Elliptical galaxies are smooth, quiescent star-piles, devoid of any of the spectacular structures found in Spiral Galaxies. They have no disk, no spiral structure and only small amounts of the gas and dust. As a result there are no obvious symptoms of continuing star formation: no H II regions or young star clusters. They are the simplest galactic systems comprising just a single component, relatively bright in the center but fading rapidly with increasing radius. Elliptical galaxies are found mostly in the denser regions of the universe, from rich clusters to small groups; truely isolated ellipticals being relatively rare. The central region of the Coma cluster is dominated by a population of ellipticals. The most luminous elliptical galaxies are amongst the brightest galaxies we know of and this, combined with their apparent simplicity, has made them targets of detailed study, not only so that we can understand their own evolution but also so that they can be used as standard candles to determine distances.

The simple structure of elliptical galaxies is reflected in their place in Hubble's Classification. They are characterized by a single number, the ellipticity ε = 10(1 − b/a), where b and a are the projected angular extent of the short and long axis of the galaxy on the sky. E5 galaxies have a long axis twice that of the short axis and are almost the flattest ellipticals. Some galaxies have been classified E6 and E7 but these have almost always been found to harbor stellar disks. Because the classification depends on the projected axial ratio it is not a physical classification. By making the simple assumption that ellipticals have the two long axes equal with the third axis shorter (i.e. they are oblate, like a tangerine) Hubble was able to invert the observed distribution of projected shapes to estimate the true distribution. He estimated that on average ellipticals have a short axis about 65% of the length of the long axis with a dispersion of about 15%. This picture remained current until the late 1970s but has been dramatically revised since then. In recent years attempts to produce a physical classification for ellipticals, and new ways of estimating their intrinsic shapes, have been developed.

This article will be limited to a discussion of classical elliptical galaxies. These range from the brightest galaxies, which are about 40 times as bright as the Milky Way, to dwarf galaxies that are 100 times fainter than the Milky Way and that are typically found as companions to more massive galaxies (e.g. M32, the companion to the Andromeda nebula, M31). I will not include DWARF Spheroidal Galaxies which are sometimes, erroneously, refered to as dwarf ellipticals. Luminous elliptical galaxies have roughly the same colors as K-giant stars giving them a yellow-orange hue; less luminous galaxies have slightly bluer colours. In appearance and spectral characteristics the bulges of spiral galaxies resemble low-luminosity elliptical galaxies and historically the bulge component of M31 was taken as typical of ellipticals. There are many parallels between ellipticals and bulges and I will compare and contrast them throughout this article.

In the next section I will draw together the ways in which we have determined the distribution of light and mass as a function of radius within elliptical galaxies, including a discussion of attempts to refine our views of the intrinsic shapes of elliptical galaxies and attempts to detect their massive dark halos. I follow that with a discussion of the special properties of elliptical galaxy cores, where we often find separate components, even rotating in the opposite direction to the bulk of the galaxy, will be discussed. I also discuss the extensive searches for black holes in the centers of ellipticals. Then I describe how we learn about the evolution of the stars that make up ellipticals and how the large-scale properties of elliptical galaxies scale with each other: luminosity, mass, surface density, color, heavy element abundance and rotation. Finally I bring together all of the facets of elliptical galaxies and present the current thinking about how elliptical galaxies formed and evolved.

Elliptical galaxies are so called because the contours of constant intensity (called isophotes) are concentric ellipses. This is a very good description of most elliptical galaxies but there are small but significant variations from this simple form. High-quality images show (a) systematic changes in ellipticity with increasing radius, the most common form being a steady increase in the flattening and (b) that the position angle of the major axis of the isophotes changes with increasing radius. In most cases these changes are small (one or two ellipticity classes or a few degrees) but in a small minority they can be dramatic.

