Introduction

I began the research I’m sharing with you today not from writing about the sociology of reading nor the crises of the humanities, but from small discoveries in two big datasets. When I first found them, I assumed that they must already be widely known. (And perhaps they are to some of the people in this room.) But, in general, that turned out to be false. I want to bring these surveys to the attention of literary studies for reasons that, I hope, will be both persuasive and obvious.

We’ll begin at the end by previewing three numbers about reading for pleasure in the US over the past thirty-ish years. Then, I’m going to explain where those numbers came from, say something about what they might mean, and make one suggestion about how literary studies might respond.

First, the numbers:

SLIDE

• Relative to 1992, the odds that US people would report reading literature in the preceding year have decreased by 44%.
• Relative to 2003, the odds that US people would report reading anything for personal interest in the preceding day have also decreased by 44%.
• In 2022, only 17% of US people reported reading anything for personal interest yesterday.

I’ll specify each of those claims and where they came from momentarily. But they make two important points obvious: As points 1 and 2 suggest, the odds of US people reading literature or anything else have gone down by almost half in the past thirty years. And, as the third point makes clear, those relative declines are coming out of leisure reading rates that weren’t very high to begin with.

Literary study’s numbers

If people in literary studies don’t know these numbers, they do know something about two other numbers that are important to the field: the numbers of academic jobs and majors. I would argue that these statistics about reading for pleasure should be a part of every conversation in which we invoke the numbers of jobs and majors because they impact both of those numbers.

The US Census Bureau has been collecting statistics about reading for pleasure since 1982 for two surveys from which I calculated these numbers: the Survey of Public Participation in the Arts and the American Time Use Survey. The Census collects this data on behalf of the National Endowment for the Arts and the Bureau of Labor Statistics, respectively.

For this research, I have compiled historical data about pleasure reading from both surveys in one place for the first time, incorporating data newly published in February, some of which these agencies do not publicize. These analyses I’m presenting are not in any government report.

When these statistics have received attention from liteary studies and the public—notably in the early 1990s and the mid 2000s—conservatives have often played the role of Cassandra, prophesying the end of reading based on these statistics. Unfortunately, I believe that this political alignment has caused literary scholars to disregard good data about changes in reading because it has been used to fight the canon wars from the right. But our interpretation of and response to the data doesn’t have to lean rightward. Moreover, intuiting a decline in reading without a sense of its rate or magnitude has prevented us from asking good questions about rapid changes in the core prerequisite for literary study: voluntary reading.

Ultimately, I’m going to suggest one way that the field can and should respond to the data: Study whether the discipline causes its students to read more for pleasure than we would expect based on their educational attainment alone.

Outline

SLIDE

As you can see here, we are going to discuss the methodologies, findings, and new analyses of three surveys that the Census Bureau conducts with questions about reading. Then, we’re going to talk about how that data relates to the throughline of John Guillory’s new book, which he terms the overestimation of literary study’s aim.

The Current Population Survey

SLIDE

In order to talk about both the Survey of Public Participation in the Arts and the American Time Use Survey, we first need to talk about the Current Population Survey, or CPS.

The CPS measures labor force participation rates. You are almost certainly familiar with its most famous statistic, which is the unemployment rate.

The CPS “is a monthly sample survey of 60,000 eligible households…conducted using a combination of live telephone and in-person interviews.”¹ They use a probability sample to estimate state and national population characteristics, weighting for factors including race, gender, and education.² Eligible survey respondents are 16 or older and neither in the military nor institutionalized.

While the primary purpose of CPS is to gather data about the labor force, it also fields a number of supplemental surveys, which measure all kinds of things like whether Americans use tobacco or have internet access at home. The data we’re considering first, from the Survey of Public Participation in the Arts, comes from one such supplement to the CPS. And the data that we’ll be looking at second, from the American Time Use Survey, selects its respondents from the CPS pool of households.

