Thank you for this important and highly relevant post. We really need to fix this problem. There are very real, current cases of "an entire area of scientific inquiry can go “off the rails” and stay off the rails for a long time, sometimes even arriving at an incorrect “consensus.”
In my own field, autism epidemiology, there is a terrible problem with "consensus" that is actually just continued repetition of speculation with no evidence. That's right, no evidence. Once an opinion, speculation or conjecture is repeated enough times it gets treated as if it were an established fact — to the point that papers can state is as a fact in an abstract with no citations and no supporting evidence. It ends up in UpToDate. Then science journalists can and do report on it to the public, falsely claiming that it's a fact. While doing all that fiddling, Rome burns.
This is beautifully argued — especially the distinction between basic (random) errors and more systematic “non-Markovian” or directional ones.
Your example of the gardening literature is spot on.
I made a complimentary argument in Nature this week: we need career paths that reward synthesis, responsiveness, error detection, and research fraud recgonition — not just output or grant dollars: https://www.nature.com/articles/d41586-025-02258-7
Thanks for continuing to push for rigor in a way that’s both clear-eyed and constructive.
Just like the idea of democracy presumes an informed and virtuous citizenry, the idea of science as a self-correcting practice presumes that a certain kind of person is more likely to become a scientist: someone dedicated (to a greater or lesser degree) to the pursuit of true knowledge. That is why it’s so important to realize that science is a social practice, done by a community of people, not an autonomous machine inexorably spinning the flax of data into the gold of truth. Like any group of people, scientists are susceptible to economic, cultural, and social forces, which affect their choice of research topics, methods, and publication venues etc etc.
Because of this, it’s very important to incentivize (financially, culturally, socially, normatively) truth-seeking amongst scientists, and strongly disincentivize political or social partisanship. Similarly, we should not tolerate the social script of scientists as “prophet-priests”, which does not conduce towards truth seeking for its own sake (I say this as a very religious person myself!).
I understand the point being made with the random walk analogy, but I believe (unless my memory fails me) that a random walk doesn’t lead you back to your starting point in the long run, or even stay in the vicinity of the starting point on average. Because if it did, it wouldn’t be random. So if we use random walk as an analogy for scientific investigations, it casts doubt on the idea that enough studies will eventually self-correct to a reliable answer.
I have made it a point of emphasis in my reading of science related publications to perform 'citation analysis'. I look for those citations that are central to the paper's conclusions or inferences and dig into them. Too often I find that (as in your examples) the citations say something less - or categorically different - than claimed by the author(s). Your highlighting this issue is important and welcome!
I love science, but I don’t trust scientists. This wasn’t always the case, but some of the nuttiest claims I’ve heard come from scientists. For instance, we were told by a well-known scientist that protesting against COVID-19 lockdowns was a threat to public health. When asked if Black Lives Matter protesters were also a threat to public health, that same person said no - these protests were good for public health because racism is a worse health threat than even COVID. Most of everything public health said about the pandemic was BS. However, I see the same issues in all areas of healthcare. Cochrane reviews indicate that fewer than 10% of common medical treatments are supported by high-quality evidence. The biggest problem I see isn’t methodology, it’s integrity. Science has been perverted by money, politics, ideology, and ego. It’s going to be very difficult to fix these.
Your assertions of lack of integrity appear more driven by your personal bias rather than science. You missed the point of the article- for science and scientists to succeed, we have to put aside our preconceived notions and examine facts without inserting our a priori assumptions and Fox News driven lack of critical thinking
I provided a concrete example to support my claim regarding integrity in healthcare science, using a meta-study of Cochrane reviews. I referenced scientific evidence. Here is the link. I have plenty more,
I understood the study, but you failed to understand my simple point that, without integrity, methodology is irrelevant. You revealed your bias by invoking Fox news. You seem to think that anyone who challenges a scientist is an ignorant right-winger who watches Fox News. This attitude only leads to more skepticism. You never learned that ad hominem attacks always backfire. To persuade people, present them with high-quality evidence. Skip the appeal to authority bit.
Polls show increasing skepticism among the public of scientific claims. This is NOT a communication problem. Why don't you try listening to why people are feeling this way instead of ascribing motives to them?
Well said, John. I have worked in preclinical and clinical research for the last 11 years both at the bench and as a study director/Global Project Lead and it has been my experience that there are many highly credentialed folks in the research industry with the same unfounded arrogance as Texas Cat Lady. It's shocking, really.