The sizes of ellipticals are measured by fitting an analytic form to the fall in brightness (averaged in concentric circular annuli) with increasing radius: the luminosity profile. The most commonly used form was proposed by Gerard de Vaucouleurs in 1948 on the basis of images taken on photographic plates. He proposed:
I(R) =
Ie −7.67[(R/Re)1/4 − 1]} 10{−3.33[(R/Re)1/4 −1]
= Ie

where Re , the effective radius, is the radius that contains half the total light of the galaxy and Ie is the surface brightness (the amount of light from a square arc second of the galaxy) at Re. This law can be integrated to give a finite total light of Itot = 7.22πRe 2Ie. de Vaucouleurs’ law, as it is known, fits remarkably well to many of the most accurately determined luminosity profiles of elliptical galaxies and provides a convenient measure of the size and total luminosity of galaxies. Effective radii range from ∼20–30 kpc for giant galaxies to hundreds of pc for the smallest. Total blue luminosities range from 100 billion solar luminosities (MB ∼−24) to 25 million solar luminosities (MB ∼−15). A generalization of this law, proposed by Sersic, which replaces the 14 power by another free parameter 1 , fits a wider range of galaxy luminosities and by allowing nn = 1 even fits the luminosity profile of spiral galaxies well. Other laws have been proposed to describe the presence of a constant surface brightness core in some elliptical galaxies and to account for the extended low surface brightness haloes found in others. Details of these are given in Binney and Merrifield (1998) which is an excellent reference for further reading.

As well as ellipticity changes and isophote twists, elliptical galaxy isophotes exhibit small deviations from pure ellipses that have become the basis for a more physical classification of elliptical galaxies. Accurate determinations of the boxy or disky nature of isophotes have been made consistently even though they arise from contributions to the light of at most a few per cent. Remarkably many properties of elliptical galaxies vary depending on the type and strength of these isophote distortions and recent work links the disky ellipticals in a sequence with lenticular galaxies leaving boxy ellipticals as a separate class. The kinematics, stellar populations, radio power and x-ray emission from elliptical galaxies all show a dependence on isophote shape and these global differences have led to sugestions that these two flavors of elliptical galaxy have different formation histories. Further clues to the formation history of elliptical galaxies are found in the faint, incomplete, shells that have been identified around many galaxies (see figure 2). These weak features may occur in as many as 30% of elliptical galaxies and are thought to be the remnant of a small disk galaxy that has fallen into an elliptical and been destroyed.

Determining the luminosity profile is the first step to calculating the mass of elliptical galaxies. The simplest assumption is that the light traces the mass and so, for a given assumed intrinsic shape, the observed surface brightness (which is the integral along the line-of-sight of the luminosity density) is related to the projected surface mass density by a constant factor, the mass-to-light ratio (M/L) of the stars.

Kinematics and intrinsic shapes

Measurements of the rotation and random motions of the stars in ellipticals are made from the shifts and width of absorption lines in the spectra of the galaxies along the major axis. The random motions are characterized by the value of the dispersion of the Gaussian that best fits the width of the absorption lines: the velocity dispersion, σ . Early measurements revealed that more luminous galaxies have broader lines such that the total luminosity L ∝ σ 4. This is known as the Faber–Jackson relation after the first people to draw attention to this in 1975. This discovery ignited an interest in using measurements of galaxy velocity dispersions to estimate the intrinsic luminosity of ellipticals and thus determine their distances. This possibility stimulated researchers to make surveys of galaxy dispersions and these produced a much better understanding of the relationships between their structure, dynamics and stellar populations that will be discussed under ‘Global properties’.

Until the late 1970s it was assumed that the flattened figures of elliptical galaxies arose because the stars in the equatorial plane are supported by their rotation about the galaxy minor axis. However, as soon as the rotation velocities of the most luminous elliptical galaxies could be measured it was discovered that they are too small to account for the observed flattened figures. Furthermore, when kinematic maps of ellipticals were made some dramatic examples of substantial rotation on the minor axis were found. This discovery led to a reassessment of the nature of elliptical galaxies. The rotation velocity, normalized to the central velocity dispersion, plotted against ellipticity for luminous elliptical galaxies, fainter ellipticals and the bulges of edge-on spiral galaxies. Also plotted is the expected relation for galaxies that are oblate and rotationally supported. Luminous elliptical galaxies are deficient in rotation by a large factor. Rotation is dynamically unimportant for luminous ellipticals; they are supported by the random motions of the stars. These random motions are not isotropic and this anisotropy produces the flattened figures observed. This discovery also challenged the assumption that ellipticals have oblate shapes and opened up the possibility that they could be prolate (two short axes equal, like a sausage) or even triaxial where all three axes have different lengths (like most potatoes).

