Twice previously, I’ve critiqued articles written by three cognitive-sciences co-authors, Eric Martinez, Frank Mollica, and Edward Gibson.
This 2022 blog post considers their article that aims to show that contracts are poorly written. That article tells us nothing we don’t already know.
And this 2023 blog post considers their article examining why lawyers write in such a convoluted way. Because of flaws in the design of the underlying experiments, that article is seriously misleading.
Now we have a third article, Even Laypeople Use Legalese, available here. (As with their previous two articles, a summary of the findings is available in a more mainstream publication—the ABA Journal (here).) As with their second article, this one too is poorly designed, rendering it seriously misleading.
The Third Article
Here’s the gist of the new article:
And although recent work has revealed that even lawyers prefer simplified legal documents over complex legal documents (Martinez et al., 2023), it remains an open question how complex features such as center-embedding make their way into legal documents in the first place.
To answer this question, we conducted two high-powered pre-registered experiments testing two leading hypotheses for why lawyers write the way that they do, including: (a) the magic spell hypothesis, according to which lawyers and lawmakers write in a convoluted manner in order to lend legal documents a ritualistic, spell-like element; and (b) the copy-and-edit hypothesis, according to which conditions and specifications are often considered only after the creation of an initial draft and are more easily embedded into the center of existing sentences as opposed to being written-out into separate sentences.
In line with the magic spell hypothesis, we found that people tasked with writing laws wrote in a more convoluted manner (i.e. more center-embedding) than when tasked with writing control texts of plausibly equivalent conceptual complexity. Contrary to the copy-and-edit hypothesis, we did not find evidence that people editing a legal document wrote in a more convoluted manner than when writing the document from scratch.
These findings suggest that lawyers and lawmakers write in a complex manner in order to confer legal documents a ritualistic, spell-like element, presenting broad-ranging implications for law, policy and human cognition.
(You might be wondering, What are center-embedded clauses? The third article says only that in center-embedded syntax, “clausal content is embedded within the center of another clause as opposed to being edge-embedded or written as a separate sentence.” For the general reader, that’s isn’t as clear as it might be. The first article offers as an example of center-embedded clauses defined-term parentheticals used to create integrated definitions (my terminology). In effect, what they mean by “center-embedding” is that the prose features sentences that are relatively lengthy.)
The Magic-Spell Hypothesis—The Experiments
Let’s consider the magic-spell hypothesis. (I added a hyphen. Because phrasal adjective.)
To test the hypothesis, the authors devised a first experiment:
In Experiment 1, we evaluated both the Magic Spell Hypothesis and Copy-and-Edit Hypothesis. To evaluate the predictions of these hypotheses, we conducted a preregistered experiment in which we asked participants (n=200) to write either a (a) legal provision prohibiting a crime; or (b) a story describing someone committing that crime.
And they devised a second experiment:
The law condition [of Experiment 2] was identical to the law condition in Experiment 1, and consisted of instructions asking participants to write a law prohibiting a crime. In the description condition, participants were asked to write an unofficial description of of a law prohibiting a crime.
Here are the results of Experiment 1 relating to the magic-spell hypothesis:
In line with the predictions of the magic spell hypothesis, participants’ responses contained a higher percentage of sentences with center-embedding in the law condition (48.1%; 95% CI: 46.0 to 51.1) compared to responses in the story condition (5.8%; 95% CI: 5.2 to 6.2).
Here are the results of Experiment 2 relating to the magic-spell hypothesis:
As in Experiment 1, in line with the magic spell hypothesis, participants were more likely to produce sentences containing center-embedded clauses in the law condition (54.6%; 95% CI: 50.3 to 59.1) than in the control condition (25.7%; 95% CI: 22.5 to 28.9).
The Magic-Spell Hypothesis—The Flaws
But as regards the magic-spell hypothesis, the article exhibits design flaws of the sort that sabotaged their second article—the results of the experiment cannot support their hypothesis. That’s because for two reasons, one can easily come up with other explanations for the results.
First, the differences noted in the results could be explained by differences in the kinds of prose involved: narrative prose and expository prose are generally less constrained than the prose used to express laws. Like the prose of contracts, the prose of statutes is limited and stylized, because it primarily regulates conduct and allocates risk.
And second, the differences noted might be explained by motives other than an urge to give a ritualistic, spell-like vibe to prose. For example, perhaps as a result of a lifetime of observing how bloated and turgid legal prose is generally, participants in the experiment thought that that’s what legal prose should look like, so they tried to achieve that effect. (For what it’s worth, I find that explanation way more likely than the magic-spell hypothesis.)
