Those of us actively trying to understand extragalactic globular clusters and their connection to galaxy formation are currently operating in a data-dominated rather than a theory-dominated field. We are accumulating a wealth of observational information that we cannot fully interpret. An urgent need exists for improvements in numerical and semi-analytic simulations to help identify GC formation sites and track their spatial, kinematic, chemical, and structural evolution. Models that can resolve the masses and sizes of a typical GC are tantalizingly close to implementation, and their advent will signal a leap in progress toward placing the formation of GCs in its proper cosmological context. With the new generation of "big telescope" wide-field multiplexing spectrographs, such as Keck/DEIMOS, Magellan/IMACS, MMT/Hectospec, and VLT/VIMOS, it is possible to study large samples of GCs in a wide variety of galaxies to carry out detailed tests of these future models, establishing ages, metallicities, and kinematics.
Other important developments will come from the community of SSP modelers. As already mentioned, there remain significant disagreements within this community on the treatment of -enhancement, horizontal branch morphology (the second parameter problem), and underlying stellar synthesis techniques. A new way forward to study extragalactic GCs in detail is the SSP modeling of high-resolution spectra (e.g., Bernstein & McWilliam 2005). This offers the possibility of estimating abundances of light, -, Fe-peak, and even some strong r- and s-process elements. With an 8-10m class telescope, this technique can be applied to significant samples of GCs in the Local Group and nearby galaxies, and even to the brightest few GCs in Virgo. This program is in its infancy, but could represent a significant leap in our understanding of the detailed formation histories of galaxies. GCs may offer the only route to measuring the abundances of interesting elements that are unobservable in massive galaxies themselves due to their large velocity dispersions. A possibility enabled by multiplexing high-resolution spectroscopy is detailed dynamical modeling of individual Galactic GCs, to determine whether any contain halo dark matter. The discovery of dark matter in GCs would be a "smoking gun" of cosmological GC formation.
With the development of new large-format detectors and CCD mosaics, wide field optical imaging is poised to address numerous outstanding issues. Obtaining global radial and color distributions for individual GC subpopulations is essential to test scenarios for GC and galaxy formation, especially as models emerge that predict these quantities in detail. Such imaging can also be used to probe the evolution of the GCLF, and provide a definitive test of the scenario in which GCs are formed with a power-law LF that subsequently evolves through various destruction processes to the log-normal LF observed for old GC systems. Since such processes are expected to operate more efficiently at small galactocentric radii, changes in the GCLF of GCs with galactocentric radius would be revealing. Such imaging has the potential to constrain the epoch of reionization from the spatial distribution of metal-poor GCs (as discussed in Section 12).
These ideas cover only a small fraction of the important advances likely to occur in the field over the next decade. The eventual availability of 30-m class telescopes and JWST may allow us to reach the inspiring goal of Renzini (2002): "To directly map the evolution of GCs in galaxies all the way to see them in formation, and eventually stick on the wall a poster with a million-pixel picture of a z = 5 galaxy, with all her young GCs around."
We thank many colleagues for reading drafts of this manuscript and for useful discussions, including Keith Ashman, Michael Beasley, Javier Cenarro, Laura Chomiuk, Juerg Diemand, Sandra Faber, Duncan Forbes, Genevieve Graves, Soeren Larsen, John Kormendy, Piero Madau, Joel Primack, Katherine Rhode, Tom Richtler, Brad Whitmore, and Steve Zepf. We also thank Michael Beasley, Andres Jordán, and Tom Richtler for permission to use their figures. Michael Beasley, Soeren Larsen, and Katherine Rhode provided data used to create other figures. Finally, we thank Takayuki Saitoh and Marsha Wolf for providing advance copies of their articles. Support was provided by NSF Grants AST-0206139 and AST-0507729, an NSF Graduate Research Fellowship, and STScI grant GO-9766.