We now believe that most luminous elliptical galaxies are triaxial and that the observed isophote twisting arises from the projection onto the sky of the changing intrinsic axial ratio with radius. This hypothesis also accounts for the rotation sometimes observed on the projected minor axis. This view has been established statistically from both the distribution of observed shapes of thousands of galaxies and from observations and models of the kinematics of many tens of galaxies. The average intrinsic shape for luminous elliptical galaxies is, however, poorly constrained. Most estimates are close to perfectly triaxial (where the intermediate length axis is the average of the long and short), e.g. a:b:c::1:0.85:0.7 with considerable scatter. These statistical studies suggest that intrinsically spherical galaxies are rare but, for example, cannot rule out a distribution of intrinsic shapes that is a mixture of oblate and prolate figures. The discovery that elliptical galaxies are not flattened by their rotation has important implications for our ideas about how they form. Models where isolated massive clouds of gas collapse dissipationally and form stars rapidly always produce galaxies which rotate faster than is observed. These results strengthened the view that elliptical galaxy formation proceeds through a sequence of mergers (of objects on misaligned orbits), to produce a galaxy with low total low angular momentum.

The less luminous elliptical galaxies and the bulges of spiral galaxies fall on or near the line describing isotropic oblate galaxies, showing that these objects have substantial kinematic similarities and are both qualitatively different from their more luminous brethren. Whereas luminous elliptical galaxies have complex kinematic structures, low-luminosity elliptical galaxies and the bulges of spiral galaxies have isotropic velocity dispersions and oblate figures flattened by rotation. There are many other characteristics of elliptical galaxies that vary with luminosity and that I will discuss in the section on ‘Global properties’.

Anisotropy and mass-to-light ratio

Precise determination of the distribution of mass within elliptical galaxies requires detailed measurements of their luminosity profile and kinematics together with self-consistent dynamical models. Both analytic models based on the Jeans equations and models that construct galaxies by selecting orbits from a library (Schwarztschild’s method) have been successfully used. For luminous ellipticals infering the distribution of mass within a galaxy on the basis of these data and models is complicated by the degeneracy between changes in M/L, and changes in the orbital anisotropy of the stars. Using only measurements of the rotation and dispersion we cannot distinguish between an increase in M/L and an increase in one component of the velocity dispersion at the expense of another. Techniques based on measuring the shape of absorption line profiles have been developed to break this degeneracy. The shape of absorption lines is determined by the luminosity distribution and the number of stars at each velocity along the line of sight. In the center of a galaxy an increase in dispersion may indicate an increase in M/L or an increase in the fraction of radial orbits. The latter will, however, produce a line profile that is less sharply peaked than a Gaussian. Thus by measuring deviations from a Gaussian profile as well as its width we can determine whether stars are predominantly on tangential or radial orbits.

This ambiguity can be avoided altogether for low-luminosity ellipticals or spiral bulges as we believe them to have isotropic dispersions so determining the mass distributions in these systems is more straightforward. Similarly if galaxies possess disks of excited or ionized gas we can trace the mass distribution in the same way as for spiral galaxies.

Dark haloes: mass at large radius

Spiral galaxy rotation curves have been used to show that they are embedded in dark haloes that contain three to ten times as much mass as is visible in stars, gas and dust. Can dark haloes be detected in elliptical galaxies? The velocity dispersion in many galaxies remains roughly constant (outside the central region) to one or more effective radii. As galaxies get rapidly fainter away from the center does this mean we are detecting a dark halo or is the fraction of stars on circular orbits increasing with radius? In the outer parts of a galaxy, if there are more stars on circular orbits the absorption lines will be less peaked than a Gaussian and have truncated wings. In a few elliptical galaxies where the measurements extend to two to three effective radii the line of sight velocity profile has been used to eliminate the possibility of tangential anisotropy and indicate the presence of a dark halo. Similar conclusions are drawn from the velocities of Globular Clusters and Planetary Nebulae at large radii in elliptical galaxies.

