The current study examines 1) the sensitivity of detection of invasive adenocarcinoma of the cervix in a routine cervical screening service, and 2) the frequency in smears of cytologic criteria previously found to be useful in diagnosis.Data on women with diagnoses of adenocarcinoma of the cervix accessioned at the Western Australian Cervical Cytology Registry during the period 1993-1998 were examined, where smears had been reported by Western Diagnostic Pathology within three years of the biopsy diagnosis. Smears and biopsy material were reviewed.Thirty-six smears from 24 women were reviewed. Of those, 58.3% had been reported as a possible or definite high grade epithelial abnormality (HGEA). On review it was thought that this could be improved to 77.8%. The screening or diagnostic error was thus 19.4% and the sampling error 22.2%. The likelihood of an individual woman receiving a report of a possible or definite HGEA in the three years before biopsy was 83.3%. In retrospect this could have been improved to 91.7%. Heavy bloodstaining with abundant abnormal glandular epithelium (14 smears) and small three-dimensional or papillary clusters (16 smears) were the most frequent clues to invasion. Tumor necrosis/diathesis was present in eight smears, but easily seen in only four, while marked nuclear pleomorphism and macronucleoli were seen in three and one smears respectively. In cases with a discrepancy between the initial and the review findings, very small amounts of abnormal material (three smears), a resemblance to endometrial cells (one smear), and an unusual appearance of folded monolayered sheets (three smears) contributed to the difficulty of diagnosis.There were significant sampling and screening/diagnostic errors (22.2% and 19.4%, respectively). Screening and diagnostic errors could perhaps be reduced by a greater awareness of the range of cytologic changes, but these may be subtle. Heavy bloodstaining with abundant abnormal glandular material may be a useful clue to invasive, rather than in situ, adenocarcinoma, even in the absence of tumor diathesis or fully malignant nuclear criteria.