Standford and colleagues, in a recent paper published in Nature Neuroscience, proved that it takes 30ms to make a decision on the basis of a perceptual event. To get to this conclusion, they devised a very smart protocol where the time for sensory processing followed the stage of planning the two possible actions.
Clearly, the time between the onset of the saccadic eye movement and the yellow to red or green target color change was critical. If that time was short, the monkeys chose one target at random because they did not have sufficient time to integrate the color information and they chose the cued target 50% of the trials. If that period was long enough, they had ample time to process the sensory information and chose the cued target on 100% of the trials. In between, the percentage of correct responses grows linearly, hitting 75% of correct responses at 130ms, i.e. if the targets turn red and green 130ms before the onset of the eye movement, the monkey will make the correct choice in 75% of the trials. The authors estimated that this period of 130ms consisted of 100ms of afferent and efferent delays (the non-decision time) and 30ms of decision time. Afferent delays correspond to whatever occurs before the integration of the information starts (around 60ms) and the efferent delays to whatever occurs between the decision time and the execution of the saccade (around 30ms).
To interpret their data, the authors suggested a race model. Typically, a race model consists of signaIs evolving towards a decision threshold. Each signal is associated with a specific action (e.g. one for choosing the left target and one for choosing the right target). As soon as one signal reaches the threshold, the associated action is elicited.
In the authors' model, the presentation of the two yellow points triggers the onset of the two different decision signals, one for each of the targets. The decision to make an eye movement to one of the two targets is taken as soon as the corresponding decision signal reaches a given decision threshold. As long as the two targets are yellow and no information about the actual target is provided, both decision signals evolves randomly towards the threshold. How fast they go towards the threshold is determined by picking a velocity for both signals randomly. Therefore, one will always move faster than the other. The faster decision signal will correspond to the default choice if the information about the actual target is not provided early enough.
As soon as information about the actual target is available, the average velocity of the decision signal associated with the cued target is increased whereas the decision signal of the non-cued target is decreased. This change in velocity is not instantaneous but occurs gradually. Therefore, if the decision signal associated with the non-cued target is close enough to the threshold, it might still reach it and the saccade will be directed to the non-cued target. If it is far enough from the threshold, the decision signal associated with the cued target will eventually evolve faster towards the threshold and it will win the race. In short, when the information is available early, it will have time to influence the race and the decision signal associated with the cued target will eventually win. In contrast, when this information is only provided after a long delay, the originally faster signal, which is determined randomly, will win the race and performance will be close to 50%.
The originality of their task is reinforced by neural recordings of neurons in the frontal eye field, an area that is critical in decision-making (Shadlen and Gold 2007). Signals recorded in that area of the brain closely resemble the decision signal of the model.
In sum, this paper shows that the brain is fairly efficient in processing sensory stimuli in order to direct our actions, however, it is fairly inefficient in conveying information. Indeed, most of the time between the cue appearance and the execution of the saccade is necessary for afferent and efferent delays.
Stanford, T. R., Shankar, S., Massoglia, D. P., Costello, M. G., & Salinas, E. (2010). Perceptual decision making in less than 30 milliseconds. Nature neuroscience, 13(3), 379-85. doi: 10.1038/nn.2485.
Shadlen MN, Gold JI (2007) The neural basis of decision making. Annual review of neuroscience 30:535-74 doi: 10.1146/annurev.neuro.29.051605.113038