An important probe of the mass distribution at large radius is the hot x-ray emitting gas that has been detected out to five to ten effective radii in a large number of elliptical galaxies. This gas is enriched in heavy elements and is thought to originate from the stellar component of the galaxies through mass loss during the late stages of stellar evolution. By assuming that this gas is isothermal and in hydrostatic equilibrium it is possible to estimate the mass out to very large radii from the distribution of x-ray emission. These measurements have confirmed that elliptical galaxies reside in massive haloes qualitatively similar to those of spiral galaxies. With the advent of more capable x-ray satellites that can determine the temperature of the x-ray gas the simple assumptions required to estimate masses have been confirmed for a small number of galaxies.

Amongst the most luminous elliptical galaxies are those that possess an extended (compared with a de Vaucouleurs law) low surface brightness halo, known as ‘D-type’ galaxies, often found in the centers of rich clusters when they are classified morphologically as ‘cD galaxies’. As well as high luminosity and extended halos these galaxies often have more than one nucleus, that is several smaller galaxies are superimposed on the image of the galaxy center. cD galaxies appear to be the product of many interactions or mergers of cluster galaxies with the central dominant galaxy. Several attempts have been made to measure the stellar kinematics in these out to large radii to establish the M/L at large radii and determine whether the stars in the halo are moving in the cluster potential. In a few such galaxies we observe the velocity dispersion rising with increasing radius; however, they do not reach the cluster dispersion, nor do they reach the velocity dispersion expected on the basis of the velocities of the multiple nuclei, so the interpretation of these motions remains controversial.

The center of a galaxy is a unique environment. The potential well is deepest at the center and any material that dissipates energy in collisions, such as dust or gas, will fall to the center relatively rapidly. For a giant galaxy this will take at most a few hundred million years. Galaxies of all types exhibit special phenomena at their centers. The high-resolution Hubble Space Telescope surveys of the centers of ellipticals often reveal patches or disks of dust and central stellar disks. The centers sometimes host powerful radio sources or even star formation. We believe that radio galaxies are powered by massive black holes, and as a result the centers of elliptical galaxies have been identified as likely sites of black holes.

Black hole searches

Studies of quasars and galaxies at high redshift reveal that active nuclei were much more common at high redshift (z ∼ 2), sufficiently common that we infer that most luminous present-day galaxies underwent an active phase and therefore should contain a massive central black hole. Searches for central black holes started in the late 1970s with observations of the luminous radio galaxy M87 (pictured in M87: The Massive Galaxy). From the rapidly increasing central velocity dispersion it was concluded that a black hole of 3 billion solar masses was present in the nucleus. The interpretation of these data was controversial because of the uncertainty introduced by the degeneracy between orbital anisotropy and mass distribution. These important observations stimulated a great deal of investigation of anisotropy and stability, new models were produced and more refined observations were made that increased our confidence in the case for a massive black hole in M87. In the 1990s measurements by the Hubble Space Telescope (HST) of the velocity of the excited gas close to the nucleus revealed a gradient in velocity of ±500 km s −1 across the central 40 pc of the galaxy implying a mass of 2.4 × 109 M(Earth) contained within a radius of 20 pc. This provided strong confirmation of the black hole hypothesis independent of the uncertainties concerning the mix of stellar orbits.

The Hubble Space Telescope (HST) has carried out extensive imaging surveys of galactic nuclei finding many galaxies that have stellar cusps (luminosity profiles that continue to increase in surface brightness to the limit of resolution of the telescope). Other galaxies have been found with resolved cores where the luminosity profile reaches a constant brightness at a finite radius. These measurements are important for evaluating the value of any central mass but a nuclear cusp alone cannot be used to conclude that a black hole is present. When combined with kinematic measurements, made mostly from the ground, we have been able to address the more general question of the frequency of black holes in elliptical galactic nuclei. More than 20 galaxies have been studied in detail. A general trend has emerged: more massive galaxies host more massive black holes. The black hole has roughly 0.2– 0.6% of the total galactic mass. This relationship is very uncertain and exhibits a factor of three scatter in the black hole mass, but if true it has wide-ranging implications. However, not all galaxies are required by the observations to host a black hole.

