The question that has been put to this conference rests on the assumption that we know what criminal defense counsel should be doing. Stephen Schulhofer’s contribution to this symposium as well as his very impressive earlier article – both of which are the more limited objects of my remarks – also leave largely unstated the values by which the criminal justice system should be measured. But we cannot know whether an institutional structure, a process, or a rule is good or bad unless we can specify the criteria that in-form such a judgment. My goals, therefore, will be to explore the range of values we might invoke for that purpose and to speculate about how these values could be advanced.

Any goal-directed social activity can be evaluated by reference to measures of input, outcome, and process. I will begin my analysis with input measures, though I think we will be forced to conclude that they are not very helpful. The input into the criminal process consists almost entirely of labor, materiel expenditures are trivial, and though some have discussed the physical format of the courtroom, few would argue that such capital investments playa decisive role. Labor can be measured quantitatively or qualitatively – by the amount of time expended or by the competence of the actors. Quantity is easy to calibrate but not very interesting, since there is no intrinsic value in having many people spend a great deal of time on the criminal process. And though it seems plausible that the quality of both process and outcome are related to the amount of time invested, we have no data on this empirical question. The quality of the labor force is more interesting but also much harder to assess.Input measures such as law school attended, class rank, or years of experience are of dubious value. Direct observational tests of past performance are difficult to devise and administer. And once again we have no data that would correlate input quality with process or outcome. Therefore, although research on the causal effects of different inputs would be highly desirable, until we obtain the results of such research we will have to look directly at outcome and process measures.