Given the design shortcomings, it’s not surprising that the conceptual underpinning of the magic-spell hypothesis is also uncertain. The article attributes a ritualistic, spell-like aspect to “performative utterances”—what I call “language of performance,” which expresses actions that happen automatically when the parties sign a contract. Here’s an example: Acme hereby transfers the Intellectual Property to WidgetCo. But language of performance is just one component of my “categories of contract language” framework; go here for my quick-reference chart. (Verb structures used in contracts are broadly comparable to those used in statutes.) Furthermore, I’ve analyzed contract verb structures extensively, and generally I’ve not encountered in language of performance much that puts me in mind of magic spells. The example in this paragraph is representative.
One kind of language performance might appear ritualized—granting language in license agreements. A license might be expressed as conveying the right to make, have made, import, use, have used, offer for sale, sell, and have sold licensed products. And a license might be modified by a string of adjectives drawn from sole, exclusive, (non)assignable, (non)transferable, (non)sublicensable, worldwide, royalty-free, royalty-bearing, paid-up, and others. See this 2016 article. But that’s just one kind of language of performance.
And that sort of thing isn’t limited to language of performance. For example, in my Drafting Clearer Contracts training I mention that the traditional recital of consideration routinely features something like this: for other good and valuable consideration, the receipt and sufficiency of which are mutually acknowledged. With its obscurity and with the rhythm afforded by the two doublets in quick succession, that does have something of an incantation about it. But it would be a stretch to think that it’s intentional. Instead, it’s simply the result of the generally acknowledged legalistic appetite for redundancy (the result of a knee-jerk aversion to risk) and appetite for obscure terms of art (the result of the copy-and-paste machine preserving outdated terminology long after it has become irrelevant). The same applies to aspects of license-granting language.
The Copy-and-Edit Hypothesis
We don’t need to spend time on the copy-and-edit hypothesis, because the authors didn’t find evidence that editing a legal document results in prose that’s more convoluted than does starting from a blank screen.
At least this hypothesis doesn’t rely on just one of several possible explanations. Instead, it relies on observable outcomes. Nevertheless, it’s counterintuitive. I’ve long been an observer of copy-and-pasting. It’s the defining feature of how contracts have been drafted. But it has never crossed my mind that revising copy-and-pasting might result in more convoluted prose. Instead, you’re pretty much stuck with what the copy-and-paste machine gives you, and usually it reflects the dysfunction of traditional contract drafting. You make whatever changes your circumstances require or permit, and you move on. (Incidentally, there’s scholarship on ways that lawyers make bespoke changes when copy-and-pasting contracts. See, for example, Stephen Choi, Mitu Gulati, Matthew C. Jennejohn & Robert Scott, Contract Production in M&A Markets, 171 U. Pa. L. Rev. 1881 (2023).)
Why Write This Post?
So one hypothesis is conceptually misguided, the second doesn’t seem to reflect the implications of copy-and-pasting, and the design of Experiment 1 is seriously flawed. And it follows that the “Discussion” part of the article can be disregarded. So why am I writing about this article?
Policing online commentary is generally futile. (The definitive take is the “Duty Calls” cartoon by xkcd. You know the one: “Someone is wrong on the internet.”) I’ve largely kicked the habit. Now, I might sporadically roll my eyes or grit my teeth, then usually I move on.
But at some point, it might be worth saying something. I can think of three reasons why that’s the case with this article.
First, a few people tagged me in posting about it, so it appears they’d be interested to hear what I think. That might be justification enough for this post.
Second, the third article has attracted attention. Someone posted about it on social media; that post prompted 560 comments, when I last checked. And there’s that article in the ABA Journal. Consider this my attempt at a corrective.
And third, in an email to me regarding a draft of my 2023 blog post, Martinez said a lot of it looked “pretty reasonable” but that he had some comments. He and co-authors ultimately decided not to offer me any comments. And they cite the second article in the footnotes to the third article, as if it weren’t subject to challenge. This post is my way of reminding the co-authors that I have no reason to think my critiques aren’t valid.
Individual Attitudes to Legalese Might Be Irrelevant
But perhaps a bigger issue is that the sort of experiments Martinez, Mollica, and Gibson describe in their articles might be doomed to irrelevance.
In the decades I’ve spent observing how people work with business contracts, I’ve seen that many are wedded to systemic dysfunction, in terms of what contracts say and how they say it. But even if individuals might be willing to make their contracts clearer, more concise, and more relevant, organizations tend to resist that sort of change. Changing a precedent-driven system is hard.
It should be possible to come up with a rigorous set of experiments that allow us to accurately assess individual attitudes to legalese and, perhaps, explain those attitudes. But legalese holds sway not because of individual attitudes to it. Instead, it’s a cultural artifact that has been built over centuries. For most of us, its momentum is unstoppable, subject only to limited tweaking.
So I don’t think further experiments will make things better. They won’t help us make business contracts clearer, more concise, and more relevant, and they won’t help make the process of drafting them more efficient. Those challenges require other solutions.