Stable numerosity representations irrespective of magnitude context in macaque prefrontal cortex.

Cognitively demanding tasks require neurons of the prefrontal cortex (PFC) to encode divergent behaviorally relevant information. In discrimination tasks with arbitrary and learned categories, context-specific parameters shape and adapt the tuning functions of PFC neurons. We explored if and how selectivity of PFC neurons to visual numerosities, a 'natural' abstract category, may change depending on the magnitude context. Two monkeys discriminated visual numerosities (varying numbers of dot items) in a delayed match-to-sample (DMS) task while single-cell activity was recorded from the lateral PFC. During a given recording session, the numerosity task was either presented in isolation or randomly intermixed with DMS tasks with line lengths and colors as discriminative stimuli. We found that the context of numerosity discriminations did not influence the response properties of numerosity detectors. The numerosity tuning curves of selective neurons, i.e. the preferred numerosity and the sharpness of tuning, remained stable, irrespective of whether the numerosity task was presented in a pure numerosity block or a mixed magnitude block. Our data suggest that numerosity detectors in the PFC do not adapt their response properties to code stimuli according to changing magnitude context. Rather, numerosity representations seem to rely on a sparse and stable 'labeled line' code. In contrast to arbitrarily learned categories, numerosity as a 'natural' category may possess a privileged position and their neuronal representations could thus remain unaffected by magnitude context.