Chomsky and Halle claim that an orthography based on their underlying phonological representations (UPR) of lexical items would be optimal for English. This paper challenges three of Chomsky and Halle's basic phonological assumptions: that their vowel shift rule is valid; that the UPR is the only sound representation to be listed in the lexicon; and that derived words do not appear as wholes in the lexicon. A less abstract phonological representation level based on the conscious perceptions of speakers, the surface phonemic (SPR), is proposed. An SPR-based orthography has advantages which a UPR-based orthography would not: it is easy to learn and teach, it can be learned at an early age, and it permits rapid detection of rhyme. It is concluded that an orthography based on SPRs, and not UPRs, would be optimal.