Forum Springer Astron. Astrophys.
Forum Whats New Search Orders

Astron. Astrophys. 319, 747-756 (1997)

Previous Section Next Section Title Page Table of Contents

3. Problems with the traditional classification scheme

The traditional classification method, by visual inspection of plates and by reference to standards for each type, has been very successful in many fields of extragalactic astronomy. However this method suffers from drawbacks, which can become serious, depending on the use one makes of the morphological types.

First, the reference to standards makes the morphological classification difficult enough that it resembles 'more an art than a physical measurement' (Buta 1990). This task is thus not accessible to most of the astronomical community, since, to accomplish it, one has to be an expert morphologist.

The strong subjectivity of the task raises the question of its reproducibility and of the consistency of the morphological types determined by different morphologists. The latter question has been addressed by Lahav et al. (1995). Less than 1 % of the large galaxies (D [FORMULA] arcmin) have the same morphological label when galaxies are classified in 16 bins by 6 well known morphologists. The cause of the disagreement is not tied to differences in the images that the morphologists studied, since they used the same images. The question of consistency will be discussed further in Sect. 4.

The fact that structural components (bar, disk, bulge, arms, etc.), which astronomers naturally think of when speaking of morphological types, are not measured quantitatively and in some cases not even detected by traditional morphologists introduces differences and biaises between the presence of such structural components and the resulting morphological types. These differences and biases could be important or negligible, depending on the study undertaken, on the fraction of galaxies for which one or several structural components were missed, and on how this fraction was classified.

Second, the uniformity of the observational material is not a desirable property when the studied sample contains galaxies at different distances, of different sizes or luminosities and/or projected at different angles on the sky.

  1. Due to the limited dynamical range of plates, images of very bright galaxies are saturated and images of faint galaxies are of too low quality to allow any classification. As repeatedly stated by morphologists, this happens very often: in one third of the cases for a sample of galaxies larger than 1.2 arcmin (Lahav et al. 1995, Naim et al. 1995) and, more generally, in 85 % of the cases (Buta 1992). This problem does not appear very often in the output catalogue; in other words there is no trace in the catalogue that some of the galaxies have classification problems. Buta (1992) stressed that 'it is important when using published morphological types to know where their types came from and their limitation'. Finally, as morphologists themselves admit (Lahav et al. 1995), they mostly classify galaxies for which they do not have suitable data. In particular for galaxies in nearby clusters, Dressler (1980) remarks that sky survey plates are not good enough for morphological classification and that Cassegrain plates (or prime focus plates from a large reflector) must be used.
  2. The morphological label attributed by morphologists to the observed galaxies unfortunately depends on the projection angle of the galaxy on the sky. It is a well known fact that face-on S0 galaxies are missing in all catalogues of galaxies, because they are misclassified as E (van den Bergh 1990). The detectability of bars, arms, disks of galaxies depends strongly on their projection angle on the sky, as well as on the resolution of the observations used to perform the classification (for disks, see e.g. Nieto et al. 1994, whereas for bars see e.g. de Vaucouleurs & Buta 1980 and Nieto et al. 1992).

The drawback of using uniform data becomes obvious in studies of galaxies in more than one cluster, when the observed galaxies are not all at the same distance. Andreon (1993) showed that the spiral fraction in nearby ([FORMULA]) clusters measured by Bahcall (1977) decreases as the redshift increases, reaching a null value at [FORMULA]. Such behavior (a sort of inverse Butcher-Oemler effect) is not intrinsic to the observed clusters, but is only a consequence of the increasing difficulty of identifying spiral galaxies as the redshift increases. A similar artificial trend has already been pointed out by Tammann (1987) for explaining the apparent rise of the Hubble constant with redshift from independent data. Such a feature is common (and has been overlooked) in the literature. Unfortunately, much of our knowledge of nearby clusters is based on such data (e.g. Sarazin 1986, Edge & Steward 1992).

Another problem linked to the resolution, which we discovered when studying a distant cluster observed with HST (Andreon, Davoust & Heim 1996), is that of the sampling of the image. The point spread functions of two images may have the same FWHM, but if the first image is oversampled and the other one has a pixel size comparable to the FWHM of the point spread function, morphological details will be lost in the latter image. Thus, as one classifies more distant galaxies, both the resolution and the sampling of the image should increase.

The use of uniform data (i.e. plates) and of images whose dynamical range is limited, is an easy way of collecting large numbers of morphological types, but it is also likely to induce misclassifications. The images should have a quality adapted to the difficulty of the galaxy classification.

Previous Section Next Section Title Page Table of Contents

© European Southern Observatory (ESO) 1997

Online publication: July 3, 1998