See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228828863

The Origin of Concepts

Article in Journal of Cognition and Development · February 2000

DOI: 10.1207/S15327647JCD0101N_3

CITATIONS READS
1,367 11,924

1 author:

Susan Carey
Harvard University
229 PUBLICATIONS 34,186 CITATIONS

SEE PROFILE

All content following this page was uploaded by Susan Carey on 26 March 2015.

The user has requested enhancement of the downloaded file.

JOURNAL OF COGNITION AND DEVELOPMENT, 2000, Volume 1, pp. 37-41
Copyright © 2000, Lawrence Eribaum Associates, Inc.

COMMENTARIES

The Origin of Concepts

Susan Carey
Department of Psychology
New York University

Jean Mandler's rich and nuanced feature article (this issue) develops a four-part
argument:

1. An adequate characterization of the adult mind must distinguish conceptual

representations from perceptual representations.
2. A priori arguments cannot decide the issue of the ontogenetic roots of each
type of representation, nor of the relations between them. In particular, there is no
convincing a priori argument that infants' representations are first perceptual and
only later conceptual. Furthermore, there is no known mechanism through which
conceptual representations may be built from perceptual representations.
3. Data from three paradigms-sequential touching, manual habituation, and
i mitation-provide convergent evidence that conceptual categories exhibit a differ-
ent course of development than do perceptual categories. Specifically, conceptual
categories emerge at the domain level (e.g., animal vs. vehicle) before the basic (dog
vs. cat) or subordinate (poodle vs. collie) levels, whereas most available evidence
suggests that categories based on visual similarity are formed at the basic level first.
4. A process of attentive perceptual analysis yields conceptual categories.

Requests for reprints should be sent to Susan Carey, Department of Psychology, New York Univer-
sity, 6 Washington Place, New York, NY 10003. E-mail: [email protected]
38 CAREY

It is always easy to quibble with a programmatic statement with the scope of

Mandler's lovely feature article, and I am not able to refrain from doing so. But
first, I would like to strongly endorse, quibbles aside, the first three parts of her ar-
gument. I reserve serious doubts only for the fourth, and offer a friendly amend-
ment to Mandler's project.
1. Distinguishing perceptual from conceptual representations. In distin-
guishing perceptual from conceptual representations, Mandler most often appeals
to a difference between what entities look like and what kinds of entities they are.
Additionally, she claims that conceptual representations differ from perceptual
representations in being consciously accessible, supporting problem solving and
inference, and being stored in long-term memory. To quibble, it is not clear that
these different properties determine a single type of representation, or whether
Mandler considers all of these to be properties of conceptual representations. Each
may characterize at least some perceptual representations. For example, some
clear cases of perceptual representations, such as a toothache or the experience of
redness, may be consciously accessible.
Nonetheless, the distinction between categories based on what entities look like
(or sound like or feel like) and categories based on what kind of things entities are
can be cashed out in the tradition of the "theory theory" of concepts (Carey, 1985;
Gopnik & Meltzoff, 1997; Keil, 1989). In this tradition, concepts such as those dis-
cussed in Mandler's feature article (tiger, mammal, animal, bird, duck, vehicle,
car, cup, container, etc.) include both core and peripheral features. The core of the
concept includes its causally deepest properties, those properties that determine
what kind of thing the entity is and its particular properties. In this tradition, con-
ceptual categories are those with cores, those for which adults take the stance
Medin and Ortony (1989) calledpsychological essentialism. Core properties, or
essential properties, are often not perceptually available. For example, in the adult
concept of a tiger, its essence is inherited from its parents, is internal, and is not ob-
servable (Keil, 1989).
If concepts' cores include nonobservable causal constructs, then concepts that
have cores have a nonperceptual component. The attribution of causality goes be-
yond spatiotemporal analysis. Even in the case of Michotte-like contact causality,
the mind contributes the causal attribution; perception merely gives us contact, si-
multaneity, and so forth. Similarly, perception gives us aspects of paths and con-
tingency; the mind attributes goals to agents.
This is a subtle point. The fact that we can see a physical event as a causal inter-
action does not make the concept of causal interaction perceptual any more than
the fact that we can see a certain building as a nuclear reactor makes the concept of
nuclear reactor perceptual. Perceptual categories, as I and, I believe, Mandler
mean them, are formed from observational properties such as red, square, dog-
shaped, and spatiotemporally specified aspects of motion.
 ORIGIN OF CONCEPTS 39

