Miller's is the most commonly used classification of gingival tissue recessions, defined as the displacement of the soft tissue margin apical to the cemento-enamel junction. However, data on the reliability of this classification are missing so far, although reliability, which reflects the consistency of repeated measurements, is regarded as a prerequisite for judging the utility of a classification. The aim of the present study was to evaluate inter- and intra-observer agreement on Miller's classification of gingival tissue recessions. Two hundred photographs (50 of each region: maxillary/mandibular anterior/posterior teeth) of gingival tissue recessions were evaluated twice by four observers with different degrees of experience in Miller's classification, gingival phenotype, tooth shape, and identifiability of the cemento-enamel junction. The following inter- and intra-observer agreements were found: Miller's classification, 0.72 and 0.73-0.95; gingival phenotype, 0.29 and 0.45-0.58; tooth shape, 0.39 and 0.44-0.59; and identifiability of the cemento-enamel junction, 0.21 and 0.30-0.59. A higher agreement was detected for anterior teeth. Further, gingival phenotype (thin-high scalloping) significantly correlated with tooth shape (long-narrow) (ρ = 0.662, p < 0.001). Miller's classification of gingival tissue recessions was evaluated by four examiners using 200 clinical photographs and yielded substantial to almost perfect agreement, with higher agreement for anterior teeth. Although limited to photographic assessment, the present study offers the so far missing proof on the sufficient inter- and intra-observer agreement of this classification.