Kinematically decoupled cores

As measurements of rotation velocities and velocity dispersions were made with greater resolution and precision an increasingly large number of galaxies were found to have central kinematics that were decoupled from those of the bulk of the galaxy. The most dramatic examples have the central region (typically 0.1–0.3Re) rotating in the opposite direction from the rest of the galaxy, some have cores rotating about an axis less misaligned, in yet others the cores have very much higher rotation and lower velocity dispersion than the main galaxy. The latter suggests that there is a central disk embedded in the elliptical. These components often occur in the less luminous elliptical galaxies together with with disky isophotes and stellar cusps. In addition to these extended components the HST surveys of galactic cores have revealed many small (100 pc) disk components in ellipticals. Decoupled kinematics are not limited to ellipticals, one early-type Disk Galaxy has been found where two identical co-extensive disks rotate with the same velocity but in opposite directions!

Although at first such decoupled systems were considered rare, taking account of projection effects and the difficulty of detecting small or weak components we now believe that between one-and two-thirds of all elliptical galaxies have extensive central components with decoupled kinematics (i.e. that are a kiloparsec or so across and have been discovered from the ground). The most likely origin for the decoupled central components in ellipticals would appear to be through the accretion of material with different specific angular momentum from that of the main body of the galaxy. The existence of decoupled cores has been widely cited as evidence that elliptical galaxies have experienced significant accretions in the past; however, the colors, metallicity and age of the stellar populations appear to be indistinguishable from those of the main galaxy itself. We will return to the possible origin of these components in the final section.

In the Milky Way we have used the concept of Stellar Populations as an important tool in unraveling our Galaxy’s history. We would like to extend these ideas to external galaxies. The colors and absorption line strengths measured in the spectra of the integrated light of elliptical galaxies are rather similar to those of the bulge of the Milky Way. Such considerations led us to believe that elliptical galaxies are old, single stellar populations devoid of star formation, which are now largely quiescent. The oldest stars in the Milky Way are found in the globular cluster population. Elliptical galaxies possess globular cluster populations in proportion to their luminosity and the colors of these populations suggest that a substantial fraction of them are as old as the globular clusters in the Milky Way, i.e. the oldest stars in elliptical galaxies are as old as the oldest stars in our galaxy.

Even the nearest luminous elliptical galaxies cannot be resolved into individual stars from the ground, and only a handful can be resolved with HST so we are forced to try to unravel the integrated light from all the stars. By analysing the integrated spectrum we aim to discover what kind of stars constitute an elliptical galaxy. We approach this by using a library of stellar spectra or model stellar atmospheres covering a range of luminosity and effective temperature. We combine these using our knowledge of the relative numbers of stars in a population of a particular age and metallicity based on evolutionary models. In this way we attempt to synthesize the observed integrated spectrum from known stellar components. In the optical (λ = 0.5 µm) we find that about half the light arises from main sequence stars and half from giant stars but that ellipticals have an excess of light in the near infrared (λ = 1–2 µm) and in the ultraviolet (λ = 0.2 µm). The former can be attributed to stars with a higher metal content than the Sun, whereas the latter are likely to arise from extreme horizontal branch stars. Both types are missing from the local stellar libraries.

Almost all the physical properties of elliptical galaxies vary smoothly with increasing luminosity. I have already discussed how the degree of rotational support decreases with increasing luminosity, a fundamental difference in the orbital make-up of faint and luminous ellipticals. This trend was only discovered through detailed study of the kinematics of tens of elliptical galaxies. While this work is becoming easier with the development of area spectrographs such as SAURON, many more galaxies have been characterized by single central spectra. These measurements have resulted in the discovery of a number of global relationships between properties that are easily measured for hundreds of galaxies. Remarkably there are tight correlations between parameters which are measured locally at the center of the galaxy, such as velocity dispersion, and global parameters such as size and brightness. I summarize these global relations here. They form much of the basis for our understanding of how elliptical galaxies formed and evolved.

