Combining prior beliefs and sensory evidence

JOVexpec_obs

If we want to try to locate a target of interest given a brief glimpse of a visual scene, then we can use at least two sources of information. Firstly, we can use any visual cues which give away the target’s location. However, in many cases the visual cues are insufficient to work out precisely where a target may be: this could be due to the presence of similar looking non-targets (distractors) for example. So in order to make a better inference about a target’s location, then we can use a second source of information: our predictions of where a target will be.

How should we combine information about the location of a target from our uncertain visual cues and our predictions? The answer comes from Baye’s equation, which tells us precisely how to make the best inference. The question is: do people behave as if they use Baye’s equation, do they make the best use of all the information available?

The data reported in this manuscript provided interesting answers. In short, it looks as though people do in fact combine their expectations and observations in a manner consistent with Baye’s equation. BUT, people’s prior predictions about a target’s location was systematically biased.

These results were seen regardless how expectations were provided experimentally to a participant. Using an endogenous cuing method, people were simply told something like “the target will be in the top-left location on 75% of trials.” Exogenous cuing was also investigated by using a peripheral pre-cue. The paper describes some specifics, but the key result was independent of method of cuing. It is interesting to see identical effects elicited by two different means of cuing as people often focus upon the differences between endogenous and exogenous attention. This work suggests that regardless of the underlying neural mechanisms, information about a target’s location was utilised optimally.

Another unexpected results was that the kinds of biases that people exhibited were very similar to the biases observed in high-level decision making tasks that Prospect Theory was designed to account for.

The paper is available (open access) from the Journal of Vision website.

Vincent, B. (2011) Covert visual search: Prior beliefs are optimally combined with sensory evidence, Journal of Vision, 11(13):25, 1-15.