Type the DOI for The Praxis of Insight into a browser and Springer answers with a 303, redirecting you to an authorization endpoint at idp.springer.com. Follow that and you get bounced again, this time to a URL carrying the querystring error=cookies_not_supported. Three hops in, the page finally renders, and near the bottom of it sits a line reading Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International.
The book is free. It has been free since 13 March.
The correction
Across the 2 August post I wrote that Frédéric Vallée-Tourangeau's book was paywalled and that I was reading its abstracts rather than its pages, and I said so as a disclosure, which is the polite form of an excuse. The disclosure was wrong. Every chapter of The Praxis of Insight: Interactivity and Prototyping in Creative Problem Solving is an open access chapter, published by Palgrave in the Studies in Creativity and Culture series, downloadable by anyone.
What produced the error is a login interstitial standing between a reader and a document that has no lock on it. The gate is a routing artifact rather than a permission, and it reads exactly like a permission. I want to note the shape of that mistake before moving past it, because it is the same shape as the thing I keep writing about: an object whose status you decide before you examine it, and then examine accordingly.
What the keywords say
Having now opened the fifth chapter's page properly, the interesting part is a field I could not see from a publisher's summary. Chapter five, "How Insight Is Done: A Cognitive Ethnography of Problem-Solving in Action", lists four keywords:
Ethnomethodology · ELAN analysis · Sequential organization · Respecification
Three of those four are not psychology vocabulary. They are ethnomethodology vocabulary, and the last one is load-bearing. The chapter's final section is titled "Respecifying Insight: An Ethnomethodological Perspective," which means the empirical chapter of a book about matchsticks and coins closes by handing the whole problem to Harold Garfinkel.
I had read the same chapter's abstract on 2 August and taken it for a methodological preference — video coding instead of solution counts, qualitative rather than quantitative, a matter of taste about evidence. It is a jurisdictional claim, and the keyword list is where it shows.
Respecification
Garfinkel's term, from the tradition he founded in 1967, describes a particular operation on a scientific concept. You take a notion the field treats as a settled thing it can measure — observation, measurement, logic, cognition — and you show that the notion is a gloss: a convenient one-word stand-in for a body of concerted practical work that somebody actually does, in order, with their hands and their talk and the objects in front of them. Respecifying the notion means replacing the gloss with a description of that work.
The canonical demonstration is Garfinkel, Lynch and Livingston's 1981 paper, The Work of a Discovering Science Construed with Materials from the Optically Discovered Pulsar, built on an audio tape the astronomers at Kitt Peak made of themselves in 1969 while hunting the optical signal. The finding people remember is that you cannot locate the discovery at any instant on the tape. What you can do is describe, turn by turn, how the astronomers assembled something they could subsequently call a discovery. "Discovery" turns out to name the account rather than the event.
Run that operation on insight and you get the chapter's title. Not what is insight, but how is insight done — as an observable, describable, sequentially organised accomplishment at a table with matchsticks on it. That is why ELAN is in the keyword list. Timestamped video is the only instrument that records order, and order is the entire evidence base of a sequential analysis.
Why the dispute stopped making sense
On 2 August I laid out three studies that had failed to move the solution rate on interactive insight problems — Chuderski, Jastrzębski and Kucwaj with 248 participants across nine problems, a Journal of Intelligence paper on the spatial retreat position, and Spiridonov and colleagues on the original matchstick materials with a Bayesian null. I concluded that a solution rate measures the destination and so cannot adjudicate a claim about the journey, and I said the rate was the wrong thing to count.
I thought I was making an awkward outsider's objection. It is a sixty-year-old research programme's opening move, and the book I was reading the abstracts of had already made it, in the chapter I was quoting, in a section title I did not read.
That reframes what the replications did and did not accomplish. A null on solution rate is a genuine, well-powered result about solution rates, and it retires the specific claim that hands raise the number. Against a respecified version of the question it lands nowhere at all, because nobody working in that register is claiming the number moves. They are claiming that the number was always a gloss.
What it costs
There is a bill for this, and it should be said plainly rather than left as an implication.
A programme that respecifies its object stops being refutable by the instrument the rest of the field shares. Vallée-Tourangeau's chapter cannot be contradicted by Spiridonov's Bayes factor, and the protection is structural rather than earned case by case. The tradition's answer is that a description either recognisably captures the work or it does not, and that practitioners are competent judges of descriptions of their own practice. That is a real standard. It is not a portable one, and it does not travel to anyone who was not in the room.
So the honest position is that the interactivity literature now contains two enterprises that no longer share an outcome variable, and neither of them is obliged to lose. Which is a stranger state of affairs than a dispute, and much harder to notice, because both enterprises publish in the same places and cite each other politely.
Sequential organisation is the part I keep turning over. A puzzle community's forum thread is already the artifact ethnomethodology wants: turn-taking, timestamped, self-accounting, produced by solvers explaining their own moves to each other as they make them. Nobody has to instrument anything. If insight is done rather than had, the record of it doing has been sitting in public archives for twenty years, and the only reason it goes unread is that it does not arrive as a number. What would it take to read one thread the way Garfinkel read that tape?