Specialized courts

Why do some disputes get their own court instead of going to the ordinary one?

A judge is supposed to be a generalist. That is close to the definition of the office: a neutral applying law to facts, competent to hear a contract case in the morning and a custody case after lunch, precisely because the competence claimed is legal rather than technical.

And yet tax disputes go to a tax court, bankruptcies to a bankruptcy court, patent appeals in the United States to one single court that hears nothing else. Germany runs five separate national court hierarchies — ordinary, administrative, labour, social and fiscal — alongside its constitutional court. If a good judge can hear anything, why carve out these subjects? And if a good judge cannot, why is the generalist court still hearing everything else?

Let us treat this as an ordinary question about dividing labour, because that is what it is. Specialisation pays when repeated exposure to a narrow class of problem makes someone materially better at it than a competent generalist could be. So ask what a judge actually gets better at by seeing only one subject.

Three things, and they are worth separating.

The first is fluency in the substrate. A patent case turns on what a person skilled in the relevant art would have found obvious; an insolvency turns on the ordering of claims across a capital structure. A generalist can learn these, but does so through the parties, one case at a time, at the parties' expense and partly under their influence.

The second is speed. A judge who has met a recurring question forty times does not restage the whole education for the forty-first litigant.

The third is subtler and, I think, the strongest: doctrinal consistency. When the same question is decided independently by a dozen courts, they diverge, and divergence in a technical field is not merely untidy — it lets a claimant choose the forum whose version of the law suits them. That was the explicit reason the US Court of Appeals for the Federal Circuit was created in 1982 and given exclusive jurisdiction over patent appeals: regional circuits had drifted so far apart on patent validity that where a case was filed substantially predicted how it came out.

Now push the other way, because a division of labour always costs something. What does a specialised bench lose?

It loses the outside view. A generalist arriving at a patent case brings a whole legal system's instincts about property, remedies and proportionality. A bench that sees only patents develops its own internal logic, and nothing routinely tests that logic against the rest of the law. The complaint that this has actually happened to the Federal Circuit is a serious one, and its most cited evidence is that the Supreme Court has repeatedly reversed it on foundational questions — on when an injunction should issue, on the standard for obviousness, on what is even eligible for patenting.

It also concentrates the audience. If one court decides every case in a field, the same small bar appears before it constantly, and the risk of capture — of the court coming to share the working assumptions of the industry it supervises — rises for structural reasons, whatever the individual judges intend.

And it manufactures a boundary. Once a subject-matter court exists, someone must decide which cases belong to it, and that question becomes litigable in its own right. A dispute that straddles two courts now costs more to resolve than it did when one generalist heard the whole thing.

So the design question is not "is expertise good" but a comparison: does this subject have enough volume, enough technical content, and enough need for uniform doctrine to outweigh the loss of the outside view?

Three refinements.

First, expertise is not the only motive, and it is polite to be honest about that. Separate courts are also created to secure faster disposal, to keep politically sensitive matters in predictable hands, or to move a class of case out of the ordinary courts' backlog. A tribunal justified as technical may be doing mostly administrative work.

Second, specialisation is a dial rather than a switch. Between a generalist bench and a dedicated court sit several intermediate settings actually in use: specialist lists within an ordinary court, such as the Business and Property Courts of England and Wales; assigning technically trained judges to a docket; appointing expert assessors to sit alongside a generalist. Each buys some fluency without fully severing the subject from the main body of law.

Third, watch the appellate layer, because it is where the design is usually rescued. Even a fully specialised trial court commonly reports to a generalist supreme court, and that arrangement is not an accident of hierarchy — it is the mechanism that restores the outside view exactly where doctrine is set.

• The Unified Patent Court, opened in 2023, as a live test of whether cross-border specialisation delivers the uniformity it promises.
• Percolation — the argument that letting several courts disagree for a while produces better law than settling it early.
• Administrative tribunals, where the specialised body sits inside the executive rather than the judiciary, and what that changes.

Every term the collection defines is gathered in the glossary.