The stimulus-evoked response is the principle measure used to elucidate the timing and spatial location of human brain activity. Brain and behavioural responses to pain are influenced by multiple intrinsic and extrinsic factors and display considerable, natural trial-by-trial variability. However, because the neuronal sources of this variability are poorly understood the functional information it contains is under-exploited for understanding the relationship between brain function and behaviour. We recorded simultaneous EEG–fMRI during rest and noxious thermal stimulation to characterise the relationship between natural fluctuations in behavioural pain-ratings, the spatiotemporal dynamics of brain network responses and intrinsic connectivity. We demonstrate that fMRI response variability in the pain network is: dependent upon its resting-state functional connectivity; modulated by behaviour; and correlated with EEG evoked-potential amplitude. The pre-stimulus default-mode network (DMN) fMRI signal predicts the subsequent magnitude of pain ratings, evoked-potentials and pain network BOLD responses. Additionally, the power of the ongoing EEG alpha oscillation, an index of cortical excitability, modulates the DMN fMRI response to pain. The complex interaction between alpha-power, DMN activity and both the behavioural report of pain and the brain's response to pain demonstrates the neurobiological significance of trial-by-trial variability. Furthermore, we show that multiple, interconnected factors contribute to both the brain's response to stimulation and the psychophysiological emergence of the subjective experience of pain.Highlights
▸ EEG–fMRI study of the origins of inter- and intra-subject pain response variability ▸ Single-trial correlation between BOLD and evoked potential responses and pain ratings ▸ Pre-stimulus DMN activity modulates subsequent pain ratings, BOLD and EEG responses. ▸ CHEPs and DMN response to pain stimulation is modulated by ongoing EEG alpha power. ▸ Pain network BOLD response to pain is predicted by the network's resting connectivity.