Movements of physical bodies, such as rocks, areexplained by external forces.Such external factors arereferred to as ‘causes’.By contrast, many human move-ments are distinguished from the movements of rocksby having explanations in terms of not just causes but‘reasons’.We describe human movements as ‘actions’directed towards a goal for a purpose,and not just as‘events’that happen through a more or less complexchain of causes. Explaining my typing of these wordsentirely in terms of a pattern of neural activity in myspinal cord, which produces muscle contractions in myfingers so that a keyboard is hit in a particular pattern,seems incomplete.A more satisfying explanation wouldinclude reasons,desires and plans – such as “I want youto know this”.A fundamental goal of cognitive neuroscience is tounderstand how mental entities like ‘reasons’ and‘desires’derive from processes in the brain.But as wecome to understand the internal factors of humanaction in terms of brain function, we must confront thefact that the brain comprises neurons and glia that fun-damentally have no interests. To paraphraseWittgenstein,what,if anything, is left if we subtractbrain processes and associated body movements ‘hap-pening’from the agent ‘acting’? To answer this question,we must understand the basic properties of decisions,choices and actions,and how they arise from brainprocesses.We are beginning to reach such a level ofunderstanding of how the brain makes decisions andgenerates actions.This review will emphasize findingsobtained in experiments in which the activity of indi-vidual neurons was monitored in specific parts of thebrain of highly trained macaque monkeys performingcertain tasks. The tasks typically presented monkeyswith a stimulus or set of stimuli that required someinterpretation to determine which of two or morebehavioural responses to produce to receive a juicereward.Although these are rudimentary tasks com-pared with human decision making under more com-plex conditions of risks, uncertainty and deadlines1,wecan be optimistic that the information gained fromthese initial neurophysiological studies will provide afoundation for future experiments that investigate morecomplex decisions.Choices, decisions and actionsThe recent literature in cognitive neuroscience refers tohuman and nonhuman subjects as ‘deciding’,‘attend-ing’,‘intending’,and so on. When a neurophysiologistuses these words,does he mean the same thing as aphilosopher,a lawyer,or the man on the street? If we areto understand the neural mechanisms of decision,choice and action,then we must use precise and effec-tive definitions that are consistent with both the casualand technical meanings.I will discuss the operationaldefinitions of these terms, which are derived fromphilosophical sources2–4.Choice.A choice is required when an organism is con-fronted with alternatives for which an action is neces-sary to acquire or avoid one or more of the alternativesbecause of a desire,goal or preference.An effective defi-nition is that a choice at the most fundamental level isan overt action performed in the context of alternativesVanderbilt Vision ResearchCentre, Department ofPsychology,VanderbiltUniversity,Nashville,Tennessee 37240, USA.e-mail: [email protected] BASIS OF DECIDING,CHOOSING AND ACTINGJeffrey D.SchallThe ability and opportunity to make decisions and carry out effective actions in pursuit of goalsis central to intelligent life. Recent research has provided significant new insights into how thebrain arrives at decisions, makes choices, and produces and evaluates the consequences ofactions. In fact, by monitoring or manipulating specific neurons, certain choices can now bepredicted or manipulated.REVIEWSNATURE REVIEWS | NEUROSCIENCE VOLUME 2 | JANUARY 2001 | 3334 | JANUARY 2001 | VOLUME 2 www.nature.com/reviews/neuroREVIEWSdecision.Foremost,whereas choice refers to the finalcommitment to one alternative,decision refers to thepreceding deliberation about the alternatives.The polarity of deciding and choosing can be high-lighted by considering a visit to a new restaurant.Everyone must enact a choice, even those who have fre-quented the restaurant.But if the restaurant is new toyou,then before you can choose (produce one of manypossible overt actions with the purpose of communicat-ing which meal you wish to receive), you must learn thealternatives,understand the differences between themand how they relate to your preferences,and deliberateabout which would be most satisfying.A defining fea-ture of decisions that distinguishes them from choices isthat decisions cannot be predicted — even by the agent.If you can say what you will decide, then you will havedecided.A corollary of this is that decisions,like percep-tions,seem just to happen; introspection cannot find thesource of the decision6.Action. The definition of an action is a complex issue7.Often one does one thing (order a meal) by doing some-thing else (point at the menu).But to point at the menu,one must do something else (move a finger).A ‘basicaction’is an action we perform without any preliminar-ies, such as moving a finger.We cannot say how wemove our finger; it just happens when we will it.Defining a body movement as an action depends oncontext.A purposeful action (pointing at the menu) isdistinguished from a mere event (a hand jostled by apassing waiter) by reference to some intelligible plan.Actions are performed to achieve a goal.Neurophysiologists have made considerable progressin characterizing the brain processes that occur whennonhuman primates make decisions and choose amongalternatives by producing specific movements with thepurpose of earning a reward8–11. This review is con-cerned with how neural activity relates to the decisions,choices and actions of monkeys. For an overview of thestructures of the brain that are implicated in choices,decisions and actions,see BOX 1.Neural correlates of choosingThe visual search paradigm has been used extensively toinvestigate visually guided choice behaviour12.In a visualsearch task,a target stimulus must be discriminatedfrom an array of distractor stimuli.Visual search for asingle target among distinct alternatives,known as ‘fea-ture search’, requires a choice that can be based entirelyon sensory processing. Search is more efficient if the tar-get is conspicuously different from distractors — forexample,a different colour or shape.Search