The Survey of Public Participation in the Arts

SLIDE

The National Endowment for the Arts (NEA) was created in 1965 and started what would become its research division in 1975.³ The research division created the SPPA to study the topics in their mandate. After the research division began issuing reports using SPPA data in the early 1980s, its work was used soon as ammunition in both the reading wars and canon wars.⁴

To the extent that literary scholars know about the SPPA, it is almost certainly by way of these research reports. The best known of these is the 2004 Reading at Risk, which showed declines in all forms of reading across all groups surveyed.⁵ The New York Times covered this report under the headline, “What Johnny Won’t Read.”⁶

Inflaming a panic that he had cretaed, Mark Bauerlein, who directed George W. Bush’s NEA Office of Research and Analysis during the writing and release of Reading at Risk, wielded it as a weapon in his denunciatory book, The Dumbest Generation.⁷ In a 2008 interview about that book, when asked who was to blame for the stupefaction of young Americans, Bauerlein—an English professor emeritus at Emory and outspoken Trump supporter since 2016—responded, “I blame my colleagues, the humanities professors.” Why? Because they taught young people an “irreverence toward historical knowledge and literary understanding.”⁸ Yet Bauerlein’s shifting of blame onto his colleagues seems especially incoherent when the statistics he published in Reading at Risk show that the sharpest declines in reading occur among those groups who have never studied at the college level. Given this, one can easily imagine why Bauerlein’s colleagues—whether humanists or statisticians—might not have expected to learn much from what he had to say.

More recent NEA reports on the SPPA have drawn less controversy than Reading at Risk.⁹ Unfortunately, a reduction in controversy has corresponded with a reduction in attention. Ignoring recent reports because they have been less polemical has been a mistake.

I want to make three points about the SPPA. The first is the most important:

SLIDE

The NEA reports are not the SPPA data. It is possible to use the data without accepting either their calculations or conclusions. Second, NEA reports on the SPPA usually compare the present survey to the immediately preceding survey or two. They do not tend to look back at all preceding surveys, as I will momentarily. Shorter time horizons show real variance. However, the overall trajectory is clear.

Third, a degree of suspicion may cloud the SPPA for critics who are aware of these histories. The association between reading statistics and reactionary politics may seem like reason enough for many to disregard them. However, justified suspicion about the motives behind certain reports may have tended toward an unjustified suspicion of their underlying data.¹⁰ Taking the data seriously does not require us to give credence to reactionary arguments that have been made about it.

Methodology

The NEA’s research divison studies the impact of public participation in the arts on learning, public health, and economic activity, among other things.¹¹ To study that, they need to figure out how many US people participate in arts activities.

SLIDE

That means they have two distinct incentives for their surveys: Keep the questions consistent from one survey to the next to measure change over time, while updating the survey methodology to accommodate changes that impact arts participation. Obvious examples in the case of reading since 1982 would include reading on the internet, e-readers, and audiobooks, among other things.

The SPPA asks respondents the following questions about reading:

SLIDE

• With the exception of books required for work or school, did you read any books during the last 12 months?
  • If yes: About how many did you read during the last 12 months?
• During the last 12 months, did you read any…?
  • novels or short stories?
  • poetry?
  • plays?

SLIDE

This reading is voluntary, not done for work or school. Religious texts “count as books” for the first question. “Reading” is not restricted to the form of the codex; it counts “regardless of whether it was in books, magazines, or newspapers in paper form or online.”¹² E-readers and audiobooks also count.¹³

Finally, the SPPA makes a distinction between books and what it calls literature. Respondents are asked if they read any books. But they have never been asked if they read “literature.” Literature is the designation that the NEA uses to refer to the set of all respondents who reported reading any novels, short stories, poetry, and/or plays.

Data

With those details in mind, let’s look at the national SPPA data for reading any book and reading literature:

SLIDE

The y-axis shows the percentage of the US population who read either type of writing over the previous twelve months.

SLIDE

The reading of literary forms once (or more) per year passed from a majoritarian to a minoritarian activity by 2012. And, for the first time in 2022—data for which were first published this February—the same is true for reading any book in the preceding year.

We can also decompose that literature category into its constituent parts:

SLIDE

As you can see, literature readers and prose fiction readers are almost but not quite equal. There are a few respondents who only read poetry and/or plays.

SLIDE

Educational attainment is the best known predictor of reading. That relationship persists even as the probability of reading a book in the preceding year declines.

SLIDE

After education, reading researchers tend to regard gender as the second best predictor of reading. Note that the SPPA does not collect data on nonbinary folks.

SLIDE

Hispanic ethnicity is also a relatively strong predictor of reading in the US. I should say here that the Census Bureau uses definitions of race and ethnicity that match those of the current census.

Of the demographic characteristics I will mention, race is both the weakest predictor of reading, and the most difficult to compare since the terms used have changed over the SPPA’s history as the Census’s representation of race has changed. The only categories that have been measured consistently over this period and that also have narrow confidence intervals are what the SPPA terms white only and Black only.

