Conformation traits are very important in the Pura Raza Español (PRE) horse. Although most of them have medium to high heritability, there is a great interest in obtaining highly reliable estimates of heritabilities. The objective of this study was to estimate the heritabilities (h²) of conformation traits in PRE horses with three different approaches. The dataset consisted of 7152 morphological records and 41,888 animals in the pedigree, obtained from the Royal National Association of Spanish Horse Breeders (ANCCE). A total of 4057 animals were genotyped with 60,136 SNPs and included in the analysis. Measurements from six zoometric traits were analysed: height at withers (HW), scapular–ischial length (SiL), length of back (LB), dorso-sternal diameter (DsD), thoracic perimeter (TP), and perimeter of anterior cannon bone (PACB). Firstly, heritabilities were estimated with a classical approach with pedigree-based REML and single-step GREML with the HiBlup v1.3.1 software. Then, SNP-based heritabilities were estimated using GREML, accounting for patterns of regional linkage disequilibrium (LD) with the LDAK 5.2 software program. The largest heritability estimates were obtained when we employed REML or ssGREML analysis in this population. Heritabilities were similar for all traits for REML and ssGREML, and their estimates oscillated between 0.33 (0.03) for DSD and PACB and 0.77 (0.035) for HW. Except for HW, the SNP-based heritabilities were smoother than those from REML and ssGREML and they varied between 0.21 (0.03) for DsD and 0.49 (0.031) for SiL. It is also notable that in the LDAK model, the h² estimation improved in most of the analysed traits, which was reflected in the decrease in the standard errors. These results suggested that in the case of these traits where there is high LD, it is necessary to adjust for the LD to correct the overestimation of heritability.