How do judges think? My short answer is: just like everyone else. While that answer may seem obvious, legal scholars have expended much effort on the question. The question is important because the results of judges' thinking-judicial decisions-can change people's lives. Scholars are concerned about judge's thinking both to understand judicial decisions and, in some cases, to redirect them.

Given the importance of the question and the amount of attention it has received, one might think that judges' reasoning processes have received careful empirical analysis. Unfortunately, that is not so. While there have been some good studies of the decision making of that other important decision maker in the court system-the jury-similar empirical studies of how judges reach their decisions are few. The main reason is access: judges are busy public officials.

So, it is not surprising that most studies of judicial reasoning have been based only on the judge's written decisions or on the author's own experience as a judge's law clerk or, in some cases, as a judge. The studies must piece together the picture of judicial reasoning using a good deal of imagination.

These efforts have led to four distinct models of judicial reasoning. These all start out as models of how judges actually think. However, advocates of each model usually also assert that the model is normatively desirable because following it would lead to better judicial decisions.

The first model, the "legal science" or the "legal formalist" model, asserts that a judge reasons like an idealized scientist. When faced with a decision, a judge first reviews all the prior cases similar to the case at hand. From the results in those cases and the reasoning in the written decisions, he or she works out a general principle from which each of the results in those cases can be derived. The general principle is then applied to the facts of the case at hand. A good written decision is an accurate, if truncated, description of the process of reasoning.

This model of reasoning is called legal science because the judge (or legal scholar) treats the prior written decisions as the empirical evidence that he or she manipulates to yield general principles. To paraphrase Christopher Columbus Langdell, the turn-of-the-century Harvard Law School dean who popularized this model, the judge's laboratory is the law library. The judge is merely a good scientist who discovers the principle contained in the cases, not the creator of it. The term "science" is used here in a very old-fashioned way. The science is normative or moral. This model of judicial reasoning assumes that the normative principle resides in the written decisions of past cases to be discovered.

The second model, pressed by the pragmatists and legal realists scholars of the 1920s and 1930s, stands this aspect of the legal science approach on its head. It suggests that a judge surveys all the facts of the case before him or her and decides which party should win. This decision is attributed variously to the political philosophy, psychology, social or economic class, or even the breakfast menu of the judge. In one famous account written by a sitting judge, the decision was said to be the product of "hunch." Having reached that decision, the judge then writes an opinion to defend the result. The written opinion may or may not reflect the true reasons for the decision. It does, however, follow the convention of asserting that the result follows from some general legal rule accepted by most people. But the actual decision precedes this defense and is not affected by it. John Dewey and others argued that this model was a more accurate representation of how judges actually think than the deductive model of legal science.

A more recent model, called "analogical reasoning;' is a product of the legal process school which flourished in the 1950s and 1960s. It is probably the model that most lawyers would identify as most descriptive of good legal reasoning, no doubt in part because their law teachers and casebooks were heavily influenced by it. This view was best expressed by a University-of-Chicago Law School dean (and former Attorney General), Edward Levi: "[Legal reasoning is] reasoning from case to case:" The judge first finds a prior case that has factual similarities to the case at hand. He then identifies the rule of law inherent in the first case. That rule is then modified to apply to the case at hand. The crux of the method is establishing and defending factual analogies between cases. If the judge finds factual reasons to distinguish the prior case from the case at hand, the rule of the prior case is not relevant; if, however, the facts of the case at hand are not "relevantly" different from those of the prior case, the rule of the prior case should be applied.

In a popular, current model of judicial reasoning that seems to meld both the realist and the analogical approach, some legal scholars have appealed to the idea of "practical reason" as a model for judicial reasoning. Leaning on Aristotle's distinction between theoretical and practical reason, this view suggests that a judge must evaluate a case in its context. The judge must both sympathetically understand the opposing litigants' positions and evaluate them with detachment. A good decision is one that balances the particulars of a case with more abstract principles. Most scholars of this genre assert that the ability to judge is learned only through experience in judging.

All of these models may seem somewhat mysterious to the layperson. That is because they all assert that judges employ a method of reasoning different from that ordinary people seem to use. While law schools would be delighted if that were s0. we should prefer a more parsimonious explanation.

My own view is that the reasoning processes that judges employ are no different from those we all use everyday. Our goal in reasoning about practical problems is to produce useful information in the most cost-effective way to assist us in achieving our desires. We rarely "reinvent the wheel" through systematic investigation and analysis; instead, we rely on our stock of knowledge, that collection of moral and empirical information about the world that we have developed from our past experiences. We tend to interpret and judge the problem before us in terms of some related element of information in that stock of knowledge, unless we have strong reasons for doing otherwise.

Judges are no different. They are under great pressures to make decisions quickly; they do not have the luxury of the systematic analysis that most of the models 1 described above suggest. Instead, judges resort to their own stock of knowledge for a wide range of information about the world and about what is right and wrong. When faced with a case, they tend to base their decisions on the knowledge that they share with the rest of us-unless there are good reasons to dig further or to rethink an issue. The very idea of precedent in law reflects this economy of information. A reason why judges would follow the ruling in a prior case is that, absent other good reasons, they have no reason to rethink a result that other judges have reached in the past.

This view provides a better description of what is happening in the realist model when a judge gets a "hunch" about the correct decision, or in the practical reason model when a judge reaches a decision by "balancing the particulars" in a case. It also suggests why the analogical judge finds certain facts more "relevant" than others and why the legal science judge sees certain "general principles" in a group of cases. This view also explains why most judges tend to be conservative in the literal meaning of the word: their decisions tend to follow the extant community standards of what is right and wrong and to reflect community beliefs about empirical facts.

Finally, this view has the virtue of driving a wedge between the normative and the descriptive accounts of judicial reasoning. As we all know, commonsense works well in some cases, but fails miserably in others. Because judges' reasoning is simply human reason applied in the legal sphere, we should not expect it to have a higher success rate.