27. M.Roederer,J.M.Brenchley,M.R.Betts,S.C.DeRosa,Clin. Immunol. 110, 199 (200 4).28. A. Perfetto, P. Chattopadhyay, M. Roederer, Nat. Rev.Immunol. 4, 648 (2004).29. The authors thank G. Church, N. Friedman, J. Albeck,P. Jasper, L. Garwin, R. Tibshirani, T. Jaakkola, D. Gifford,and D. Koller for helpful discussions or readings of themanuscript and R. Balderas and BD-PharMingen Bio-sciences for reagents. K.S. and D.A.L. were supportedby the National Intitute of General Medical SciencesCenter of Excellence in Complex Biomedical Systemsat MIT. O.D.P. was supported as a Bristol-Meyer SquibbIrvington Fellow and a Dana Foundation human immu-nology award. D.P. was supported by a PhRMA Centerof Excellence in Integration of Genomics and Infor-matics grant, an NIH Center of Excellence in GenomicStudies grant to G. Church, and an NSF PostdoctoralResearch Fellowship in Biological Informatics. G.P.N.was suppo rted in this work by NIH grants P01-AI39646and AI35304, a grant from the Juvenile Diabetes Foun-dation, and National Heart, Lung, and Blood InstituteProteomics Center contract N01-HV-28183I.Supporting Online Materialwww.sciencemag.org/cgi/content/full/308/5721/523/DCMateria ls an d Met ho dsFigs. S1 to S4Table S1Data Sets S1 to S14References29 September 2004; accepted 19 January 200510.1126/science.1105809Parallel and Serial NeuralMechanisms for Visual Searchin Macaque Area V4Narcisse P. Bichot,1*Andrew F. Rossi,2,3Robert Desimone1,4To find a target object in a crowded scene, a face in a crowd for example, thevisual system might turn the neural representation of each object on and offin a serial fashion, testing each representation against a template of thetarget item. Alternatively, it might allow the processing of all objects in par-allel but bias activity in favor of those neurons that represent critical fea-tures of the target, until t he target emerges from the background. To testthese possibilities, we recorded neurons in area V4 of monkeys freely scanninga complex array to find a target defined by color, shape, or both. Throughoutthe period of searching, neurons gave enhanced responses and synchronizedtheir activity in the gamma range whenever a preferred stimulus in their re-ceptive field matched a feature of the target, as predicted by parallel models.Neurons also gave enhanced responses to candidate targets that were s e-lected for saccades, or foveation, reflecting a serial component of visual search.Thus, serial and parallel mechanisms of response enhancement and neuralsynchrony work together to identify objects in a scene.In a crowded visual scene, we typically focusour attention on behaviorally relevant stimuli.When subjects know the location of a relevantobject, the brain mechanisms that guide theirspatial attention to the object largely overlapwith those for selecting the targets for eyemovements (1). The outcome of this selectionfor attention or eye movements is to enhancethe responses of visual cortex neurons to therelevant object, at t he expen se of distracters(2–6). As a result, object recognition mecha-nisms in the temporal cortex are typically con-fronted with only a single relevant stimulusat a time (7). However, in most common vi-sual scenes, people rarely know the specificlocation of the relevant object in advance—instead, they must search for it, based on itsdistinguishing features, such as color or shape,which is commonly termed visual search. Along-standing issue has been whether objectselection in visual search is also mediated byneural mechanisms for spatial attention, whichscan the objects in the scene sequentially untilthe target is identified (serial search), whetheror not eye movements are made. If so, thenvisual attention could be broadly served by aunitary mechanism, linked to the neural sys-tems that control gaze. Alternatively, searchmay be mediated by nonspatial attentionalmechanisms that are sensitive to featuressuch as color and shape and that bias visualprocessing in favor of neurons that representthe target features throughout the visual field,all at once (parallel search) (7). Search couldalso be mediated by hybrid mechanisms suchas guided search (8).Previous studies of visual search (9–11)and attention to stimuli with particular fea-tures (12–14) in brain area V4 have foundthat neuronal responses to attended targetstimuli were enhanced over time, but the studieswere not designed to test whether the targetswere Bfound[ by serial or parallel neural mech-anisms. In one of these studies, monkeys didnot search for a specific feature but insteadsearched for a singleton (i.e., popout) stimulusin one of two feature dimensions (11). In an-other stu dy that used backgrounds of naturalscenes , the average neural activity throughoutthe trial varied according to the searched-fortarget features, but the authors could not ruleout that these effects were due to differencesin eye scan paths across the scene for differ-ent targets rather than feature-selective effectson neuronal responses (10).We tested for parallel and serial attentionalmechanisms in area V4 in monkeys perform-ing a search task with free gaze. We recordednot only neuronal responses but also the syn-chrony between neur onal responses and thelocal field potential (LFP) (15, 16), because V4neurons synchronize their activity when atten-tion is directed to their receptive fields (RFs)(17), similar to neurons in parietal cortex dur-ing a memory-saccade task (18). Such syn-chrony, especially in the gamma frequencyrange, could potentially amplify their effecton postsynaptic neurons, similar to increasesin firing rate (19). The monkeys freely scannedmultielement arrays composed of coloredshapes to find a target defined by color orshape (20). During color feature search (Fig.1A), the cue was a colored square, and themonkey was rewarded for fixating the stim-ulus in the array that matched the cue color.During shape feature search (fig. S1), the cuewas a gray shape, and the monkeys were re-warded for fixating the stimulus in the arraythat matched the cue shape. When shape wasrelevant, color was irrelevant, and vice versa.We selected two colors and two shapes as cuesfor each recording session, on the basis of ini-tial recordings in which we determined a pre-ferred (strong response) and nonpreferred (weakresponse) color and shape for a given neuron.Overall, monkeys performed similarly dur-ing color and shape search, finding the targeton 86% and 91% of the trials, respectively.Both tasks were