In South Africa health care practitioners are commonly professionals who speak only one, or at most two, of the languages spoken by their patients. This provides for language provision challenges, since many patients are not proficient in English or Afrikaans and ad hoc and haphazard arrangements are made for interpreting by untrained personnel. As part of a larger study (conducted in 2010) in a public psychiatric hospital, we report here on the potential consequences for diagnostic assessments of 13 psychiatric evaluations mediated by ad hoc interpreters who were employed as health care workers and household aides. The psychiatric evaluations were recorded and transcribed verbatim. The first author checked for accuracy of transcription and translations, and the two members of the author team who are both senior African language academics rechecked transcription and translation. We used the typology developed by Vasquez and Javier (1991) to study interpreter errors (i.e. omissions, additions and substitutions). All errors were independently rated by a senior psychiatrist and a senior clinical psychologist to determine whether the errors were likely to have a bearing on clinical decisions concerning the patient and to rate whether errors deemed clinically significant contributed to making the patient appear more ill psychiatrically, or less ill. Of the 57 errors recorded, 46% were rated as likely to have an impact on the goal of the clinical session. Raters concurred that the clinically significant errors contributed towards potentially making the patient look more psychiatrically ill. Detailed analyses of evaluations demonstrate the complexity of informal interpreter positioning regarding issues of diagnosis and cultural factors in illness. Evaluations conducted where clinicians and interpreters are not trained in language and interpreting issues may create a distorted picture of the patients' mental health conditions.