Encoding multiple words simultaneously in reading is implausible.

Several prominent models of reading posit that attention is distributed to support the parallel lexical processing of multiple words. We contend that the auxiliary assumptions underlying this attention-gradient hypothesis are not well founded. Here, we address three specific issues related to the ongoing debate about attention allocation during reading: (i) why the attention-gradient hypothesis is widely endorsed, (ii) why processing several words in parallel in reading is implausible and (iii) why attention must be allocated to only one word at a time. Full consideration of these arguments supports the hypothesis that attention is allocated serially during reading.