A model consistent with the data thus far is that WMs transferred between hemispheres recruit the same neural ensembles as memory traces activated by feedforward visual inputs into the same cortical hemisphere. Under this generic ensemble model (Figure 6A), when a given object—say, a banana in the upper location—is held in WM within a cortical hemisphere, it uses the same neural ensemble (i.e., the same pattern of spiking across our electrode arrays) whether it arrived there via feedforward inputs from visual cortex (Figure 6A, left) or via interhemispheric inputs from the contralateral hemisphere (Figure 6A, right). We could test this because the saccade on swap trials brought the remembered sample location to the same retinotopic coordinates where it appeared on the no-swap trials. An alternative model is motivated by the fact that unique prefrontal ensembles are activated by different combinations of input features and task contexts (Rigotti et al., 2013). Perhaps the same information arriving via different circuits—feedforward (Figure 6C, left) versus interhemispheric (Figure 6C, right)—also activates different ensembles. We call this alternative the novel ensemble model.
