Repetition suppression: a means to index neural representations using BOLD?

Abstract

Understanding how the human brain gives rise to complex cognitive processes remains one of the biggest challenges of contemporary neuroscience. While invasive recording in animal models can provide insight into neural processes that are conserved across species, our understanding of cognition more broadly relies upon investigation of the human brain itself. There is therefore an imperative to establish non-invasive tools that allow human brain activity to be measured at high spatial and temporal resolution. In recent years, various attempts have been made to refine the coarse signal available in functional magnetic resonance imaging (fMRI), providing a means to investigate neural activity at the meso-scale, i.e. at the level of neural populations. The most widely used techniques include repetition suppression and multivariate pattern analysis. Human neuroscience can now use these techniques to investigate how representations are encoded across neural populations and transformed by relevant computations. Here, we review the physiological basis, applications and limitations of fMRI repetition suppression with a brief comparison to multivariate techniques. By doing so, we show how fMRI repetition suppression holds promise as a tool to reveal complex neural mechanisms that underlie human cognitive function.This article is part of the themed issue 'Interpreting BOLD: a dialogue between cognitive and cellular neuroscience'.

Keywords: functional magnetic resonance imaging adaptation; neural computation; neural representation; repetition suppression.

Figure 1.

Suppression in neural activity in response to repeated stimuli. (a) Experimental timeline of suppression experiment reported in [35]. From 0 to 15 min a 9 µs light stimulus was presented every 2 s. From 15 to 30 min it was switched off, before flashing was resumed at the same frequency from 30 to 34 min. Panels (b– e) show post-stimulus histograms of a single neuron in the primate IT cortex over the 1 s period following the onset of a light stimulus, at different times during the experiment: (b) 0–4 min; (c) 5–9 min; (d) 10–14 min; (e) 30–34 min, adapted from [35]. (f) Timeline of a trial in the serial recognition task used to probe recency responses in the macaque entorhinal cortex. Pictures of naturalistic scenes or objects were presented in random order, with a variable number of pictures between the repetition of any given picture. In every trial, the animal indicated by button press whether a picture was novel or familiar. Correct responses were rewarded. (g) Responses of neurons in the entorhinal cortex to the first presentation of a novel picture, shown as peristimulus histograms and raster plots for 10 trials. Bin width is equal to 100 ms. (h) Responses of neurons in the entorhinal cortex when a novel picture is repeated. Panels (f–h) adapted from [36].

Figure 2.

Effect of interleaving stimuli on repetition suppression. Mean percentage change in neuron response to visual stimulus presentation relative to spontaneous activity as a function of the number of interleaved stimuli in (a) area TE, (b) perirhinal cortex, and (c) entorhinal cortex. With an increasing number of intervening trials between the first and second presentation of the stimulus, a decrease in the repetition suppression effect is observed. Asterisk indicates significant repetition suppression effect. N: first presentation of a stimulus. Adapted from [36].

Figure 3.

Schematic illustration of the principle underlying fMRI adaptation. (a) The raw BOLD signal measured in conventional fMRI paradigms provides only a measure of the mean activity of the population of neurons within a given voxel. In this example, the raw BOLD signal in response to stimuli X, Y and Z is the same, because the average neural activity within the voxel is comparable for all three stimuli. The raw BOLD signal alone is therefore invariant to the relationship between representations of stimuli X, Y and Z. (b) In fMRI adaptation paradigms, the relationship between different stimulus representations X, Y and Z may be indirectly measured. If stimulus X is preceded by stimulus X (X→X), then the fMRI signal in areas encoding features particular to stimulus X is suppressed. If stimulus X is preceded by stimulus Y (Y→X), the response to stimulus X should not show any suppression, as the representations for X and Y are not overlapping. If stimulus X is preceded by stimulus Z (Z→X), the response in areas encoding the features that are shared between X and Z should show suppression which scales with the amount of overlap between representations of X and Z.

Figure 4.

Repetition suppression and expectation suppression occur at different times in cortical processing. (a) Repetition of an auditory tone results in reduced neural signal during the early time bin following stimulus onset (green: repeated; black: alternating). (b) Expectation of an auditory tone results in reduced neural signal during intermediate time bins following stimulus onset (blue: expected; red: unexpected). Both plots: both repetition and expectation of an auditory tone give reduced neural signal during late time bins following stimulus onset. Time is denoted along the x-axis and periods during which a significant effect was observed are shown in grey’. The average auditory evoked response measured using MEG is shown on the y-axis. Adapted from [103].

Figure 5.

Using repetition suppression to measure grid cells in the human entorhinal cortex. (a) Spatial autocorrelogram of a typical grid cell in the rat entorhinal cortex, constructed from the single-unit firing rates within a box shaped environment. The grid cell's firing fields are arranged symmetrically at an angle of 60°. (b) fMRI adaptation was measured as a function of the angle between running directions. (c) Speed-dependent adaptation can be observed within the entorhinal cortex (ERH) when participants run at 60° from their previous direction. (d) The speed-dependent adaptation effect was specific to running angles of 60° and was not observed at running angles of 90° or 45°. (e) The same brain region showed a modulation of the raw activity consistent with a sixfold rotational symmetry. (f) Visualization of the sixfold rotational symmetry with running direction in the raw fMRI signal extracted from the entorhinal cortex. Adapted from [135].

Figure 6.

Reptition time-lag and exposure: factors to consider when designing an fMRI repetition suppression experiment. (a) regions of interest in the visual cortex and orbitofrontal cortex from which repetition suppression was measured, adapted from [145]. (b) An interaction was observed between the brain regions shown in (a) and the time-lag across which repetition suppression was observed. In both the visual cortex and OFC, suppression was observed if a stimulus was repeated after 400 ms. However, only the OFC showed repetition suppression if a stimulus was repeated after 6000 ms, suggesting a difference between anterior and posterior brain regions, adapted from [145]. (c) Repetition suppression in the left (L) and right (R) hippocampus and the parahippocampal gyrus (PHG) declines linearly as a function of the number of presentations or repetitions of a stimulus, adapted from [147].