SLIDE

While these graphs all give us a sense of directionality, it can be hard to reason about the magnitude of these changes in percentages alone.

SLIDE

Odds ratios can help control for differences in absolute participation—such as the small number of play readers compared to novel readers—while enabling us to see the change over time by comparing subsequent years to the 1992 baseline. Previous studies have used odds ratios to evaluate changing participation rates in the SPPA, and I have adapted their methods to the questions about reading.¹⁴

SLIDE

You can read this graph of odds ratios (and the ones that follow) by applying the following sentence structure to each value: “The odds that a respondent read literature within the past year were 44% lower in 2022 than they were in 1992.” The story is clearly one of decline—every value is negative relative to 1992—but not in a uniform way. The odds of reading of poetry and plays, for example, has vacillated.

Some would view these trends not as a deleterious effect of new technologies, but as a return from a state of exception to the status quo ante.¹⁵ The sociologist of reading Wendy Griswold argues that “most people in advanced industrial and post-industrial countries are not and will not be readers.”¹⁶ Instead, Griswold draws a distinction between what she terms a reading class and a reading culture. All societies with writing have reading classes, but few societies have ever had what Griswold calls reading cultures. The reading class is defined by its professional and nonprofessional reading, as well as its elite socioeconomic status. A reading culture, by contrast, “is a society where reading is expected, valued, and common.”¹⁷ In Griswold’s terms, panicky NEA reports and these declining odds suggest the further erosion of the US reading culture, which Griswold argues peaked from the late-nineteenth to the mid-twentieth century. If there is any truth in her distinction, it seems noteworthy that modern literary studies is an institutional artifact of that era, and the field may increasingly need to reconceptualize itself in relation to a reading class without a reading culture.

What remains unclear from the SPPA figures is the extent to which the rate of decline is increasing. The SPPA has been taken at uneven intervals of approximately five years, so drops have been jagged. Yet the ceiling is lower than it used to be. One of the values of the American Time Use Survey, to which we now turn, is that it gives us more detailed information about who is reading and the extent of their reading habit on the scale not of the previous year but of the previous day.

The American Time Use Survey

SLIDE

The ATUS is more recent and more frequent than the SPPA, having been conducted annually since 2003, except 2020 due to the pandemic.¹⁸ Work on the ATUS originally began in the wake of a bill proposed in 1991 that called for a survey to calculate the monetary value of unpaid work like homemaking.¹⁹ The ATUS appears to be even less well-known to literary studies than the SPPA.

The ATUS’s reading statistics differ from those of the SPPA in two key ways, one advantageous and the other disadvantageous. The advantageous difference is that the ATUS is more precise than the SPPA, recording the number of minutes respondents read yesterday as opposed to whether the respondent recalled reading anything within the preceding year. The disadvantageous difference is that, unlike the SPPA, the ATUS makes no distinction between literary forms.²⁰ That said, the reading measured by the ATUS includes nonprofessional literary reading as a subset of all reading it reports.

Methodology

SLIDE

Like the SPPA, the ATUS is also conducted on a representative subset of the Current Population Survey. One person from each CPS household is randomly selected to respond to the ATUS, and they report how they used their time over one predetermined day.²¹ Respondents are notified of their selection in advance and are instructed to keep a time use diary in minutes describing their activities that day. Each respondent is only interviewed one time, so year-over-year changes do not reflect longitudinal changes in individuals’ time use.²² Interviews are conducted on every day of the week in every month, so even if reading had seasonal variations, that would be accounted for in the data collection process. Census workers then aggregate specific time use reported by respondents in their diaries to more general categories.

SLIDE

We are concerned here with the ATUS’s questions about reading, specifically what they term “reading for personal interest.” The ATUS provides contemporary and historical code books for its datasets, showing through both abstract rules and concrete examples how different work and leisure activities diarists report have been classified over time.²³

SLIDE

One important type of reading this list may seem to exclude is the pervasive reading that N. Katherine Hayles has discussed under the rubric of “hyper reading.”²⁴ Some hyper reading may indeed fall under the ATUS category of “Computer use for leisure (exc. Games).” Yet it is noteworthy that, as with “reading for personal interest,” Census coders also specify and attempt to classify computer use in other categories the survey focuses on like work, volunteering, education, and household management before categorizing the remainder as computer use for leisure.

