This study compared dentine demineralization induced by in vitro and in situ models, and correlated dentine surface hardness (SH), cross-sectional hardness (CSH) and mineral content by transverse microradiography (TMR). Bovine dentine specimens (n = 15/group) were demineralized in vitro with the following: MC gel (6% carboxymethylcellulose gel and 0.1 m lactic acid, pH 5.0, 14 days); buffer I (0.05 m acetic acid solution with calcium, phosphate and fluoride, pH 4.5, 7 days); buffer II (0.05 m acetic acid solution with calcium and phosphate, pH 5.0, 7 days), and TEMDP (0.05 m lactic acid with calcium, phosphate and tetraethyl methyl diphosphonate, pH 5.0, 7 days). In an in situ study, 11 volunteers wore palatal appliances containing 2 bovine dentine specimens, protected with a plastic mesh to allow biofilm development. The volunteers dripped a 20% sucrose solution on each specimen 4 times a day for 14 days. In vitro and in situ lesions were analyzed using TMR and statistically compared by ANOVA. TMR and CSH/SH were submitted to regression and correlation analysis (p < 0.05). The in situ model produced a deep lesion with a high R value, but with a thin surface layer. Regarding the in vitro models, MC gel produced only a shallow lesion, while buffers I and II as well as TEMDP induced a pronounced subsurface lesion with deep demineralization. The relationship between CSH and TMR was weak and not linear. The artificial dentine carious lesions induced by the different models differed significantly, which in turn might influence further de- and remineralization processes. Hardness analysis should not be interpreted with respect to dentine mineral loss.