UC Berkeley

Our perception is of a stable visual world, yet we make several saccadic eye movements per second – to say nothing of the motion of objects in the world or our own motion through the physical world. Even standing on a stage, looking out at an audience, we have the perception of the crowd in detail, in spite of our less than accurate peripheral input. Standing on this stage, it is a trivial matter to make a series of saccades to the faces of various audience members to find one particular person – how can we do this, since visual crowding makes these faces unidentifiable? How much information is available from peripheral vision prior to a saccade? While making saccades to individual members of the crowd is useful for identifying them, what about gaining a sense of the audience as a whole? We can perceive the ensemble emotion of the audience without examining each and every member, but is this solely possible with peripheral information, the details of which we cannot otherwise access, or does it require foveal information?

While a great deal of information is available from peripheral vision, identifying individual objects is difficult due to visual crowding; we solve this problem by simply making a saccade to the object we wish to identify. The ability to saccade to a crowded object implied that a saccade to a crowded object might reduce crowding before the object was foveated, essentially unlocking peripheral information that is otherwise inaccessible. We performed a series of experiments where subjects made saccades to crowded faces and were not permitted to foveate them; we found that crowding was diminished presaccadically and that this was more effective than covert attention alone. Our crowding result shows that saccade planning can diminish crowding, but what information becomes available to conscious vision? To determine this, we performed a series of experiments to determine whether saccadic remapping acquired object representations or merely constituent features from saccade targets. We found that saccadic remapping was object-selective and tightly tuned to the target of the saccade, suggesting that, prior to a saccade, detailed information is made available to conscious vision. However, there simply is not time to take advantage of this unlocking for each face in a crowd – can the visual system use this inaccessible information to generate a percept of the group as a whole – an ensemble code – and is foveation of individuals required? We occluded foveal input and determined that it was unnecessary for ensemble perception of emotion. In all, these three studies demonstrate that peripheral input is remarkably detailed and that the visual system can use this information to both facilitate identification of individual group members as part of saccade planning, and to provide a useful assessment of the crowd as a whole