A revised version of the G-mode multivariate statistics has been used to classify the 438 asteroids for which the eight-color photometric data and IRAS albedo are available. At a confidence level of 99.7%, seven taxonomic units of asteroids are separated by the method, while with higher values of the confidence level no separation occurs in the adopted sample. Decreasing the confidence level (i.e., accepting a higher probability of a wrong decision in classifying the asteroids), we obtain a more detailed grouping, which results in a successive subdivision of the first units found. At a confidence level of 97.5%, two groups are added to the original ones: C asteroids are in fact subdivided on the basis of their albedo into three different units. In total, 18 groups of objects can be distinguished, showing small deviations from the trend of the original groups. The method gives a quantitative estimate of the significance of the variables used. A comparison with the results obtained by D. J. Tholen (1984, Asteroid Taxonomy from Cluster Analysis of Photometry, Doctoral thesis, University of Arizona) shows that the nine principal units coincide with the Tholen's A, S, C, D, M, E, BU, G, (Q, R, S) classes, while the other groups are subdivisions of these classes, particularly of the S, D, and B ones. The method allows us to interpret the obtained results in terms of natural processes which characterized the history of asteroid population.