The aim of this study was to elucidate the word decoding strategies of Hebrew readers with pre-lingually-acquired, severe hearing impairments as opposed to regular Hebrew readers, and to determine their efficiency. The research paradigm used to clarify these issues asked participants to perform a sequence of three learning tasks, calling for the association of digits with (1) written Hebrew nouns, (2) with pseudo-homophones (PH) of these nouns, and (3) with nonsense words (NW) comprised of the same graphemes as the nouns. Ninety-two secondary school students, half of them hearing impaired (HI), performed the tasks in succession, with the word/digit association task (the primer) always being administered first. For two sub-groups, however, comprised of approximately half of the members of each of the groups, the order of the latter two tasks was reversed. As such, 23 of those who were HI and 24 of those who were not performed the PH/digit association task after the word/digit association task (primed), whereas for the remainder (unprimed), the word/digit association task was immediately followed by the NW/digit association task. It was expected that, for hearing readers but not for HI readers, the existence of homophony between the real words and the PH's in the primed condition would facilitate learning associations for the latter. This hypothesis, however, was not confirmed. Neither of the two primed groups exhibited sensitivity to homophony. This finding is particularly surprising with respect to the primed hearing group, since it totally contradicts current popular theories that assign phonology a central role in the processing of written materials by unimpaired readers.