Decoding location-specific and location-invariant stages of numerosity processing in subitizing

Extracting the number of objects in perceived scenes is a fundamental cognitive ability. Number processing is proposed to rely on two consecutive stages: an early object location map that captures individuated objects in a location-specific way and a subsequent location-invariant representation that captures numerosity at an abstract level. However, it is unclear whether this framework applies to small numerosities that can be individuated at once ("subitized"). Here, we reanalyzed data from two electroencephalography (EEG) experiments using multivariate pattern decoding to identify location-specific and location-invariant stages of numerosity processing in the subitizing range. In these experiments, one to three targets were presented in the left or right hemifield, which allowed for decoding target numerosity within each hemifield separately (location specific) or across hemifields (location invariant). Experiment 1 indicated the presence of a location-specific stage (180-200 ms after stimulus), followed by a location-invariant stage (300 ms after stimulus). A time-by-channel searchlight analysis revealed that the early location-specific stage is most evident at occipital channels, whereas the late location-invariant stage is most evident at parietal channels. Experiment 2 showed that both location-specific and location-invariant components are engaged only during tasks that explicitly require numerosity processing, ruling out automatic, and passive recording of numerosity. These results suggest that numerosity coding in subitizing is strongly grounded on an attention-based, location-specific stage. This stage overlaps with the subsequent activation of a location-invariant stage, where a full representation of numerosity is finalized. Taken together, our findings provide clear evidence for a temporal and spatial segregation of location-specific and location-invariant numerosity coding of small object numerosities.