This study is the first stage of a wider enquiry into alternative ways of assessing the readability of specialist texts. The interest in assessing these texts arose from the need to grade 60 medical journal articles for an individualized English-as-a-Foreign-Language reading scheme for doctors. The study reports on an investigation of subjective judgements of difficulty by "expert" raters. This involved the identification of possible components of difficulty and their independent assessment and scoring by five raters. Subsequent analysis focused on the structure and reliability of these judgements. Preliminary results of the data analysis indicate that four out of the seven components possessed satisfactory levels of inter-rater reliability and that syntactic and lexical difficulty as assessed by the raters may be the best predictors of overall difficulty. Finally, there is statistical evidence that the putative judgement by the raters of seven discrete components may be more adequately modelled as the assessment of two "latent" components of difficulty. (Author/JL)