Humanities in Crisis - Dissertation Trends and Washington Times Comparison

Parent: Source Audits & Media · Topic: Leftism & Woke

Assessment

The Washington Times captures the report’s main descriptive themes and includes meaningful counterarguments. Its numerical presentation needs correction, and the report itself contains inconsistencies. Most importantly, the headline 49% decline compares five-year archive totals, not annual national production. The study documents changing vocabulary and rhetorical context; it does not demonstrate that identity politics caused the humanities’ contraction, that scholars have abandoned truth, or that dissertations have become specifically anti-American after the report’s additional controls.

Sources and versions

• Circulation report: Ober, George, and Kalinin, Humanities in Crisis: A Study of 65 Years of American Dissertation Trends, American Academy of Sciences and Letters, October 2026. Complete 67-page PDF (file not included in this release) · Publisher’s PDF.
• Technical companion: Humanities in Crisis: Quantitative Analysis of 65 Years of American Dissertation Trends. Complete 135-page PDF (file not included in this release) · Publisher’s technical PDF. This companion resolves the distinction between annual counts and five-year aggregates and supplies the sentiment robustness checks.
• Article: Sean Salai, The Washington Times, October 21, 2026. The displayed headline when read was Humanities dissertations increasingly critical of Western culture; the URL retains an earlier formulation about dissertation decline and identity politics. Article.
• Access and scope: Full article read in the browser on October 22, 2026; circulation report reviewed, with relevant technical methods, appendix tables, and robustness checks cross-checked. PDF page numbers below match the printed page numbers. The underlying ProQuest corpus and analysis code were not independently rerun.

Summary of the report

Research question and dataset

The authors examine how American doctoral scholarship changed between 1960 and 2025, using the full text of 443,543 dissertations from 678 institutions in ProQuest. They compare Humanities and Arts with Social Sciences and divide the material into History, Literature/Language, Philosophy/Theology, Interdisciplinary Studies, Social Science I, and Social Science II. The last two groupings broadly separate Economics/Political Science from Anthropology, Psychology, Communications, Sociology, and related fields. These are the authors’ groupings, not a universally agreed classification of disciplinary methods. Data and coverage, p. 15 (file not included in this release) · Classification, p. 16 (file not included in this release).

Five researcher-selected word baskets track Facts/Truth Claims, Aesthetics, Power, Identity, and Evaluation. Counts are normalized by total words. A separate sentiment analysis scores the language surrounding nine focal terms, including Truth, Western, American, European, Identity, and Activism. University fixed-effects models compare changes within institutions, and a balanced set of ten universities checks the broader, changing institutional sample. Word baskets, pp. 17–18 (file not included in this release) · Models, pp. 49–51 (file not included in this release).

Main findings

Humanities doctoral output contracts more sharply than social sciences in this archive

The clearest institutional finding is a change in the balance of the archived dissertations. Humanities/Arts falls from 33,473 dissertations in 2011–2015 to 16,951 in 2021–2025, a 49.4% decline. Social Sciences falls from 31,416 to 29,997, only 4.5%, while the combined total falls 27.6%. Social Sciences therefore occupies a larger share of the combined corpus. These are five-year totals of archived documents, not annual national degree counts. Five-year totals, Tables A.1–A.2, pp. 63–64 (file not included in this release).

The authors see a loss of institutional position for traditional humanities departments alongside the relative resilience of Social Science II, their grouping of anthropology, communications, psychology, sociology, and related areas. They suggest that some humanistic questions may be finding a home in these fields. This is a proposed interpretation of changing disciplinary shares; the study does not track individual students transferring between programs. Recent archive registration delays also exaggerate the apparent endpoint contraction. Coverage limits, p. 15 (file not included in this release) · Institutional interpretation, p. 54 (file not included in this release).

Identity vocabulary is the fastest-growing language category

Identity-related language rises far more rapidly than any of the other four word baskets. The technical appendix’s annual measure increases from 0.000434 in 1960 to 0.004348 in 2025, approximately tenfold. In more familiar units, that is roughly 43 to 435 matches per 100,000 words, or 0.0434% to 0.4348% of words. The increase concerns the frequency of selected language, not the proportion of dissertations endorsing identity politics. Annual normalized frequencies, Table A.23, p. 84 (file not included in this release).

