In this study, 31 stem taper functions from three model groups (simple polynomial, segmented and variable-form stem taper models) were analysed for Bornmullerian fir (Abies nordmanniana subsp. bornmulleriana Mattf.). Diameter at breast height, total height and diameters outside bark at intervals of 1 m from the stump to the top of the tree were measured on 115 sample trees. Taper functions were fitted with data from 85 trees (76% of total data set) using a non-linear least squares method. Data from the other 30 trees (24% of total data set) were used as the validation data. Four statistical criteria (per cent variation explained, bias, absolute bias and mean squared error) were used to determine model accuracy. Box plots of the residual distributions were also produced to compare statistical accuracy by relative height class. Minimum residuals were obtained with the variable-form stem taper models at all stem parts. When the models were compared within their respective groups, the most accurate simple polynomial, segmented and variable-form models were Demaerschalk-II (1972), Farrar (1987) and Kozak (1988), respectively. Based on the statistical and graphic analysis, the model by Kozak (1988) was recommended for A. nordmanniana subsp. bornmulleriana in the Black Sea region of Turkey. The Demaerschalk-II (1972) model could also be used for practical purposes. [ABSTRACT FROM AUTHOR]