Main

Plasticity in adult V1 has been demonstrated by multiple, independent lines of evidence from more than twenty studies in three species (see refs 2–6, for example). Physiologically, the evidence derives from measurements of lesion-induced shifts in the locations of V1 neuronal receptive fields. By plotting receptive fields before and at various points after making retinal lesions, it was shown that the affected cortex — with receptive fields originally inside the lesion — develops new, shifted receptive field positions after recovery. These shifts are cortically mediated because the lateral geniculate nucleus, the source of thalamic input to V1, shows limited reorganization7. All these measurements were made using suprathreshold spiking neuronal responses: this is an important point as the neuronal output from V1 to subsequent cortical stages is carried entirely by spikes. Any measure of V1 reorganization — and consequent functional remapping of visual information — therefore needs to assess the effect on spiking activity. These physiological results are buttressed by anatomical findings showing a selective increase in the density of axon collaterals in reorganized cortex8, and the sequential expression of biochemical markers9,10.

By contrast, the primary evidence for lack of plasticity offered by Smirnakis et al. is the observation that the V1 ‘silent zone’, mapped with BOLD fMRI immediately following a retinal lesion, did not change over time. There are plausible reasons why fMRI maps may fail to change, despite re-emergent neuronal activity. The reorganization of cortex is believed to be mediated by long-range horizontal connections within V1 (refs 8,11,12). In normal V1, these connections mediate subthreshold modulation. Following retinal lesions, horizontal connections stretching from ‘normal’ cortex into the lesion projection zone (LPZ) are believed to strengthen their synapses — but not to change anatomical extent. They therefore induce re-emergent spiking activity, but only in neurons lying within their target zone in the silenced cortex.

Such re-emergent activity could involve reduction of inhibition as much as an increase in excitation. As the BOLD signal probably reflects synaptic input into a region rather than spiking output, the ‘silent zone’ observed by Smirnakis et al. immediately following a lesion may mark not the edge of the real LPZ but the inner edge of subthreshold activation spreading into the LPZ through horizontal connections. In subsequent measurements, the BOLD signal would continue to show the unchanging position of this inner boundary while being blind to synaptic reorganization, which would lead to re-emergent spiking activity over the extent of the horizontal connections. The single set of electrode recordings by Smirnakis et al. after months of recovery might simply show the extent of largely completed recovery and, not surprisingly, produce a border in register with the edge of the BOLD signal.

Furthermore, BOLD gives a local measure of the total cortical activity, a significant component of which comes from thalamocortical inputs, the contribution of which is probably further accentuated by the disproportionately high vascularization of layer 4, the cortical input layer. However, the neurons showing recovery may reside primarily in the superficial layers12, which receive the long-range horizontal connections, as opposed to layer 4.

Owing to these uncertainties about the validity of BOLD fMRI as a yardstick of functional reorganization in V1, we believe that Smirnakis et al. do not present a convincing contradiction to the body of earlier evidence indicating substantial receptive field plasticity in adult animals following retinal lesion. The recovered activity demonstrated in the earlier studies has a likely corollary in the recovery of visual perception: human subjects suffering from macular degeneration, or with artificially induced retinal lesions, show improved perceptual fill-in over time after the lesions13,14,15.