Top-down predictions in the cognitive brainIntroductionA primary function of the brain is to predict its environmentThe efficiency of visionThe visual pathwaysFeedforward and feedback views of visual recognitionAnatomical and functional differences between top-down and bottom-up connectionsTheoretical and computational modelsThe role of neural synchrony in top-down/bottom-up interactionsEvidence for top-down facilitation in visual recognitionA model for the triggering of top-down facilitation in object recognitionTesting the top-down facilitation modelActivity in OFC is initiated earlyOFC responds preferentially to LSFsMagnocellular projections trigger top-down facilitationOrbitofrontal activity correlates with predictive power of LSF stimuliSummary of supporting studies for the top-down facilitation modelLinks to priming, synchrony, and contextual facilitationTop-down facilitation of object recognition is modulated by experiencePriming enhances cortical communicationTop-down facilitation by contextual informationLinks to emotion and action preparationCooperation of OFC and the amygdala in processing of emotional stimuliPotential role of OFC and the amygdala in action preparationLinks to schizophrenia and dyslexiaThe pathways involved in top-down facilitation are affected in schizophreniaThe M pathway abnormalities in dyslexiaDiscussionAcknowledgmentsReferencesTop-down predictions in the cognitive brainKestutis Kveraga, Avniel S. Ghuman, Moshe Bar*Martinos Center for Biomedical Imaging at the Massachusetts General Hospital, Harvard Medical School,149 Thirteen Street, Charlestown, MA 02129, USAAccepted 7 June 2007AbstractThe human brain is not a passive organ simply waiting to be activated by external stimuli. Instead, we propose that the brain con-tinuously employs memory of past experiences to interpret sensory information and predict the immediately relevant future. The basicelements of this proposal include analogical mapping, associative representations and the generation of predictions. This review concen-trates on visual recognition as the model system for developing and testing ideas about the role and mechanisms of top-down predictionsin the brain. We cover relevant behavioral, computational and neural aspects, explore links to emotion and action preparation, and con-sider clinical implications for schizophrenia and dyslexia. We then discuss the extension of the general principles of this proposal to othercognitive domains.Ó 2007 Elsevier Inc. All rights reserved.Keywords: Vision; Visual pathways; Magnocellular; Parvocellular; Object recognition; Computational methods; Priming; Synchrony; Contextual facili-tation; Orbitofrontal cortex; Neuroimaging; Magnetoencephalography1. Introduction1.1. A primary function of the brain is to predict itsenvironmentAs military tacticians have known for millennia, surpriseand uncertainty are costly in term s of time and energyexpenditures (Tzu, 2006). Sharing a view expressed in thepast, we propose that a fundamental function of the brainis to predict proximate events, which facilitates interactionswith external stimuli, conserves effort, and ultimatelyincreases the chances of survival. Useful predictions typi-cally do not arise de novo. Even innate abilities of thehuman brain, such as vision or language, usually requiredevelopment, during which experience-bas ed mapping ofsensory inputs to their identities, and to the appropriateresponses, takes place. For example, it is only after yearsof training that a baseball player is able to effectively antic-ipate a particular type of pitch, predict the trajectory of theball by combining top-down information about a particu-lar pitch with bottom-up perception of the rotation andspeed of the ball, and correctly decide whether and howto swing at it—all in about half of a second. This abilityto rely on stored knowledge and learned modes of behaviorreduces the need to consider a large number of potentialcauses or courses of action, which enables quicker interpre-tation of endogenous and exogenous events, and faster,more precise, and less effortful responses.Recent computational and theoretical work demonstrateshow predictions can be integrated with sensory input toreduce the computational demands in perception (Bar,2007; Engel, Fries, & Singer, 2001; Grossberg, 1980; Ullman,1995). However, the neuroanatomical mechanism underly-ing the generation and efficient representation of pr edictionsin the brain is not completely understood. In this review, wewill focus on how the brain generates predictions about thevisual world and uses these predictions to allow an efficientand accurate interpretation of the environment. Thoughwe largely concentrate on vision, the same general principlescan be applied to other sensory modalities, as well as to morecomplex cognitive domains.0278-2626/$ - see front matter Ó 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.bandc.2007.06.007*Corresponding author.E-mail address: [email protected] (M. Bar).www.elsevier.com/locate/b&cAvailable online at www.sciencedirect.comBrain and Cognition 65 (2007) 145–1681.2. The efficiency of visionPicture an impatient driver attempting to overtake aslow-moving vehicle on a busy two-lane road. Much infor-mation has to be acquired and computed very quickly bythe driver’s brain to avoid a collision. The visual systemmust parse the constantly changing two-dimensional imageat the retinae into a coherent three-dimensional representa-tion of the scene, in which the road, the objects on and sur-rounding it, and the position of the driver’s car areaccurately identified. The velocities and trajectories of eachvehicle in the vici nity must be estimated and changes inthese parameters anticipated. Anticipating the movementof other vehicles on the ro ad also involves judging the roadconditions, the state of the traffic signals ahead, as well asthe other drivers’ intentions based on their interpretationsof these factors and their internal states. Many of thesejudgments are based not only on visual, but also on audi-tory, proprioceptive, and even olfactory and social cues(e.g., ‘‘will the driver of that junker in front of me emanat-ing noxious smoke and loud music allow me to pass, or tryto keep me in the opposite lane?’’). Lastly, the forcesapplied to the car’s controls must be computed correctlyand adjusted instantly based on both internal (propriocep-tive) and external (the vehicle’s characteristics and