One critical question for the front end of models of visual-word recognition and reading is whether the stage of letter position coding is purely orthographic or whether phonology is (to some degree) involved. To explore this issue, we conducted a silent reading experiment in Japanese Kana-a script in which orthography and phonology can be easily separated-using a technique that is highly sensitive to phonological effects (i.e., Rayner's (1975) boundary technique). Results showed shorter fixation times on the target word when the parafoveal preview was a transposed-mora nonword (a.ri.me.ka [(sic)-a.me.ri.ka (sic)) than when the preview was a replacement-mora nonword (a.ka.ho.ka (sic) -a.me.ri.ka [(sic). More critically, fixation times on the target word were remarkably similar when the parafoveal preview was a transposed-consonant nonword (a.re.mi.ka [a,cent a integral not signa integral Ya,<<]-a.ri.me.ka (sic)) and when the parafoveal preview was an orthographic control nonword (a.ke.hi.ka [(sic)]-a.me.ri.ka [(sic)). Thus, these findings offer strong support for the view that letter/mora position coding during silent reading is orthographic in nature.