Performance was examined in a task requiring the reconstruction of spatial locations. Previous research suggests that it may be necessary to differentiate between memory for smaller and larger numbers of locations (Postma & DeHaan, 1996), at least when locations are presented simultaneously (Igel & Harvey, 1991). Detailed analyses of the characteristics of performance showed that such a differentiation might also be required for sequential presentation. Furthermore the slope of the function relating each successive response to accuracy was greater with 3 than with 6, 8, or 10 locations that did not differ. Participants also reconstructed the arrays as being more proximal than in fact they were; sequential presentation eliminated this distortion when there were three but not when there were more than three locations. These results support the idea that very small numbers of locations are remembered using a specific form of representation, which is unavailable to larger numbers of locations.