Uri Maoz

Decades of research have shown that early ramping activity is a very reliable antecedent of self-initiated movement. However, the dominant paradigm has been to only analyze data epochs culminating in movement, which discounts how often slow ramping might occur when no movement follows. To address this, we introduced a matched control condition that did not culminate in movement. We recorded electroencephalography data and compared these two conditions using a variety of powerful machine-learning classifiers. Although early ramping activity was present, it did not predict movement. Instead, classification accuracy rose abruptly about 100 ms before movement. By contrast, the traditional approach reproduced the spurious impression of early predictability. Our results resolve a long-standing controversy, showing that the neural commitment to act is a late-stage event consistent with the timing of conscious intention and the ability to inhibit movement. More broadly, our results show that a reliable antecedent is not necessarily a good predictor, underscoring the need for proper control conditions in time series analyses.

Respiratory processes are increasingly implicated in shaping neural activity and behavior. Recent studies have reported a coupling between respiratory phase and the cortical readiness potential (RP), suggesting that breathing may modulate the neural processes preceding voluntary action. Here, using electroencephalography recordings in humans, we re-examine this claim using the original dataset and a new independent dataset, as well as in simulated data with no coupling. We show that the reported association arises from a confound: both RP amplitude and respiratory phase are coupled to movement onset. The original analysis does not control for this dependency, leading to a spurious effect that is also observed in simulated data. When trials are instead grouped by respiratory phase at the time of movement, thereby controlling for this confound, the apparent coupling disappears. Across datasets, Bayesian analyses provide evidence for the absence of an effect under natural breathing conditions. These findings indicate that respiratory phase does not directly modulate RP amplitude during spontaneous behavior. More broadly, they reveal a shortcoming of phase-amplitude coupling analyses applied to epoched data with slowly varying signals, and highlight the importance of controlling for shared dependencies when interpreting physiological–neural relationships.

Objective. Accurate and reliable neural decoding of locomotion holds promise for advancing clinical applications such as rehabilitation and prosthetic control, as well as for understanding neural correlates of action. Recent studies have demonstrated successful decoding of locomotion kinematics across species in motorized treadmill settings. However, efforts to decode locomotion speed directly and continuously in more natural contexts—where pace is self-selected rather than externally imposed—are scarce, and those that exist generally achieve only modest accuracy and require intracranial implants. Here, we aim to decode self-paced locomotion speed non-invasively and continuously using cortex-wide electroencephalography (EEG) recordings from rats. Approach. We introduce an asynchronous brain–computer interface (BCI) that processes a continuous stream of 32-electrode skull-surface EEG recordings (0.01–45 Hz) to decode instantaneous speed readouts from a non-motorized treadmill during self-paced locomotion in head-fixed rats. Using recurrent neural networks and a large dataset comprising over 133 h of recordings, we trained decoders to map ongoing EEG activity to treadmill speed. Main results. Our decoding methodology achieves a correlation of 0.88 ( R 2 = 0.78 ) for speed, primarily driven by visual cortex electrodes and low-frequency ( < 8 Hz) oscillations. Moreover, pre-training on a single recording session permitted decoding on other sessions from the same rat, suggesting the presence of uniform neural signatures of locomotion that generalize across sessions but fail to transfer across animals. Finally, we found that cortical states not only carry precise information about the current speed, but also about future and past dynamics, extending up to 1000 ms. Significance. These findings demonstrate that self-paced locomotion speed can be decoded accurately and continuously from non-invasive, cortex-wide EEG. Our approach may provide a useful framework for developing high-performing, non-invasive BCI systems for locomotion and contribute to understanding distributed neural representations of action dynamics.

The distinction between intention and foresight is critical for law and morality. For example, a firefighter who forces a hose through a parked car to reach a hydrant may foresee the damage without intending it—unless the car belongs to a rival, in which case the damage may be intended after all. This study aimed to disentangle the neural correlates of intention and foresight using a novel paradigm. Twenty-five participants completed a task inspired by the game Frogger, where hitting objects could be desirable (earning points) or inconsequential. Outcomes varied by likelihood (high vs. low) and actuality (hit vs. miss), resulting in a 2×2×2 design: intention (high/low) × outcome (hit/miss) × expectedness (expected/unexpected). Electroencephalography (EEG) revealed that feedback-related negativity (FRN) amplitude was larger for participants during high-intention than low-intention outcomes while the P3 potential was larger for unexpected than expected outcomes. However, neither signal exclusively reflected intention or foresight; the FRN also indexed expectedness, and the P3 reflected intention. To address this overlap, we used exploratory effect-matched spatial (EMS) filtering, which identified a widespread brain-activity pattern that better dissociated intention from foresight. These findings enhance our understanding of the neural processes underlying complex decision-making, with potential applications in ethical and legal frameworks.

