2.6. An improved Hubble system?
Most observers would agree that (locally) the Hubble system supplies a fairly complete framework for classifying galaxies in the field - it seems that over 90% of local field galaxies find a natural home within the system. Furthermore, discussion of the previous section argues rather forcefully for a close connection between the morphological bins of the Hubble system and an underlying physical order. The major criticism I would level against the ranking implicit in the Hubble sequence is that the classification of the early-type portion of the diagram (based on apparent eccentricity) seems to me rather unphysical. Furthermore the system is biased heavily in favour of luminous giant and supergiant galaxies (which dominate the galaxian mass content of the Universe, but which do not dominate in terms of total numbers seen in volume-limited samples). My own view is that both of these criticisms would be greatly alleviated if luminosity were included as a classification criterion. This is not really practical at the present time, since redshifts are rarely available for random galaxies on CCD images. But at least as an organizational structure, it seems to me that a three-dimensional system such as that proposed by van den Bergh (1998) has much to commend it (Figure 3), summarizing pictorially many of the issues dealt with so far in this lecture.
Figure 3. The three-dimensional tuning fork proposed tentatively by van den Bergh . Note how the the early-type sequence has been replaced in all panels by more physically meaningful classifications. As one descends the luminosity scale of the three-dimensional tuning fork, the conventional forms of late-type galaxies on the Hubble sequence disappear, and become replaced by lower-luminosity systems with more ragged morphology.
On balance, I would argue that according in terms of the first two of the three (rather strict) criteria for a useful classification scheme espoused in Section 2.1, the Hubble system is a resounding success. However, because of the rather depressing results from controlled comparisons between independent classifications made by expert morphologists , I have rather grave doubts about the third criterion (observer-to-observer consistency in visual classifications - I have no doubt that an expert morphologist can make classifications that are, internally, highly consistent). Recent years have seen great advances in the usefulness of objective, machine-based morphological classifications [27, 65, 2, 68]. None of these automated classification schemes is able to reproduce the Hubble system in detail, although the classifications do track the system in a crude sense. At high redshifts, where low signal-to-noise and subtle selection effects play havoc with visual classifications, simple and objective classifications that can be calibrated by simulations are essential.