Daun Jung¹ (), Junghyun Park¹, Sang Chul Chong¹; ¹Yonsei University

Ensemble perception enables efficient extraction of statistical structure. Prior work has suggested that explicit and implicit ensemble tasks rely on distinct representations (Hansmann-Roth et al., 2021), but this claim was based on paradigms that differed in both stimuli and computational demands. Using identical stimulus arrays and a large sample, we tested whether the two tasks relied on a shared representation. Ninety-six participants completed the explicit and implicit tasks in separate sessions spaced 1–3 days apart, with session order counterbalanced across participants. Each pair consisted of a Priming and a Following streak (5–7 trials each). Priming distractors were sampled from either a broad Gaussian or a uniform distribution and Following distractors from a narrower Gaussian with nine levels of mean shift (−80° to 80°). In the implicit task, we replicated the characteristic RT–distribution-distance function reported by Chetverikov et al. (2016), confirming robust sensitivity to distractor statistics. More importantly, in the explicit task the accuracy of judging ‘whether a change occurred’ showed an intersecting pattern as a function of |Δμ| depending on the form of the priming distribution (Uniform vs. Gaussian). This hit-rate curve closely resembled a theoretical distribution-overlap distance, suggesting that explicit judgments are sensitive not only to mean differences but also nonlinear similarity arising from the overall distribution shape. We also computed within-participant correlations between explicit and implicit reaction times. After Fisher-z transformation, the mean correlation was significantly negative, revealing weak but reliable covariation, consistent with partially shared noise sources. These results challenge claims of dissociating explicit and implicit ensemble representations. Explicit judgments also incorporate information from the full feature distribution. Overall, the findings indicate that both tasks rely on overlapping probabilistic representations, suggesting that explicit and implicit tasks are two different ways of accessing the same ensemble representation.

Acknowledgements: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT)(RS-2022-NR070542).