From cs240@cs.rochester.edu Tue Nov 4 18:00:31 2003 Date: Tue, 4 Nov 2003 17:59:05 -0500 From: cs240 To: tetreaul@vienna.cs.rochester.edu Name: - Sachin Degwekar
email: - degwekar@cs.rochester.edu
CS 240W Assignment number: - 6
Date: - 24th October, 2003
News and Views on "Subtask sequencing in the primary visual cortex" by Roelfsema et al

News and Views on "Subtask sequencing in the primary visual cortex" by Roelfsema et al


This paper talks about the article "Subtask sequencing in the primary visual cortex" by Pieter R. Roelfsema, Paul S. Khayat, and Henk Spekreijse and presents the author's views on the article. The article describes an experiment done by the scientists to decide whether subtasks of complex visual tasks are processed in serial or in parallel, in the primary visual cortex.

The experiment

Before the experiment was conducted, the scientists trained two macaque monkeys (called B and G) for it by teaching them to carry out some tasks. The experiment itself consisted of making the monkeys perform two tasks. In either task, a trial was started as soon as the monkey's eye position was within a 1 degree X 1 degree square window centered on a 0.3 degree fixation point. The stimulus appeared after an interval of 300 ms. In the first (simple) task, the monkey had to trace a target curve connected to the fixation point to locate a circle at its other end and to make a single eye movement to this circle. There was also a distractor curve unconnected to the fixation point and it could be ignored. The color of the fixation point was blue on half of the trials and yellow on the other half, so as to discount for the exogenous and undesirable effect of color on reaction times. It was found that the color did not matter.

The second task was more complex, involving the first as a subtask. It required the monkey to fixate on a gray fixation point. After a delay of 300 ms, the color of the fixation point changed to blue or yellow, and at the same time two curves and two colored markers appeared on the screen. The monkey had to search for a marker with the same color as the fixation point and then trace out the target curve associated with the marker. Thus, this task involved both visual search and curve tracing. As in the first experiment, the monkey had to trace the curve to its other end to locate a larger circle that was the target for an eye movement. Also, each monkey was allowed to make only a single eye movement from the fixation point to the circle. The left marker was yellow on half of the trials and blue on the other half, with the right marker having the other color, to discount for the color effect as in the first task.

The simple and composite tasks were interleaved in blocks of 60-72 trials, and each session was made of three to nine blocks per task. In each trial, the reaction times of the two monkeys were noted and the neuronal activity in the V1 area of their brains was studied.

Finally, the trials were repeated with an extra stimulus. An additional contour segment was placed in the receptive field of a group of V1 neurons. The segment had a color that either matched or mismatched the target color. This segment was immaterial for the tasks.

Results and inferences

A comparison between the average time taken for the two tasks, coupled with the percentage of tasks done correctly supports the thesis that subtasks in a composite task are carried out in a structured and sequential manner. The average reaction times in the first task for B and G were respectively 251 and 222 ms (with a performance better than 99% correct). For the second task, B's reaction time increased by 141 ms (94% correct) and G's time increased by only 48 ms (but the performance dropped down to 78%). This shows that the two subtasks, namely visual search and curve tracing, were done in a sequential manner. This makes intuitive sense, since the monkey would first have to search for the marker with the correct color in order to trace the correct curve. However, the monkey could have done both the tasks in parallel and in the end focussed on the circle at the end of the correct curve. This possibility needed ruling out.

The study of the neuronal activity confirms the sequential-processing hypothesis. Observations of the location of the receptive field of a group of V1 neurons relative to the stimuli of the simple curve tracing task show that the receptive field was either on the target curve, connecting the fixation point to the target circle for an eye movement, or on the distractor curve, but the contour segment inside the receptive field was identical for all stimuli. Activity was pooled across the stimuli for which the receptive field focussed on the target curve and for which it was on the distractor curve, with a different color for the fixation point. . It was found that responses to the target curve were stronger than responses to the distractor curve. There were similar observations for the composite task, with the neurons enhancing their response significantly if their receptive field fell on the target curve.

The trials with the added contour segment showed similar evidence. Responses averaged across the stimuli with the segment's color matching and with the ones where the color did not match the target color were enhanced transiently if the color in the receptive field matched the target color. In the simple task, the strength of the neuronal responses was independent of the color of the fixation point matching the color in the receptive field. This gives evidence to the presence of visual search in the complex task, which appeared to be implemented by a color-selective feedback signal to retinotopic areas.

The results support Ullman's theory that complex visual tasks are managed by so-called "visual routines". These routines are made up of a limited collection of attentional operators, such as visual search and curve tracing. It can be inferred from the experiment that it is possible to observe the progression of such visual routines.

It is noteworthy that the results of the experiment support neither a strictly parallel-processing nor a strictly serial-processing paradigm. It is evident from above that the former is clearly out of the question. As for the latter, it was found that successive attentional operators have a considerable temporal overlap. Roelfsema et al suggest that the overlaps are important because they allow transfer of information between successive operators. If these operators are thought of as methods in a computer program, then the information exchange analogous to parameter transfer takes place in one of two ways, which are the visual cortex equivalents of "call by value" and "call by reference".

Personal opinion

The Roelfsema article is interesting. It is clear that the scientists have taken the utmost care to filter out external factors affecting the results. For instance, repeating trials after switching colors between the target curve and the distractor curve rules out incorporating the event that the monkeys could trace out a target curve better if the fixation point was of a particular color. (This can well be the case, but it is not pertinent to the experiment.) Also, the sample space of trials is large enough to get a clear relationship between recording times from the two tasks, despite the noise that must necessarily be.

It can be said, however, that the experiment assumes that each trial is independent of any previous trial and in turn, will not affect any subsequent trial. This is somewhat of a big assumption. It is conceivable that after a certain number of truly independent trials, the monkeys got used to the timing of their tasks and performed later trials mechanically. This could explain why G had a low performance percentage- he adapted to doing the tasks faster, though more erroneously. This problem could have been overcome were there a large number of monkeys, each doing the tasks only once or twice. But this approach is not an improvement. Such trials are better suited to thought experiments- it is quite impractical to train a large number of monkeys for only a trial or two. Besides, the observations from the neuronal activity study confirm the evidence from the reaction times, so there seems to be very little doubt about the conclusion.

In summary, the Roelfsema article is a good article to read. It describes an experiment to determine the nature of subtask sequencing and concludes from results and observations of that experiment that subtasks of composite tasks are carried out in a structured and sequential manner that can be monitored in the primary visual cortex.



Bibliography: -
"Subtask sequencing in the primary visual cortex" by Pieter R. Roelfsema, Paul S. Khayat, and Henk Spekreijse