Purpose: Several classification systems have been proposed in the literature, but none of them has been validated by means of an appropriate statistical analysis and, therefore, it is unknown whether these systems work in the same manner among different clinicians. The aim of the present study is to investigate the intra- and inter-rater agreement of a new system to classify gingival recession defects and to evaluate its agreement among different clinicians. Materials and methods: A new classification system was proposed and tested considering the following three factors: the amount of keratinised tissue (2 or >=2 mm), the presence/absence of noncarious cervical lesions and the presence/absence of interproximal attachment loss. Kappa-statistics were performed to analyse the intra-rater and inter-rater agreement among three blinded examiners. Results: A total of 120 gingival recessions were evaluated using the new classification system. The intra-rater agreement ranged from 0.74 to 0.96 for the variable keratinised tissue, from 0.67 to 0.94 for the variable non-carious cervical lesions and from 0.70 to 0.92 for the variable interproximal attachment loss. The inter-rater agreement values ranged from 0.70 to 0.85 for keratinised tissue, from 0.54 to 0.59 for non-carious cervical lesions and from 0.54 to 0.77 for the interproximal attachment loss. Conclusions: Based on the results of this study, the proposed new classification system showed moderate to substantial agreement among investigators when used in the present study population and was therefore able to diagnose the severity of gingival recession defects. Keywords: classification, diagnosis, gingival recession