Beyond the evoked/intrinsic neural process dichotomy

Abstract

Contemporary functional neuroimaging research has increasingly focused on characterization of intrinsic or "spontaneous" brain activity. Analysis of intrinsic activity is often contrasted with analysis of task-evoked activity that has traditionally been the focus of cognitive neuroscience. But does this evoked/intrinsic dichotomy adequately characterize human brain function? Based on empirical data demonstrating a close functional interdependence between intrinsic and task-evoked activity, we argue that the dichotomy between intrinsic and task-evoked activity as unobserved contributions to brain activity is artificial. We present an alternative picture of brain function in which the brain's spatiotemporal dynamics do not consist of separable intrinsic and task-evoked components, but reflect the enaction of a system of mutual constraints to move the brain into and out of task-appropriate functional configurations. According to this alternative picture, cognitive neuroscientists are tasked with describing both the temporal trajectory of brain activity patterns across time, and the modulation of this trajectory by task states, without separating this process into intrinsic and task-evoked components. We argue that this alternative picture of brain function is best captured in a novel explanatory framework called enabling constraint. Overall, these insights call for a reconceptualization of functional brain activity, and should drive future methodological and empirical efforts.

Keywords: Enabling constraint; Intrinsic activity; Neural variability; Synergy; Task-evoked activity.

Figure 1.. Conceptual illustration of the linear superposition principle. In this example, a participant is presented two sequences of fixation (rest) and visual checkerboard blocks, summarized by the block structure below the stimuli. The observed signal from a task-responsive brain region exhibits a predictable increase in BOLD activity during task blocks, followed by a decrease in activity during fixation/rest blocks. The linear superposition principle states that this observed signal is linearly composed of unobserved or latent intrinsic and task-evoked activity (along with possible random measurement error), whose contribution to the magnitude of the BOLD signal is illustrated by green and red arrows, respectively. During the task block, where we expect the task-evoked signal to be dominant, task-evoked activity contributes most to the observed signal, and during the rest/fixation period, intrinsic activity contributes most to the observed signal. Importantly, the linear superposition principle claims that these two signals linearly sum together to form the observed signal.

Figure 2.. Temporal activity space and task response. Illustration of the temporal activity space approach used by previous research (He, 2013) to study a task-responsive brain region’s (e.g., dorsal anterior cingulate cortex; dACC) temporal trajectory across a task scan. (A) The time series of a hypothetical brain region (e.g., dACC) exhibits a consistent increase in signal amplitude to the task blocks. (B) The constraining of dACC’s temporal trajectory across the task scan is illustrated by the three-dimensional scatterplot. Using the same approach as He (2013), we plotted the BOLD amplitude at three randomly chosen successive time points—Bold(t), Bold(t + 1), and Bold(t + 2)—during the off-task (red) and on-task (blue) blocks. As can be observed, the dACC’s trajectory tightens during the on-blocks (i.e., smaller volume of space) and expands during the off-blocks (i.e., larger volume of space), as indicated by the length of the lines or whiskers next to the scatterplot.

Figure 3.. The reflexive and predictive accounts of brain function. (A) According to the reflexive, stimulus-driven processing account, environmental inputs (represented by the arrow pointing to the brain) are received by the brain for neural processing, with the subsequent production of a behavioral response (represented by the arrow leaving the brain). Thus, the brain’s interaction with the environment is governed by the onset of stimuli. (B) According to the predictive-processing account, the brain is neither a passive recipient of external stimuli nor exclusively dedicated to internal processing, but is constantly active, continually trying to predict the stream of sensory stimulation it receives from the environment (represented by the “prediction error” and “prediction” arrows moving to and from the brain and environment). Of note, behavioral output in many predictive-processing theories is intimately related to the sensory input (and prediction errors) received by the system and would be represented in the following diagram by feedback relationships from behavioral output to the brain and environment.