2. A priori relations between conceptual and perceptual categories. Here I

have no quibbles with the points Mandler develops in her feature article and wish
only to add the following observation. If causal analysis is at the core of at least
some conceptual representations, then a relevant literature as to the origin of con-
ceptual representations is the literature on infants' appreciation of causality
( Gergeley, Nadasdy, Csibra, & Biro, 1995; Leslie, 1988; Spelke, Phillips, &
Woodward, 1995). As I read this literature, at least by the time infants are 7 to 12
months of age, the earliest ages at which Mandler's studies reveal conceptual rep-
resentations of domain-level concepts, infants appreciate both Michotte-like con-
tact causality and aspects of agency. Furthermore, both types of causality are
attributed to entities on the basis of analysis of their action, not on the basis of what
they look like. Gergeley et al. (1995) showed that 12-month-old infants attribute
goal directness to dots that appear to chase other dots through gaps in walls or ap-
pear to jump over barriers to reach each other. Johnson, Slaughter, and Carey
(1998) showed that 12-month-olds follow the focus of attention of faceless, amor-
phous robots, as long as those robots interact contingently with the baby. Although
perceptual information is required for the infant to identify cases of contact causal-
ity or intentional goal-directed activity, the causal attributions go beyond the per-
ceptual information.
3. For conceptual categories, domain-level distinctions precede, developmen-
tally, basic-level distinctions. Whether a task reflects perceptual or conceptual
categorization is both an empirical and a theoretical matter. It depends on how
we draw the perceptual-conceptual distinction, and then it depends on what
drives the behavior in question. A minor quibble concerning this part of
Mandler's argument is that the tasks in which she and her colleagues have
shown domain-level distinctions to precede basic-level distinctions (sequential
touching, manual habituation, and imitation) do not transparently reflect concep-
tual rather than perceptual categorization. It is an open question, for example,
whether infants of the age Mandler has studied would habituate to a manually
presented class of objects united by a clearly perceptual property. Van de Walle
(1999) showed that they will, as 9-month-olds presented with a series of red
horses recover interest when allowed to play with a yellow horse, and they gen-
eralize habituation when offered a red pig. That the basic-level distinction did
not determine recovery of interest is, of course, consistent with Mandler's own
data from this age. What is new is that the color distinction did. Manual habitua-
tion at 9 months can reveal perceptual categorization.
This is not to say that the domain-level distinctions reflected in these tasks do
not reflect categories with conceptual cores. Rather, my point is simply that the na-
ture of the representational distinction subserving some behavioral distinction
cannot simply be ascertained by considering the nature of the task itself. That be-
ing said, I do agree that the imitation-induction paradigm in particular seems on its
40 CAREY

