Dimensions of cognition in infancy

Abstract

The present study explored the dimensionality of infant cognition by factor analyzing measures from a battery of tasks administered to a large cohort of 7-month-old preterm and full-term infants (N=203). A principal axis factor analysis yielded three factors accounting for 37% of the variance. There was one attention factor (look duration and shift rate), one speed factor (reaction time), and one memory factor (immediate recognition). The results indicate that infant cognition is better characterized as multidimensional than as unitary, and support the idea that many of the tasks used in infancy tap discrete elementary abilities.

Introduction

Information has been accruing about the existence and persistence of individual differences in infant cognition, especially for measures of information processing. Significant reliability and/or stability has been reported for a number of infant measures including attention, processing speed, memory, and those capturing representational competence (such as object permanence, symbolic play, cross-modal transfer, categorization—for an overview, see Rose & Tamis-Lemonda, 1999). These information processing measures all show strong age-related change, differentiate risk groups from their age-matched nonrisk peers, and predict later cognition. Literature reviews indicate that habituation and visual recognition memory, two of the most widely assessed infant abilities, predict later IQ with a median correlation of r=.45 (McCall & Carriger, 1993). This level of prediction, although modest, is better for information processing measures than for traditional psychometric infant tests, which generally have little predictive validity before 2 years for either normal or risk groups Kopp & McCall, 1982, McCall et al., 1972. A critical question arises as to whether these newer measures are predictive because they tap a single underlying ability or, rather, as is more likely, they tap discrete abilities that are among the building blocks of later cognition (Rose & Tamis-Lemonda, 1999).

In trying to understand the basis for these predictive relations, questions about the dimensionality (i.e., factor structure) of infant cognitive abilities arise. While questions about dimensional structure play a large role in studies of adult cognition Cattell, 1971, Guilford, 1967, Horn, 1968, Spearman, 1923, Sternberg, 1985, this issue has been explored very little in infancy. In fact, only three studies have even touched on the problem. In one of these studies (Rose & Feldman, 1995), where a canonical correlation was used to assess the relation between several different measures of information processing in infancy and at 11 years, a single canonical variate emerged. Given the pattern of loadings at both ages, and the fact that many of the 11-year measures were timed tests, the variate was interpreted as reflecting processing speed. In the second study (Rose & Feldman, 1997), where path analysis was used to relate infant visual recognition memory to later IQ, memory, as well as speed, played a role, suggesting that there is more than a single dimension to infant cognition. In the third study (Jacobson et al., 1992), the single-factor analytic study of infant cognition, four factors explained 60.6% of the variance among 13 measures obtained between 6.5 and 13 months. Speed and memory emerged as two of the factors, along with developmental level and one uninterpretable factor.

The present study is a further exploration of the dimensionality of infant cognition. Here, the measures were drawn from four domains: memory, processing speed, attention, and representational competence. These domains not only figure prominently in models of adult human cognition, but in infant cognition as well. Indeed, it has been suggested that individual differences in these domains underlie the predictive relations between infant cognition and later language and IQ Colombo, 1993, Rose & Tamis-Lemonda, 1999. In addition, in view of the dynamic nature of early development, an effort was made to restrict the assessments of cognitive performance to a single age (7 months). In this way, we could avoid complicating any conclusions about the dimensionality of infant cognition with changes in cognitive structure that might occur over the first year of life. Multiple measures of each domain were obtained from a battery of tasks administered over two sessions.

The first domain, memory, is recognized as fundamental to learning and development. One type of memory, visual recognition, was one of the first aspects of infant cognition to show predictive validity for later cognitive ability Bornstein & Sigman, 1986, Fagan & Detterman, 1992, McCall & Carriger, 1993. We have included measures capturing additional aspects of memory in infancy, some of which were developed in the course of this project. One, a span task, assesses short-term memory capacity, an important component of working memory. Another, a task examining retention over successively longer delays, assesses more long-term memory.

Speed of processing is another aspect of infant cognition considered to have important implications for later intellectual functioning. The concept of processing speed has served as the cornerstone in many theoretical formulations about the nature of cognition Detterman, 1987, Jensen, 1992, Vernon, 1987 and has been found to be the central limiting factor in a wide range of diverse cognitive achievements in childhood and adolescence Hale, 1990, Kail, 1991. In the present study, reaction time (RT), one aspect of processing speed, was assessed in infancy with the Visual Expectation Task-VExP (Haith, Hazan, & Goodman, 1988). This task gauges ocular RT to a fast-paced sequence of computer-generated pictures that appear in a predictable (or unpredictable) spatiotemporal order. In addition to the VExP task, we included a continuous familiarization task that assesses a second aspect of processing speed—encoding time. In this task, infants are presented with a series of paired stimuli (in this case, pictures of faces), one of which remains the same from trial-to-trial, while the other changes Rose et al., 2002, Rose et al., 1999, Rose et al., 2002. Trials continue until the infant recognizes the unchanging picture, as gauged by the emergence of a consistent preference for the novel target.

Attention, a precondition for any cognitive activity, is frequently studied in infancy by examining look duration and shift rate. Growing evidence indicates that shorter looks and higher shift rates are indicative of superior attention. Look durations decrease and shift rates increase dramatically with age Colombo et al., 1987, Frick et al., 1999, Rose et al., 2001a, while, within an age, more mature patterns of attention (short looks and more shifts) are associated with better information processing and visual recognition memory Colombo, 1993, Colombo et al., 1991, Freeseman et al., 1993, Jankowski & Rose, 1997, Jankowski et al., 2001, Rose et al., 1999. Longer looks are characteristic of risk infants Carlson & Werkman, 1996, Jacobson et al., 1993, Reisbick et al., 1997, Rose et al., 2001a, Rose et al., 1988 and are associated with lower IQ Colombo et al., 2001, Rose et al., 1986, Sigman et al., 1991, Sigman et al., 1997.

Representational competence refers to the ability to extract, from concrete individual instances, information that is common to them and to represent this commonality in some abstract or symbolic form. The possibility that this type of mental representation could support some cognitive continuity was suggested by a study relating habituation in infancy to a latent dimension of representational competence in later childhood, comprised of language comprehension and play sophistication (Tamis-Lemonda & Bornstein, 1990). In the present study, tactual-visual cross-modal transfer and anticipations represented this domain. Cross-modal transfer involves visual recognition of a shape based on a representation formed from tactual experience. This infant ability has been found to relate to language and IQ in later childhood Rose & Feldman, 1995, Rose et al., 1988, to show continuity from infancy to 11 years (Rose, Feldman, Futterweit, & Jankowski, 1997) and to relate to later visuospatial skills (Rose & Feldman, 1995). Anticipations, or expectations about future events, require that the infant detect the underlying regularity in a sequence of events and extrapolate from past experience to the future. The measure of anticipations used here was taken from the VExP task. This measure also shows considerable stability during the latter half of the first year Canfield et al., 1997, Rose et al., 2002.