SLIDE

As you can see from this desk aid that Census workers use when classifying reading activities, “reading for personal interest” is a remainder category after other time use categorizations for reading have been attempted. Those categories include work, volunteering, education (including reading to children), and religious reading.

Any reading not done for these reasons is reading for personal interest, which is conceptually similar to the SPPA’s reading. To invoke a more familiar concept to literary criticism, we might instead think of this as reading for pleasure.²⁵

Data

SLIDE

In his 1910 guide to The Harvard Classics—the “five-foot shelf” of texts he deemed essential to a liberal education—Charles Eliot suggested that even if someone could “spare but fifteen minutes a day for reading,” they could educate themselves with these volumes.²⁶ As it happens, fifteen minutes a day is about the average amount of time US people spent reading anything for personal interest yesterday in 2022.

You see two opposite trends in this figure.²⁷ The average amount of reading time in the population has declined, yet the average amount of reading among readers (i.e., participants) has increased. If this sounds contradictory, it isn’t: A smaller proportion of the population is now reading, but those who still read spent more time reading yesterday than those who have lost the habit did.

As we did with the SPPA, let us also look at the odds ratio of whether people read for personal interest relative to the earliest value, which in this survey is 2003.

SLIDE

In 2004, the peak year for reading in the ATUS, about 28% of the population read for personal interest for more than zero minutes yesterday. In 2022, about 17% did. Since 2006, the odds ratios have all been negative, with an especially large and persistent decrease from 2010. By 2022, the odds of respondents reading for personal interest yesterday were 44% lower than they were in 2003.

Given the persistent decline after 2010, you’re no doubt wondering about the role played by reading’s most oft-cited enemies: television, games, and, above all else, the internet, especially via smartphones. The ATUS collects data on the role played by these during leisure time.

SLIDE

This chart shows the participation rates in leisure activities viewed as competition to reading. Note that, contrary to a popular perception of ever-increasing busyness, people in the US reported 5% more total leisure minutes in 2022 than they did in 2003. This means that people have an absolutely greater amount of time for leisure, which means they could theoretically participate in more distinct activities.²⁸

Now, we all know our own screen time and can dimly imagine our students’, so the low percentage of computer use for leisure is surely setting off some alarm bells. Two points on this: First, when reporting time, diarists are asked to identify what their primary activity was. If you watched TV for three hours but you were on your phone the whole time, you could report your primary activity as watching TV, and the ATUS would only count that. Second, as with reading, the ATUS works hard to classify computer use in other activity categories before putting the remainder in leisure. The most important such category would be Communicating/Socializing (120101). Interviewers prompt respondents to specify activities in their diaries, so if a diarist said “I was on my phone for an hour,” the interviewer would ask them to specify what they did during that time in order to apportion activities to the correct categories.

SLIDE

The proportion of the population reading has declined, and the proportions playing games and using computers for personal interest have both increased. To be clear, we can only say that these changes correlate. We can’t say definitively that people are trading reading time for computing or gaming time.

Reading by education

While all of the ATUS values presented thus far are weighted to represent the population as a whole, they can be decomposed into demographic groups.

When we break out the ATUS data by educational attainment, we see three noteworthy patterns.²⁹

SLIDE

First, even as reading for personal interest declines, education continues to strongly predict reading. The higher your education, the higher your chance of reading yesterday. Second, the odds of participation have declined for all groups since 2005. The third point is a bit subtler, but perhaps the most important: Since the ATUS began in 2003, none of these educational groups has been more likely than not to read for personal interest on any given day. At best, slightly over one-third of people with BA reported reading for personal interest yesterday. Not reading is the default.

Odds ratios reveal how dramatic this story is. But they also suggest how different educational groups’ odds of reading are declining at rates that are more similar to each other than their baseline reading rates.

SLIDE

In 2022, people with a BA had a 44% lower chance of reading anything for personal interest than they did in 2004. But educational attainment still matters a great deal: Those with less than a high school education had an 81% lower chance of reading in 2022 than they did in 2004.

Reading by gender

SLIDE

After education, gender is the best predictor of reading. Women tend to read more and more often than men. This is not a phenomenon that is specific to the US.

This chart shows the different participation rates between US men and women reading for personal interest. Like the SPPA, the ATUS does not have data on nonbinary respondents. As with education, both groups decline, albeit unevenly. Women in 2022 read for personal interest at rates comparable with men in 2009.

What none of this data tells us is whether there is a floor for reading today, or what that floor might be.

