Developmental Psychology 1997, Vol. 33, No. 3, 423-428 Copyright 1997 by the American Psychological Association, Inc. O012-l649/97/$3.00 Numerical Abstraction in Infants: Another Look Kelly S. Mix, Susan Cohen Levine, and Janellen Huttenlocher University of Chicago This article examines an important finding from the literature on infant numerical competence. The finding, reported by P. Starkey, E. S. Spelke, and R. Gelman (1990), was that infants looked longer toward a visual display that was equal in number to an auditory set. In Experiment 1, when the procedures described by P. Starkey et al. were followed and duration was held constant across auditory sequences that varied in number, infants looked longer toward the display that was not numerically equivalent to the auditory set. In Experiment 2, when the rate and duration of the auditory sequences were varied randomly within infants, no significant preference for either the equivalent or nonequivalent visual display was shown. These results raise questions about P. Starkey et al.'s claims that infants can represent the numerosity of sets in different modalities and then perform one-one correspondence computations over them. Interest in the origins of number concepts has led to several studies focusing on numerical competence in infants. Most of these have evaluated the ability of infants to discriminate be- tween small set sizes by using the habituation procedure. In these studies, it was found that infants* looking time decreased after being shown several arrays with the same small number of items (usually two or three) but that looking time recovered when a novel number of items was shown (Antell & Keating, 1983; Starkey & Cooper, 1980; Strauss & Curtis, 1981). The same result was found when experimenters in these studies var- ied the displays to control for differences in brightness, density, line length, contour, area, and homogeneity of set items, indicat- ing that the infants used numerical information, rather than these other factors, as the basis for making discriminations. Although the habituation studies provided evidence of a sensi- tivity to numerical information during infancy, they gave little indication of what particular process might underlie the infants' performance. Starkey, Spelke, and Gelman (1990) argued that although infants might possess central numerical processes, such as one-one correspondence, this had not been clearly demon- strated because the use of other processes (e.g., use of configural information or visual subitizing) could explain infants' looking preferences in the visual habituation experiments. They pro- posed that if infants could recognize numerical equivalence be- tween visual displays and sounds, this would demonstrate an Kelly S. Mix, Department of Psychology, University of Chicago; Susan Cohen Levine and Janellen Huttenlocher, Departments of Psychology and Education, University of Chicago. The study presented in this article was supported by a grant from the University of Chicago School Mathematics Project. Portions of this research were presented at the annual meeting of the Midwestern Psycho- logical Association, Chicago, Illinois, May 1994. We gratefully acknowl- edge Rebecca Compton, Katchen Locke, and Kirste Moline for their assistance with data collection and data entry. Correspondence concerning this article should be addressed to Kelly S. Mix, who is now at the Department of Psychology, Indiana University, Bloomington, Indiana 47405. Electronic mail may be sent via Internet to kmix@indiana.edu. ability to establish one-one correspondence because these other processes could not be applied to temporally distributed sets. To test whether infants could recognize auditory-visual nu- merical correspondences, Starkey et al. (1990) designed a series of experiments in which 6- to 8-month-olds were shown pairs of visual displays that included one display of two objects and one display of three objects. While the displays were still visible, infants heard either two or three drumbeats. Measurement of looking time revealed that the infants looked significantly longer toward the display that matched the number of sounds. Starkey et al. interpreted these results as evidence that infants can per- ceive the number of distinct entities both in a sequence of sounds and a visual display and can relate these sets to one another in terms of numerical equivalence. They contended that to detect such relations, infants must make use of a process involving both one-one correspondence and the abstraction principle (i.e., knowledge that any discrete element, including sounds, can be enumerated; Gelman & Gallistel, 1978). They further proposed that the emergence of these abilities is dependent on neither the acquisition of language nor a culture-specific count- ing system. This interpretation attributes to infants a competence far be- yond that shown in previous studies of infant number concepts. As such, it has had a strong influence on theories of numerical development (e.g., Gelman & Brenneman, 1994; Starkey, 1992; Wynn, 1992). However, there are several reasons that the claims advanced by Starkey et al. (1990) warrant further investigation. First, the results they reported for infants have not been repli- cated. In the only replication attempt to date (Moore, Benenson, Reznick, Peterson, & Kagan, 1987),' infants showed a signifi- cant preference for one of the visual displays but in the opposite direction of that reported by Starkey et al. That is, infants looked longer at the display that was not equivalent to the number of sounds, Starkey et al. attributed the discrepancy between their findings and those of Moore et al. to differences in statistical 1 This replication attempt was conducted in response to an earlier publication that reported a portion of the studies described in Starkey et al. (1990; see also Starkey, Spelke, & Gelman, 1983). 423