Current theoretical accounts of perception and high-level cognition suggest that awareness plays an active role in disambiguating incoming sensory information. However, the relationship between ambiguity resolution and conscious access remains unclear, partially due to a lack of quantifiable measures of ambiguity. Here, we describe a novel paradigm designed for testing whether more ambiguous stimuli would enjoy preferential access to awareness, as indexed by the time it takes them to break interocular suppression in the breaking continuous flash suppression paradigm. In a series of three experiments, we found that stimuli’s mid-level perceptual features (most likely, visual symmetry levels), rather than their ambiguity, facilitated access to awareness. We therefore propose that such features can drive preferential access to awareness and hypothesize that the potential effect of symmetry might be driven by information redundancy due to the invariance of symmetric patterns under geometric transformation.

Historically, voluntary action and volition more generally have been investigated through the lens of meaningless decisions. Importantly, these findings have been used in the debate about key notions like free will and moral responsibility. However, more recent claims have challenged the possibility of generalizing findings from a meaningless context to a more meaningful one. The current study investigates the markers of volition, specifically comparing meaningful and meaningless decisions. In an effort to maximize their monetary gain, 50 participants repeatedly deliberated between two options, making either rewarded choices—hard-deliberation decisions (where the options differed along two dimensions) or easy-deliberation decisions (where the options differed along a single dimension)—or unrewarded choices, a.k.a. arbitrary decision. This enabled us to contrast rewarded and unrewarded decisions as well as the degree of deliberation between easy- and hard-deliberation choices. We found evidence that rewarded and unrewarded decisions differed along several measures related to volition: participants reported a higher sense of volition, exhibited a stronger Readiness Potential, had increased temporal binding (mostly inconclusive), and demonstrated increased Effort Exerted in the rewarded condition. In contrast, we found evidence for similarity across these measures between easy-deliberation and hard-deliberation conditions. Our results suggest that it is not the complexity of the deliberation process prior to the action that makes it more volitional, but rather that the decision serves a meaningful goal. Our study also introduced a new implicit measure of volition— effort exerted—that well aligned with other measures of volition and should therefore prove useful in future studies.

Artificial intelligence (AI) systems assist humans to predict future outcomes by analyzing large, heterogeneous data of past patterns. In particular, Decision Support Systems (DSSs) leverage AI to analyze vast amounts of information and provide insights quickly, making them particularly attractive to professionals who routinely make high-stakes decisions under time pressure. DSSs are used in various fields—including medical diagnostics, insurance, and criminal justice—and are increasingly incorporating Large Language Models (LLMs). But for AI-powered DSSs to be effective in assisting professionals, their outputs as well as inner workings must be intelligible. In this paper, we argue that “Interpretable” AI models, i.e. models whose inner workings are simple enough for humans to comprehend, are better suited to provide the understanding users need compared to more complex and opaque models, whose outputs are ‘explained’ through so-called “Explainable AI” (XAI) tools. Professionals need to be able to communicate the reasons for their decisions to patients, clients, etc., and while explanations constructed through XAI techniques do not necessarily provide users with the kind of information required for reason-giving, interpretable models do.

Historically, voluntary action and volition more generally have been investigated through the lens of meaningless decisions. Importantly, these findings have been used in the debate about key notions like free will and moral responsibility. However, more recent claims have challenged the possibility of generalizing findings from a meaningless context to a more meaningful one. In this context, the current study investigates the markers of volition, specifically comparing meaningful and meaningless decisions. In an effort to maximize their monetary gain, 50 participants repeatedly deliberated between two options, making either rewarded choices—hard-deliberation decisions (where the options differed along two dimensions) or easy-deliberation decisions (where the options differed along a single dimension)—or unrewarded choices, a.k.a. arbitrary decision. This enabled us to contrast rewarded and unrewarded decisions as well as the degree of deliberation between easy- and hard-deliberation choices. We found evidence that rewarded and unrewarded decisions differed along several measures related to volition: participants reported a higher Sense of Volition, exhibited a stronger Readiness Potential, had increased Temporal Binding, and demonstrated increased Effort Exerted in the rewarded condition. In contrast, we found evidence for similarity across these measures between easy-deliberation and hard-deliberation conditions. Our results suggest that it is not the complexity of the deliberation process prior to the action that makes it more volitional, but ratherthat the decision serves a meaningful goal. Our study also introduced a new implicit measure of volition—exerted effort—that well aligned with other measures of volition and should therefore prove useful in future studies.