Testing brain-time theoretic links between neural and behavioural estimates of the fidelity of human temporal perception

Daily life plays out against a backdrop of sequential events in which causes precede consequences. An inability to distinguish causality is a hallmark of catastrophic breaks of human wellbeing such as psychotic episodes. More prosaically, errors of perceived causality might underlie milder delusional beliefs, such as fleeting impressions of clairvoyance.

Because causes invariably precede consequences, an accurate sense of relative timing is key to making causal sense of our world. However, how the human brain represents the timing of events is controversial. We need to correct this deficit in knowledge if we are to understand situations where the processes of timing and causality perception fail.

We will address this gap in knowledge by having people judge pairs of events such as sequential flashes and beeps. When presented with such stimuli, the human brain generates a response which can be recorded using encephalography (EEG). Interestingly, if we repeat the exact same presentation, the brain response is similar, but not identical.

These differences in brain response are termed "noise". Noise plays a fundamental role in psychological models of perception. Indeed, people's judgments in psychological experiments are often analysed via models which then imply particular degrees of noisiness accompanying perception.

However, there has been little work attempting to draw precise correspondences between the noise measured in neural recordings and the noise implied by the judgments people make while being recorded - their behavioural noise.

In this application, we propose the use of a novel form of analysis building on an established statistical technique known as bootstrapping. When applied to human brain recordings, this allows us to specifically quantify noise relating to the brain's temporal fidelity. This is distinct from other sources of noise (for example in the intensity, rather than timing, of the brain's response).

According to a leading account of relative-timing perception, which we term the brain-time account, variability in temporal judgments should specifically reflect temporal brain noise. By isolating temporal noise, we can compare it to behavioural noise. According to the brain-time account, they should match.

Our first aim is to verify a predicted correlation between neural and behavioural measures of noise. Essentially, people who are good at relative-timing tasks should have lower levels of neural latency noise compared to those who perform less well. This relationship should be highly specific, so we will take additional measurements (of other kinds of brain noise, and other kinds of task) to demonstrate such selectivity.

Next, we will test predictions stemming from the notion that human perception is "Bayesian" - the brain weights different sources of information in a very particular way to optimise behaviour. This well-established framework allows us to predict biases in temporal perception based on the noisiness with which different events are encoded. Unlike past research, we will use neural measures of noise in place of behavioural ones to simultaneously probe both the Bayesian-brain and brain-time accounts.

Finally, we will test further implications of the brain-time account, by attempting to influence noise through both experimental manipulations and direct stimulation of the brain. Any changes we induce should be evident in both neural and behavioural estimates of noise. Hence, overall, we propose several complementary tests of the brain-time account.

We hope to improve understanding of how humans determine event timing, and thus support future applied work relating to pathologies of causal perception.