Massive Fraud In Nutrition Science

#1

Public health policy over decades has been driven by massive fraud. Universities which were supposed to be guardians of scientific integrity instead were complicit in covering up the fraud (Cornell). And there were no consequences for the figure at the center of this particular fraud, other than reputational damage.

The “Food Psychologist” Who Fooled The World (via @Falsumdoc)

Summary added by RapAdmin:

I. Executive Summary

The investigated video deconstructs the rise and fall of Brian Wansink, a consumer behavior and marketing researcher who founded the Food and Brand Lab at the University of Illinois and later expanded operations at Cornell University. Over a three-decade career culminating in more than 250 published papers, bestselling trade books (Mindless Eating, Slim by Design), and policy appointments including the Executive Directorship of the USDA Center for Nutrition Policy and Promotion, Wansink advanced the behavioral economics thesis that human caloric intake is governed primarily by subconscious environmental architecture rather than conscious willpower or physiological homeostatic drive. Key concepts—including the “bottomless soup bowl,” portion reduction via the Delboeuf optical illusion (smaller plates), and cafeteria nudges deployed in over 30,000 schools via the Smarter Lunchrooms Movement—became structural fixtures of institutional public health.

The documentary traces the empirical unraveling of this body of work following Wansink’s self-incriminating 2016 blog post detailing post-hoc data dredging (“salami-slicing” null datasets) directed at a visiting scholar. Independent forensic analysis using statistical verification algorithms—specifically the GRIM (Granularity-Related Inconsistency of Means) and GRIMMER (Granularity-Related Inconsistency of Means Mapped to Error Repeats) tests developed by Brown and Heathers, 2017 and Jordan Anaya, 2016—exposed pervasive mathematical impossibilities, impossible standard deviations, and anomalous test statistics across his published literature. Subsequent investigative journalism by Stephanie M. Lee at BuzzFeed News unearthed internal communications confirming routine p-hacking, retrospective hypothesis generation (HARKing), multiple-testing manipulation without family-wise error rate control, and publication targeting based on editorial leniency.

The institutional investigation by Cornell University resulted in formal findings of academic misconduct—specifically data misreporting, problematic statistical manipulation, failure of archival data preservation, and inappropriate authorship—culminating in Wansink’s forced retirement and at least 18 formal retractions and 15 corrections tracked by Retraction Watch. Rigorous independent systematic reviews and preregistered clinical trials have since demonstrated that environmental nudges such as plate-size manipulation produce negligible or inconsistent effects on caloric consumption, demonstrating how uncorrected statistical malpractice can misdirect national nutritional policy.

