Human decision making is typically studied in tasks with a rigid trial structure, offering the choice between two (or more) discrete choice options, a narrow response window, and feedback at the end of every trial. Such paradigms contrast with real-life situations in which humans do not face multiple options, but only a single option (or “offer”) at a time, having to decide whether to pursue or abandon it at any moment—a situation studied as “patch leaving” or “foraging” in animal ecology. We designed such a task for humans. Participants entered patches and decided whether to stay and exert effort for the chance of a reward, or to move on to the next patch. Patches varied in quality, but depleted over time. The global distribution of patch quality varied across blocks. Based on past outcomes and block type, participants needed to infer momentary patch quality and decide when to leave. I will show how we capitalized on the rapid series of events in this task both to elucidate neural correlates of hierarchical outcome processing as well as altered effort-persistence coupling in depression. Results from an 7T-fMRI study (N = 30) showed how outcome processing of single reward events is modulated by global environmental statistics. Results from a large online study (N = 461) showed how the amount of effort exertion is coupled to past rewards and used to guide future leave decisions, a link that leads to higher performance and is stronger in individuals with high depressive symptoms.

The holy grail for neurocognitive experiments is a task that is intuitive to play, fast to complete, and yet provides meaningful and reliable readouts of a relevant cognitive process and its neural correlates. We have recently demonstrated how we can get closer to this goal by abandoning the traditional trial structure: using a gamified continuous (trial-free) predictive inference task, we were able to measure EEG signatures of belief updating with high test-retest reliability in as little as 6min of task performance. In this talk, I will present a novel task that similarly uses such a continuous, time-efficient design, but critically to study emotional inference in volatile environments. N=30 (+N=200 online) participants accumulated evidence for changes in affective states expressed by noisy emotional stimuli presented in rapid succession. We found that participants reliably detected changes in valence and arousal levels and were further able to flexibly adapt their integration strategy to the level of volatility and noise in the stimulus stream. Two hours after an intravenous infusion of ketamine, a rapid acting antidepressant, this flexibility in emotional processing was significantly enhanced. Our task is well suited to study emotional inference in changing environments and with interventional designs.

Psychophysical methods are a cornerstone of experimental psychology, widely used to quantify perceptual behaviour. Unfortunately, they can be time-consuming and tedious for participants. Continuous psychophysics changes this by replacing forced-choice decisions with an intuitive tracking task. Since this approach needs fewer instructions and yields faster measurements, it promises to be applicable with untrained participants and in clinical settings. However, the more naturalistic task introduces additional cognitive and motor processes that are typically controlled in classical psychophysics. Rather than treating this as a complication, continuous psychophysics presents an opportunity to jointly study these processes involved in the perception-action loop. The first part of this talk introduces an inverse optimal control framework for fitting optimal feedback control models to continuous psychophysics data, allowing to jointly quantify visual uncertainty, motor variability, behavioural costs, and internal models of target dynamics. The second part extends continuous psychophysics to multisensory integration, asking whether the optimality of visuo-proprioceptive integration generalizes from static position judgments to continuous movements. Participants tracked a randomly moving target using vision, proprioception, or both, while we manipulated visual uncertainty and visuo-proprioceptive alignment. The observed behaviour was consistent with predictions from a multisensory optimal feedback control model: proprioception outperformed vision when cues were aligned, and behaviour was biased toward the more reliable cue and when a discrepancy was introduced. Together, these studies illustrate how continuous psychophysics and inverse optimal control can jointly quantify visual, multisensory, and motor processes from a naturalistic tracking task, offering a powerful and efficient alternative to classical trial-based paradigms.

Goal-directed actions, such as reaching for and grasping a pencil, are fundamental to natural behaviour. Traditionally, visuomotor control of goal-directed behavior has been studied by isolating distinct components of the action task, such as movement planning and execution. Experimentally, the computational and neural mechanisms underlying the control of goal-directed action have been studied in trial-based paradigms, in which movement targets are repeatedly presented in highly reduced virtual environments. However, in naturalistic settings, visuomotor control is shaped by both external stimuli and internal movement goals, requiring the continuous coordination of perceptual, motor, and cognitive processes. In my talk, I will present three different tasks that allow us to study visuomotor control in a continuous way. The challenge in these continuous tasks is to identify and define behaviour-relevant events that can be linked to the observed patterns in movement recordings. Here, a key insight is to shift our view from a stimulus-centred to an action-centred reference frame. Studying goal-directed action in continuous tasks promises to reveal general principles of visuomotor control and further our understanding of visuomotor function in natural behaviour.