Lunedì 27 aprile ore 10.00
Sala Lauree di Psicologia Edificio Agorà U6 piano 3
Dr.ssa Alessia Tonelli
Perception relies on contextual information to stabilize inherently noisy sensory inputs. A central question is whether these contextual influences reflect shared mechanisms across senses or emerge from modality- and domain-specific processes. This seminar presents a series of behavioral and modeling studies investigating how different forms of contextual information shape perception across sensory modalities and perceptual domains. Two major sources of context are considered: statistical context, reflected in central tendency, and temporal context, reflected in serial dependence. According to studies, these effects show clear dissociation. Central tendency systematically scales with sensory uncertainty, consistent with Bayesian predictions, whereas serial dependence operates independently and does not track uncertainty, indicating distinct underlying mechanisms. Crucially, the expression of these effects depends on both sensory modality and task domain. In temporal and spatial estimation, contextual influences follow task-dependent sensory dominance: audition drives temporal estimates, while vision dominates spatial processing, with no evidence for a global supramodal prior. In speed perception, this organization becomes more complex: central tendency shows cross-modal consistency, suggesting partially shared mechanisms, whereas serial dependence remains modality-specific.
Finally, studies in visually impaired individuals provide insight into the developmental origins of contextual mechanisms. Central tendency is preserved even in early blind individuals, indicating that statistical contextual effects can emerge independently of visual experience. However, these individuals do not show optimal, Bayesian use of prior information, suggesting a dissociation between the presence of contextual biases and their functional benefit.
In contrast, serial dependence shows a different pattern: it is robust in simple spatial tasks but selectively altered under specific conditions, while central tendency emerges primarily when sensory uncertainty increases. Together, these findings further support a dissociation between statistical and temporal contextual mechanisms.
Together, these findings support a framework in which perceptual inference relies on multiple interacting priors that are flexibly weighted depending on sensory reliability, task demands, and developmental history.