Temporal coding carries more stable cortical visual representations than firing rate over time

The brain's ability to stably represent recurring visual scenes is crucial for behavior. Previous studies have used slow dynamic (1-5 seconds) rate code measurements to study visual tuning, revealing varying degrees of gradual activity changes over time or "representational drifts." However, it remains unclear if there is an underlying neural code that maintains the encoding of information stable over time. In this study, we extracted structures in fast (tens of milliseconds) temporal responses and explored the role of such temporal codes in supporting the stability of visual representations. We tracked the spiking activity of the same visual cortical populations in male mice for 15 consecutive days using custom-developed, large-scale, ultraflexible electrode arrays. Across various types of stimuli, we found that neurons exhibited varying degrees of day-to-day stability in their firing rate-based tuning. The across day stability correlated with tuning reliability. Notably, accounting for spiking temporal dynamics increased single neuron tuning stability, especially for less reliable neurons. Temporal coding further improved population representation discriminability and decoding accuracy. The stability of temporal codes was more correlated with network functional connectivity than rate coding. These results show that temporal coding is crucial for stably encoding sensory stimuli, suggesting its significant role in ensuring consistent sensory experiences.