September 07, 2026

Cass Sunstein (Harvard) has posted “On Not Being Trapped in Amber” on SSRN. The essay has the dual virtues of being interesting and short, making it well worth reading in its entirety. Sunstein’s topic is the use of analogical reasoning, or judgments of “relevant similarity,” in cases such as United States v. Rahimi and Trump v. Barbara (his two principal case studies).

Sunstein argues that “whenever analogical reasoning is in play, judges cannot avoid . . . to some extent normative choices.” This is because “a judgment of relevant similarity is not, in the legal context, a judgment about a fact. It is not a matter of history or of original public meaning. It is a normative judgment.”

I think that this sweeps too broadly. Sometimes a judgment of relevant similarity is a normative judgment, but sometimes it is a descriptive or “factual” judgment. Questions of relevant similarity arise when it is given that Y is F and the question is whether X is “relevantly similar” to Y such that, if Y is F, then X is also F. Judgments of relevant similarity are normative if and only if F is a normative feature and descriptive if and only if F is a descriptive feature.

For example, where it is given (by the results of clinical trials, say) that rats can ingest Experimental Drug without thereby experiencing increased blood pressure, one might ask whether humans are relevantly similar to rats. Here, a judgment of relevant similarity is a descriptive judgment: a judgment that humans are relevantly similar to rats such that, if rats can ingest Experimental Drug without thereby experiencing increased blood pressure, then so can humans.

I have suggested elsewhere such judgments can be understood in terms of Bayesian updating. But what matters here is that they can be either normative or descriptive (or, for complex judgments, a mix of both), depending on the feature with respect to which the alleged analogs are claimed to be similar.

Now consider Sunstein’s two case studies. In the context of Second Amendment history-and-tradition analysis (Rahimi), it is plausible that Sunstein is right: judgments of relevant similarity are normative judgments. Plausibly, the Second Amendment was intended to constitutionalize something like a general-law right subject to reasonable regulation, with courts filling out what counts as “reasonable” in a common-law fashion that accepts as precedents not only prior judicial decisions but also sufficiently longstanding or widespread regulatory traditions. This is view is elaborated and defended by Sherif Girgis in a forthcoming Harvard Law Review article entitled “Preexisting Rights: ‘History and Tradition’ and Levels of Generality” (alas, not yet on SSRN); for a contrasting view, see here. If Girgis is right, then–subject to complications that Girgis discusses but that we can bracket here–whether the Rahimi majority was right to treat the regulation at issue as “relevantly similar” to longstanding surety and “going armed” laws turns on whether, taking as given that surety and “going armed” laws are reasonable, one should deem the regulation at issue in Rahimi reasonable too. To answer that question, one must construct the most plausible theory of what counts as a “reasonable” regulation that respects the constraint that surety and “going armed” laws are reasonable and ask whether, on that theory, the regulation at issue in Rahimi is reasonable too. (This is of course very Dworkinian, as Sunstein notes, and Girgis does too.) Because being reasonable is a normative feature, this analysis is indeed “inescapably normative” as Sunstein suggests (see also here).

But in the debate about United States v. Wong Kim Ark’s exceptions to birthright citizenship (Barbara), it is plausible that judgments of relevant similarity are descriptive. Here, the relevant feature is being “subject to the jurisdiction” of the United States within the meaning of that phrase in section 1 of the Fourteenth Amendment. One must construct the most plausible theory of what this phrase means that respects the constraint that those born in the United States who fall into one of Wong Kim Ark’s exceptions are not subject to the United States’ jurisdiction and ask whether, on that theory, those covered by the executive order at issue in Barbara also are not subject to the United States’ jurisdiction. On the plausible assumption that being “subject to the jurisdiction” of the United States within the meaning of that phrase in the Fourteenth Amendment is a descriptive feature, this analysis is not “inescapably normative” as Sunstein suggests. And for the originalist, the analysis is a matter of “original [] meaning.”

The same reasoning applies to questions about what constitutes a Fourth Amendment search, a topic that Sunstein mentions briefly. Plausibly, whether something constitutes a “search” within the meaning of that term in the Fourth Amendment is a factual question. It certainly is for the originalist. Reasoning by analogy to activities that are given to be “searches” in this sense is therefore a descriptive rather than a normative exercise.

Three additional points are worth mentioning. First, even when judgments of relevant similarity are normative, it does not follow that they are unconstrained in the sense that the judge is free to decide as she thinks best. A judge might consistently believe that (1) surety and “going armed” laws on the one hand, and the regulation at issue in Rahimi on the other, are all unreasonable; but (2) assuming the former are reasonable, the most plausible explanation for why implies that the latter is reasonable too. Left to her priors, such a judge would invalidate the regulation at issue in Rahimi. But constrained to update her priors around the assumption that surety and “going armed” laws are reasonable, she would (assuming she is honest) uphold the regulation. Girgis emphasizes this point, and I suspect that Sunstein would agree. Of course, the same point applies to the originalist’s descriptive judgments about, say, the meaning of “search” in the Fourth Amendment or “jurisdiction” in the Fourteenth: the originalist may be constrained by precedent to reach a conclusion that is contrary to what she believes is true as a matter of first principles.

Second, Sunstein characterizes judgments of relevant similarity as “highly discretionary.” In one sense, he is right. To make a judgment of relevant similarity often is to exercise “weak discretion” in Dworkin’s sense: the judge must consult her own, perhaps debatable or controversial priors when constructing what she takes to be the best theory that respects what she must take as given. But it does not follow that to make a judgment of relevant similarity is to exercise “strong discretion” in Dworkin’s sense, that is, to render a decision where the law has “run out” without directing the judge one way or the other. The judge constrained to take as given that Y is F must construct the best theory of why and ask whether, on that theory, X is F too. The exercise of constructing and applying the best theory may involve weak discretion, but—assuming that there is, in fact, one best theory and that it has something to say about whether X is F—it involves no strong discretion. Sunstein notwithstanding, there is a real sense in which the exercise involves “finding a principle” of decision rather than making one. For more on this point, see here.

Third, regardless of whether you agree with his analysis, Sunstein does us all a service by attempting on the Justices’ behalf to cash out the metaphor “trapped in amber.” As Michael Ramsey puts it:

[Uncashed out] metaphors such as trapped in amber are unhelpful and distract from analysis.  If you can’t express your analysis without a metaphor, likely you are eliding a flaw in your analysis. And if you could, but aren’t, expressing your analysis without a metaphor, then the metaphor is a rhetorical device to strengthen an otherwise weak argument. I would just say “fixed” or “not fixed”, and be suspicious of people who don’t.

 

Posted at 3:05 PM