Anaerobic threshold (AT) determination by gas exchange measurements has to date been performed mostly by visual inspection from plots of ventilatory and gas exchange indexes. Owing to its subjective nature, this approach appears to be somehow arbitrary. Moreover, it still remains undetermined which of the commonly used gas exchange indexes is most reliable in detecting the anaerobic threshold. In the present study we have compared the AT values provided by the gas exchange indexes in common use, each one processed by a computerized method, based on a 2-segments linear regression model. A quantitative analysis of their reproducibility has also been performed. By comparison, the same has been done by using the conventional technique based on visual inspection of plots of the same gas exchange indexes, made by 2 independent observers. The aim of the study was to evaluate whether an objective method of AT detection improves the reliability of anaerobic threshold determination by gas exchange measurements. Eight healthy male subjects performed 2 incremental exercise tests at exhaustion with a 2-week interval. Respiratory exchange data were collected on a breath-by-breath basis. The following gas exchange indexes have been employed: V-slope; VE vs VO2; VE/VO2 vs VO2 or time; PetO2 vs VO2 or time. The latter 2 have been examined coupled with the VE/VCO2 and PetCO2 curves, respectively. By employing the computerized method, there were no significant differences in AT detection among the 4 indexes. The best agreement was between V-slope and VE vs VO2 (delta 15.8 ml/min).(ABSTRACT TRUNCATED AT 250 WORDS)