Neural Substrates and Models of Omission Responses and Predictive Processes

Abstract

Predictive coding theories argue that deviance detection phenomena, such as mismatch responses and omission responses, are generated by predictive processes with possibly overlapping neural substrates. Molecular imaging and electrophysiology studies of mismatch responses and corollary discharge in the rodent model allowed the development of mechanistic and computational models of these phenomena. These models enable translation between human and non-human animal research and help to uncover fundamental features of change-processing microcircuitry in the neocortex. This microcircuitry is characterized by stimulus-specific adaptation and feedforward inhibition of stimulus-selective populations of pyramidal neurons and interneurons, with specific contributions from different interneuron types. The overlap of the substrates of different types of responses to deviant stimuli remains to be understood. Omission responses, which are observed both in corollary discharge and mismatch response protocols in humans, are underutilized in animal research and may be pivotal in uncovering the substrates of predictive processes. Omission studies comprise a range of methods centered on the withholding of an expected stimulus. This review aims to provide an overview of omission protocols and showcase their potential to integrate and complement the different models and procedures employed to study prediction and deviance detection.This approach may reveal the biological foundations of core concepts of predictive coding, and allow an empirical test of the framework's promise to unify theoretical models of attention and perception.

Keywords: animal research; corollary discharge; feedforward inhibition; interactionism; mismatch negativity; predictive coding; stimulus specific adaptation (SSA).

Figure 1

Model of corollary discharge microcircuitry before and after action-stimulus association. (A) In the networks’ baseline state, motor input (red arrows) results in inhibition of sensory responses in all excitatory populations (E_1,2). Such generalized inhibition, independent from sensory input (blue arrows) is mediated by all inhibitory populations (I_1,2). (B) After repeated paired presentations of motor and sensory input (i.e., stimulus 1), the motor input results in stimulus-specific inhibition for the paired sensory input. Specific inhibition is likely mediated by long-term potentiation of motor-to-sensory synapses (thick red arrows), which results in increased activation of I₁ interneurons that are tuned to the action-paired stimulus. Circles indicate neural populations and lines indicate their connectivity. Arrows and filled circles indicate excitatory and inhibitory synapses, respectively. Thickness is used to indicate the strength of synaptic connectivity (arrows and filled circles) and stimulus response (circles and lines). Excitatory and inhibitory populations that respond preferentially to the same stimulus are grouped and surrounded by light pink rectangles.

Figure 2

Hypothetical model of mismatch negativity microcircuitry based on electrophysiology and tracing studies in mice. (A) In its baseline state interconnected excitatory populations (E_1,2) respond similarly to the stimuli (black and green bottom-up arrows) they are selective for, with no relevant modulation from local interneurons (which are not stimulus selective). (B) After repeated stimulation the standard-selective pyramidal population (E₁) reduces its excitability due to inherited adaptation from the upstream sensory input, and the deviant-selective population (E₂) increases its excitability due to NMDA-mediated (purple triangles) synaptic potentiation. Circles indicate neural populations and lines their connectivity. Arrows and filled circles indicate excitatory and inhibitory synapses, respectively. Thickness is used to indicate the strength of synaptic connectivity (arrows and filled circles) and stimulus response (circles and lines). I_SST and I_PV indicate somatostatin and parvalbumin interneurons, respectively. Excitatory and inhibitory populations that respond preferentially to the same stimuli are grouped and surrounded by light pink rectangles. Only the relevant connectivity is shown. (C) Typical activity traces for these neural populations as they transition from baseline to adapted activity (E₁, I_PV), including responses (E₂) to a deviant during an oddball paradigm. Grey and green squares indicate the stimulus sequence. Adapted from Ross and Hamm (2020).

Figure 3

Summary of omission response protocols and response examples. (A) The omission protocol consists of a series of auditory stimuli (black squares) presented at regular intervals. The omission response (green arrow) is observed at the end of the stimulus train, at a fixed latency from the onset of the expected but missing stimulus (green dashed square). Adapted from Andreou et al. (2015). (B) The button-pressing symbol indicates that the auditory stimuli are self-initiated. This protocol consists of self-paced button pressing to which a stimulus is associated but randomly withheld. The omission response is observed when the action is performed but the action-associated stimulus is withheld. The black arrow indicates the response to the standard self-initiated stimulus. Adapted from SanMiguel et al. (2013a). (C) The orange dashed square indicates the omission of the first stimulus in a pair. This protocol consists of a series of stimulus pairs of which the first or second stimulus is randomly omitted. The omission response is observed when the second stimulus of the pair is omitted. Adapted from Bendixen et al. (2009). (D) The blue squares represent visual stimuli. This protocol consists of a series of paired presentations of a visual stimulus and an auditory stimulus, in which the auditory stimulus is randomly omitted. The omission response is observed in auditory areas when the auditory stimulus is omitted. Adapted from Stekelenburg and Vroomen (2015).

Figure 4

Computational model of deviance detection nodes and simulated MEG signals. (A) Architecture of generic deviance detection circuit with two reciprocally connected stimulus-selective nodes (pink squares) that can function as deviance detectors for each other’s activation state. The synaptic weights of the inter-node connections (magenta) are the free parameters of the model. Circles indicate neural populations, excitatory (E_1,2) and inhibitory (I_1,2), and lines indicate their connectivity. Arrows and filled circles indicate excitatory and inhibitory synapses, respectively. (B) Simulated magnetoencephalography (MEG) signals of a network composed of multiple such nodes include responses (arrows) to the omission (at time 0) of a stimulus from a regular input sequence. The responses, and their dependence on stimulus onset asynchrony, closely resemble those observed experimentally in recordings from human listeners. Adapted from Andreou et al. (2015). (C) Simulated MEG signals from a network composed of the same nodes shown in (A). As the input is switched from a regular to a random sequence, an MMN-like increase in response amplitude is observed in response to the first deviant of the sequence (arrow and dashed square). The time series closely resembles those observed in recordings from human listeners (Barascud et al., 2016) with the same stimulation protocol. Adapted from Chien et al. (2019).