The seventh sentence of the book under review (opening of the Preface, p. vii) is a question: “Why do we run toward people we love, but merely walk toward others?” The sentence is copied 332 pages later to launch the book's Conclusion, p. 323. A reader might expect, then, that that embedded proposition, that “we run toward people we love … but only walk toward others,” is pertinent to Vigor's actual empirical content. But a word-by-word scan of the book's electronic version uncovers only eight appearances of the word “love” (also three of “loved one,” but that is a different concept) of which four are from Shakespeare, one is from a fourteenth-century poet named Hafez, and two others are from the duplicate sentence I've already quoted. Only one of the eight actually deals with the notion of “love” at all: “when we are walking to meet someone we love, … reward is computed via its neural proxy, dopamine” (p. 15), and that is another proposition for which no evidence is offered.

The grammar of embedded sentences such as those lulls us into glossing them as factual, but they could as easily be myth or exaggeration as generalization. Noting the lack of discussion, a skeptical reader might well conclude that the book won't answer that “why” opening teaser, inasmuch as it has no intention of answering the simpler embedded question: Do we move faster toward people we love? On this specific point, the book offers no discussion at all. The answer would, of course, have been “Well, it depends” – countless other factors affect speed of ambulation: stature, altitude, rate of climb or descent, age, social context, crowding, health, weather, footwear, category of love, and so on, and the answer to that “Why” query would depend on all those other factors. Although the book refers to the opening statement as “our airport example,” p. 14, it presents no data relevant to this fable, from airports or anywhere else. Therefore, the authors’ initial Why question is incoherent: there is not yet any phenomenon to be explained.

The critique instigated by this casual aspect of the book's introduction applies across all the domains of “neuroeconomics of movement control” that its authors purport to consider. Let me illustrate lexicometrically. The word “vigor” itself, title of the book, appears 521 times; “dopamine,” that proxy for “reward,” 468 times; “capture rate,” a general term from “optimal foraging theory” defined as “the sum of all rewards acquired minus all efforts expended, divided by total time” (p. 331), 285 times; “utility,” the unobservable core of economic theory, 431 times. But then explore the glossary of inference. The words “prove,” “proof,” “disprove,” and so on never appear here at all! Evidently the authors do not claim to have actually settled any scientific debate. The word “proxy” itself, which surely is relevant to any interdisciplinary argument as complicated as theirs, appears 24 times but is never defined: apparently this is a term of art. The trope of “consistent,” as in “being consistent with,” appears 46 times, and “correspond,” 40 times; but the total number of correlation coefficients printed here is a mere three, and the concept of “regression” arises only in connection with two of the figures, 5.16 and 5.19. The appearances of the word “suggest,” a vestige of the weakest form of scientific inference, total 118, but the number of inferences, or rather the count of appearances of the word “infer” and its derivatives, is only 19, less than one-sixth as many.

Now let us dig more deeply into the methodology of this volume. The Preface summarizes its purpose on p. xi: “In this book, our goal is to consider a simple question: why do we move faster toward things we value more?” A quick retort would be “Who says?” or “Oh, do we?” A more reasoned response would note that no such “Why?” question can be answered until there are quantifications of how much – how much faster, how much more valued. Therefore, a book like this cannot validly persuade unless its methodology is explicitly written out – how every discussion like “why greater values of A sometimes accompany greater values of B” is to be pursued in a logically rigorous way that convinces readers not already indoctrinated into its esoterica. Vigor, alas, eschews any such prolegomenon. The phrase “how much” appears 17 times in this book, but the usage is always merely qualitative: The probe is never responded to with an actual quantity, let alone a regression slope, as on p. viii, where the phrase “how much more you prefer one dessert to the other” is simply left hanging. Although every chapter ends with a discussion under a heading of “Limitations,” the roster of these taps not the logic of interdisciplinary inference but instead diverse aspects of experimental design. In these commentaries, too, all comparisons are of the purely qualitative semantics I've been criticizing.

How ought one to carry out a study bridging economics to other disciplines? A good example comes from the classic topic of how environmental tobacco smoke (ETS, passive smoking) affects health, as uncovered by a research consensus in the late 1990s reviewed on pp. 253–262 of my textbook on quantitative reasoning (Bookstein, Reference Bookstein2014). The study precised there demonstrated the truth of the proposition that ETS causes excess cases of ischemic heart disease (Law, Morris, & Wald, Reference Law, Morris and Wald1997) or lung cancer (Hackshaw, Law, & Law, Reference Hackshaw, Law and Law1997) by finding that the increased rates of prevalence among nonsmoking wives of smokers match known direct causal analyses numerically – dose–response models from other studies – and are stable against challenge by a great variety of competing explanations, such as diet (which is where aspects of economics come in, namely, home economics). The basic logic of inferring causation from extended chains of mixed-modality quantitative data is by now a well-established branch of applied statistics where the fundamental notion linking data to theory is a particularly fruitful diagram style, the directed acyclic graph (cf. Pearl, Reference Pearl2009). Hyphen-economics fields themselves have celebrated this same basic logic, as reviewed, for instance, in Wold and Jöreskog (Reference Wold and Jöreskog1982) and polished in the method of structural equations analysis. In this mature methodology, causation, measurement proxies, and competing inferences are all handled at once.