Growth becomes sustained from around 1980 and is especially pronounced in Literature/Language, Interdisciplinary Studies, and Social Science II, with a further acceleration in recent years. The institutional models also identify Identity as the fastest-growing basket in both the complete 678-university sample and the ten universities with continuous coverage. The report summarizes fitted growth rates as approximately 5–8% annually, implying doubling on the order of 8–12 years. Those are model-based summaries, not a claim that every field increased at the same constant rate. Timing and field variation, pp. 28–33 (file not included in this release) · Institutional models, pp. 49–50 (file not included in this release).

Power language expands while aesthetic language loses prominence

Across the full corpus, Power vocabulary rises sharply around the late 1980s and 1990s, then broadly levels off after 2000. Literature/Language has a particularly strong early rise and a further increase around 2015–2020. The authors associate this pattern with the spread of poststructuralist, postcolonial, and related frameworks, although the timing alone cannot establish that intellectual transmission caused the measured changes. Full-corpus trends, pp. 27–30 (file not included in this release).

The descriptive series shows Aesthetics vocabulary declining from around 1988, including a reduction within literary studies. That matters because literature historically gave particularly strong emphasis to aesthetic language. The authors interpret the changing balance as greater emphasis on power and identity relative to beauty, form, and artistic evaluation. Importantly, Power and Facts/Truth language initially rise together: the new vocabulary does not simply replace all evidence-oriented language. Parallel growth and aesthetic decline, pp. 27–30 (file not included in this release).

Truth remains a prominent and positively framed term

The Facts/Truth basket remains the most frequent overall and grows across much of the period. The individual word Truth also has the most positive surrounding language of the nine focal terms, with a corpus-wide score around +0.35 on the authors’ −1 to +1 sentiment scale. Its positive framing is broadly stable rather than collapsing as identity and power language become more common. Sentiment levels and differences, p. 36 (file not included in this release) · Model results, pp. 51–52 (file not included in this release).

This coexistence is central to the report’s interpretation. It portrays a change in the subjects and cultural orientation of scholarship alongside continued rhetorical commitment to truth. It does not demonstrate that scholars have abandoned truth, and the persistence of truth language does not independently establish how rigorously their arguments are tested. Interpretation and limits, pp. 54–55 (file not included in this release).

The disciplines show substantially different trajectories

Sources: Vocabulary by field, pp. 28–33 (file not included in this release) · Field sentiment, p. 36 (file not included in this release) · Social-science and interdisciplinary sentiment, p. 45 (file not included in this release). The fields are broad author-defined groupings, not internally uniform methodological traditions.

Western becomes less positively framed, but the geopolitical results diverge under controls

In the raw descriptive series, American, European, and Western all appear in less positive contexts over time. This means their surrounding language becomes more critical or neutral; it does not necessarily become negative in absolute terms. Similarly, Power can become more frequent while being less positively framed, because more discussion of domination or coercion can increase its frequency and lower its sentiment at the same time. Frequency versus tone, p. 36 (file not included in this release).

The stronger result concerns Western. Its decline survives the authors’ adjustment for general negativity across academic writing. European weakens to statistical insignificance under that adjustment. American reverses sign, leading the authors explicitly to reject citing its raw decline as evidence of anti-American sentiment. Their findings therefore support the narrower claim that Western-related language becomes more critically framed more clearly than a generalized anti-American interpretation. Qualified geopolitical findings, p. 51 (file not included in this release) · Baseline-adjusted results, Table C.3, p. 132 (file not included in this release).

Institutional composition does not explain away the main vocabulary finding

Comparing universities with their own earlier records helps test whether changes merely reflect new institutions entering the archive. Identity remains the fastest-growing basket in the continuously observed ten-university group as well as the full sample. This strengthens the descriptive conclusion that language changed within established institutions, rather than only because the database’s institutional coverage changed. It does not eliminate changing topics, subfields, archive availability, or other explanations within institutions. Design and results, pp. 49–50 (file not included in this release).

The authors’ overall interpretation

The authors describe institutional contraction alongside a continuing cultural and rhetorical transformation. Traditional humanities lose relative ground in the archive; identity and power frameworks become more prominent in several fields; aesthetic language weakens; and some inherited cultural terms receive more critical framing. Yet truth and evidence remain prominent in the same texts. They suggest that some humanistic inquiry may be relocating into social sciences rather than disappearing altogether. Conclusions, pp. 53–56 (file not included in this release).

