The aim of the study was to compare the accuracy of self-administered vaginal tampon (VT) specimens for the detection of human papillomaviruses (HPVs) with that of cervicovaginal lavage specimens (CVL). Two hundred seventy-four paired VT and CVL specimens were collected prospectively from women at risk of sexually transmitted diseases. Specimens were treated and amplified with the polymerase chain reaction (PCR). Each woman served as her own control. One hundred and forty-four of 272 (52.9%) CVLs and 159 of 271 (58.7%) VTs contained HPV DNA sequences (correlation of 88%). The sensitivity and specificity of vaginal tampons reached 93.9% (138/147) and 80.5% (99/123), respectively. HPV typing results were concordant for 99 negative paired samples and 114 paired samples positive for the same type(s) (correlation of 78.9%). It is concluded that these sampling methods collect cells from different areas of the genital epithelium, highlighting the importance of further assessment of the comparative predictive value of HPV detection in each sample.