To investigate the use of several preferred retinal loci (PRL) when attempting to read, two patients with bilateral central scotomas were asked to decipher single words, successively projected onto the retina using a scanning laser ophthalmoscope (SLO). Video-recordings of the fundus image, on which the projected targets were superimposed, were analyzed frame by frame. One patient used two PRL in association and the other used three, each PRL having a specific function. Single word reading made it easier than with full texts to correlate the images parts scrutinized and the retinal areas involved. Then, as patients were unable to describe their reading behavior, the examiner monitored refixation movements using the SLO and asked questions to help them to become aware of their reading behavior. Eventually, they could localize their PRL, describe their specific functions, and switch at will between them.