A scalable touchscreen-based spatial working memory task for cross-species research.

The common marmoset is rapidly emerging as a powerful nonhuman primate model in neuroscience, yet the development of scalable, mechanistically informative cognitive paradigms has lagged behind advances in neural recording and genetic tools. Here, we introduce and validate a touchscreen-based spatial working memory task designed for direct cross-species translation between marmosets and humans. The paradigm independently manipulates retention delay and spatial separation between test choice stimuli, enabling parametric control over maintenance and interference demands within a single framework. Twelve marmosets and seventy-one human participants performed a Delayed Non-Match-to-Position task in which memory delay (1, 5, 10 s) and angular separation between target and distractor locations were systematically varied. Across species, accuracy declined as delay increased and as spatial separation decreased, demonstrating robust sensitivity to both maintenance demands and similarity-based interference. Critically, delay and separation interacted in both species, indicating that these had additive effects. Choice latency analyses further supported interpretation of performance, with slower responses on incorrect trials in both groups. Together, these findings establish a scalable and translationally aligned spatial working memory paradigm that captures interacting maintenance and interference processes. This task provides a powerful platform for circuit-level investigation and offers a sensitive cognitive assay for future studies of aging, neurodegenerative disease, and therapeutic intervention in the marmoset model.