Internal noise in contrast discrimination propagates forwards from early visual cortex

Abstract

Human contrast discrimination performance is limited by transduction nonlinearities and variability of the neural representation (noise). Whereas the nonlinearities have been wellcharacterised, there is less agreement about the specifics of internal noise. Psychophysical models assume that it impacts late in sensory processing, whereas neuroimaging and intracranial electrophysiology studies suggest that the noise is much earlier. We investigated whether perceptually-relevant internal noise arises in early visual areas or later decision making areas. We recorded EEG and MEG during a two-interval-forced-choice contrast discrimination task and used multivariate pattern analysis to decode target/nontarget and selected/non-selected intervals from evoked responses. We found that perceptual decisions could be decoded from both EEG and MEG signals, even when the stimuli in both intervals were physically identical. Above-chance decision classification started <100ms after stimulus onset, suggesting that neural noise affects sensory signals early in the visual pathway. Classification accuracy increased over time, peaking at >500ms. Applying multivariate analysis to separate anatomically-defined brain regions in MEG source space, we found that occipital regions were informative early on but then information spreads forwards across parietal and frontal regions. This is consistent with neural noise affecting sensory processing at multiple stages of perceptual decision making. We suggest how early sensory noise might be resolved with Birdsall's linearisation, in which a dominant noise source obscures subsequent nonlinearities, to allow the visual system to preserve the wide dynamic range of early areas whilst still benefitting from contrast-invariance at later stages.

Dynamic properties of internal noise probed by modulating binocular rivalry

“…It has also been suggested that monocular noise might be increased in the affected eye of individuals with amblyopia [43]. Finally, we have recently shown [44] using a contrast discrimination paradigm that EEG and MEG data are consistent with both an early (~100ms post stimulus onset) noise source in low level visual areas, and a later noise source in more frontal and parietal brain areas, both of which affect perceptual decisions. All of these results are therefore consistent with an early monocular source of internal noise, as included in our model, but do not preclude the addition of later sources of noise which we do not consider here.…”

Power contours: Optimising sample size and precision in experimental psychology and human neuroscience.

“…We next analyzed event-related potentials (ERPs) from a contrast discrimination experiment reported by Vilidaite, Marsh, and Baker (2019) , recorded using a 64-channel EEG cap. The stimuli were sine wave gratings with a contrast of 50%, presented sequentially in pairs for 100 ms each, with an interstimulus interval of 400–600 ms.…”

The forgotten history of signal detection theory.

“…As they put it: “That false alarms evoked more activity than misses indicates that activity in early visual cortex corresponded to the subjects’ percepts, rather than to the physically presented stimulus” (p. 414). Although not all neuroimaging studies have detected elevated neural signals in sensory pathways associated with false detection (e.g., Hulme, Friston, & Zeki, 2009; Mostert, Kok, & de Lange, 2015), many have (e.g., Pajani, Kok, Kouider, & De Lange, 2015; Ress & Heeger, 2003; Watkins, Shams, Tanaka, Haynes, & Rees, 2006; Vilidaite, Marsh, & Baker, 2019). In addition, still other neuroimaging studies have provided evidence that activity in higher (nonsensory) regions supports the false subjective experience of perception (e.g., Lloyd, McKenzie, Brown, & Poliakoff, 2011; similar conclusions about the role of higher brain regions were reached in single-unit work by de Lafuente & Romo, 2005, 2006).…”