These rules of interdisciplinary causal reasoning lie nowadays at the core of every graduate curriculum in statistics or biostatistics; but they make no appearance in Vigor. The word “effect” appears 150 times, and “affect” (as a verb) another 133; but the technical term “effect size,” which is universally acknowledged the currency according to which magnitudes should be exported across disciplines, never appears here at all. (In linear models, effect size is a standardized path coefficient; in grouped or experimental data, it is the ratio of a group difference to the pooled within-group standard deviation. It has hardly anything to do with statistical significance or p-values.)

The authors of Vigor have ignored these and many other tenets of interdisciplinary quantitative method. This is in spite of the massive menu of quantitative diagrams on offer – the book's 134 figures span a total of 331 named panels, many of which themselves involve multiple parts (alternative experimental conditions or measures). But no tables present any roster of alternative explanations compared among themselves for the power with which they manage to explain the same empirical pattern (“strong inference,” Platt, Reference Platt1964). The comparisons reported from experimental data all come in the form of one average being “greater” than another, or one movement “faster,” or one chemical titre “higher,” all these comparisons “consistent with” or “corresponding to” the capture-rate theory, but never is there any more convincing language of agreement. Theory is never called upon to predict how much higher or faster or greater the one empirically grounded quantity should be than the other. But all the great examples of interdisciplinary quantitative inference take the form of agreement of such quantities across multidisciplinary testing, the methodology Wilson (Reference Wilson1998) famously reminded us is called consilience (see my review in Ch. 3 of Bookstein, Reference Bookstein2014).

“Because the neurobiology of vigor is based on many of the same neurotransmitters that malfunction in diseases such as Parkinson's and depression,” the authors say on p. 327, “tracking vigor may provide a real-time proxy for the state of these chemicals. Thus, vigor assessment could aid in the administration of interventions and provide an objective measure of treatment efficacy.” Here, the casual informality with which this book tosses around the undefined concept of proxy is worse than unconvincing – attributing any credibility to so untested a claim of proxy outcomes would actually be dangerous. In any context of medical intervention it is unethical simply to declare one quantity to be a proxy of another. Enormous additional effort needs to be invested first in all the factors that link the proxy to the variable it is supposedly proxy for. (This is a main theme of today's evidence-based medicine: see e.g., Higgins & Green, Reference Higgins and Green2008.) No such effort is evident in any of Vigor's examples, and so any talk of applications to Parkinson's disease is vastly premature; similarly, the promise of other payoffs such as the various erosions of privacy suggested as future marketing applications on pp. xii–xiii. For an opposite example, one where this methodological mandate for interdisciplinary inference has been accepted, see Atwater et al. (Reference Atwater, Musumi-Rokkaku, Satake, Tsuji, Ueba and Yamaguchi2005) on the causal analysis of a tsunami following the great Seattle earthquake of 1700.

In closing, I return to my critique of the factoid about when we run that both opens and closes this book, along with the tacit assumption that it can be embedded inside a “why” query without fallacy. A treatise claiming to bridge multiple disciplines of mismatched intellectual impedance (from EEG voltages to “love” is indeed quite a range of observation styles!) requires sustained methodological clarity, particularly in respect of its ultimate component, the conversion of a quantitative comparison into an explanation. A rhetoric of “how much more” is mandatory for any methodology that claims to be explaining “more” across disciplines. One must not settle for the purely qualitative polarity, “more” versus “less,” that is encountered throughout this book.

What emerges from Vigor's failure to constructively critique its own methodology or even to set it out in propositions is a book only for the True Believer, or maybe someone worried about Parkinson's. You can't build an interdisciplinary theme grounded in laboratory-derived data (as were all the data in nearly all the figures of this book) if your only arguments are about the signs of differences or slopes from a roster of experimental measurements untethered to any quantitative theory, any metrological chain. But the book's “Conclusions” chapter, pp. 323–327, in spite of its dozens of qualitative terms (“greater,” “lower,” “faster,” “shorter,” “biased,” etc.) per page, presents the reader not a single decimal quantity at all. Incomprehensibly, the book under review neither explicitly sets down nor implicitly acknowledges any methodology for quantitative inference from instrument readings to human actions, and its absence severely (in my view, fatally) impugns the authority of its narrative. The concern about love with which the book begins and ends is itself no model of inquiry – it brackets neither valid inferences about causal chains in neuroeconomics nor valid explanations of human action from laboratory evidence. The rest of the purported “explanations” here are lost in the fog of proxy measurement, a domain the authors never subjected to any formal discipline. “Proxy” here is a Humpty Dumpty word, meaning whatever the authors choose it to mean; but one cannot persuasively disseminate an interdisciplinary claim when crucial concepts such as this are left undefined, which is to say, unconstrained by method.