The most defensible synthesis is that the archive records a substantial change in the language and disciplinary location of doctoral scholarship. Whether that change improves or damages scholarship, and whether it explains declining enrollment or program resources, requires evidence beyond the word counts and sentiment models. The numerical inconsistencies in some published prose, detailed below, also make the appendix units essential when quoting particular figures.

Comparison with the Washington Times article

What the article handles fairly

Salai reports vocabulary growth and continued truth-related language, acknowledges archive coverage and registration delays, and includes explicit warnings against causal inference. He also includes opposing interpretations involving funding, employment, and student demand. It would be inaccurate to say the article suppresses all qualifications or presents ideological causation as an established statistical finding.

Where the presentation needs correction

Correct units for the 49% decline

The technical appendix labels the relevant observations 5-Year Aggregates. The periods marked 2015 and 2025 correspond to 2011–2015 and 2021–2025. This is independently consistent with the companion’s complete annual document counts: those years sum to 64,889 and 46,948 across both broad fields. Period totals, p. 63 (file not included in this release) · Annual totals, pp. 83–84 (file not included in this release).

By contrast, the circulation report’s annual appendix lists Humanities/Arts counts of 6,575 in 2015 and 2,286 in 2025. Their raw archived decline is about 65.2%, a different comparison. This larger percentage is not a better estimate of the national decline: the incomplete 2025 endpoint is particularly vulnerable to registration lag. Annual counts, Table 2, p. 65 (file not included in this release).

The omitted anti-American qualification

The article does not explain RC-4’s distinction among Western, European, and American. The technical report adjusts the focal terms against a baseline derived from other sentiment series. Under that specification, the late-period American estimate moves from −0.007 in the raw model to +0.006; European attenuates from −0.032 to −0.010; Western remains around −0.031. These are model coefficients relative to the baseline year, not percentages of scholars expressing hostility. The appendix reports mean coefficients and mean standard errors; this note has not recomputed their joint statistical uncertainty. Specification, pp. 131–132 (file not included in this release).

The distinction is important for evaluating anti-American claims, but the article’s displayed headline concerns Western culture. That narrower headline is closer to the authors’ qualified result than a blanket claim about hostility toward America would be. Even then, less-positive word context is not a direct measurement of an author’s political allegiance or attitude toward a civilization.

Problems inherited from the report

• Top five institutions: The narrative gives approximately 19%; the table lists 11,025 + 10,570 + 10,486 + 9,476 + 8,661 = 50,218, or 11.32% of 443,543. That published table does not reproduce the narrative figure. The article’s near-20% formulation inherits the problem. Narrative, p. 46 (file not included in this release) · Table 3, p. 66 (file not included in this release).
• History plus Literature shares: The narrative’s roughly 30%-to-below-11% claim cannot be reproduced from the displayed annual or five-year tables. Annual table values yield 38.9% in 1965 and 13.3% in 2025; five-year totals yield approximately 35.8% and 16.3%. Unless another denominator is supplied, treat the narrative percentages as unresolved. Narrative, p. 22 (file not included in this release) · Annual table, p. 65 (file not included in this release) · Five-year table, p. 64 (file not included in this release).
• The reported 70-fold expansion of Social Science II: The comparison mixes 283 dissertations in the single year 1960 with 19,940 in the five-year period ending in 2020. The annual 2020 count is 3,845, approximately 13.6 times the 1960 count. Neither ratio is a clean national growth estimate because early archive coverage is incomplete. Narrative, p. 22 (file not included in this release) · Period counts, p. 64 (file not included in this release) · Annual counts, p. 82 (file not included in this release).
• Sentiment window: The executive summary says five words on either side, but the methods and sentiment sections specify ten on either side. The technical methods also specify ±10. The latter is the more detailed stated method, but the inconsistency should be corrected by the authors. Summary, p. 7 (file not included in this release) · Methods, p. 18 (file not included in this release).

Methodological assessment

Useful descriptive evidence. Full texts are richer than titles or abstracts, a large longitudinal corpus permits within-field comparisons, normalized frequencies avoid confusing more documents with more frequent vocabulary, and the authors explicitly investigate institutional composition and alternative sentiment baselines. These are real strengths.

