Microsaccades mediate a bottom‐up mechanism for cross‐frequency coupling in early visual cortex (Commentary on Lowet et al .)

Billions of neurons whisper or shout, chant or speak simultaneously in what may seem like a big cacophony. Yet, somehow, they dynamically cluster into local assemblies and distributed networks to produce coherent behaviour. How this magic of flexible organization (‘binding’) comes about is one of the most important questions in neuroscience. A compelling hypothesis is that active communication channels between neurons are formed by timing presynaptic action potentials such that the relevant postsynaptic neurons will be in optimal conditions to respond (Fries, 2015). Such temporal relations are facilitated if the membrane potentials of different neurons oscillate, because then the phase of the oscillation (i.e. how depolarized the neuron is) at each moment is predictable. It is not surprising then that neuronal oscillations and their interactions have become a prime target in neuroscience. Presumably, if oscillatory activity should provide phase predictability, its frequency should be quite specific (narrowband) and stationary. In a new study, Lowet et al. (2015) found a narrowband gamma oscillation (NGO; ~ 35–45 Hz) in early visual areas V1 and V2 of three monkeys, yet its peak frequency varied over time, violating the stationarity requirement. This could have been a serious problem for the theory of communication-by-coherence. However, Lowet et al. show that the frequency variability is itself oscillatory, rising and falling by a few Hz at a rate of ~3 Hz (theta rhythm), with a more or less similar phase in V1 and V2. Hence, phase predictability is regained by a special instance of cross-frequency coupling (CFC; Canolty et al., 2006), in which not only the amplitude but also the peak frequency of one oscillation is modulated by the phase of a lower frequency. Moreover, Lowet et al. show convincingly that the change in NGO frequency and amplitude (Bosman et al., 2009) is time-locked to the onset of microsaccades, occurring at a theta rate of 3–4 Hz. Microsaccades are very small, normally imperceptible eye movements occurring involuntarily while the monkey (or a human) tries to fixate (for recent reviews, see Martinez-Conde et al., 2013; Hafed et al., 2015). They are generated at the level of the brainstem, by mechanisms similar to those generating regular saccades (Hafed et al., 2009, 2015). How do microsaccades affect the NGO to produce the observed CFC? There are two cardinal options – an indirect, retinally mediated effect, or a direct ‘extra-retinal’ effect. First, like any change of visual input, the change in the retinal image induced by eye movements evokes a neural response along the visual pathway (for a summary, see fig. 4 of Martinez-Conde et al., 2013). These evoked responses can be clearly seen in figs 2f, 3e and 4e in the study by Lowet et al. Each rather sharp evoked potential includes broadband power extending into the gamma range. This is evident in the response to both the stimulus onset and the microsaccades in fig. 4 of Lowet et al. The theta oscillation reported in V1/2 is likely a smoothed version (narrowly filtered at the theta band) of the periodic evoked potentials. Thus, a parsimonious account for the results is that following a microsaccade, the power spectrum of the local field potential reflects the spectral profile of this bottom-up evoked potential, which sums up or otherwise interacts with the ongoing NGO. The combined response is manifested as a periodic increase in power and frequency of the steady-state NGO, i.e. as CFC. These ‘resetting signals’, which travel down the visual pathway, may resynchronize the different nodes, allowing feedforward and feedback communication. This scenario presents a unique case of CFC, which is mediated by periodic movement of the eye, rather than by direct communication between two oscillating neuronal assemblies or processes. The second source of visual cortex modulation could be ‘extra-retinal’ signals (e.g. efferent copies of the oculomotor activity), which are independent of the change in retinal input. To show that the CFC reported by Lowet et al. is caused by low-frequency extra-retinal signals will require either experiments run in complete darkness (Rajkai et al., 2008) or using specialized noise stimuli, which can rule out the effect of retinal input change (McFarland et al., 2015). Unfortunately, these requirements make it unlikely that narrowband gamma activity will be recorded, considering the rather limited conditions in which this activity has been found. In fact, whether NGO is a ubiquitous phenomenon underlying perception, as suggested by the communication-by-coherence model, is under fierce debate. Several studies in monkeys and in humans showed relatively narrow gamma band activity elicited in early visual cortex by stationary or moving gratings, depending on their size and contrast (Hoogenboom et al., 2006; Perry et al., 2013; Roberts et al., 2013; Brunet