Çiçek Güney¹ (), Yasemin Gunindi¹, İlayda Efsane Algın¹, Nihan Alp¹; ¹Sabanci University, Istanbul

Statistical learning enables the visual system to extract regularities from probabilistic structure in the visual environment. When an object repeatedly occurs in a specific place, the visual system can form object-place associations. Yet underlying neural correlates of such statistical regularity remain elusive. Here, we investigate this by separating item-level and pair-level regularities and examining how integration processes evolve across repetitions (1st-10th). To this end, we used frequency-tagging EEG paradigm (object and place tagged at fundamental frequencies: f1 = 7.5, f2 = 6 Hz) and measured Steady-State Visual Evoked Potentials (SSVEP). Participants (N = 20) viewed objects and places, either as repeating object-place pairs (pair-level regularity) or as independently repeating items in different pair combinations (item-level regularity), followed by memory tasks. We quantified Signal-to-Noise Ratios (SNRs) at fundamental frequencies and their harmonics as an index of item-level processing, and their intermodulation (IM: e.g., nf1±mf2) frequencies as an index of pair-level integration processing. As expected, behaviorally, participants showed higher accuracy for pairs learned under pair-level regularity, indicating stronger associative learning. SSVEP results revealed that place-related (mf2) SNRs decreased more strongly within individuals across repetitions in item-level than in pair-level condition, indicating greater item-specific adaptation. Similarly, IM SNRs decreased only in item-level regularity condition. Moreover, across participants, memory association was significantly correlated with IM SNRs only in the pair-level regularity condition, suggesting a link between behavioral performance and SSVEPs. Contrary to our expectations, this correlation was negative (r = -0.53, p = .017). It may suggest that better performers integrate information more efficiently and require less neural effort, resulting in lower overall IM SNRs while maintaining stable IM SNRs across repetitions. Overall, these findings suggest that repeated exposure to item-level regularity may induce neural adaptation that weakens object-place integration. Accordingly, pair-level regularities seem to buffer against the adaptation effects induced by repeated items.

Acknowledgements: N.A. and Ç.G. are supported by Scientific and Technological Research Council of Turkey (TUBITAK 1001) under the Grant Number 122K922. The authors thank to TUBITAK for their support.