face to reflect conceptual categorization. Affording "keying" or "drinks from a

cup" is not an observation property, and the infants' generalization patterns were
unaffected by perceptual similarity.
The convergence of developmental pattern from the three paradigms-much
earlier evidence of domain-level distinctions before basic-level distinctions-is a
very important set of findings in the developmental literature. For the sake of argu-
ment, let us accept these findings as reflecting a distinct course of development of
perceptual and conceptual categorization. I certainly believe Mandler is right on
this point, even though it is difficult to establish this beyond doubt.
4. A process of attentive perceptual analysis yields the first conceptual catego-
ries. Mandler does not think that the concept of animal, uniting birds, fish, toads,
dogs, horses, snakes, and so forth, is innate. She does think that by 7 to 9 months of
age, infants have formed such a category, and this domain-level category has a
conceptual core (self moving agent). The problem, then, is how the infant goes
from the state of not having the domain-level concept of animal to having one. The
answer Mandler gives is that the infant produces conceptual categories from per-
ceptual input through a process of active, attentive, perceptual analysis. Visual in-
formation is redescribed into a simpler and explicitly realized form, most likely in
the format of an image schema, but perhaps in a more abstract format. The content
of these explicit conceptual relations includes paths objects take, plus various rela-
tions among objects such as containment, support, contact, and contingent
relations.
I see several problems with this proposal. Most trivially, Mandler offers no evi-
dence for the "active, attentive" character of the process of perceptual analysis.
From adult studies of Michotte (1963) on contact causality, it is clear that some
conceptual attributions that go beyond the perceptual input are automatically in-
voked. Although there is no evidence on this point, I would think this is also likely
to be true for infants in the case of contact causality, as well as in the case of the at-
tribution of agency and goal directedness to the interactions of moving entities.
Much more important, for reasons that Mandler clearly lays out in her sections
on why there is no known mechanism for deriving conceptual categories from per-
ceptual ones, there is equally no known way that perceptual analysis could do the
trick on its own. Where do the categories represented in the image schematic
meanings themselves come from? If one cannot derive causality from
spatiotemporal descriptions, or agency from spatiotemporal descriptions (even
those that provide the necessary input for attributions of each type of causality),
then the problem of how these concepts arise has not been solved.
In sum, although endorsing the first three parts of Mandler's argument, I do not
believe that perceptual analysis can be the mechanism through which conceptual
representations are formed. I offer a friendly amendment to Mandler's story and
suggest she look to the literature on core knowledge (cf. Carey & Spelke, 1994,
1996) for the origin of the cores of conceptual categories. By hypothesis, core
 ORIGIN OF CONCEPTS 41

knowledge derives from innate learning mechanisms in at least two domains: intu-
itive mechanics, with the concept of an object and contact causality at its core, and
intuitive psychology, with the concept of an agent and intentional causality at its
core. Core knowledge is appropriately abstract, takes perceptual information as its
input, and outputs the event-level descriptions Mandler needs for her very persua-
sive account of the distinct developmental course of perceptual and conceptual
categorization to fly.

REFERENCES

Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: MIT Press.

Carey, S., & Spelke, E. (1994). Domain specific knowledge and conceptual change. In L. Hirschfeld &
S. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture (pp. 169-200).
Cambridge, MA: Cambridge University Press.
Carey, S., & Spelke, E. (1996). Science and core knowledge. Journal ofPhilosophy ofScience, 63, 515-
533.
Gergeley, G., Nadasdy, Z., Csibra, G., & Biro, S. (1995). Taking the intentional stance at 12 months of
age. Cognition, 56, 165-193.
Gopnik, A., & Meltzoff, A. N. (1997). Words, thoughts and theories. Cambridge, MA: MIT Press.
Johnson, S. C., Slaughter, V., & Carey, S. (1998). Whose gaze would infants follow: The elicitation of
gaze following in 12-month-olds. Developmental Science, 1, 233-23 8.
Keil, F. C. (1989). Concepts, kinds, and cognitive development. Cambridge, MA: MIT Press.
Leslie, A. (1988). The necessity of illusion: Perception and thought in infancy. In L. Weiskrantz (Ed.),
Thought without language (pp. 185-2 10). Oxford, England: Clarendon.
Medin, D., & Ortony, A. (1989). Psychological essentialism. In S. Vosniadou & A. Ortony (Eds.), Simi-
larity and analogical reasoning (pp. 179-195). New York: Cambridge University Press.
Michotte, A. (1963). The perception of causality. London: Methuen.
Spelke, E. S., Phillips, A., & Woodward, A. L. (1995). Infants' knowledge of object motion and human
action. In D. Sperber, D. Premack, & A. J. Premack (Eds.), Causal cognition: A multidisciplinary
debate (pp. 44-78). Oxford, England: Clarendon.
Van de Walle, G. (1999, April). Nine-month-olds show manual habituation to shared object color. Pa-
per presented at the Society for Research in Child Development, Albuquerque, NM.

View publication stats