Contextual Scaling of Representational Uncertainty for Size and Number

Accurate discrimination of magnitudes, such as object size and quantity, is crucial for effective action selection and decision-making. A fundamental principle in magnitude discrimination is that discriminability is determined by the ratio of the stimulus values being compared, as formalized by Weber-Fechners law. Here, we show evidence that the discriminability of numerosity is not as robust as previously thought; rather, it is influenced by task-irrelevant environmental statistics. We measured discrimination thresholds in 24 participants performing a numerosity discrimination task in which pairs of dot arrays were presented either simultaneously or sequentially. Stimulus values were sampled from either narrow or wide ranges. Our findings revealed that, although the magnitude pairs were identical, discrimination performance was significantly better (i.e., lower threshold) when stimulus values were sampled from a narrow-range compared to a wide-range. Strikingly, this improvement was observed only when the stimuli were presented sequentially. A similar pattern emerged in a size discrimination task with analogous manipulations. A control experiment ruled out potential confounding variables related to stimulus variety, emphasizing the importance of the stimulus range in gaining this benefit. Together, these findings suggest that perceptual variability is flexibly scaled according to context, particularly when the representations are less reliable. Our study highlights the adaptive nature of stimulus encoding strategies, whereby the brain minimizes representational uncertainty by dynamically leveraging contextual information.