Thunderstruck: The ACDC model of flexible sequences and rhythms in recurrent neural circuits.

Adaptive sequential behavior is a hallmark of human cognition. In particular, humans can learn to produce precise spatiotemporal sequences given a certain context. For instance, musicians can not only reproduce learned action sequences in a context-dependent manner, they can also quickly and flexibly reapply them in any desired tempo or rhythm without overwriting previous learning. Existing neural network models fail to account for these properties. We argue that this limitation emerges from the fact that sequence information (i.e., the position of the action) and timing (i.e., the moment of response execution) are typically stored in the same neural network weights. Here, we augment a biologically plausible recurrent neural network of cortical dynamics to include a basal ganglia-thalamic module which uses reinforcement learning to dynamically modulate action. This "associative cluster-dependent chain" (ACDC) model modularly stores sequence and timing information in distinct loci of the network. This feature increases computational power and allows ACDC to display a wide range of temporal properties (e.g., multiple sequences, temporal shifting, rescaling, and compositionality), while still accounting for several behavioral and neurophysiological empirical observations. Finally, we apply this ACDC network to show how it can learn the famous "Thunderstruck" song intro and then flexibly play it in a "bossa nova" rhythm without further training.

Context-dependent modulation of cognitive control involves different temporal profiles of fronto-parietal activity.

To efficiently deal with quickly changing task demands, we often need to organize our behaviour on different time scales. For example, to ignore irrelevant and select relevant information, cognitive control might be applied in reactive (short time scale) or proactive (long time scale) mode. These two control modes play a pivotal role in cognitive-neuroscientific theorizing but the temporal dissociation of the underlying neural mechanisms is not well established empirically. In this fMRI study, a cognitive control task was administered in contexts with mainly congruent (MC) and mainly incongruent (MI) trials to induce reactive and proactive control, respectively. Based on behavioural profiles, we expected cognitive control in the MC context to be characterized by transient activity (measured on-trial) in task-relevant areas. In the MI context, cognitive control was expected to be reflected in sustained activity (measured in the intertrial interval) in similar or different areas. Results show that in the MC context, on-trial transient activity (incongruent - congruent trials) was increased in fronto-parietal areas, compared to the MI context. These areas included dorsolateral prefrontal cortex (dlPFC) and intraparietal sulcus (IPS). In the MI context, sustained activity in similar fronto-parietal areas during the intertrial interval was increased, compared to the MC context. These results illuminate how context-dependent reactive and proactive control subtend the same brain areas but operate on different time scales.

The unfolding action model of initiation times, movement times, and movement paths.

Converging evidence has led to a consensus in favor of computational models of behavior implementing continuous information flow and parallel processing between cognitive processing stages. Yet, such models still typically implement a discrete step between the last cognitive stage and motor implementation. This discrete step is implemented as a fixed decision bound that activation in the last cognitive stage needs to cross before action can be initiated. Such an implementation is questionable as it cannot account for two important features of behavior. First, it does not allow to select an action while withholding it until the moment is appropriate for executing it. Second, it cannot account for recent evidence that cognition is not confined prior to movement initiation, but consistently leaks into movement. To address these two features, we propose a novel neurocomputational model of cognition-action interactions, namely the unfolding action model (UAM). Crucially, the model implements adaptive information flow between the last cognitive processing stage and motor implementation. We show that the UAM addresses the two abovementioned features. Empirically, the UAM accounts for traditional response time data, including positively skewed initiation time distribution, functionally fixed decision bounds and speed-accuracy trade-offs in button-press experimental designs. Moreover, it accounts for movement times, movement paths, and how they are influenced by cognitive-experimental manipulations. This move should close the current gap between abstract decision-making models and behavior observed in natural habitats. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Losing the boundary: Cognition biases action well after action selection.

For selecting an action, traditional theories suggest a cognitive architecture made of serial processing units. Others suggested that action selection emerges from the parallel implementation of and competition between multiple action plans. To disentangle these 2 hypotheses, we created a reaching task assessing the temporal dynamics of action selection. Crucially, our design did not force action selection processes to operate in parallel, allowing an informative comparison between the hypotheses. We manipulated the probability of congruence between a cue and a delayed reach target to investigate, in an unbiased way, whether congruence probability interacts with reach trajectory. Our results show that reach trajectories are modulated by the probability of congruence. Hence, action selection is temporally spread, continues after movement onset, and emerges from a competition between multiple afforded action plans, in parallel biased by relevant task factors (e.g., probability of reach).

Setting the stage subliminally: unconscious context effects.

An important approach to understand how the brain gives rise to consciousness is to probe the depth of unconscious processing, thus to define the key features that cause conscious awareness. Here, we investigate the possibility for subliminal stimuli to shape the context for unconscious processing. Context effects have generally been assumed to require consciousness. In the present experiment, unconscious context processing was investigated by looking at the impact of the context on the response activation elicited by a subliminal prime. We compared the effects of the same subliminal prime on target processing when the prime was embedded in different unconscious contexts. Results showed that the same prime can evoke opposite responses depending on the unconscious context in which it is presented. Taken together, the results of this study show that context effects can be unconscious.