Inside and Outside the Focus of Attention

rons and a corresponding convergence of information A wide expanse of the visual environment, typically clutas an increasing number of objects can fit within each tered with visual information, is encoded by the retinae receptive field. This confluence would seem to lead inand passed to the central visual system. How do we variably to an inseparable mixture of information from integrate and coordinate the different pieces of informadifferent objects. Does this mixture represent the nertion provided by this onslaught of information to achieve vous system’s composition of the visual scene, or does our coherent visual experience? The key to this problem the information about individual objects somehow reseems to reside in two factors, attentive selection and main separate despite the convergence? Apparently, capacity limitation. Attentive selection refers to our abilthey remain separate. Directing attention to a particular ity to concentrate our perceptual effort on a selected object alters the convergent balance in favor of the atportion of the available sensory information so as to tended object and suppresses the neural response to achieve a clear and vivid impression of the environment. other objects within the neuron’s receptive field (Moran It has long been recognized that despite our experience and Desimone, 1985; Treue and Maunsell, 1996). of a wide visual field, actual object analysis is not uniform The important questions posed by Kastner and colacross the visual field but occurs within the relatively leagues (1998) were whether the suppression observed restricted region in which we attend. The capacity limits at the single neuron level would actually be preponderof this restricted region usually have little to do with ate enough to be reflected in the fMRI image of the acuity limits set by the retinal receptor density gradients activity of large cortical areas. As such, these studies but instead are governed by the competitive interference mark a significant departure for fMRI studies of visual between objects that arises in relation to the density attention from a strict localization question to one of and similarity of the objects. functional differentiation. Significant progress has been made in the last few The first study asked whether fMRI could detect the decades in our understanding of the behavioral concompetitive interference between visual objects prestraints of attentive selection, and a beginning has been sented outside the region of focal attention. The submade in uncovering the neural correlates and mechajects’ attention was fully engaged at the point of fixation nisms of attention. Two recent reports highlight several by using a task of counting Ts or Ls presented in a 4 important pieces of the attention puzzle regarding the Hz serial stream. While attention was thus engaged, neural processing of visual information within the region a set of four complex objects were presented in the of focal attention (Kastner et al., 1998) and our ability to quadrants of a square cluster centered at an eccentricity process information outside the region of focal attention of about 88 in the upper right visual field. The objects (Braun and Julesz, 1998). were presented either simultaneously (SIM) for 250 ms Kastner and colleagues (1998) report on two funcrepeated every second for 18 s, or each object was tional magnetic resonance imaging (fMRI) studies of the presented sequentially (SEQ) in one of the four quadhuman visual system. The first investigates whether rants within the cluster on each 250 ms interval for 18 the overall activation of visual cortical areas reflects the s. The 18 s blocks of SIM or SEQ conditions were sepacompetitive interference between visual objects that has rated by 18 s blank periods and were presented in an been observed to occur within the receptive fields of overall block sequence of SEQ–SIM–SIM–SEQ. Activity single neurons recorded in extrastriate cortical areas. differences between stimulus and blank periods and The second study investigates whether directed attenbetween SIM and SEQ presentations were averaged for tion can isolate the attended object from the competitive each cortical area and condition. interference of other objects. For each area, the SEQ condition produced a larger As background for this study, let us recognize that response than the SIM condition, even though the total visual information is dispersed from primary visual corvisual stimulus integrated over time was the same. This tex through extrastriate cortex along two main routes. result supported the hypothesis that competitive interOne leads ventrally toward anterior temporal cortex, the ference suppressed activity producing an overall weaker other dorsally into parietal association cortex. The venresponse: the whole was less than the sum of the parts. tral stream progression (V1–V2–V4–TEO) portrays a sysKastner and colleagues noted that the magnitude of tem devoted to object analysis and, in anterior temporal the effect appeared to increase from V1 to TEO and areas, clearly represents a stage where sensory prosuggested that this was related to the increasing size cesses have merged with systems associated with obof receptive fields from V1 to TEO. As receptive fields ject recognition and memory. As visual processing proincrease, more objects fit within them and the degree gresses across these cortical areas, each level appears of competitive suppression increases. A control experito add a higher level of analytic abstraction, such that ment in which the quadrants of the stimulus cluster were