Classification in Astronomy: Past and Present

Astronomers have always classified celestial objects. The ancient Greeks distinguished between asteros, the fixed stars, and planetos, the roving stars. The latter were associated with the Gods and, starting with Plato in his dialog Timaeus, provided the first mathematical models of celestial phenomena. Giovanni Hodierna classified nebulous objects, seen with a Galilean refractor telescope in the mid-seventeenth century into three classes: "Luminosae," "Nebulosae," and "Occultae." A century later, Charles Messier compiled a larger list of nebulae, star clusters and galaxies, but did not attempt a classification. Classification of comets was a significant enterprise in the 19th century: Alexander (1850) considered two groups based on orbit sizes, Lardner (1853) proposed three groups of orbits, and Barnard (1891) divided them into two classes based on morphology.

Aside from the segmentation of the bright stars into constellations, most stellar classifications were based on colors and spectral properties. During the 1860s, the pioneering spectroscopist Angelo Secchi classified stars into five classes: white, yellow, orange, carbon stars, and emission line stars. After many debates, the stellar spectral sequence was refined by the group at Harvard into the familiar OBAFGKM spectral types, later found to be a sequence on surface temperature (Cannon 1926). The spectral classification is still being extended with recent additions of O2 hot stars (Walborn et al. 2002) and L and T brown dwarfs (Kirkpatrick 2005). Townley (1913) reviews 30 years of variable star classification, emerging with six classes with five subclasses. The modern classification of variable stars has about 80 (sub)classes, and is still under debate (Samus 2009).

Shortly after his confirmation that some nebulae are external galaxies, Edwin Hubble (1926) proposed his famous bifurcated classification of galaxy morphologies with three classes: ellipticals, spirals, and irregulars. These classes are still used today with many refinements by Gerard de Vaucouleurs and others. Supernovae, nearly all of which are found in external galaxies, have a complicated classification scheme:Type I with subtypes Ia, Ib, Ic, Ib/c pec and Type II with subtypes IIb, IIL, IIP, and IIn (Turatto 2003). The classification is based on elemental abundances in optical spectra and on optical light curve shapes. Tadhunter (2009) presents a three-dimensional classification of active galactic nuclei involving radio power, emission line width, and nuclear luminosity.

These taxonomies have played enormously important roles in the development of astronomy, yet all were developed using heuristic methods. Many are based on qualitative and subjective assessments of spatial, temporal, or spectral properties. A qualitative, morphological approach to astronomical studies was explicitly promoted by Zwicky (1957). Other classifications are based on quantitative criteria, but these criteria were developed by subjective examination of training datasets. For example, starburst galaxies are discriminated from narrow-line Seyfert galaxies by a curved line in a diagramof the ratios of four emission lines (Veilleux and Osterbrock 1987). Class II young stellar objects have been defined by a rectangular region in a mid-infrared color-color diagram (Allen et al. 2004). Short and hard gamma-ray bursts are discriminated by a dip in the distribution of burst durations (Kouveliotou et al. 2000). In no case was a statistical or algorithmic procedure used to define the classes.