2011/att11/EyeTrackingAttention

Attention during an Eye-Tracking Task

Participants

  1. Francisco Barranco (Granada) - Lead
  2. Julio Martinez-Trujillo (McGill University) - Mentor
  3. Nuno Vasconcelos (UCSD) - Mentor

Purpose

Eye tracking experiments on attention and reference frames

The idea of this experiment is to find out whether attention works preferentially in a retina centered or in a space centered frame.

Julio's Proposal

Method:

The subject is instructed to pursue a dot that moved up and down on the screen with a speed of 3 degrees/sec (speed could change to adjust task difficulty). We could them position targets on both sides of the screen, at about 10 degrees from the central moving dot. The targets would preferentially be vertically oriented stimuli and they need to be two symmetric rows at both sides (e.g. 10 vertical gratings on each side). The task for the subjects is to make a saccade toward the target that changes orientation (from vertical to horizontal). Observe that changes are orthogonal to the movement of the eyes to minimize the potential confounding effect of velocity vectors on the retinal.

The following links list the videos for each one of the stimuli:

*  Stimuli: Retina-centered slow *  Stimuli: Space-centered slow

*  Stimuli: Retina-centered fast *  Stimuli: Space-centered fast

*  Stimuli: Space-centered fast, audio reinforcement cues *  Stimuli: Space-centered fast, audio distraction

*  Stimuli: Retina-centered fast, audio reinforcement cues *  Stimuli: Retina-centered fast, audio distraction

Examples of the stimuli

Example of the stimuli

Example of the stimuli with the target changing orientation

Experiment 1

Targets move on the screen at the same speed as the central dot. Here these targets are retina-centered and change position in space. We will measure the speed and accuracy of saccades toward the target.

Targets remain stationary. Here the targets remain space-fixed. We will also measure the speed and accuracy of saccades toward the target. Note that because the change could randomly occur in any target, the subject cannot ‘deviate’ the eye position in advance toward any side. Also because the distances from the moving dot to the targets is the same in both experiments we will be able to compare saccades that have the same metric.

Hypothesis:

If attention works preferentially in a retina-centered frame, we should obtain shorter saccade latencies in the retina-centered condition relative to the space-centered condition.

Results: The following images show the evolution of our gaze following the dot which goes down, up, down, up (from the top of the screen to the bottom). The fixation in the space-centered seems to be faster than in the retina-centered frame. The blue arrows indicate the position of the saccades and the green ones indicate blinks.

Retina-centered slow

Space-centered slow

In the following graphic we show the temporal analysis of the saccades occurrence. As we can see, the saccades in the space-centered frame happen some time before than in the retina-centered frame.

Attention reference frame

Some parameters influence the results: Y-coordinate of the central dot, speed, left/right change of the orientation. In the following images, the subject performed the same experiment but, in this case, the speed is x3.

Retina- and space-centered speed x3

As we can see, now the fixation is harder and actually, the subject missed some of the changes in the orientation of the targets.

Experiment 2

Here the goal is to find out whether when conflicting cues are given to the subjects (auditory vs. visual), it will interfere with the latency and accuracy of saccades. For example, we could position speakers on the left and right hand sides and repeat the experiments using combinations of left, right auditory cues given at different times from target change onset and left and right targets. We could also position speakers up and down and examine whether the latency changes for targets changes up and down with the same hemifield. Prediction: non-conflicting cues should produce shorter latencies than conflicting ones. This will argue in favor of a combination of multimodal cues for the allocation of attention. On the other hand, if visual overrides auditory we should obtain no differences between conflicting and non-conflicting cues.

In this case, we have included a beep sound in the previous videos (see attached videos in first section). We included a beep in the correspondent speaker (left or right) depending on the side in which the targets changed their orientation. In this way we made the videos with the reinforcement auditory cues. We also made the videos with auditory cues as distractions. The results are shown in the following graphics (only for the space-centered frame).

Space-centered audio

As seen, the distraction cues influences the fixation tasks (is the worst case in all the experiments). On the other hand, we cannot conclude that the auditory cues reinforce the fixation cues since the results in this last two graphics are not categorical.

Conclusions

As is seen in the paper referenced, it can be intuitive allocate the attention in the space-centered reference frame. In tasks as ours, following to a stationary object in the periphery during the subject is moving requires attention to change the position dynamically in a retinotopic map following the retinal displacement of the target representation. In this case, it would be much more efficient a space-centered frame for the target object, avoiding in this way the repetitive shifts across the retinotopic map.

On the other hand, the same authors also propose that humans can divide proportionally the attention between targets in different frames of reference or different retinal velocities, during smooth pursuit and fixation.

References for programming the stimuli and task Niebergall et al. Journal of Vision, 2010. It also contains a comprehensive review of the literature. These experiments were only visual.

Eye Tracking Software

Attachments