In the study of perceptual decision making, it has been widely assumed that random fluctuations of motion stimuli are irrelevant for a participant’s choice. Recently, evidence was presented that these random fluctuations have a measurable effect on the relationship between neuronal and behavioral variability, the so-called choice probability. Here, we test, in a behavioral experiment, whether stochastic motion stimuli influence the choices of human participants. Our results show that for specific stochastic motion stimuli, participants indeed make biased choices, where the bias is consistent over participants. Using a computational model, we show that this consistent choice bias is caused by subtle motion information contained in the motion noise. We discuss the implications of this finding for future studies of perceptual decision making. Specifically, we suggest that future experiments should be complemented with a stimulus-informed modeling approach to control for the effects of apparent decision evidence in random stimuli.

Les études neurophysiologiques de la prise de décision, traditionnellement ancrées dans des principes neuro-économiques, ont évoluées pour inclure une variété d’aires du cerveau. Partant d’abord du lobe frontal associé aux jugements de valeur, le champ s’est élargi pour inclure d’autres types de décisions incluant les décisions perceptuelles et les décisions incarnées qui impliquent notamment les aires sensorimotrices du cerveau. La théorie moderne de la prise de décision modèle l’activité neurale dans ces régions comme une compétition entre les différents stimuli et actions considérés par un individu. Cette compétition est résolue lorsque l’activité neurale associée à un stimulus ou une action choisie atteint un seuil critique. Toutefois, il reste à éclaircir comment ce modèle s’applique aux décisions effectuées alors que l’individu est déjà engagé dans une activité. Dans ce mémoire nous examinons ce type de décision chez des sujets humains dans une tâche de suivi continu. Des cibles « choix » apparaissaient sur un écran pendant que le sujet suivait de la main une cible qui se déplaçait doucement en continu. Le sujet pouvait ignorer ces cibles choix, ou abandonner la cible suivie pour toucher une cible choix, dans quel cas la cible sélectionnée devenait la nouvelle cible à suivre du doigt. Tel qu’attendu, nous avons observé que les sujets favorisaient les cibles plus proches, plus grandes, et les cibles alignées avec l’axe du mouvement. Toutefois nous avons été surpris de constater que les sujets ignoraient les coûts énergétiques du mouvement, tel que modélisés. Un biais pour minimiser les coûts du mouvement fut réintroduis lorsque la tâche fut divisée en séries de mouvements point-à-point, plutôt qu’un mouvement continu. Même si nous ne pouvons expliquer ce résultat surprenant, nous espérons qu’il inspire de futures études utilisant le paradigme expérimental de décision durant l’action. / Neurophysiological studies of decision-making have expanded over decades to involve many brain areas. The field broadened from neuroeconomics, mainly concerned with frontal regions, to perceptual or embodied decision-making involving several sensorimotor areas where neural activity is linked to the stimuli and actions necessary for the decision process. Current models of decision-making envision this neural activity as a competition between different actions that is resolved when enough activity favors one over the other. However, it is unclear how such models can explain decisions often present in natural behavior, where deliberation takes place while already engaged in an action. In this thesis, we examined the choices human subjects made as they were engaged in a continuous tracking task. While they were manually tracking a target on a flat screen, subjects were occasionally presented with a new target to which they could freely choose to switch, whereupon it became the new tracked target. As expected, we found that subjects were more likely to move to closer targets, bigger targets, or targets that were aligned to the direction of movement. However, we were surprised that subjects did not choose targets that minimized energetic cost, as calculated by a biomechanical model of the arm. A biomechanical bias was restored when the continuous movement was broken up into a series of point to point movements. While we cannot yet explain these findings with certainty, we hope they will inspire further studies using decide-while-acting paradigms.

