Citation Asymmetry

What it is

The mechanism is simple and it is worse than an uncited claim. An uncited claim announces itself: a reader meeting the words studies show knows they have been given nothing and can discount accordingly. A reader meeting an endnote goes to the back of the book, finds a real citation in a real journal with volume and page numbers, concludes the sentence was checked and stops there. The apparatus has terminated the enquiry it existed to enable. The better it looks, the more effectively the asymmetry hides.

The category is distinct from its nearest siblings. Source-class inflation is about what class of thing a citation is. Borrowed certainty is about a book with no apparatus to be one-sided with. Asymmetry requires real citations sitting on one side of a live dispute.

The test is one question, asked construct by construct: for each contested claim in the book, could any source cited in support of it possibly disagree with the author. If the answer is no for every contested claim, the apparatus is decorative on exactly the points where it was needed.

In effect

The defining instance is Cal Newport's Deep Work, which carries twenty-three pages of endnotes and roughly 215 entries. Four contested constructs, four one-sided sets of notes. Ego depletion is stated as fact and cited to Roy Baumeister and co-authors and a trade book, with four of the five self-control sources in the first section authored or co-authored by Baumeister, and no replication failure or publication-bias material anywhere in the apparatus. Deliberate practice is cited to Anders Ericsson three times, to Ericsson's own website, to two popularisers and to Ericsson's own rebuttal of his critics, with no meta-analysis and no critic. Unconscious thought is cited to Ap Dijksterhuis and colleagues alone. Media multitasking is cited to a radio interview rather than to the paper underneath it. In every case the contrary literature existed before the book was published and none of it appears.

Zelana Montminy's Finding Focus is the second instance and the more instructive one, because its apparatus is about seventy-one per cent peer-reviewed with full bibliographic detail. The longhand note-taking advantage is cited correctly to the original study and the large preregistered replication that did not reproduce it is absent. The parole judges finding is cited correctly and presented as settled, with the non-random case ordering objection and the implausible effect size objection both missing. The phone and conversation finding is presented as established with a mixed replication record unmentioned. Brain training is credited to a researcher whose commercial interest goes undisclosed and whose contested far-transfer literature goes unmentioned.

One entry in that book shows the asymmetry is a choice rather than an accident. She cites the correction putting working memory nearer four items in the back, and prints seven plus or minus two in the front. The correcting source was in her hands.

The counter-example is The Elephant in the Brain, audited a day after this category was created. Its fourth chapter uses the watching-eyes effect as support for a claim, and the endnote attached to that sentence discloses that two recent meta-analyses cast doubt on all eye-cue effects. The charity chapter warns the reader off a priming study and then explains why the authors still think their contrast holds, with the failure-to-replicate paper sitting in the references alongside the supporting ones. The medicine chapter twice announces its own omissions, concedes that a large insurance experiment was underpowered for mortality, and reports a nearly significant result suggesting free medicine harmed a subgroup, which cuts against the chapter's own framing. About one note in six concedes something, qualifies the text or gives a rival view its due. That settles a question this category left open: the defect is a choice rather than a genre constraint.

Three proxies make the test quick. Count the surnames, because if most sources on one mechanism share an author you have a closed loop rather than a literature. Look for the defendant's reply, since a dispute settled by citing the disputed party's own rebuttal is a nameable special case. And check the dates against the replication record: a 2016 book citing a 2006 finding is not at fault for a 2018 failure, and is at fault for a 2014 one.

What it does not say

It does not rest on research. This is a category this publication defines and applies, not a proposition anyone has tested, and each instance was established by reading the book's own apparatus.

It does not say authors must adjudicate every dispute. No trade writer is obliged to produce a literature review, and a book that stopped to settle every contested construct would be unreadable. The obligation is not to settle the dispute but to say that one exists. A single sentence costs an author nothing and restores the reader's ability to judge.

It does not ignore the dating defence. Nobody can cite what has not appeared. Against that, the corrections on both of Deep Work's load-bearing mechanisms were available while it was being written, and the book has been reprinted continuously since.

It does not follow that provenance discipline and evidential quality are the same virtue. Deep Work passes every first-author check it takes, discloses its discovery chains and credits the right originator where others do not. It is scrupulous about where a claim came from and silent about whether the claim is any good.

It does not apply to every book with a thin apparatus. The Art of Clear Thinking has no citations to be one-sided with, and The Modern Mind's apparatus flags contrary scholarship repeatedly and passes this test on a book of more than eight hundred pages. Those belong to different categories in this vault.

It is not a criticism of authors. Every writer named here cites better than most of this shelf, which is the point.

---