II. Insight Bullets

  • Brian Wansink, trained as a marketing specialist rather than a nutritional biochemist or medical scientist, founded the Food and Brand Lab to apply consumer behavior theories to dietary habits.
  • Wansink authored more than 250 peer-reviewed papers, published international bestsellers (Mindless Eating), and accumulated thousands of citations by framing eating behavior as an unconscious response to environmental cues.
  • The “bottomless soup bowl” experiment argued that visual cues of food availability override physiological interoceptive satiety signaling, causing individuals to overeat without conscious awareness.
  • Wansink’s “plate size” hypothesis claimed that smaller plates exploit the Delboeuf optical illusion to reduce perceived portion deficits, mechanically curbing caloric intake.
  • Wansink served as Executive Director of the USDA Center for Nutrition Policy and Promotion, directly translating his behavioral nudge theories into the 2010 Dietary Guidelines for Americans.
  • The public appeal of Wansink’s paradigm stemmed from an externalized attribution model that absolved individuals of personal willpower deficits during the modern obesity epidemic.
  • In November 2016, Wansink published a blog post (“The Grad Student Who Never Said No”) explicitly detailing instructions to a junior researcher to salvage statistically null data by iteratively torturing post-hoc subgroups.
  • External researchers identified this practice as an egregious combination of p-hacking, data dredging, and hypothesizing after the results are known (HARKing).
  • Independent researchers Nick Brown and Jordan Anaya applied mathematical anomaly-detection algorithms to five papers originating from that specific dataset.
  • The analytical tool GRIM (Granularity-Related Inconsistency of Means) identified dozens of reported means that were mathematically impossible given the discrete integer response scales and reported sample sizes (Brown and Heathers, 2017).
  • The algorithmic extension GRIMMER demonstrated that reported standard deviations and variances were mathematically irreconcilable with the underlying granular distributions.
  • Initial institutional inquiries at Cornell University cleared Wansink of misconduct, characterizing fatal mathematical contradictions as benign administrative errors.
  • Investigative reporter Stephanie M. Lee obtained years of internal Food and Brand Lab correspondence through Freedom of Information Act (FOIA) requests directed at state-affiliated collaborators.
  • Leaked correspondence exposed Wansink and co-authors explicitly discussing p-hacking targets, including running hundreds of uncorrected bivariate regressions to force a p-value below 0.05.
  • Internal emails confirmed that Wansink managed the laboratory as a media-centric enterprise, prioritizing viral press headlines and sensationalist book titles over biological truth.
  • Collaborators explicitly acknowledged data quality deficits while deliberately routing compromised manuscripts to second-tier journals known for permissive peer-review standards.
  • In September 2018, Cornell University reversed its prior stance, formally concluding that Wansink engaged in academic misconduct, including data falsification, improper statistical techniques, and authorship manipulation.
  • Following the misconduct findings, Wansink was removed from all research and teaching activities and entered mandatory retirement in 2019.
  • According to the database maintained by Retraction Watch, Wansink has accumulated at least 18 formal retractions and 15 errata, including 6 simultaneous retractions by the JAMA Network.
  • The celebrated “Elmo sticker study,” which asserted that branding carrots or apples with cartoon decals doubled pediatric fruit selection, contained fabricated or impossible participant demographic data and was formally retracted.
  • The Smarter Lunchrooms Movement, an intervention implemented across 30,000 American schools using USDA grant funding, was built primarily upon these compromised and unreplicable behavioral datasets.
  • Independent school-based evaluations have failed to replicate the purported magnitude of Smarter Lunchroom interventions, showing no statistically significant changes in real fruit and vegetable intake when measured against plate-waste metrics.
  • Large systematic reviews, including meta-analyses by Robinson et al., 2014, revealed that dishware size manipulations fail to produce reliable, clinically meaningful reductions in ad libitum food intake.
  • Modern nutritional science categorizes uncredentialed behavioral influencers and marketing-driven researchers as “nutrition gurus” who leverage academic branding to monetize unverified lifestyle claims.
  • The Wansink scandal exposed profound systemic vulnerabilities in the peer-review process, which historically relied on institutional trust rather than automated computational data audits.
  • The absence of legal or punitive sanctions beyond reputational loss demonstrates a lack of regulatory accountability for academic fraud funded by federal taxpayer grants.
  • Following his forced retirement, Wansink continued to publish unvetted lifestyle and dietary manuscripts in open-access and predatory journals without standard institutional oversight.
  • The broader scientific field must transition from narrative-driven “nudge” psychology to preregistered clinical protocols, raw data transparency, and objective metabolic endpoints.