Vocabulary is not endorsement. The Identity basket mixes ordinary research topics such as disability, diaspora, and ethnicity with more theory-specific expressions. It includes language used by proponents, critics, and historians. A dissertation criticizing identity politics may score highly. Likewise, counts of war, evidence, law, statistics, or truth do not independently establish intellectual rigor. The baskets contain overlaps, including race, gender, oppression, and colonialism, so they are not mutually exclusive partitions of a dissertation’s content. Definitions, pp. 16–18 (file not included in this release).

Local sentiment is an imperfect attitude measure. A negative context may arise because an author describes atrocities, quotes an opponent, discusses historical prejudice, or criticizes an institution. The method cannot reliably separate the author’s view from the material being described. Criticism can be accurate scholarship; positive rhetoric can coexist with poor evidence. The study therefore supports claims about measured discourse more directly than claims about ideological allegiance or scholarly quality. Sentiment limits, p. 34 (file not included in this release).

The archive is not a national census. Older digitization is uneven; ProQuest participation varies; recent deposits are incomplete; and the corpus is concentrated at research universities. The independent federal benchmark discussed in RC-1 ends in 2016–2017, so it cannot by itself validate the precise magnitude of a decline through 2025. Its grouping also places history within Social Sciences, unlike the main corpus classification, further limiting a direct numerical comparison. Archive limitations, p. 15 (file not included in this release) · Benchmark definitions and period, pp. 126–127 (file not included in this release).

Temporal association does not identify causes. The report does not link an individual student’s ideological vocabulary to enrollment decisions, hiring outcomes, or program closures. Fields can face different funding and employment pressures, so unequal declines do not rule out economic explanations. Nor can a corpus excluding STEM directly establish where students who leave humanities programs go. The authors’ stronger institutional interpretations should be treated as hypotheses requiring additional evidence.

Two defensible readings. Critics can reasonably point to a substantial change in scholarly vocabulary and a narrowing institutional position for traditional humanities disciplines. Defenders can reasonably point out that truth-related rhetoric persists and that financial and employment constraints remain plausible explanations. Neither side can use these word counts alone to establish the actual quality of scholarship or the cause of contraction.

Comparison with the Vanderbilt–WashU commission

The State of Scholarship report adds a different kind of evidence: an expert assessment of scholarly standards, professional norms, and illustrative disputes, supplemented by field-specific reports. It distinguishes politically motivated inquiry from political control over acceptable answers. By its own criteria, rigorous research about identity or inequality is not defective merely because it addresses those topics. Distinctions, pp. 10–12 (file not included in this release).

The sources can coexist without validating one another’s strongest claims. Continued truth vocabulary does not establish fair treatment of contrary evidence; a rise in identity vocabulary does not establish a decline in rigor. The commission’s examples do not establish prevalence, and Rubin’s later sociology report explicitly distinguishes stronger evidence of politicized norms from weaker evidence of downstream effects on research quality. Commission limits, p. 5 (file not included in this release) · Norms versus outcomes, p. 6 (file not included in this release).

Boghossian's July essay clarifies and defends the commission’s argument; he chaired it, so this is a participant’s interpretation. Neither source independently establishes that identity politics caused the numerical contraction described here. The commission was submitted in April 2026, before the October AASL report, and cannot be presented as a response to that later study.

Vault connections

• Grievance Studies Affair concerns actual journal acceptance decisions under a hoax design; this dissertation study measures vocabulary and context. Neither result substitutes for the other kind of evidence.
• Postmodern provides conceptual context, but a shared vocabulary does not establish endorsement of a particular philosopher or rejection of truth.
• The Identity Trap offers a book-length framework to compare with the study’s operational definition of Identity.
• Pinker content-analysis audit raises a parallel problem: observed word or topic frequencies do not automatically establish authorial intent, bias, accuracy, or effects on readers.

Evidence status and follow-up questions

This is a comparison of the supplied circulation report, selected technical evidence, and the article as displayed on October 22, 2026. Calculations above were independently checked against published appendix cells; they are not a replication of the authors’ underlying NLP analysis. The report’s empirical claims, interviewees’ interpretations, and this note’s methodological assessment are separated deliberately.

Before using the strongest political claims, ask for: corrected annual-versus-period labeling; reconciled institutional and subfield percentages; the actual sentiment-window implementation; validated coding of endorsement versus criticism; recent comparable federal degree data; and linked enrollment, funding, hiring, or placement records capable of testing causal explanations. No finding in this note establishes that identity politics caused humanities programs to contract.