Roughness-tactile perception is part of activity assessment in initial-caries-lesions. Hypothesizing that a probe's design influences this examiner's assessment, four probes were designed. The aims of this study were to select the probe with highest inter-/intra-examiners' roughness-assessment agreement and to determine its diagnostic accuracy on artificial initial-caries lesions. A pilot study was conducted with trained dentists to select one controlled-pressure probe design (n = 4) by assessing roughness on known-roughness metal plaques with 5-point Likert scale. Diagnostic accuracy of roughness assessment was conducted with the selected controlled-pressure probe and the WHO-probe on sound and artificial initial-caries-lesion (n = 20) human enamel blocks. Intra-class correlation coefficients (ICCs) and quadratic weighted-Kappa scores were used to assess examiners' reproducibility and Multilevel Poisson models to determine diagnostic accuracy between both probes controlling for confounding variables. The probe design with the highest inter/intra-examiner's agreement (ICC = 0.96) was selected for subsequent analyses. Unadjusted sensitivity, specificity and accuracy values were for the controlled-pressure and the WHO probes: 71.1%,90.6%,81.2%, and 67.4%,84.6%,75.8%, respectively (p > 0.05). Examiner remained the most important factor influencing diagnostic accuracy. While this study did not show significantly higher diagnostic accuracy of the designed controlled-pressure vs. the WHO-probe when used by trained dentists, all over roughness-assessment accuracy and reproducibility were high.