The goal of this study was to evaluate the accuracy of several well known tonalness metrics : Tone-to-Noise ratio, Prominence Ratio and Aures subjective model. The first experiment used a magnitude estimation method to evaluate the subjective tonalness of sounds made of a background noise and an emerging tone with different frequencies and emerging levels. An analysis of variance of data showed that each factor was significant, while the interaction between them was not. Aures' model proved to give accurate prediction of the subjective results as well as Prominence Ratio (but this was no longer true for Tone-to-Noise Ratio). In a second experiment, subjects were asked to evaluate the unpleasantness of each sound, using the same procedure. Subjective values could be reasonably well described by adding the DINpenalty to the A-weighted level, though a more accurate model could be built from a linear combination of Aures' tonalness and sharpness. These two experiments were repeated using pluri-harmonic sounds. In that case, DIN DL failed to predict tonalness, while Aures' model was still valid. Once more, this metric, together with sharpness, was also able to predict annoyance. The conclusion of this study is therefore to promote this model instead of TNR and PR.