Characterization of Brain Signals Across Scales Using Temporally Modulated Visual Stimuli

Abstract

Electrical signals from the brain can be recorded at several different scales, ranging from spiking activity to local field potential (LFP) in animals to scalp electroencephalogram (EEG) in humans. Each signal represents a progressively larger level of neural integration and is thought to reflect different attributes of the underlying neural population. In this work, we characterized the relationships between these signals using a common paradigm of temporally modulated visual stimuli, which are known to engage the underlying neural activity in distinct ways. In the first part, we asked whether the LFP reflects the input or the output of a cortical area around the recording microelectrode. Using chronically implanted arrays in the primary visual cortex (V1) of awake behaving macaque monkeys, we recorded spikes and LFP in response to drifting sinusoidal gratings of varying temporal frequency. Previous reports have shown that the primate lateral geniculate nucleus (LGN) which projects to V1 has a higher temporal frequency cutoff than V1, such that at higher drift rates, visual input to V1 persists but V1 output ceases, permitting partial dissociation. Using an adaptive decomposition technique called Matching Pursuit (MP) to generate the time-frequency spectrum of the LFP at high resolution, we show that distinct frequency bands in the V1 LFP are tuned differently to temporal frequency, such that the lower frequencies of the LFP (up to ~50 Hz) likely represent the input, the gamma band of the LFP (~30-80 Hz) likely represents local cortical processing and the high-gamma band (above ~80 Hz) represents the output. In the second part, we asked whether steady-state visually evoked potential (SSVEP) tags used in “frequency tagging” EEG studies of visual attention are independent or not. In these studies, it is observed that paying attention to one stimulus increases the amplitude of the SSVEP at the tagging frequency of that stimulus and simultaneously decreases the SSVEP amplitude at the unattended frequency. This has been explained using a “push-pull” or “spotlight” mechanism of attention. However, it is unclear whether changes in SSVEP amplitude could arise due to the presence of competing temporal frequencies without any top-down cognitive modulation, and whether this depends on the separation between the tagging frequencies or the features of the stimuli such as their orientations. To address these questions, we recorded spikes and LFP from V1 as well as EEG from awake behaving macaque monkeys while they passively fixated plaid stimuli whose components counterphased at different temporal frequencies. We observed reliable SSVEP response suppression, but the suppression was much greater for lower competing temporal frequencies than for higher ones. Further, the strength of this asymmetry depended on the relative orientation difference between the plaid components, with similar orientations causing significant suppression and orthogonal orientations causing little or no suppression. In the third part, we show that the well-known normalization model, adapted to SSVEP responses, provides a good account of temporal frequency suppression as a function of the difference in temporal frequencies and orientation. Our results provide evidence for interaction between temporal frequencies independent of effects of cognitive modulation and suggest exercising caution in the interpretation of frequency tagging studies.