Abstract A test frequently used to complement endodontic diagnoses is the cold test; however, in the last 20 years, authors have reported incorrect results within pulp sensitivity tests. Specifically, a high frequency of false results in posterior teeth, were found. The aim of this study was to identify the most appropriate site for the cold test in molar teeth with a need for endodontic treatment, calculating predictive values, accuracy and reproducibility. A cross-sectional study was performed, evaluating 390 subjects. A total of 152 subjects of both genders from the ages of 15-65 years old participated. The ideal standard was established by direct pulp inspection, and the cold test agent used was 1,1,1,2-tetrafluoroethane. The patients were divided into four groups in relation to the molar tooth: (1) mandibular first molar, (2) mandibular second molar, (3) maxillary first molar, and (4) maxillary second molar. 169 teeth and 676 sites were studied. (a) The most appropriate sites for cold test were the middle third of the buccal surface and cervical third of the buccal surface in the mandibular molars with the following results: Middle third of the first molar: Accuracy 0.93, positive predictive value 0.90 and negative predictive value 0.96. Middle third of the second molar: Accuracy 0.93, positive predictive value 1.00 and negative predictive value 0.90. In relation to third cervical the results were: First molar: Accuracy 0.93, positive predictive value 0.89 and negative predictive value 0.97 y second molar: Accuracy 0.93, positive predictive value 1.00 and negative predictive value 0.90. (b) The highest reproducibility was observed in the middle third of the buccal surface with cervical third of the buccal surface in the mandibular second molar (1.00). The most appropriate site and reproducibility of the sites are auxiliary to complement endodontic diagnose with the cold test.