Chapter 23 — Further Reading

A note before the list, because this chapter's reading problem is unusual. Most further-reading sections send you to sources that will answer your question. This one mostly sends you to sources that will let you characterize the gap — which is the honest deliverable when a literature is inaccessible. Where a specific paper is not named below, that is deliberate: this book does not cite work it has not verified, and a search strategy you can run yourself is more useful than a citation you would have to take on trust.

Tier 1 — Start here

Chapters 5 and 6 of this book. Reread §5.5 (blinding and functional unblinding) and §5.10 (multiplicity and outcome reporting) before anything else on this list. Nearly every judgment in Chapter 23 is one of those two sections applied to a harder case. Chapter 6's evidence hierarchy is the source of the six questions.

Chapters 21 and 22. Chapter 21 for the oxytocin replication story, which is this chapter's symmetry argument in its most concrete form; Chapter 22 for what is and is not established about intranasal delivery to the brain. You cannot read a Semax or Selank trial correctly without Chapter 22 in hand.

The Cochrane Library. Search Cerebrolysin. Cochrane has published systematic reviews of Cerebrolysin in acute ischaemic stroke and in vascular dementia, and reading the reviews alongside their plain-language summaries is the single most efficient way to see what "genuinely mixed" looks like in practice. Pay particular attention to the risk-of-bias tables and to what the reviewers say about heterogeneity and adverse events — those sections carry more information than the pooled estimate does.

The CONSORT Statement (consort-spirit.org). The reporting checklist for randomized trials. Read the item list once, slowly. It is essentially the §23.9 specification written by a committee, and having it in memory turns "this paper feels thin" into "items 6, 12, and 17 are absent."

ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP). Both are free, both are searchable by compound name, and ICTRP aggregates national registries beyond the U.S. Searching a compound in both is a five-minute exercise that changes what you can honestly say about it.

Tier 2 — Going deeper

Richard Feynman, "Cargo Cult Science" (Caltech commencement address, 1974). Short, free online, and the standing text for §23.7. Feynman is describing physics, and every paragraph transfers.

John Ioannidis, "Why Most Published Research Findings Are False," PLoS Medicine (2005). The paper that made the structural argument mainstream. Read it as a description of pressures that operate in every research system, not as an indictment of any one of them — that reading is what makes §23.3's symmetry argument concrete rather than diplomatic.

Simmons, Nelson, and Simonsohn, "False-Positive Psychology," Psychological Science (2011). The clearest available demonstration of how flexibility in analysis manufactures significant results. It is the mechanism behind "four of eleven measures improved," and once you have read it you will not read a results-section list the same way again.

Ben Goldacre, Bad Pharma (2012). Book-length treatment of publication bias, selective reporting, and industry funding effects in the Western pharmaceutical literature. Included here specifically as the antidote to the "Western good, other bad" framing; it is not a book about anyone else's problems.

The Cochrane Handbook for Systematic Reviews of Interventions. Free online. The chapters on searching (including non-English sources and gray literature) and on assessing risk of bias are directly relevant to Case Study 23.1's procedure, and they are written by people who face the access problem professionally.

The PRISMA statement for reporting systematic reviews. Its search-reporting requirements are the reason a good review can tell you the size of your own access gap.

On practice effects: search for practice effects neuropsychological assessment meta-analysis and retest effects cognitive testing. There is a substantial measurement literature quantifying how much scores rise on repeated administration, broken down by instrument. Reading two or three of these makes §23.7's central claim quantitative rather than rhetorical.

On baseline dependence and inverted-U effects: search for baseline dependency cognitive enhancement and inverted-U dopamine working memory. The finding that a compound's effect depends on where a participant started — sometimes to the point of reversing — is well established and rarely mentioned in consumer coverage.

Tier 3 — Specialist and primary sources

Regulatory assessment documents. The EMA publishes European Public Assessment Reports; the FDA publishes review documents through Drugs@FDA. Read one, for any approved drug, all the way through. It is the best possible calibration exercise: it shows you what a full evidentiary dossier looks like when it is public, and therefore what you are missing when it is not.

Non-English and non-default databases. PubMed and Embase are not the world. eLIBRARY.RU is Russia's national scientific electronic library; regional indexes such as LILACS, SciELO, and the various national registries aggregated by ICTRP cover literatures that default searches miss. You may not be able to read what you find. Finding it anyway is what converts "there is nothing" into "there is something I cannot assess," which is a different and more honest sentence.

Machine translation, used with explicit limits. Modern translation is good enough to extract a population, a design, a sample size, and an endpoint from an abstract. It is not good enough to let you assess nuance, statistical description, or the meaning of an unfamiliar reporting convention. Use it for the first job. Do not pretend it did the second.

Primary literature on the compounds themselves. Search Semax, Selank, and Cerebrolysin in PubMed and in ICTRP; note what you find, what is available in full text, and what exists only as a translated title. Then write down the counts. That written record is Ledger B from Case Study 23.1, and it is worth more than any summary someone else produces for you.

Retraction Watch and the retraction and correction notices attached to any paper you rely on. A one-minute check, applied to everything, in every language.

If you only do one thing

Take one compound — from this chapter or from your own dossier — and run the seven-step procedure in Case Study 23.1 all the way through, including step 2 (write down your prior first) and step 5 (keep two separate ledgers).

It takes about an hour. What you will produce is a paragraph that states a position, distinguishes it from a judgment you are declining to make, labels the rating's type, and names two separate conditions that would change it — one requiring new science and one requiring only a translation.

Almost nobody writing about these compounds, in any language, on any side, produces that paragraph. Producing it once is how you find out that careful engagement with hard-to-evaluate evidence is a procedure you can run rather than a temperament you either have or lack.