Great information on science. Simple summary is that most of the "scientific errors" for many reasons were not truly based on scientific methods. Significant revelation that oversight is performed by part time people who lack the knowledge and experience to perform that duty.
I really enjoyed this. I've never really bought the idea that science is self-correcting, at least in so far it will just happen, just as the universe proceeds to entropy. Perhaps over lengthy time periods. I almost wonder whether self-correcting is the wrong term and we need something more aspirational, like driven-correcting. Well that's the best I can come up with on limited coffee.
Hey, Guys - As long as we have people abusing the methods of science to exonerate widespread exposures as their singular goal, and as long as well allow them to label their fraud as science, science will have a terrible, well-earned reputation. We need to hold our colleagues to the proper standards we learned in graduate school. It's simple. #BeBrave https://popularrationalism.substack.com/p/yet-another-epidemiological-hatchet
Excellent article! One general cause of the problem is not adhering to principles of measurement, experimental design, and analysis. If we could better teach core principles I think researchers would do better research. For the statistics part I made an attempt of listing principles at https://fharrell.com/post/principles (input would be appreciated). On another dimension the famous NIH-HCI statistician-epidemiologist Nathan Mantel once quipped “There’s nothing wrong with bad cancer research that a little lack of funding woudn’t fix.”.
And then there is the politically motivated climate science graveyrtrain.
...the number of articles indexed under "climate change" in the Web of Science Core Collection increased from an annual average of almost 16,000 between 2014 and 2018 to over 33,000 annually between 2019 and 2023...
One wonders how much of this is replicatable, and how is it collated and curated with any meaningful utility?
Thank you for this important and highly relevant post. We really need to fix this problem. There are very real, current cases of "an entire area of scientific inquiry can go “off the rails” and stay off the rails for a long time, sometimes even arriving at an incorrect “consensus.”
In my own field, autism epidemiology, there is a terrible problem with "consensus" that is actually just continued repetition of speculation with no evidence. That's right, no evidence. Once an opinion, speculation or conjecture is repeated enough times it gets treated as if it were an established fact — to the point that papers can state is as a fact in an abstract with no citations and no supporting evidence. It ends up in UpToDate. Then science journalists can and do report on it to the public, falsely claiming that it's a fact. While doing all that fiddling, Rome burns.
This is beautifully argued — especially the distinction between basic (random) errors and more systematic “non-Markovian” or directional ones.
Your example of the gardening literature is spot on.
I made a complimentary argument in Nature this week: we need career paths that reward synthesis, responsiveness, error detection, and research fraud recgonition — not just output or grant dollars: https://www.nature.com/articles/d41586-025-02258-7
Thanks for continuing to push for rigor in a way that’s both clear-eyed and constructive.
Science is rigor and if you do it in Bayesian fashion you're guaranteed to always approach truth.
Doug Altman wrote this in 1994 in a BMJ, "As the system encourages poor research it is the system that should be changed.
We need less research, better research, and research done for the right reasons.
Abandoning using the number of publications as a measure of ability would be a start."
Even after 20 years, little has changed!
Amyloid Beta and alzheimers anyone?
Feeding their families is a scientist's primary concern.
Everything else is way down the list.
What an excellent article.
Just like the idea of democracy presumes an informed and virtuous citizenry, the idea of science as a self-correcting practice presumes that a certain kind of person is more likely to become a scientist: someone dedicated (to a greater or lesser degree) to the pursuit of true knowledge. That is why it’s so important to realize that science is a social practice, done by a community of people, not an autonomous machine inexorably spinning the flax of data into the gold of truth. Like any group of people, scientists are susceptible to economic, cultural, and social forces, which affect their choice of research topics, methods, and publication venues etc etc.
Because of this, it’s very important to incentivize (financially, culturally, socially, normatively) truth-seeking amongst scientists, and strongly disincentivize political or social partisanship. Similarly, we should not tolerate the social script of scientists as “prophet-priests”, which does not conduce towards truth seeking for its own sake (I say this as a very religious person myself!).
I understand the point being made with the random walk analogy, but I believe (unless my memory fails me) that a random walk doesn’t lead you back to your starting point in the long run, or even stay in the vicinity of the starting point on average. Because if it did, it wouldn’t be random. So if we use random walk as an analogy for scientific investigations, it casts doubt on the idea that enough studies will eventually self-correct to a reliable answer.
