The main outstanding question for interaction induced star formation is the same as for any field of star formation: is the IMF universal, and if not how is it different in these violent environments? Are there upper and lower mass limits? Is the IMF top heavy? All of these effects have been claimed [30, 62, 63], but the observations allow for a wide range of parameters. A major problem with these determinations is that the observered bursts are both temporally and spatially variable, and as a result there will be a mix of burst populations of varying strength and ages spread throughout the merger. An IMF measured globally will reflect the luminosity averaged IMF over the region observed, which will be different in different wavebands. Careful UV and/or NIR spectroscopy of individual knots in conjunction with dynamical modeling should help separate these effects. Such observations should also help constrain the past star formation history in the systems, allowing one to map the age distribution of the star forming episodes.
Another major question is how does the energy injected back into the surrounding gas via the winds from massive stars and SNe change the gas dynamics and burst properties? This process, referred to as "feedback", is seen in its most extreme form in mergers, as evidenced by the galactic scale superwinds emerging from many ULIR systems (see Heckman, these proceedings). It may simply be an interesting side-effect of the circumnuclear star bursts, or it may dramatically affect the physics of star formation, for example by raising the lower-mass cutoff of the IMF or changing its high-end slope [64], or by regulating the star formation rate at some critical value [65].
There is much hope for further progress to be made in these areas in the future. It is becoming feasible to model a multi-phase ISM, which should help assess the importance of feedback. Further numerical trade studies should help decide how the different parameters interact and predict relationships between gaseous and star forming regions. Careful observations can test these predictions, and in this way we can discriminate between the different numerical formalisms. Not only will this provide a deeper understanding of interaction induced star formation, but ultimately it should be possible to decide which of the two histories depicted in Fig. 4 a merger follows, and which of these lead to an ULIR phase. Since the ULIR systems may be the closest analogs to the star forming galaxies seen at high redshift (see Madau, these proceedings), this understanding will provide valuable insight into the major processes at work during the epoch of galaxy formation.
I wish to thank J. van Gorkom and W. Vacca for comments on this manuscript; to M. Yun, C. Mihos, S. Aalto and W. Vacca for permission to reproduce their figures; and to the organizing comittee for setting up such an interesting meeting. This work is supported by Grant HF-1059.01-94A from STScI, which is operated by AURA, Inc., under NASA contract NAS5-26555.