The most-assigned STEM textbooks at 12 top US universities

We maintain a catalog of canonical STEM textbooks and the university courses that assign them. While building out the course mappings, one question kept nagging: across the top US programs, is there actually a consensus on which textbook to use — or does every department pick its own?

So we counted. Here's what ~60 canonical courses across 12 universities actually assign.

Methodology (read this before you argue with the numbers)

Be skeptical of any "most popular textbook" list that doesn't tell you how it was built. Here's ours, honestly:

Scope: ~60 evergreen, canonical courses across 12 institutions — MIT, Stanford, Berkeley, CMU, Caltech, Princeton, Cornell, Georgia Tech, UIUC, Michigan, UT Austin, and UW.
Source: public course pages, OpenCourseWare, and published syllabi. These are the evergreen textbook–course pairings (the book a course has used consistently), not a real-time scrape of this semester's reading list.
Counting rule: a textbook scores one point per distinct course that lists it, whether as required or recommended/reference.
Known bias: the catalog skews toward CS, math, and physics. It is not exhaustive, and a course using a book we haven't mapped won't show up. Treat this as a strong directional signal, not a census.

With that said — the headline result is not subtle.

The single most-assigned STEM textbook in America

It isn't close.

Rank	Textbook	Distinct courses	Universities
1	Cormen, Leiserson, Rivest, Stein — Introduction to Algorithms (CLRS, 4th)	15	10 of 12
2	Strang — Introduction to Linear Algebra	6	6
3	Bishop — Pattern Recognition and Machine Learning	5	5
3	Murphy — Probabilistic Machine Learning: An Introduction	5	5
3	Sipser — Introduction to the Theory of Computation	5	4
6	Hastie, Tibshirani, Friedman — The Elements of Statistical Learning	4	4
6	Aho, Lam, Sethi, Ullman — Compilers (the "Dragon Book")	4	4
8	Young & Freedman — University Physics	3	3
8	Lay — Linear Algebra and Its Applications	3	3
8	James et al. — An Introduction to Statistical Learning (ISL)	3	3

CLRS appears in 15 distinct courses across 10 of the 12 universities we mapped — MIT 6.006 and 6.046, Stanford CS 161, Berkeley CS 170, Princeton COS 226, two courses each at Cornell, Georgia Tech, and UIUC, plus Michigan, UT Austin, and UW. The second-place book has fewer than half as many. If there is a single book a US computer science student is statistically guaranteed to be assigned, it is this one.

The more interesting finding: which subjects have a "canonical" book — and which don't

The ranking is a fun trivia answer. The structure underneath it is the actually-useful insight. Some subjects have converged on one definitive text; others are a genuine fight.

Algorithms: a monarchy

One book, CLRS, owns the entire field. There is no #2 algorithms text with meaningful share. Forty years in, the discipline has a settled canon, and every department teaches from the same 1,300-page brick. If you're taking an algorithms course — MIT 6.006, Stanford CS 161, Berkeley CS 170 — buy it once, used, and keep it. It will not be replaced.

Linear algebra: a required subject with no agreed book

Linear algebra is the most universally required subject in the dataset — nearly every institution mandates it — yet there's no consensus text. Strang leads with 6 courses, but Lay (3) and Axler's Linear Algebra Done Right (2) split the rest along a clear line: Strang and Lay for computational/applied sections, Axler for proof-based honors tracks. Same requirement, three philosophies. See the full split on the mathematics page.

Machine learning: a four-way deadlock

This is the most contested slot in all of STEM. Bishop (5), Murphy (5), Elements of Statistical Learning (4), and ISL (3) are in a genuine four-way tie, often with two of them assigned in the same course (one as the rigorous reference, one as the accessible companion). Contrast this with algorithms: ML is young enough that the field hasn't agreed on what the canon even is. Browse the contenders on the statistics page.

Theory of computation: a quiet near-monopoly

Sipser (5) is the standard almost everywhere, with Hopcroft, Motwani & Ullman as the legacy alternative at a couple of schools. Closer to the algorithms monarchy than the ML free-for-all.

Compilers: a 20-year-old book still winning

The Dragon Book (4) — first edition 1986 — remains the assigned text at Stanford, MIT, Berkeley, and CMU. Compilers as a field moved on; the canonical textbook didn't.

Physics: fragmented by design

Intro physics is the opposite of algorithms. Young & Freedman's University Physics (3) is the most-shared intro text, but honors sequences deliberately diverge — Caltech leans on the Feynman Lectures, MIT's honors mechanics uses Kleppner & Kolenkow, Cornell uses Purcell for E&M, Princeton uses Taylor for classical mechanics. Physics departments treat textbook choice as pedagogy, not logistics.

What this means if you're buying textbooks

Three practical takeaways fall directly out of the data:

For monarchy subjects (algorithms, theory, compilers): buy used and keep it. These books don't churn editions to kill the resale market — the canon is stable. CLRS 4th will be assigned unchanged for the next decade. Buying used is a one-time cost, not a recurring tax.
For ML: don't buy four books. Your course will likely list two or three. Borrow or rent the "accessible companion" and only buy the one you'll actually reference for years (usually Bishop or Murphy). We break down which is which on each textbook page.
For commercial intro texts (the fragmented subjects): rent or buy international. This is where publishers churn editions fastest — see our guide to textbooks that hold their value and whether international editions are legal.

Caveats, restated

This is a directional study of a curated catalog, not a national survey. The dataset over-represents CS/math/physics and under-represents engineering, chemistry, and biology, where the sample is too thin to draw conclusions. A book missing from this list might be dominant in a field we haven't mapped deeply. We'll update as the catalog grows — and if you spot a course-textbook pairing we've gotten wrong, that's exactly the kind of correction we want.

The one conclusion that survives every caveat: if you are taking an algorithms course at a top US university, you are buying CLRS. You may as well find the cheapest copy now.

Want the cheapest price on any book mentioned here? Every title links to its page, where we compare prices across Amazon, eBay, AbeBooks, Alibris, and Pearson. Start from the full catalog.