Great analysis. Therefore happy now also to be a paid subscriber.
Q: I'd like to see older 'subscriber only' posts. Can you unlock them?
You should be able to. Please open a recent "Don't be Fooled" Wednesday video to check.
Thnx. 🤔 I thought the same. And like to learn. So I did check. See images for the result.
It seems my paid subscription is somewhere stuck? See images. I can’t open paid episodes - But I can add comments as paid subscriber.
I have made it a point of emphasis in my reading of science related publications to perform 'citation analysis'. I look for those citations that are central to the paper's conclusions or inferences and dig into them. Too often I find that (as in your examples) the citations say something less - or categorically different - than claimed by the author(s). Your highlighting this issue is important and welcome!
I love science, but I don’t trust scientists. This wasn’t always the case, but some of the nuttiest claims I’ve heard come from scientists. For instance, we were told by a well-known scientist that protesting against COVID-19 lockdowns was a threat to public health. When asked if Black Lives Matter protesters were also a threat to public health, that same person said no - these protests were good for public health because racism is a worse health threat than even COVID. Most of everything public health said about the pandemic was BS. However, I see the same issues in all areas of healthcare. Cochrane reviews indicate that fewer than 10% of common medical treatments are supported by high-quality evidence. The biggest problem I see isn’t methodology, it’s integrity. Science has been perverted by money, politics, ideology, and ego. It’s going to be very difficult to fix these.
Your assertions of lack of integrity appear more driven by your personal bias rather than science. You missed the point of the article- for science and scientists to succeed, we have to put aside our preconceived notions and examine facts without inserting our a priori assumptions and Fox News driven lack of critical thinking
I provided a concrete example to support my claim regarding integrity in healthcare science, using a meta-study of Cochrane reviews. I referenced scientific evidence. Here is the link. I have plenty more,
https://pubmed.ncbi.nlm.nih.gov/35447356/
I understood the study, but you failed to understand my simple point that, without integrity, methodology is irrelevant. You revealed your bias by invoking Fox news. You seem to think that anyone who challenges a scientist is an ignorant right-winger who watches Fox News. This attitude only leads to more skepticism. You never learned that ad hominem attacks always backfire. To persuade people, present them with high-quality evidence. Skip the appeal to authority bit.
Polls show increasing skepticism among the public of scientific claims. This is NOT a communication problem. Why don't you try listening to why people are feeling this way instead of ascribing motives to them?
Well said, John. I have worked in preclinical and clinical research for the last 11 years both at the bench and as a study director/Global Project Lead and it has been my experience that there are many highly credentialed folks in the research industry with the same unfounded arrogance as Texas Cat Lady. It's shocking, really.
Great information on science. Simple summary is that most of the "scientific errors" for many reasons were not truly based on scientific methods. Significant revelation that oversight is performed by part time people who lack the knowledge and experience to perform that duty.
I really enjoyed this. I've never really bought the idea that science is self-correcting, at least in so far it will just happen, just as the universe proceeds to entropy. Perhaps over lengthy time periods. I almost wonder whether self-correcting is the wrong term and we need something more aspirational, like driven-correcting. Well that's the best I can come up with on limited coffee.
Hey, Guys - As long as we have people abusing the methods of science to exonerate widespread exposures as their singular goal, and as long as well allow them to label their fraud as science, science will have a terrible, well-earned reputation. We need to hold our colleagues to the proper standards we learned in graduate school. It's simple. #BeBrave https://popularrationalism.substack.com/p/yet-another-epidemiological-hatchet
Excellent article! One general cause of the problem is not adhering to principles of measurement, experimental design, and analysis. If we could better teach core principles I think researchers would do better research. For the statistics part I made an attempt of listing principles at https://fharrell.com/post/principles (input would be appreciated). On another dimension the famous NIH-HCI statistician-epidemiologist Nathan Mantel once quipped “There’s nothing wrong with bad cancer research that a little lack of funding woudn’t fix.”.
Now do "climate science."
And then there is the politically motivated climate science graveyrtrain.
...the number of articles indexed under "climate change" in the Web of Science Core Collection increased from an annual average of almost 16,000 between 2014 and 2018 to over 33,000 annually between 2019 and 2023...
One wonders how much of this is replicatable, and how is it collated and curated with any meaningful utility?
Geez, I thought with the computer age and A/i there would be no more errors. What happened is that they are compounded at light speed.