Galaxy classification in practice is difficult for several reasons. First and foremost is that not every galaxy presents a favorable orientation to the line of sight. High inclination makes it difficult to (1) estimate a type consistent with low inclination galaxies, and (2) recognize bars (end-on ones especially), rings, or other disk details. In the case of spirals, the first problem arises because three criteria (bulge-to-disk ratio, the degree of resolution, and the degree of openness of the arms) can be used for typing face-on examples, while only one criterion (bulge-to disk ratio) generally can be used for edge-on examples. This can lead to problems since Hubble's three classification criteria for spirals have been known for a long time to be inconsistent in some galaxies (Sandage, 1961). The second difficulty means that the statistical frequency of important features such as bars and rings will be underestimated in highly inclined galaxies (see, e.g., de Vaucouleurs and Buta, 1980).
Another problem which makes classification difficult is that the image material often used for types (e.g., the small-scale sky surveys) is entirely inadequate for some types. As noted by Sandage and Brucato (1979), "the classification of E, S0, and early Sa galaxies is often confused" when the types are based on dense, overexposed plates or paper prints or on underexposed, small-scale plates taken with short focal length telescopes. On overexposed images such as are often found on the SERC IIIa-J southern sky survey films, high surface brightness bars, rings, or lenses are easily missed. Even on Palomar Sky Survey prints, overexposure of high surface brightness galaxies can lead to very misleading classifications; one interesting case, NGC 3928, looks like type E0 on the PSS but appears as a small, tight spiral on a CFHT prime focus plate (van den Bergh, 1980b; see also Taniguchi and Watanabe, 1987).
At the other extreme, underexposed images may cause important faint details to be completely missed. For example, low surface brightness rings, spiral patterns, disks, or other features (e.g., shells) can be difficult to detect on PSS prints or ESO-B films but clearly distinguishable on the SERC and Palomar II surveys or on processed or amplified images. Failure to detect these features could lead to serious misclassification or misinterpretations in some cases, or even to uncataloguing of an object in a diameter-limited survey (Bothun et al., 1987, 1990).
Another problem is that no classification system is perfect enough to encompass all galaxies. Many galaxies cannot be easily fit into one or another system. As emphasized by Sandage and Binggeli (1984), Hubble's classification system encompassed mainly giant, high luminosity galaxies whose forms did not readily extend to much lower luminosities. Galaxies near or outside the fringes of the old classification systems have often been peculiar (colliding or merging systems).