Abstract Friedman (1995) presents a review of 6 phonological dyslexic cases in which she suggests that this reading impairment is divisible into subtypes on the basis of two straightforward behavioural profiles: (1) cases with poor function word reading and normal nonword repetition (3 cases) and (2) cases with good function word reading and poor nonword repetition (3 cases). This evidence is interpreted as support for a reading model that does not posit a non-lexical reading route (Harm and Seidenberg, 2001, 2004; Plaut, et al. 1996), with the pattern of impairment in case (1) being due to disruption of direct connections between orthography–phonology and in case (2) being due to a ‘generalised’ impairment of phonology. The current review re-evaluates the validity of Friedman's original criteria as a basis for identifying ‘subtypes’ of phonological dyslexia with a larger sample of 38 documented cases. On balance it is found that the majority does not easily fit either subtype on the basis of the expected behavioural profiles. Thus the present survey would suggest that Friedman's criteria for classification of phonological dyslexia is not particularly useful for the general population of such cases. However, subsequent correlational analyses did find evidence in support of a relationship between nonword reading impairment and nonword repetition/function word reading impairment. This finding is largely consistent with the principles of the type of reading model outlined by Friedman (1995), although other potential accounts cannot be ruled out. These findings are discussed in greater detail with relation to present models of reading and potential avenues for further research are proposed.