas change of state causality has not been not shown before 8 months of age, and not even then unless a hand is involved. In infancy research, this is a sizeable gap. Leslie's experiments differ in detail from his definitive work with 6-month-olds, but the outcomes at 3 to 4 months are essentially the same.
Further, there is the dynamic aspect of causality. If forceful causality can be learned, then why cannot change of state causality also be learned, leaving motion transfer sufficient for core cognition? A way to do this (Mandler Reference Mandler2008; Reference Mandler2010) is by an innate conceptual primitive of “make move” based on seeing motion transfer from one object to another, with force added to the concept only when infants begin to move themselves around in the world and experience their own exertions in manipulating objects. Three-to-four-month-olds have little, if any, such experience. Once they do in the second six months, there is an already organized representation of caused motion available to be integrated with feelings of bodily exertion. Change of state causality is apt to be conceptualized even more slowly than adding force, because although infants may notice the relevant correlations, they need a more complex chain of associations to reach the core “make move” concept. Even adults often misconstrue change of state causality when it is not associated with motion transfer; it is not obligatory in the same way.
Carey also rejects my single analyzer approach (p. 195) because she says there is no known way that Perceptual Meaning Analysis could transform spatio-temporal properties into representations of intentional agency. But agency (goal-directed behavior) can be defined in spatial–temporal vocabulary, and there is evidence that this is indeed how it begins in infancy, as the observation of repeated paths of motion taking the most direct possible paths to the same end point (e.g., Csibra Reference Csibra2008). Infants learn early on that people are the most likely agents, but they accept inanimate boxes as agents too if they follow contingent paths. Even adults sometimes do, suggesting the obligatory character of this core concept. Understanding agency in terms of mental intentions is not part of core cognition but a late development, requiring infants' own attempts (and failures) to reach goals to become associated with the earlier established representations of agency in terms of paths of motion. Associating eyes (or head turns) with goal-directed paths is easy enough to learn, but mental intentionality is difficult and may even require language to become established.
Another concern is how Carey's core cognition enables the recall of event sequences and mental problem solving that have been demonstrated in the second six months. Such mental activities require explicit concepts, but the latter are not part of her core cognition and in her account explicit concepts appear to require language (Ch. 1). Therefore, although concepts are defined as units of thought, it is not clear how preverbal infants manage such thoughtful processes as recall and problem solving. An advantage of a mechanism such as Perceptual Meaning Analysis is that it creates iconic representations enabling imaginal simulations that even preverbal infants can use for thought.
I agree with much that Carey proposes but suggest that a single innate mechanism may suffice as the origin of core concepts.