The color–magnitude relation

The most luminous elliptical galaxies are slightly redder than the faintest ellipticals. This is a small effect but it has very small scatter (a few hundredths of a magnitude) and is well established in nearby clusters of galaxies. There has been a debate over the origin of this effect, is it due to a higher average metal content in the luminous galaxies or are they on average older? The color–magnitude relation has been observed as a tight correlation in clusters at redshifts in excess of 0.5 and this makes the interpretation in terms of age changes unlikely. This result is established in clusters, the simplest assumption to make is that the color–magnitude relation is universal, but this is not yet proven. We believe that the color–magnitude effect arises because the deeper potential wells of luminous galaxies can retain the metals produced in supernova explosions more effectively.

The Mg–σ relation

This correlation involves the strength of the magnesium absorption features at a wavelength of 0.52 µm and the stellar velocity dispersion σ. Galaxies with high-velocity dispersions have stronger magnesium absorption lines. (This is the best developed correlation but other absorption line strengths have also been used.) This relation also exhibits very little scatter, <0.02 magnitudes, and because it is a distance-independent correlation galaxies from a wider range of environments can be included. Once again the correlation could arise from changes in age or metallicity, but measurements in rich clusters out to z =0.4 are consistent with the local relation changed only through the stellar evolution that occurs in that time interval (about 4 Gyr). This suggests that the correlation arises because high-σ galaxies are metal rich. Tests of this relation in different galactic environments have failed to find any systematic differences.

The Faber–Jackson relation and the Fundamental Plane

Earlier I introduced the Faber–Jackson relation: the increase in central velocity dispersion with luminosity, L ∝σ4. In the 1980s surveys of elliptical galaxies were made to measure the non-Hubble motions of galaxies. Using the high-quality data collected it became apparent that an extra parameter, the galaxy surface brightness, increased the precision of the L, σ4 relation as a distance indicator. Thus a new relationship, which has become known as ‘the Fundamental Plane’ superceded the L, σ4 relation. The Fundamental Plane relation is:
log Re =1.25 log σ +0.33(SBe)+ constant.
where SBe is the effective surface brightness Ie expressed in magnitudes and (SBe)is the average surface brightness within Re so that (SBe)=3.61SBe.

In the size, velocity dispersion, surface brightness space elliptical galaxies are not distributed uniformly, they fall on the above plane with relatively small intrinsic scatter of about 15% in Re. The Fundamental Plane (FP) is close to what would be expected on the basis of the virial theorem if all galaxies have the same M/L ratio. The difference with the observed relation can be used to restrict the possible range in M/L of ellipticals to a factor of three. Similarly, the width of the plane can be used to restrict the scatter in M/L at any point in the plane to be less than 12%.

The scatter in the FP and the Mg–σ relations are sensitive to the range in both the age and metallicity of the stellar population. However, the scatter in the FP is more sensitive to the range of ages, whereas the scatter in the Mg–σ relation is more sensitive to the range of metallicities. Using these two together the range in the age and metallicity of ellipticals has been restricted to 30%, i.e. luminous ellipticals vary from say 10–13 billion years old and have metallicities in the range 3 ± 1 times solar abundance.

Structural properties

Many structural properties vary smoothly with luminosity or central velocity dispersion. I have discussed most of them already but bring them together here. Statistically, large, luminous, high-velocity dispersion elliptical galaxies have higher x-ray luminosity, host more powerful radio sources, have triaxial figures and anisotropic velocity distributions and have boxy isophotes and flat central luminosity profiles. Alternatively the small, low-luminosity ellipticals and spiral bulges have lower x-ray and radio luminosities, have oblate figures and isotropic velocity distributions, and have disky isophotes and cuspy central luminosity profiles. These differences have been used to argue that luminous-boxy and faintdisky galaxies formed and evolved differently and should be part of separate morphological sequences. This hypothesis is attractive but has yet to address how two physically distinct classes obey the same color–magnitude, FP and Mg–σrelations if indeed careful study proves that they do.

