American readers, especially those of a textualist or originalist persuasion, will likely be familiar with the idea of corpus linguistics. As one well-known article promoting its use explains, it involves searching “for patterns in meaning and usage in large databases of actual written language” in order to clarify the meaning of legal texts that would otherwise be ambiguous or vague. But can techniques involving computer analysis of text help us understand unwritten rules―say, the United Kingdom’s constitution?
In his article, The Changing Concepts of the Constitution, Alex Schwartz sets out to do just that. Using big-data wizardry, Schwartz explores the way in which members of the UK Parliament speak about the constitution and about constitutional concepts such as parliamentary sovereignty, human rights, and the rule of law, and discerns changes in their use over time. The exercise is enlightening both to those studying the UK constitution itself and also, I submit, to those who are interested in the project of understanding constitutions, in all their diversity.
The UK constitution, one wit has said, “is no more and no less than what happens. Everything that happens is constitutional. And if nothing happened that would be constitutional also.” This is often paraphrased as “the constitution is just what happens.” But that is not quite right. Although legally free from binding constraint, the UK Parliament is nonetheless enmeshed in a web of ideas and understandings, some admittedly vague, about what the constitution allows and what it requires. These ideas and understandings can always be contested, and they change over time, partly in consequence of such contestation and partly, indeed, in consequence of “what happens.”
And a lot has happened in the last few decades: the UK’s marriage of convenience to and tumultuous divorce from the European Union; the consummation of a love-hate relationship with the European Convention on Human Rights, which begot the Human Rights Act 1998―which is now in strong danger of being disowned; Scotland and Wales becoming more autonomous with devolution, and in Scotland’s case, trying to decide whether to move out entirely. It seems plausible, and a number of scholars have argued, that these events and the debates surrounding them have changed the old understandings of the UK constitution, making it more legal and less political, more focused on rights and less on traditional notions of parliamentary sovereignty. Vernon Bogdanor, for one, has argued that the constitution has changed beyond all recognition.
Schwartz tests these claims by examining parliamentary debates about constitutional issues. American readers, mindful of the shape of political debate in recent years, may wince at the suggestion that this has anything to do with actual thinking about the constitution. Yet since Parliament has and frequently exercises the power to change the UK’s constitution, what is said there matters. As Schwartz puts it, “Changes in how political elites speak about the relevant concepts may have practical implications for how the boundaries of constitutional propriety are understood and enforced.”
But it is Schwartz’s method that makes his work especially interesting, both for its own sake and for the kind of results it enables him to get. The method is called “word embedding” and it produces a calculation of the degree to which various concepts are associated with that of the constitution itself, what he calls their “constitutional resonance.” More specifically, the more two words or phrases appear in similar contexts, the more closely connected they must be. If a word or phrase tends to be used in a similar context to “constitution,” the concept to which it refers must be closely related to that of the constitution in the minds of the speakers.
Such an analysis only becomes truly meaningful if applied to sufficiently large bodies of text. In this way, “the estimated change in constitutional resonance will reflect a change in the typical discourse surrounding that concept,” and not be influenced by what was said on a single occasion or even by the frequency with which a concept is debated. This is where big data techniques help. Schwartz was able to analyze the entirety of the parliamentary debates since Margaret Thatcher’s government took office in 1979 and identify all those in which reference was made to a range of constitutional concepts. The resulting corpus contained just over 200 million words, equivalent to roughly 750 Ulysses-length novels.
Separating it into sub-corpora according to time periods, one can also observe the changes, if any, in the constitutional resonance of the relevant concepts. As Schwartz explains, these changes are significant because “[a]s a concept acquires greater constitutional resonance—that is, as its meaning becomes more bound up with the meaning of the constitution—we would expect it to play a weightier role in contesting or determining prevailing notions of constitutional propriety,” and vice versa. The magnitude of these changes taken together also makes it possible to pass a judgment on claims, such as Bogdanor’s, about the UK constitution’s sub silentio amendment or even replacement.
Schwartz’s findings suggest no such dramatic change. While the resonance of “unwritten constitution” has steadily declined over the decades, it easily remains the most constitutionally resonant concept of all those he examines. The decline, to be sure, seems logical in light of constitutional changes such as the enactment of the Human Rights Act and of the (now-repealed) Fixed-term Parliaments Act 2011, as well as devolution. Logically too, “devolution” rose in constitutional resonance with the reforms of the New Labour government, but its rise has slowed considerably since then. Meanwhile, the constitutional resonance of “parliamentary sovereignty” declined―but it remains high, and the decline seems to have been arrested since 2015.
Other shifts in constitutional resonance that one might have expected seem not to have occurred at all. Notably, both “separation of powers” and “the rule of law” became less, not more, constitutionally resonant in the early 2010s, although they may have since recovered much of the lost ground. Surprisingly, too, “human rights” did not become more constitutionally resonant following the enactment of the Human Rights Act, although this may now be starting to change―perhaps, Dr Schwartz (plausibly) suggests, in response to repeated proposals for the Act’s repeal and/or replacement. In short, the way in which members of the UK Parliament speak about the constitution and related concepts has changed rather less than one might imagine since 1979.
What are we to make of this potentially surprising finding? One thing to note, as Schwartz does, is that members of Parliament are not the only people who speak about the constitution. Judges and legal academics do too, of course, and their constitutional discourse may diverge from that of members of Parliament, given the different institutional settings in which it occurs. Schwartz does not mention this, but one might also usefully take an interest in the discourse of journalists, and indeed―if perhaps with more practical difficulties―that of the general public. In the absence of a legally binding constitutional text authoritatively interpreted by the one institution (or, for that matter, even in its presence, albeit to a lesser extent), these discourses are all important in their own way. Nonetheless, as noted above, the parliamentary view of the constitution matters a great deal, and to the extent that it comes as a surprise to others who study and discuss the UK constitution, they may need to ask themselves why that is the case.
Beyond this, there are lessons to be learned for all those who take an interest in constitutions, and perhaps especially in comparing “written” and “unwritten” constitutions. Schwartz’s findings support the view that “unwritten” constitutions, despite their superficial flexibility, can remain quite stable, at least so far as their fundamental structure is concerned. (I have elsewhere made this case in reference to New Zealand.) At the same time, one must keep in mind that the stability of background constitutional understandings is compatible not only with significant constitutional reform, which has occurred during the period studied by Dr Schwartz, but also with what has been described—rightly, in my view, as “constitutional vandalism,” such as Prime Minister Boris Johnson’s unsuccessful effort to suspend Parliament in 2019. To be sure, polities with “written” constitutions are not immune to “switches in time” either. But it may well be the case that the temptation to accept that “everything that happens is constitutional” is stronger in those with unwritten ones―and at the same more difficult to guard against.