Statutory interpretation canons

Why do judges settle an ambiguous statute with shared rules of thumb rather than by asking what the legislature meant?

A statute says "dogs, cats and other domestic animals". A ferret arrives in court. The obvious way to settle it looks obvious: find out what the legislature meant — read the debates, the committee report, the minister's explanation.

Instead a judge is likelier to reach for a maxim, that a general term following a list takes its colour from the list, or that mentioning some things implies the exclusion of others, and settle the ferret with a rule of thumb nobody who voted ever wrote. Why prefer a canon to the actual evidence of intention?

Start by asking what "the legislature meant" would have to be. A body of hundreds voted. Some read the clause and wanted it broad, some narrow, most did not read it, and a few voted for the bill to get something else in the same package. What they jointly did was enact a text. There is no shared mental state to recover, and no procedure ever aggregated individual states into one intention — the votes aggregated preferences over the whole bill, not over the word "domestic". A legislature, as Shepsle put it, is a "they", not an "it" (attribution recalled).

That reason is negative: the evidence you want may not exist. It does not yet explain why canons rather than the judge's own view of the sensible result.

So ask what the court is for here. It is the second half of a two-party system: somebody drafts a text, somebody else reads it later, and neither is in the room with the other. What that pair needs is not accuracy about the past but predictability about the future. If the drafter knows in advance how a word will be read, the drafter can write the sentence that produces the intended result — and only then does the text carry meaning reliably at all.

Notice the shape. This is a coordination problem, and coordination problems are solved by conventions, not by truth-seeking. A canon need not be the most probable reading — it needs to be the known one. That is why canons are stated as flat maxims rather than probabilistic guidance, and why courts apply them to statutes drafted before the canon was announced.

Follow the loop and it strengthens itself. A court applies a canon. Legislative counsel notice and write the next statute to fit it, and drafting manuals in several jurisdictions codify exactly this. The convention becomes more reliable, so the court has more reason to keep it. Nobody chose it for its accuracy and it becomes accurate by being followed.

Now the price, which is large. Llewellyn showed in 1950 that the canons come in opposing pairs: follow the plain meaning, but avoid absurd results; expressing one thing excludes others, but a list may be merely illustrative. If the choice among canons is unconstrained, the discretion the canons were meant to remove has moved up one level, and is now exercised invisibly, under the appearance of technique.

Two further costs. Substantive canons — lenity, presumptions against retroactivity, clear-statement rules protecting federal structure — are not conventions about language at all but policy thumbs on the scale. Critics hold they let a court impose a value under cover of a drafting rule; defenders reply that a predictable thumb beats an unpredictable one and that a legislature can always speak clearly enough to override it. That dispute is live and I leave it so. The apparatus also advantages the repeat player: a well-resourced drafter knows the conventions, an occasional party does not, and pays for the ignorance.

The comparison is evidence, not decoration. Civil-law courts consult preparatory materials far more freely, as ordinary evidence of purpose. English courts long refused Hansard entirely, then admitted it in narrow circumstances (Pepper v Hart, 1993, recalled). European Union law is enacted in many equally authentic language versions, so a court cannot rest on the plain meaning of one text and is pushed toward purpose by the structure of the system itself. None of these is the default case.

Three refinements hold it together.

First, the coordination account does not make canons arbitrary. Many textual canons are compressed generalisations about how competent English is written — a general term at the end of a list usually is meant in the spirit of the list. Their authority is partly descriptive and partly conventional, and separating the strands is much of what interpretation scholarship argues about.

Second, this is a different question from whether to draft a rule or a standard in the first place, worked through in bright-line-rules.md. There the choice is made ex ante by a drafter deciding how much discretion to hand out. The fixed point of difference: canons operate ex post, on a text whose author has left, and their function is to make the reader predictable rather than the rule precise.

Third, the load-bearing empirical claim: coordination requires the drafters actually to know the canons. Gluck and Bressman's interview study of congressional counsel, published around 2013, found drafters well aware of some conventions and unaware of others, with substantive canons and deference doctrines faring worst (stated as recalled). That is the shape a coordination theory predicts if it is only partly true — conventions circulating through drafting manuals are real coordination, and ones invented by courts and never transmitted are something else wearing the same clothes.

Which gives the test. If canons work by coordination, their use should track transmission: those taught in drafting manuals should predict outcomes well, while canons known only to courts should show no drafting response and should correlate with the deciding judge rather than the text. The refuting observation: untransmitted canons predicting outcomes just as reliably, with drafters conforming anyway — which would mean they track something in the language itself, not a convention between two institutions.

Every term the collection defines is gathered in the glossary.