The Arabic version of the Hospital Anxiety and Depression (HAD) scale was retested and cut-off points determined in a sample of 217 patients attending a primary health care centre in Al Ain, United Arab Emirates (U.A.E.). Subjects were screened using the HAD scale and all patients were then interviewed by a single consultant psychiatrist. The scale scores were assessed against the psychiatrist's clinical evaluations. The study furnished evidence that the Arabic version of the HAD scale is a valid instrument for detecting anxiety and depressive disorders in primary health care settings. Spearman rank correlations of all items of the scale were significantly above zero. The butterflies item of the anxiety subscale had the lowest correlation coefficients. The overall Cronbach alpha measures of internal consistency were 0.7836 and 0.8760 for anxiety and depression, respectively. The cut-off points that produced a balanced combination of sensitivity and specificity appropriate for referral to a psychiatric facility by the general practitioner were 6/7 for anxiety and 3/4 for depression. Almost all other similar studies have determined a single cut-off point for both subscales of the HAD. This study also indicated that the HAD depression subscale is more consistent and more predictive than the HAD anxiety subscale. Moreover some of the problems arising from applying psychiatric research instruments across cultures are highlighted by this study.