Rat superior colliculus encodes the transition between static and dynamic vision modes

Abstract

The visual continuity illusion involves a shift in visual perception from static to dynamic vision modes when the stimuli arrive at high temporal frequency, and is critical for recognizing objects moving in the environment. However, how this illusion is encoded across the visual pathway remains poorly understood, with disparate frequency thresholds at retinal, cortical, and behavioural levels suggesting the involvement of other brain areas. Here, we employ a multimodal approach encompassing behaviour, whole-brain functional MRI, and electrophysiological measurements, for investigating the encoding of the continuity illusion in rats. Behavioural experiments report a frequency threshold of 18±2 Hz. Functional MRI reveal that superior colliculus signals transition from positive to negative at the behaviourally-driven threshold, unlike thalamic and cortical areas. Electrophysiological recordings indicate that these transitions are underpinned by neural activation/suppression. Lesions in the primary visual cortex reveal this effect to be intrinsic to the superior colliculus (under a cortical gain effect). Our findings highlight the superior colliculus' crucial involvement in encoding temporal frequency shifts, especially the change from static to dynamic vision modes.

Fig. 1. Behavioural results.

A Behaviour box schematic. Water-deprived rats were placed in a dark box with three poking ports: a middle port to initiate trials and two lateral ports for continuous or flicker light reports. B Schematic of the task: Animals start each trial by poking in the central port. An overhead LED would turn on displaying either continuous light or flickering light at various frequencies. Rats were rewarded for poking to one side if the stimulus was continuous, and to the other side if the stimulus flickered. Incorrect responses triggered a noise burst and a time penalty. C Percentage of reports to the flicker port. Thin grey dashed lines reflect the performance of each individual animal (n = 7) while blue circles correspond to the averaged individual performances. As the frequency increases the animal reports less often to the flickered port signalling a shift towards the dynamic vision mode. The calculated FFF threshold proxy at “chance level” is 18 ± 2 Hz. Source data are provided as a Source Data file.

Fig. 2. Whole pathway fMRI results (total n = 18, for every individual frequency n ≥ 5).

A TOP: Stimulation paradigm used in the fMRI. The stimulation paradigm consisted of 15 s stimulation at different frequencies followed by 45 s rest; Bottom: Schematic of the animal’s position in MR bed. B fMRI t-maps for representative visual stimulation frequencies and atlas overlapped. As the stimulation frequency increases, transitions from PBRs to NBRs are observed first in VC and subsequently in SC. C Mean ± s.e.m. of fMRI signal temporal profiles across animals (n = 18). Fine structure appears in the fMRI responses. Onset and offset peaks are evident in the SC profiles from 15 Hz onwards, along with a “steady-state” (black arrows). D Correlation of behaviour reports with fMRI “steady-state” signals. Coloured circles represent the average response of behavioural and fMRI sessions and error bars represent the standard deviation across runs/sessions. Only the SC shows a clear transition from PBR to NBR that correlates with the behaviourally measured FFF threshold surrogate (R(4) = 0.95, P = 0.004, 95% CI = [0.61,1.00], two-tailed Pearson’s correlation). Weaker skewed correlations were found for the VC and LGN (R(4) = 0.75, P = 0.09, 95% CI = [−0.16,0.97] and R(4) = 0.85, P = 0.03, 95% CI = [0.11, 0.98], respectively, two-tailed Pearson’s correlation). E fMRI-derived neurometric curves generated from individual trial data reveal that only SC tracks the psychometric curve obtained from the behavioural experiments, reporting a “chance level” threshold of 20.0 ± 2.8 Hz. PBR positive BOLD responses, BOLD blood oxygenated level dependent, NBR negative BOLD responses, VC visual cortex, SC superior colliculus. Source data are provided as a Source Data file.

Fig. 3. Electrophysiology results (total n = 20, for every individual frequency n ≥ 4).

A Probe insertion schematic and fluorescence microscopy image. B Median LFP traces. LFP traces show individual flash-induced LFP oscillations for the 1 Hz condition and onset and offset peaks for the higher frequencies. Zoomed LFP plots show the first second of responses where flash-induced LFP (with increasingly reduced amplitude) can be observed until the 25 Hz stimulation regime. C Spectrograms between 1 and 50 Hz for 1, 15, 25 Hz and continuous light stimulation regimes. D “Steady-state” LFP-driven neurometric curves. The dots represent mean flicker reports for each animal while the error bars represent the 95% confidence interval. For frequencies 1, 2, 8, 12.5, 15, 20, 25, 40 Hz and continuous light, the used number of trials was 440, 380, 230, 230, 210, 380, 210, 230 and 180, respectively. The fitted sigmoids reveal “chance-level” thresholds of 18.0 ± 1.7 and 17.7 ± 2.4 Hz, for the early and late “steady-state” intervals, respectively. These values are in accordance with the ones obtained in the SC fMRI-driven neurometric curves and behavioural psychometric curve. E Mean ± s.e.m MUA relative power plots across animals. These plots reveal a decreased power during stimulation as the frequency of stimulation increases. Offset signals become evident for the high frequencies. The two highest frequency regimes show power decreases during stimulation below baseline level. F Correlation between “steady-state” MUA relative power and fMRI percent signal change. Coloured circles represent the average response of electrophysiological and fMRI sessions and error bars represent the standard deviation across runs. A high correlation coefficient of R(4) = 0.98, (P = 0.0005, 95% CI = [0.84,1.00], two-tailed Pearson’s correlation) shows a tight relationship between the two measurements. NBRs at high stimulation frequencies correlate with strong MUA power reductions. LFP local field potential, MUA multi-unit activity. Source data are provided as a Source Data file.

Fig. 4. fMRI ibotenic acid lesions results.

A Schematic of the visual pathway after V1 lesion highlighting reduced cortical feedback (TOP); histological and structural MR images (BOTTOM) confirming the lack of brain tissue in the lesion site. B fMRI t-maps for representative frequencies (n = 13). As the frequency of stimulation increases, transitions from PBRs to NBRs appear in SC but not in VC. Secondary visual cortical areas show enhanced PBRs at 1 Hz compared to the control regime, probably reflecting plasticity events that took place between the lesion induction and the fMRI experiments. C Mean ± s.e.m. SC temporal fMRI profiles across animals (n = 13). SC temporal profiles reveal marked positive to negative BOLD shifts and the onset and offset signals remain present after the lesion. Reduced amplitude of both positive and negative SC BOLD responses highlights the gain effect from V1 feedback projections. D Electrophysiology results. Electrophysiological signals were recorded in lesioned animals (n = 6) to confirm the presence of the SC oscillatory responses. From the left and middle plots, single flash-evoked responses are observed until 25 Hz. Onset and offset signals are still observable with the latter only present for the higher stimulation conditions. MUA power plots show a power reduction with stimulation frequency. E Hypothesized decomposed SC response. A hypothesized mechanism is proposed based on the fMRI results. The SC response has at least two different contribution sources: a contribution within SC (novelty and a constant frequency discrimination perception) and a cortical contribution acting as gain control. Possible contributions from other structures within the visual pathway cannot be ruled out. V1 primary visual cortex, PBR positive BOLD responses, BOLD blood oxygenated level dependent, SC superior colliculus, VC visual cortex, MUA multi-unit activity. Source data are provided as a Source Data file.