A study of the relations between the perceptive similarity of voices and physical measures of the speech signal has been carried out. The goal of this study was to identify a set of acoustic parameters which contribute most to the voice discrimination process. We worked on a corpus of logatoms CVCV. Ten male speakers uttering nine items were considered and a single sample of each item was used. We collected dissimilarity judgments (direct estimation) for every distinct pair of speakers out of sixteen listeners. The experimental procedure was designed to study the effects of item and intrapair order as well as listener effect. In addition, we measured thirteen parameters on the speech signal: fundamental and formant frequencies of the vocalic segments of each item, along with the durations of each different segment. The correlation and regression analyses of the three INDSCAL dimensions - which explained 81.5 % of the perceptive data variance - based on the acoustic variables reveal that the first perceptive discriminant axis is associated with the three formant frequencies of speech vocalic segments as well as with durational parameters. The second perceptive axis is associated with the fundamental frequencies at the center of each vowel. Finally, a third stage in the perceptive discrimination involves duration measures again.