Die Fähigkeit, Zustände in der Außenwelt zu beurteilen und zu kategorisieren, wird unter dem Oberbegriff „perzeptuelles Entscheiden“ zusammengefasst. In der vorliegenden Arbeit wurde funktionelle Magnetresonanztomografie mit multivariater Musteranalyse verbunden, um offene Fragen zur perzeptuellen Entscheidungsfindung zu beantworten. In der ersten Studie (Hebart et al., 2012) wurde gezeigt, dass der visuelle und parietale Kortex eine Repräsentation abstrakter perzeptueller Entscheidungen aufweisen. Im frühen visuellen Kortex steigt die Menge entscheidungsspezifischer Information mit der Menge an verfügbarer visueller Bewegungsinformation, doch der linke posteriore parietale Kortex zeigt einen negativen Zusammenhang. Diese Ergebnisse zeigen, wo im Gehirn abstrakte Entscheidungen repräsentiert werden und deuten darauf hin, dass die gefundenen Hirnregionen unterschiedlich in den Entscheidungsprozess involviert sind, je nach Menge an verfügbarer sensorischer Information. In der zweiten Studie (Hebart et al., submitted) wurde gezeigt, dass sich eine Repräsentation der Entscheidungsvariable (EV) im fronto-parietalen Assoziationskortex finden lässt. Ferner weist die EV im rechten ventrolateralen präfrontalen Kortex (vlPFC) einen spezifischen Zusammenhang mit konfidenzbezogenen Hirnsignalen im ventralen Striatum auf. Die Ergebnisse deuten darauf hin, dass Konfidenz aus der EV im vlPFC berechnet wird. In der dritten Studie (Christophel et al., 2012) wurde gezeigt, dass der Kurzzeitgedächtnisinhalt im visuellen und posterioren parietalen Kortex, nicht jedoch im präfrontalen Kortex repräsentiert wird. Diese Ergebnisse lassen vermuten, dass der Gedächtnisinhalt in denselben Regionen enkodiert wird, die auch perzeptuelle Entscheidungen repräsentieren können. Zusammenfassend geben die hier errungenen Erkenntnisse Aufschluss über den neuronalen Code des perzeptuellen Entscheidens von Menschen und stellen ein vollständigeres Verständnis der zugrundeliegenden Prozesse in Aussicht. / Perceptual decision-making refers to the ability to arrive at categorical judgments about states of the outside world. Here we use functional magnetic resonance imaging and multivariate pattern analysis to identify decision-related brain regions and address a number of open issues in the field of perceptual decision-making. In the first study (Hebart et al., 2012), we demonstrated that perceptual decisions about motion direction are represented in both visual and parietal cortex, even when decoupled from motor plans. While in early visual cortex the amount of information about perceptual choices follows the amount of sensory evidence presented on the screen, the reverse pattern is observed in left posterior parietal cortex. These results reveal the brain regions involved when choices are encoded in an abstract format and suggest that these two brain regions are recruited differently depending on the amount of sensory evidence available. In the second study (Hebart et al., submitted), we show that the perceptual decision variable (DV) is represented throughout fronto-parietal association cortices. The DV in right ventrolateral prefrontal cortex covaries specifically with brain signals in the ventral striatum representing confidence, demonstrating a close link between the two variables. This suggests that confidence is calculated from the perceptual DV encoded in ventrolateral prefrontal cortex. In the third study (Christophel et al., 2012), using a visual short-term memory (VSTM) task, we demonstrate that the content of VSTM is represented in visual cortex and posterior parietal cortex, but not prefrontal cortex. These results constrain theories of VSTM and suggest that the memorized content is stored in regions shown to represent perceptual decisions. Together, these results shed light on the neuronal code underlying perceptual decision-making in humans and offer the prospect for a more complete understanding of these processes.

Perceptual decision making can be described as a process of accumulating evidence to a bound which has been formalized within drift-diffusion models (DDMs). Recently, an equivalent Bayesian model has been proposed. In contrast to standard DDMs, this Bayesian model directly links information in the stimulus to the decision process. Here, we extend this Bayesian model further and allow inter-trial variability of two parameters following the extended version of the DDM. We derive parameter distributions for the Bayesian model and show that they lead to predictions that are qualitatively equivalent to those made by the extended drift-diffusion model (eDDM). Further, we demonstrate the usefulness of the extended Bayesian model (eBM) for the analysis of concrete behavioral data. Specifically, using Bayesian model selection, we find evidence that including additional inter-trial parameter variability provides for a better model, when the model is constrained by trial-wise stimulus features. This result is remarkable because it was derived using just 200 trials per condition, which is typically thought to be insufficient for identifying variability parameters in DDMs. In sum, we present a Bayesian analysis, which provides for a novel and promising analysis of perceptual decision making experiments.

Perceptual decision making can be described as a process of accumulating evidence to a bound which has been formalized within drift-diffusion models (DDMs). Recently, an equivalent Bayesian model has been proposed. In contrast to standard DDMs, this Bayesian model directly links information in the stimulus to the decision process. Here, we extend this Bayesian model further and allow inter-trial variability of two parameters following the extended version of the DDM. We derive parameter distributions for the Bayesian model and show that they lead to predictions that are qualitatively equivalent to those made by the extended drift-diffusion model (eDDM). Further, we demonstrate the usefulness of the extended Bayesian model (eBM) for the analysis of concrete behavioral data. Specifically, using Bayesian model selection, we find evidence that including additional inter-trial parameter variability provides for a better model, when the model is constrained by trial-wise stimulus features. This result is remarkable because it was derived using just 200 trials per condition, which is typically thought to be insufficient for identifying variability parameters in DDMs. In sum, we present a Bayesian analysis, which provides for a novel and promising analysis of perceptual decision making experiments.