SLIDE

Griswold has argued that,

“professional members of the reading class are evangelists, fighting at the front line of culture to convert people to reading. Examples include teachers, professors, writers, editors, publishers, journalists, and…librarians.”³⁰

Yet, in my experience, there is little discussion within literary studies as to whether—to stick with Griswold’s terms—the field recruits people to the reading culture, or preaches to the converted of the reading class. This is not discussed in part because, without knowledge of these statistics, it would be hard to see that there might be a difference between the two.

On the overestimation of aim

Literary studies’ relative disinterest in nonprofessional reading is a byproduct of its overriding interest in professional reading.³¹ Elizabeth Long has described literary studies’ reflections on reading as premised on the tacit “assumption that everyone reads (or ought to) as we do professionally,” to which point I seek to add one key word: everyone who reads.³² As Deidre Lynch has suggested, “loving literature” is a personal and professional demand of people in literary studies, one that necessarily begins with “the psychological intensities that reading creates” and the attendant formation of a literary subjectivity.³³ Yet if Lynch’s literary subjectivity only emerges in a subset of all readings, absolute declines in nonprofessional reading matter to literary studies because they are a precondition for professional reading.

There’s much more that could be said on this point, but I would like to suggest one final possibility as to why these data from the SPPA and ATUS have been professionally repressed, unlike our valedictory jobs and majors numbers. In Professing Criticism, John Guillory described this possibility as literary criticism’s overestimation of its aim. Guillory suggests that the difficulty of specifying the research aims of literary studies within the paradigm of the modern research university causes the discipline to overstate its effects.³⁴ In Guillory’s telling, literature becomes the occasion and vehicle for a criticism of society, which justifies the profession. He makes his point about this mercilessly:

“the absurdity of the situation should be evident to all of us: as literary study wanes in public importance, as literature departments shrink in size, as majors in literature decline in numbers, the claims for the criticism of society are ever more overstated.”³⁵

“Overstated” in what sense? In that, “criticism, as many understand it, is an Archimedean lever for moving the world.”³⁶

As these statistics about reading for pleasure suggest, the fulcrum for that Archimedean lever would increasingly be in its effects on Griswold’s reading class. If reading well is to make a difference, it must do so through the practice of reading. If the extent of literary studies’ influence is not bounded by the extent of reading, what else could plausibly be its limiting factor? To presume that it is unimpeded by declines in reading is to overestimate literary studies’ aim in a new dimension.

Conclusion

Before we conclude, I should emphasize that I don’t think that any of these figures represent the capital-T truth about reading for pleasure in the US. To take up one obvious criticism, knowing in advance that you will be keeping a time diary about your day almost certainly changes your behavior on that day.

A less obvious complicating factor is something survyeors call social desirability bias.³⁷ By way of example, survey respondents underreport their use of pornography because most people would prefer to be perceived by the surveyor as a person who doesn’t watch a lot of porn. That’s social desirability bias.

In our context, imagine an SPPA interviewer asking if you had read one book in the preceding year. Maybe the decline we observe indexes both changes in reading habits and a change in how socially undesirable it is to admit that you do not read. But I’ve come to the conclusion that, for my purposes here, this is a distinction that doesn’t make much difference. I think literary studies would clearly benefit if people view reading as socially desirable behavior.

The important thing is not the precise values for a given year, but that the consistency of survey methodology allows us to observe the directionality, rate, and magnitude of the change.

Let’s return now to those opening numbers.

SLIDE

I think that these numbers are profoundly difficult for professional readers to internalize. A large proportion of our friends and family are likely among the 17% who read for personal interest yesterday.

What, if anything, is to be done? I have no cure, but I do have one suggestion: We know that educational attainment continues to be the best predictor of reading. But we do not know whether literary study has a greater impact on reading for pleasure than what we would expect based on educational attainment alone. As Griswold suggested, reading class professionals work to “convert” new members to the reading class. If that seems at all objectionable, consider the opposite proposition: What literary scholar would hope that her students, as a result of taking her courses, would read less?

Literary studies should research the differential impact that the field has on reading for pleasure. We may assume that it is obvious that students of literature would do more reading for pleasure than students who do not study literature. But this is presumptuous. One null hypothesis would be that literary studies has no impact on reading for pleasure beyond what is explained by educational attainment. If literary studies knew its impact on reading, we could better estimate our aim. If we read to encourage and sustain others in their reading, we should want to know if we are succeeding.

Thank you.

Works cited