The second 'review' of 'Functional specialization in the visual cortex?'

Below is the full text of the second review of 'Functional specialization in the visual cortex?' . I have added the numbers of the comments in my response at the appropriate position, in square brackets.


During their scientific careers, researchers learn to use shorthand assumptions to stand in place of gaps in their knowledge, and become adept at manipulating these vicarious concepts in formulating the arguments and discussions put forward in papers. Thus concealed, the gap in scientific knowledge can recede from consciousness, or at least from everyday thought. Challenges to an argument that derive from challenging the shorthand assumptions nay then be misunderstood, or simply not recognized. I think that Harpaz's criticism are along these lines. I Have tried to consider them with an open mind.

1. Concerns the "higher level" emotional or conceptual ('cognitive') properties of visual stimuli. The underlying idea is serial processing: several visual stages, culminating in recognition of familiar object, followed by associations with the object that represents its significance to the perceiver [1].

The design of the experiment attempts to minimize the emotional or conceptual properties [2]. The optical flow presents a featureless surface; the running man is not a videotaped runner, but a set of moving dots whose pattern of motions stimulates the pattern of joint (knee, elbow etc.) movement during running. Viewers recognize the pattern of human running motion; whether they are additionally prompted to imagine running, or runners, or humanity in general is harder to assess. The potential for such 'knock-on' effects should not be dismissed, but is universal in brain scanning experiments [3].

The images produced by the scanner are assumed to show all the resulting activity (visual and cognitive). More accurately, to show all the difference between the test and control states of activity [4]. It is anticipated that some early visual stages (measuring speed and direction of individual dots) may cancel out, leaving only the later stages of processing visible. Examination of the results shows that this is the case for optical flow and biological motion.

The large, occipital, patch of activity for biological motion coincides with part of the larger territory activated by coherent motion. The specialization of this territory for motion analysis is supported by a vast body of evidence from experimental work on primate brains [5]. It would nothing but perverse to attribute this activity to a "higher level" cognitive factor [6]. The smaller patch of activity on the temporal lobe is harder to interpret. The authors struggle to do so, but select comments that relate to what the cortex in this region is known to do - mediate hearing and speech. They might have speculated that the activity represents subvocalisation "it's a man!" by the subjects. But many readers, editors and referees will frown upon such unsubstantiated speculation. The authors chose to confine themselves to vaguer interpretations [7].

Overall I do not see much weight in this criticism. The doubts attach mainly to one small part of the response to one condition[8].

2. The criticism here - that "the bulk of information in the brain is not coded in level of activity, but in the pattern" [9] - is fundamental, as it essentially implies that all scanning experiments are worthless [10]. However no evidence is cited to substantiate this criticism and it becomes vague and self contradictory [11].

There is some, very recent evidence that signal processing, or the control of behaviour, may be mediated by changes in the correlated firing pattern of neurons without any change in the overall firing rate of those neurons. This evidence in no way invalidates 30 years worth of single unit recording studies from the brain demonstrating overwhelmingly that a change in firing rate is a common consequence of signal processing [12]. The scanning evidence is also beginning to build a self-consistent picture, based on the interpretation that increased blood flow corresponds to increase activity corresponds to useful signal processing [13]. It seems more likely that scanning may give too exuberant picture of activity, as there is some evidence that synaptic activity can be decoupled from changes in spiking behaviour; this is the opposite criticism to that made by harpaz so I will not pursue it further [14].

In his initial comments Harpaz writes "processing input means changes in pattern of activity, not level of activity". One sentence later "..peaks of differential activity can arise from other reasons, including (1) high level PROCESSES.." This is a clear contradiction [15]. The other reasons given, (b) and (c) are incomprehensible to be (what on earth can a direct effect of an input be, as if it were somehow to escape processing?) [16].

It is worth noting that harpaz, in his response to the referee, adjusts his position to concede that " we would still expect to find few cases where activity and functional significance coincide". The 'few' is an advance upon the blanket ban of the original critique; can we expect further developments? [17].

I agree with the final conclusion of the referee and would add that Harpaz pays no heed to the body of evidence that accrued in nonhuman primates that guides the conclusion arrived at by Howard et al [18].

3. I have much more sympathy for this point. Scientists like to polarize arguments, and specialization of an area of the cortex might result purely from internal structure and physiology, or if cortical structure and physiology were absolutely uniform, purely from the cortical wiring that brings a different set of signals to an area. Very likely both factors are at play [19], but Harpaz wants to define 'specialization' as a term that refers only to the first of this [20].

This might be a reasonable definition but it is not the one used in cortical physiology [21]. The older term 'functional localization' was avoided by Zeki, on his discovery of this property in primate visual cortex, because a statement like "there is a functional localization of motion in the mid-temporal lobe" could be taken to mean that motion is not localized anywhere else in the brain. The alternative "there is functional specialization for motion in the mid-temporal lobe" is neutral in this respect [22].