The morphology–density relation

Elliptical galaxies are gregarious, they are most often found in regions of high galaxy density. The mix of galactic morphological types is a strong function of environment, in rich clusters approximately 40% of the galaxies are ellipticals and another 50% are lenticulars with the remaining 10% being spiral galaxies. In less dense environments the fraction of ellipticals falls dramatically while the spiral fraction increases. In the lowest density environments less than 10% of galaxies are classified as elliptical with spiral and lenticular galaxies being found in roughly equal proportions. The same analysis has been carried out for clusters of galaxies at z∼ 0.4 imaged by the HST. The ellipticals still dominate the galaxy population in rich clusters at z ∼ 0.4 but the fraction of lenticular galaxies has dropped to 20%, with spirals accounting for 30% of the rich cluster population. It is clear that in rich clusters the elliptical galaxy population is already in place by z∼ 0.4 but that there are very many fewer early type galaxies with disks than we find in rich clusters today.

We believe that the distribution of galaxies traces out the large-scale structure of dark matter and that the dark matter haloes of galaxies assemble hierarchically with the most massive haloes forming most recently. Only through understanding the behavior of the baryons within these haloes can we hope to understand how the luminous galaxies we observe evolve.

The high central densities and relaxed structure of elliptical galaxies are taken as strong evidence that they underwent substantial gaseous dissipation during formation. Furthermore the small scatter observed in their global relations suggests that they are old (with a formation redshift greater than z = 2–3) and coeval. These considerations have led to models in which elliptical galaxies form through the dissipative collapse of a massive gas cloud resulting in a huge starburst which turns almost all the gas in a halo into stars simultaneously, say 10 billion years ago. The most recent measurements at sub-millimeter wavelengths have detected galactic masses of cool dust at z= 2–3 that are forming stars at a very high rate. These sub-millimeter sources have been identified as the progenitors of the most luminous cluster ellipticals we see today. However, models show that the monolithic collapse of a massive gas cloud produces an unrealistically rapid rotation in the end product and furthermore the collapse of massive objects at high redshift appears to contradict the hierarchical assembly of galaxies.

These drawbacks have led to alternative suggestions that ellipical galaxies formed through a sequence of galactic mergers. There is much empirical evidence for this: low rotation velocities, decoupled cores and shells are all a natural consequence of accretions and mergers. In the alternative model disk galaxies merge together to generate elliptical galaxies during the formation of groups and clusters of galaxies. Theoretical investigations of galaxy mergers using n-body techniques have successfully demonstrated that disk galaxies undergoing mergers will produce elliptical galaxies with de Vaucouleurs’ law luminosity profiles and low rotation velocities. In cosmological simulations the bulk of the stars that comprise present day ellipticals form in merger events that occur typically at z = 1–2. The galaxies continue to grow in mass and luminosity through mergers until they fall into massive clusters where the typical velocities of galaxies get so large that mergers are inhibited. This scenario accounts for why ellipticals are predominantly found in high-density regions and predicts that on average ellipticals in rich clusters will be older (in the sense of a luminosity weighted age) than those in less dense regions.

These two formation mechanisms predict different age distributions of elliptical galaxies, particularly as a function of galactic environment. We would like to infer the star formation history of ellipticals from their integrated spectra, in particular to distinguish between the hypotheses that the stars formed in a single event at high redshift, z>2 − 3 or that star formation was much more extended with perhaps some star formation occurring at z << 1. Such tests are under way. It has been suggested that the two flavors of elliptical galaxy, luminous-boxy and faint-disky, may have different evolutionary histories with luminous-boxy galaxies formed predominantly through a dissipationless merging sequence that produce triaxial anisotropic galaxies. The luminosity weighted ages of the two flavors of ellipticals have yet to be determined.

The powerful telescopes at our disposal now, in particular the HST, have enabled us to study the evolution of elliptical galaxies directly. We have known for more than two decades that the fraction of blue galaxies in rich clusters at high redshift (z = 0.5) is much higher than we find in local clusters. HST has revealed this blue excess as disk galaxies and has established that at z= 0.5 elliptical galaxies have already assembled. It seems very likely that these massive cluster ellipticals did form rapidly and early at z>3 and they ceased star formation before z= 1. Perhaps ellipticals in lower-density environments will be found to have experienced a more extended and recent series of mergers, giving them younger mean ages. Such a formation scenario takes ingredients from both the monolithic collapse and merger hypotheses for the origin of ellipticals but already faces apparent contradictions with the empirical measurements.