Borderline nuclear change; can a subgroup be identified which is suspicious of high-grade cervical intraepithelial neoplasia, i.e. CIN 2 or worse? Only 10% of first borderline smears are associated with a histological high-grade (HG) abnormality, i.e. CIN 2,3, invasive malignancy or glandular neoplasia on subsequent investigation. The advantages of highlighting this subgroup are obvious but is this possible? From 1996 and 1997, 242 borderline smears with histological follow-up were examined by two independent experienced observers (observer 1 and 2) without prior knowledge of further investigation results. For each smear a profile of nuclear details was produced, also noting the type of cell mainly affected by the process; then the observers were asked to assess the degree of worry of HG disease for each smear i.e. whether the smear fell into group 1 borderline changes indicative of low-grade (normal, inflammatory, CIN1/HPV) disease (BL/LG) or group 2 difficult borderline smear, HG disease (CIN 2,3, invasive neoplasia or glandular neoplasia) cannot be excluded (BL/HG). Observer 1 selected a group of BL/HG with a PPV for HG disease of 38%, with observer 2 having a PPV of 50%; this compared with the overall laboratory HG disease PPV for borderline smears of 14%. Both observers found the most useful criterion to be the increase in nuclear:cytoplasmic ratio. Our results show that it is possible to separate a small group of borderline smears which should be classified as 'borderline/high grade lesion difficult to exclude' (BL/HG). Both observers had some success in arriving at this classification although their method of selecting out this group was quite different.