Persuasion — Yascha Mounk’s new media venture, which bills itself as “determined to defend free speech and free inquiry against all its enemies” — has published an analysis of historical surveys on self-censorship. The surveys ask respondents whether they feel free to speak their mind, and the authors — James Gibson and Joseph Sutherland — detect a significant upward trend. In 1954, only 13% of respondents said they self-censor; last year, that number was at 40%.

I think the way this analysis is being used to warn of a growing “culture of orthodoxy” is overly simplistic, but first I would like to spotlight a few technical concerns.

Technical issues

Here I’ll limit myself to pointing out technical issues that the authors themselves seem to acknowledge in the paper their post is based on, even if these hedges all happen to be absent in the Persuasion post.

For one thing, the authors concede that “Some might object to this database because the years coverage of surveys in the period prior to 2005 is sparse indeed.” This is easy to miss in Persuasion’s chart of self-censorship survey results where the X axis skips directly from 1954 to 1973 to 1987 to 2005. But displayed to scale, the data actually looks something like this:

Can we derive much historical analysis given these massive decade-long data gaps? Is the recent cluster of 40+ scores part of a gradually rising trend, or a historical hiccup that is already receding? Hard to say.

Elsewhere in the paper, the authors make another concession: “these data are not, strictly speaking, directly comparable owing to different modes of survey administration (and other factors).” Because this is a compilation of surveys that were all administered in different ways, one can imagine systematic biases creeping in that might have skewed earlier numbers downwards and more recent numbers upwards. The authors dismiss this issue by insisting that

the trend in these data is completely obvious and is strong enough to defeat any explanation of the trend in the data based on survey mode.

But their post is about more than just a trend; it is about the magnitude of a trend, trumpeting that the “percentage of Americans afraid to share their political views has tripled.” Since this shift consists of a 27 point increase since the 1954 data point and a mere 8.2 point jump from the 2005-2011 average to the 2013-2019 average, even relatively modest polling biases could undermine the flashy “tripled” headline while leaving a slighter trend intact.

Early on in their paper, Gibson and Sutherland write that

Our analysis is assuredly not comprehensive, but …our findings are too important to ignore. Thus, our imperative in this paper is to spur additional research…

Read enough of this sort of literature and you’ll be familiar with this rhetorical move: what it often means is something like “huge, if true.” Like the authors, I also think that this data probably demonstrates a measurable shift in reported self-censorship, particularly beginning in 2005. But the data before then is sparse, and the comparative problems may very well be significant enough that I would hesitate to make the dramatic and unhedged claims that their post makes.

Self-censorship and intellectual freedom

Setting all of these technical concerns aside, I’d like to consider a more profound question: what is the relationship between self-censorship and intellectual freedom? To see where I am going with this, let’s distinguish between three different scenarios:

1. Conscious and deliberate self-censorship;

2. Never engaging in self-censorship because cultural hostility towards taboo ideas was too intense for you to ever even consider adopting them;

3. Never engaging in self-censorship because taboo ideas have been so thoroughly exiled from an intellectual culture that you were never even aware of them.

All three of these scenarios illustrate cultures that are hostile to free thought and expression. And in fact, one could argue that among these, the culture of self-censorship is the most free — because at least there, one had the opportunity to consider taboo ideas and adopt them, even if one does not feel free to talk about them. To put this in Orwellian terms, (1) is where Winston Smith spends much of Nineteen Eighty-Four, guilty of thought crime but engaged in self-censorship; (2) is closer to crimestop, the unthinking hostility that citizens have developed towards subversive thought; and (3) resembles the world of ideas imposed by Newspeak and the Ministry of Truth, constrained so rigidly that citizens rarely even encounter subversive ideas in the first place.

These variations on the same kind of problem are worth distinguishing because the Gibson-Sutherland analysis really only measures (1) — even though their commentary often proceeds as if they’re talking about all three. Here for example is where they arrive at the end of their Persuasion post:

…high levels of self-censorship should be treated as an ominous warning sign. They signal the development of a culture of orthodoxy that is animated by a false sense of certainty about what is true and what is false—and a proud intolerance of those who might dare to voice an opinion that conflicts with the mainstream.

Is this actually true? Consider the problem of McCarthyism. It’s entirely possible that people consciously and deliberately self-censor regarding their views on socialism more than they did in the 1950s. But is this because there was less of a “culture of orthodoxy” back then than there is now? Or is it because red scares and decades of anti-socialist propaganda in the US had imposed such rigid ideological orthodoxy on the population that no ever even considered saying something about socialism that they might have to self-censor?

Suppose that the cultural and institutional mechanisms that encourage self-censorship about socialism were just as potent in the mid-twentieth century as they are today — but that public support for socialism has significantly increased, as polling seems to indicate. Would we not, then, expect to see an increase in reported self-censorship, not because our culture has become less intellectually tolerant, but rather because people dare to consider ideas that they would not even consider before? If so, then we may indeed see an increase in scenario (1), but only because scenarios (2) and (3) are less common.

When there are more people than there used to be who say that they are self-censoring, a simple way to interpret this is just to do what Persuasion has done and conclude that intellectual freedom is on the decline. But one can also read this trend quite differently, I think: one can also conclude that people are thinking about things they simply never thought about before, and that they are paying more attention to the problem of self-censorship than they have in the past. The real state of affairs, I suspect, is probably some combination of these two interpretations — but to really understand it, we’d need much better polling, and a much more sophisticated theory of what cultural orthodoxy really means.