Cortical responses to visual motion in alert and anesthetized monkeys

To the editor

Cortical area MT of macaque monkeys contains direction-selective neurons. Some respond only to the components of complex moving patterns, whereas others compute true pattern motion^1,2. In Nature, Pack et al.³ reported that most MT neurons compute pattern motion in alert macaques but signal only component motion under anesthesia, but we have found the prevalence of component- and pattern-selective neurons to be unaffected by anesthesia^1,2,4. Pack et al. conclude that motion integration circuits are impaired by anesthesia³. We believe their results can be explained by their choice of stimuli and anesthetic.

A neuron is classified as component-selective if its responses to 'plaid' patterns (Fig. 1c) are proportional to the sum of its responses to the plaids' components¹ (Fig. 1a and b). However, the plaids used by Pack et al.³ (Fig. 1d) are not the sum of the components they used. Where the gratings' bars intersect, the luminance of an additive plaid is doubled, but their stimulus had uniform luminance. A third component (Fig. 1e) must be subtracted from the gratings to create their non-additive plaid. Classification using these stimuli requires that the components of the plaid move in a different direction than the plaid itself, but this third component moves in the same direction⁵ (Fig. 1e). The third component would elicit pattern-like responses from all MT cells, and thus component-selective cells tested with this non-additive plaid would masquerade as pattern-selective. It is therefore inappropriate to use non-additive plaids for this classification method, and the use of these patterns led Pack et al. erroneously to claim that most MT neurons in alert animals are pattern-selective.

Figure 1: Additive and non-additive plaids.

(a, b) Grating components of (c), an additive plaid. (d) The non-additive plaid used by Pack et al.³. (e) The component subtracted from (c) to create (d).

Why might anesthesia change pattern-like into component-like responses? The third component is low in contrast compared to the grating components. Neurons in MT are very sensitive⁶, so such a low-contrast stimulus component would normally be effective. However, Pack et al. used isoflurane anesthesia, which substantially reduces contrast sensitivity in cortical and thalamic neurons while only modestly reducing responses to high-contrast stimuli^7,8. Isoflurane also potentiates GABA-mediated inhibition⁹, and thus strengthens the cross-orientation suppression characteristic of cortical responses¹⁰. These factors would selectively weaken responses to the third component (Fig. 1e). Thus, isoflurane would effectively convert non-additive plaids into additive plaids by selectively attenuating the third component. This is relevant only because Pack et al. used non-additive plaids; classification using standard additive plaids¹ is unaffected by anesthesia.

We therefore suggest that the findings of Pack et al.³ do not reflect any fundamental property of cortical motion detection, but result instead from unfortunate choices of stimulus and anesthetic. The stimulus led Pack et al. to misclassify their neurons in alert animals, and the anesthetic caused this classification to change. These factors explain the apparent difference between their results³ and ours^1,2,4.

See Reply to "Cortical responses to visual motion in alert and anesthetized monkeys" by Pack, Berezovskii and Born.