Epiluminescence microscopy (ELM), or dermatoscopy, is a non-invasive technique for the diagnosis of cutaneous melanoma that may play a role in the non-invasive, preoperative assessment of melanoma thickness. This study investigated the correlation between the frequency of appearance of some standard ELM criteria and the histological thickness of melanomas. In addition, the possible role of the total dermoscopic score (TDS) according to ABCD rule of dermoscopy as a predictor of melanoma thickness was analysed. The dermoscopic images of 84 cutaneous melanomas were retrospectively investigated to evaluate the presence of 10 standard ELM criteria, and for each lesion the TDS was established (with observers blinded as to the tumour thickness). A statistically significant association was found between the presence of an irregular pigment network and melanomas with a Breslow index equal to or lower than 0.75 mm (positive predictive value of 68%); in contrast, radial streaming, atypical vascular pattern and grey-blue areas were associated with melanomas > 0.75 mm (positive predictive values of 77%, 65% and 70%, respectively). Of the melanomas thinner than 0.76 mm, 82% showed a TDS lower than 6.80 (optimized cut-off point), while 79% of melanomas thicker than 0.75 mm had a TDS higher than 6.80 (chi2 = 30.815, P < 0.001); the positive predictive value of a TDS > 6.80 in the detection of lesions thicker than 0.75 mm was 79%. In conclusion, a statistically significant correlation does exist between some dermoscopic features and melanoma thickness. Both the mostly used dermoscopic methods (standard ELM pattern analysis and the ABCD rule of dermatoscopy) may provide useful information in the non-invasive assessment of melanoma thickness. However, their diagnostic performance is far from 100%; further studies are needed to investigate whether the combination of dermoscopy with other non-invasive approaches (e.g. sonometry) may result in an overall improvement in the diagnostic performance.