III. Adversarial Claims & Evidence Table

Claim from Video / Wansink Lab Speaker’s Evidence Scientific Reality (Current Data) Evidence Grade (A–E) Verdict
Individuals consume significantly more food from “bottomless” self-refilling soup bowls without increased perceived fullness Retracted laboratory eating experiment using rigged tables (Wansink et al., 2005). Retracted by Obesity Research / JAMA family due to data integrity failures. Pre-registered replication attempts show that visual cues modify immediate bite kinetics, but internal gastric distension and neurohormonal satiety peptides (CCK, PYY, GLP-1) rapidly compensate, preventing massive unperceived caloric overconsumption. Level B(Compromised RCT / Inability to Replicate) Unsupported(Retracted / Artifact of Manipulation)
Reducing plate and bowl diameter mechanically decreases caloric intake via the Delboeuf Illusion Multiple Food and Brand Lab publications asserting optical size contrast dictates portion estimation. Comprehensive Cochrane systematic review (Hollands et al., 2015) and meta-analysis (Robinson et al., 2014) show that tableware size has no consistent, statistically significant effect on ad libitum food consumption. Preregistered laboratory trials demonstrate plate-size changes alter intake by under 3% or induce subsequent meal compensation (Kosīte et al., 2019). Level A(Systematic Reviews & Meta-Analyses) Unsupported
Applying cartoon character stickers (e.g., Elmo) to fruit doubles healthy selection among elementary students Wansink’s 2012 Archives of Pediatrics & Adolescent Medicine paper on child snack choice. Retracted. The dataset contained misreported child ages (reporting preschool cohorts as older children), impossible participant tallies, and invalid statistical distributions that failed both GRIM and GRIMMER algorithmic verification. Level B(Retracted Human Trial) Unsupported(Falsified / Retracted)
Choice architecture in school cafeterias (“Smarter Lunchrooms”) reliably reverses pediatric poor diet quality Non-randomized observational pilots and scorecard evaluations from the Cornell Food and Brand Lab. Large-scale independent field trials and regression analyses of Smarter Lunchrooms Scorecards demonstrate that cafeteria layout adjustments do not yield statistically significant increases in student fruit/vegetable intake or reductions in waste (Michigan School Cafeteria Evaluation, 2020). Level C (Field Trials / Implementation Cohorts) Unsupported
Post-hoc iterative data slicing (“p-hacking”) can uncover legitimate biological relationships from null trials Wansink’s explicit defense that extensive subgroup exploration reveals hidden scientific truths. Violates fundamental frequentist probability theory. Testing 20 independent null hypotheses at alpha = 0.05 yields a 64% family-wise false-positive probability (1−(1−0.05)20). Running 400 models guarantees spurious associations via random noise. Level A(Mathematical / Methodological Consensus) Unsupported(Methodological Malpractice)
Environmental cues completely override human metabolic homeostatic drive in the development of obesity Popularized central thesis of Mindless Eating (2006). Human energy balance is regulated primarily by central leptin-melanocortin pathways, vagal mechanosensation, and neuroendocrine signaling (Schwartz et al., 2017). Environmental micro-nudges exert marginal, transient effects that are rapidly overridden by homeostatic appetite drives. Level A(Neuroendocrine Meta-Analyses / Consensus) Unsupported
5 Likes
#2

Sad to witness the decline in standards. Scientific fraud has always been with us but at the margin. The shift began when careers and opportunities to make a great deal of money money were attached to findings.

I’m less certain about the adjective. I continue to see and believe that most scientists, especially those who are scientists at “heart” – if you will – are honest. The core of good science is curiosity and it tends not to coexist with greed – again, IMO.

2 Likes
#3

Yes, obviously tons of scientists are honest and interested in the thruth. But I think people underestimate the amount if fraud out there, we only find the tip of the iceberg, we only know of those few we find. I think at this point we must be reflexively skeptical and tentative about each and every study, no matter how prominent or famous are the scientists involved.

1 Like
#4

There is a difficulty that even if people are not fraudulent they are often quite wrong.

5 Likes
#5

I have to come back to this to acknowledge our own complicity. The number of fraudsters peddling phony pea soup is virtually zero. Health and life extending fraud persists because its fruits – both promised to the deceived and delivered to the fraudster-- are prized. We are by no means immune from the colorations delivered by hopes and desires in this very forum.

3 Likes
#6

To bastardize a line from one of my favorite visionaries, It is the business of science to make mistakes.

2 Likes
#7

That may be true, but it is quite extreme.

#8

Lol this literal serial fraudster tricked me, plate size: food ratio and food intake, I think I saw that from Greger.

This type of fraud might’ve been complicit why nothing in diet and lifestyle ever made a dent on population scale obesity, or it was a supply for a demand when nothing worked.

When proper double blind placebo controlled trials with high compliance (GLP-1a) arrived did population obesity change. That type of study isn’t fabricated when seeking FDA approval because of the costs and incentives.

I was going to say RCT’s are difficult to fabricate, but there were tons of such fraudulent studies for ivermectin and covid.

