La Cour, c’est qui?

Peter McCormick, University of Lethbridge

The Supreme Court’s recent decision in Comeau has definitely put the judicial cat among the federalist pigeons.  At first glance – we have all seen the headlines – the case is about bringing cases of cheap beer into New Brunswick (“Free the Beer!”).  On a closer look, the already enfeebled Section 121 of the Constitution Act 1867 has been effectively gutted, taking with it any realistic prospect of a major shift toward greater intra-Canadian free trade.  Along the way, the sort of trial judge’s revisiting of precedent that was so highly lauded in Bedford has been severely chastised.  An interesting case, therefore, on several levels.

The decision took the somewhat infrequent form of a “By the Court” judgment – one that is both unanimous and anonymous – which arguably makes it more emphatic while coyly veiling the identity of the judge who did the drafting.  But the curtain of anonymity can be brushed aside to identify the lead author, or at least to establish solid relative probabilities.  That identity will come as no surprise, but the methodology I will describe takes it some distance beyond simple conjecture.

That methodology is function word analysis.  Function words are the words that express grammatical or structural relationships between other words (prepositions, pronouns, determiners, conjunctions, auxiliary verbs and particles), as distinct from the content words (nouns, adjectives, verbs and adverbs) that convey more concrete meaning.  Function words are the words that everybody uses, but different people use with different frequencies and proportions, so much so that these frequencies and proportions provide a literary fingerprint.  There are about 300 function words in the English language; my more focused function word list is drawn from the literature, and modified to reflect the actual usages of the Supreme Court over the last twenty years.  It involves the 44 most frequently used function words, some of which are totals for related words such as “a” and “an”, or the different tenses of common auxiliary verbs like “to be” or “to have” or “to do”; together, these words accounted for a rather remarkable 40% of the total word count – overall and for every one of the judges.  This was used to generate a word-usage profile from the written reasons attributed to each judge, and these in turn can be compared with the parallel profile of any specific anonymous decision.  (The logic and procedures of the methodology are described at length in my article in the Dalhousie Law Journal.[1])  The point is to calculate a “Similarity Index”, summing for the 44 words the absolute value of the differences between that judge’s word-use frequencies and those that appeared in Comeau.  The lower the score, the more likely it is that the particular judge was the lead writer.

Language is a possible problem – because counting words within even a superbly translated version will tell us as much or more about the translator than about the original writer – but the Supreme Court Reports assures us (by describing the English reasons as “the judgment” and the French reasons as the “version francaise”) that the original language of the Comeau decision was English.  This also limits the number of “suspects” for the lead writer; I am assuming that Gascon and Cote would have written in French, such that the French language text would have been “le jugement,” and the English language text the “English version”.

Quotations are also a problem – extensive direct quotations distort the word counts by reflecting the usage patterns of the quoted writer, rather than those of the immediate writer.  My solution is to delete all direct quotations from the examined text.  Some Supreme Court justices quote very extensively, to such an extent that quotations can make up a quarter or more of the total word count.  For the Comeau decision this proved to be a negligible factor, reducing the word count by less than 4%.  As I will indicate below, this unusually low quotation count is itself a pointer to the identity of the lead writer.

Law clerks can be a problem, because they may have contributed early drafts for at least the more routine parts of the judgment.  My solution was to eliminate these more routine parts (the introduction, the background, the decisions of the lower courts) and focus only on the much longer analysis section.  This further reduced the word-count by about 20%, but it left 11,000 words and this is easily enough for the function word analysis to operate with credibility.

An adequate comparison basis is a problem; both Brown and Rowe have been appointed recently enough, and have had such a limited opportunity to write judgments or minority reasons, that there is not a large enough body of words to provide a reliable basis for comparison.  Seniority is a large enough factor in decision assignment generally, especially for major cases and especially for constitutional cases, that it would in any event have been unlikely that either of these more junior members of the Court would have been doing the lead writing.

Finally, the “circulate and revise” process pursued by the Supreme Court can be a problem.  All indications are that the other members of the panel take this very seriously, such that the lead writer’s initial draft can undergo significant revision as a result.   My “fingerprint” metaphor above should be qualified to recognize that what is available for analysis may be a smudged rather than a perfect fingerprint.  However, checking results back against the handful of By the Court decisions whose authors have actually been identified in judicial biographies has validated the methodology even for reasons that are described as having undergone major revisions. (Most dramatically, it revealed the “did not participate” Le Dain as having been the initial lead author of Ford and Devine, a finding that has been confirmed by both the Dickson biography and a recent CBC radio documentary).

Running this process for the Comeau decision, restricting the enquiry to the five senior judges who normally write in English, yields the following results:

Lower scores pointing to a more likely author, function word analysis points to McLachlin.  Readers may initially be disappointed because the spread between individual judge’s scores are modest, but the tug of ingrained writing habits makes this meaningful.  A smoking gun this may not be, it provides a rank ordering for the likelihood of lead authorship, and McLachlin is clearly indicated.

Moreover: the middle row in the table is significant in way that allows us to ratchet up the language with which to describe the findings.  This provides the similarity score the word by comparison with an all-judge figure based on a combined total of four million words over a twenty-year period.  Karakatsanis, Moldaver and Wagner are less like Comeau than is that all-judge figure; McLachlin – and only McLachlin – is significantly closer to Comeau than is the all judge figure.  This makes the findings more decisisve than might have appeared at first glance.

Further: I mentioned earlier that eliminating direct quotations from Comeau reduced the total word count by only about 4%.  For the McLachlin Court’s constitutional cases more generally, the average figure for such quotations was 13.5%.  But this, too, is a distinctive and persisting characteristic of individual judges:  some quote extensively and some do not.  Abella, for example, frequently uses direct quotations, accounting for fully one-quarter of the words in her constitutional decisions, almost double the average.  McLachlin, however, does not; direct quotations account for only 6.5% of the total words in her numerous constitutional decisions, less than half the all-Court average.  This reinforces the suggestion of the similarity scores that McLachlin is the most likely lead writer of the Comeau judgment.

It is somewhat frustrating that one can create a large data-base, run detailed calculations, generate complex indices – and then wind up with a conclusion that simply confirms what was the most obvious guess from the beginning.  (Who needs science when hunches work so well?) Beverley McLachlin has led the Court for almost 20 years, longer than any other Chief Justice in the Court’s history.  During that time, she has delivered a disproportionate share of the Court’s constitutional decisions, and this statement remains true even if one pro-rates the counts to accommodate the fact that no other member of her Court has served the full eighteen years.   Comeau is one of the last major constitutional decisions with which she will have been involved, and arguably the most significant federalism case of her Chief Justiceship; if there is any surprise, it is that she chose to write behind the veil of “By the Court” rather than over her own name.

[1] Peter McCormick, “Nom de Plume: Who Writes the Supreme Court’s ‘By the Court’ Judgments?” Dalhousie Law Journal, Vol. 39 (2016) 77

Author: Leonid Sirota

Law nerd. I teach public law at the University of Reading, in the United Kingdom. I studied law at McGill, clerked at the Federal Court of Canada, and did graduate work at the NYU School of Law. I then taught in New Zealand before taking up my current position at Reading. View all posts by Leonid Sirota