But that seems fairly obvious, at least to me. Face recognition seems to be an extremely fundamental capability that has probably existed in some form almost as long as brains and image-forming eyes themselves. In RFS’s case, his brain’s ”face module” spotted the faces, his damaged ”number module” couldn’t converge to anything useful, and the jumbled mess for whatever reason overrode the faces in higher-level perception fusion processes. Possibly because the faces were so small so were lost in the noise, so to speak. Would be interesting to know how RFS would have perceived a face with small numbers drawn on it.
In any case, it is pretty well understood that our conscious visual perception is almost entirely unlike the raw stream of nerve signals emitted by the retina. We’re completely unaware of the huge amount of filtering, denoising, smoothing, stabilization, spatiotemporal fusion, extrapolation, interpolation, object detection/classification/conceptualization/contextualization/prioritization, and other processing that is done to create what we consciously perceive.