1 Like
#9

Agree. The RCT is not immune from fraudulent activities. In some ways, its rigid structure makes it easier to identify the “best” way to commit fraud. A much mor common way that RCTs fail is when they are employed in the wrong research context where, even if rigorously true, they have traded so much realism in a misguided attempt to gain precision that the findings – while valid – are not generalizable to the contexts in which you and i live. The RCT is one category of tools in service of scientific progress, it is far from the only one and is too often applied in the health sciences.

1 Like
#10

Why Most Published Research Findings Are False

A Short Note on P-Value Hacking

Interesting observations on interpreting results of RCTs and biomedical literature in general.

#11

We simply need most studies to be done as pre-registrations, in the way that human clinical trials are.

  1. Submit your research plan to your target journal - including the proposed methods, analyses, statistical tests, sample sizes etc.
  2. They peer review the plan, provide suggestions and agree to accept the study once it’s done (or reject, if they think the study isn’t interesting, novel enough etc etc)
  3. You carry out the experiments, following the plan
  4. The journal peer reviewers only need to check that the proposal from step 1 was actually followed, and not much else.

Nature journals are already doing this for some types of study, and I hope it spreads to others.

It removes so much of the disincentive to p-hack, delete outliers and basically tweak the data to find something significant. It rewards people being innovative and taking risks to try new things, test new hypotheses, without having the pressure to find statistically significant results.

1 Like
#12

I would add one more item to the structure. If they reject it though the methodology was followed, it should still be published in a public and indexed archive. Researchers could refer to studies that led to a rejection of the hypothesis. It would be very helpful heuristic tool.

1 Like
#13

My first rule as to the credibility of an RCT is are they trying to disprove their theory or prove it if it’s the ladder then it passes the first test

1 Like
#14

It can be useful broaden the idea of fraud to separate the overall problem of invalid research into a few distinct types. Any brush that attempts to paint all of them in a single stroke is of little value. If you go through this exercise, you may conclude that constructing a taxonomy is difficult when the focus is behavioral research on humans, including medical and health research.

One major category is that of intentionality from which culpability derives and the ‘fraud’ label is appropriate. Did the researcher (here, I’ll denote a person but it could be a team, which introduces other dimensions) have a desired endpoint in mind and therefore at some point decide to ensure that it was achieved? Did he invent data or merely populate missing values in ways that landed in a desired direction. Perhaps he chose a data acquisition strategy that might lean in that direction. Perhaps the statistical tools he chose would be more or less sensitive to certain directions. Opportunity to misrepresent occurs even in the interpretative writeup. Peer review – when it operates as intended – will usually detect cases in which there is a deliberate intention to misrepresent. This is the category, with all of its subcategories, that I believe is still relatively small but has grown in proportion to the financial rewards attached to publication.

A larger major category is misrepresentation that occurs through the application of subconscious, tacit, or unrecognized biases. We might not call such cases outright fraud but the validity of their findings is compromised in ways the reader may not be able to detect. These biases can influence everything from problem conceptualization to the discussion of conclusions and implications. Off the books p-value hunting without the required statistical adjustments is a common way that bias creeps in. Several strategies mitigate against researcher bias. The first is requiring diversity in research teams and ensuring that team members have a voice at each stage of the research process. A description of the research team and its decision-structure should be included in published research. Typically, it is not. Peer review, when it operates as designed, can also detect researcher bias.

The largest major category, and the one with the most subcategories is almost ubiquitous in research, steps out of the dimension of intention and asks a key question: “Are the findings generalizable to the implied context?” Misrepresentation here is more likely to do with incompetence than outright fraud. I mention it here because of how common it is. A study on the effects of vitamin D on the walking gait balance of women over 65 who were provided, and reported taking, 400 IU of D3 daily for 30 days with no pre- or post-measurement of (25(OH)D) is not valid in (i.e., does not generalize to) any ordinary human context. One is tempted to say the study was worthless but it is always possible that there is a valuable fragment of information in such a study. While this example may be obvious, thousands of studies on human health and longevity have been conducted in a way that their findings cannot be generalized to the implied audience. Incidentally, much of the current research on dietary protein falls into this category.