Interhemispheric integration of visual concepts in infancy

Interhemispheric integration of visual concepts in infancy Kimberly M. Scott (kimscott@mit.edu) Department of Brain and Cognitive Sciences, MIT Cambridge, MA 02139 USA Laura E. Schulz (lschulz@mit.edu) Department of Brain and Cognitive Sciences, MIT Cambridge, MA 02139 USA Abstract Abstraction often requires appropriate integration of more con- crete representations. During development, the more specific or localized representations may arise first. Here we study the special case of integration of visual representations from the left and right hemispheres during infancy. We present failures of interhemispheric integration in two domains, form percep- tion and approximate number, in infants ranging from 8 to 18 months of age. In Experiment 1, infants succeeded in repre- senting equality of two shapes only when both shapes were presented in the same visual hemifield. In Experiment 2, in- fants represented 16 when shown 16 dots in one hemifield but not when shown 8 dots in each hemifield. We argue that inter- hemispheric integration poses a particular and unusually late- resolved challenge in infant vision. Keywords: visual invariance; interhemispheric integration; corpus callosum; split-brain; approximate number system Introduction Any latent cause in the world can lead to myriad patterns of observations. Recognizing an object in the visual world requires achieving invariance over transformations–including translations, rotations, scaling, and lighting changes–that can drastically affect its retinal projection. Similarly, understand- ing more abstract concepts often requires recognizing the identity or shared causes of superficially disparate manifes- tations: three voices, three bears, or three hugs; perceptual causality or statistical covariance; a frown and a raised voice. Assuming that in some cases these specific “views” may be represented before the more general concepts, how do infants integrate these distinct representations? Newborns have already been demonstrated to have sev- eral forms of visual invariance such as size constancy (Slater, Mattock, & Brown, 1990) and recognition of invariant posi- tional relations (Antell & Caron, 1985; Milewski, 1979), al- though these progress with visual experience during the first year (McKenzie, Tootell, & Day, 1980; Granrud, 2006; Gliga & Dehaene-Lambertz, 2007). Visual adaptation and after- effects suggest that high-level visual concepts including ap- proximate number (Ross & Burr, 2010) and facial identity (Webster & Maclin, 1999) may pose similar learning chal- lenges, as they are first represented by spatially localized de- tectors. Split-brain patients likewise provide dramatic evi- dence for distinct spatially localized representations of high- level visual concepts (for review see Gazzaniga, 2005). Regardless of the exact computations involved, the rela- tionships among localized detectors need to be learned or refined to produce integrated representations. We propose to study the special case of integration of visual concepts from the left and right hemispheres. Late myelination of the corpus callosum, which connects the two cereberal hemi- spheres (Yakovlev & Lecours, 1967), its continued develop- ment through adolescence (Giedd et al., 1999; Salamy, 1978), and the separate critical period for the corpus callosum to af- fect visual development (Elberger, 1984) make this plausi- bly challenging in early childhood. Indeed, interhemispheric transfer of a visual rule learned in a single hemisphere does not occur spontaneously before 4 months of age (de Scho- nen & Bry, 1987), and children under 24 months have diffi- culty integrating information about shape across hemispheres (Liegeois, Bentejac, & de Schonen, 2000). However, this ef- fect is specific to face stimuli; the operant conditioning task introduced additional demands; and the bilateral and unilat- eral presentations differed in visual angle from the fovea, al- lowing several alternative explanations of the apparent fail- ure. We first sought to confirm infants’ difficulty comparing shapes from opposite visual hemifields. In Experiment 1, we attempted to familiarize infants with matching shapes either unilaterally or bilaterally by briefly presenting matching pairs of shapes while infants fixated on a small video. This famil- iarization period was designed to affect their preference for looking at matching shapes, which we measured before and after familiarization. Experiment 1: Is a square on the left the same shape as a square on the right? Methods Participants Infant subjects were recruited at the Boston Children’s Museum and parents provided informed consent to participate. 48 infants between 8 and 14 months of age (mean age 11 months 2 days) participated in this study. An additional 33 infants were excluded due to fussiness, inatten- tion, or experimenter error. Procedure Each infant sat on a parent’s lap for the dura- tion of the study, 1.5 m from a large monitor used to dis- play all stimuli. Subjects were videorecorded using a cam- era positioned directly above the monitor. The experimenters were positioned behind the monitor, hidden from the view of the infant, and monitored the infant using a webcam posi- tioned above the monitor while controlling the progression of the study using Psychtoolbox extensions (Brainard, 1997) in