Tony: You’re right—I covered that in an earlier essay. Here, I’m focusing on how journalists, geneticists, and other scientists often reinforce the reflexive dismissal of observational research.
"Correlation may not prove causation, but it is a good hint!"
Observations which involve noting correlations is the beginning of the scientific process. It is hair-pulling lunacy to ignore correlation. It's like my former MIL telling me that the fact that I got violently ill every time after eating certain foods didn't mean anything and I should continue to eat those foods. Yeah, she was a bitch and so is anyone who tells you to continue to consume poison because it isn't "proved" yet.
Maggie: I agree—scientific insight often begins with an anecdote, a careful case study, or an observed correlation. Those early signals shouldn’t be dismissed simply because they are preliminary or incomplete. They are how hypotheses are born. At the same time, we distinguish between signals that generate questions and evidence that settles them. Some correlations—like the population-level patterns that revealed herd immunity—become powerful tools when they are consistent, biologically plausible, and repeatedly confirmed. Others remain clues that require deeper investigation.
The was an interesting Cochrane Review article several years ago on sources of lead poisoning in children. They concluded paint was not a source of lead exposure to children because they excluded any studies that were not controlled. All of the studies of children's blood lead levels going down after lead paint abatement didn't have a control group where there was a group of children with no intervention. Of course it would have been unethical to have such a control group.
Holly: Do you mean the 2020 Cochrane review which found that environmental, educational, or combined interventions, such as cleaning up dust or removing paint, often fail to significantly reduce lead levels in children's blood? They didn't dismiss that lead paint was a potential source, but they did conclude the evidence supporting specific household interventions to reduce lead exposure was weak. We conducted a RCT of lead hazard reduction and found some evidence that blood lead levels declined in the more heavily exposed children, but lead exposure was relatively low. You can read the trial here: https://jamanetwork.com/journals/jamapediatrics/fullarticle/2696978
Thank you, this is such a crisp diagnosis of a deadly double standard.
I love the Jebel Irhoud contrast because it exposes the epistemic asymmetry in one move: we’ll revise human origins on fragments of bone, yet when bodies, at population scale, show consistent, biologically coherent signals of harm from involuntary exposures, we suddenly demand an impossible bar (randomized trials of pollution, absolute certainty, “no confounding ever”). That isn’t scientific rigor; it’s selective skepticism with predictable winners and predictable casualties. 
Your point about the “convenient confusion” around correlation is also exactly right. Good epidemiology doesn’t stop at association, but it stress-tests it: timing, dose–response, replication across settings, and mechanistic plausibility. Meanwhile, other correlational designs (twin studies, anecdotes, fossil fragments) are routinely treated as insight-generating and actionable. 
And the policy takeaway is the part that should be stapled to every chemical regulatory meeting: when exposure is widespread, persistent, and non-consensual, waiting for “perfect proof” isn’t neutrality, but it’s a decision to keep running the experiment on the public while industry externalizes the costs. Lead, PFAS, tobacco, asbestos—same script, same delay, same downstream harm.
You wrote: "Nowhere is this blind spot more consequential than with pollution and chemicals in commerce."
Fluoride is a major industrial pollutant. During the period after World War 2 there were more lawsuits over fluoride air pollution than all over air pollutants combined.
Fluorides are also vital industrial chemicals. We need them for our advanced civilization. So we have to accept some fluoride exposure, but we don't have to accept it being added to our tap water.
Science has consistently found harm from fluoride exposure starting before World War 2, during the 20th century and ever earlier. This explains why there was a lot of resistance to the first proposal to add fluoride to tap water in 1939. And this explains why most of the world has never adopted this practice or stopped doing it entirely.
With the now admitted (by the EPA) neurotoxic effect of fluoride ingestion, we see pro-fluoridation studies that look at the IQ of adults long after they had their fluoride exposure when young. They claim there is no negative effect from early fluoride exposure in adults , but we really don't know how much each individual ingested in these studies.
It seems these studies are being done because they can't produce science showing children do not have negative brain effects from fluoride exposure.
I had always though that if the correlations met Bradford Hill's criteria, causation is highly likely and governments should act immediately to limit exposure of the population to the toxicant. But Greg Martin says more needs to be done. https://www.youtube.com/watch?v=0JV1orh1QGE In my opinion, proving plausibility is crucial, requiring biochemical evidence of damage. Epidemiologists often ignore the biochemical evidence because they do not have enough experience in understanding the in vitro, animal and human biochemical studies.
Hardy: I agree that government agencies should act once there is sufficient evidence. I indeed, I am not convinced they should wait until all of Bradford Hill’s criteria are met. I also agree that identifying biochemical effects or biomarkers—whether in laboratory or human studies—is valuable. Observational findings are harder to dismiss when there is supporting evidence for a specific mechanism. That said, plausibility does not require a defined mechanism. Plausibility asks whether an observed association is consistent with existing biological knowledge—whether it makes biological sense given what we know at the time. Mechanistic evidence goes further: it provides direct evidence of how an exposure produces an effect by identifying specific pathways, targets, or intermediate processes.
Hardy, we've discussed this issue before. I disagree that "proving plausibility is crucial". And what does it mean to "prove" plausibility? Plausibility is not much different than "possibility" and does not confer certainty. Also, it is too much to require proof of biochemical damage. In many cases, the biological mechanism for harm of environmental agents is not well known by the time it is obvious there is harm being caused by the agent. Lead and IQ loss is a good example. Observational epidemiology and animal toxicology studies can be black boxes as far as the biological mechanism. Sure, it can strengthen the case for causality when you also have evidence of a biological mechanism that explains how the harm is caused, but it is not necessary, and waiting to have any certainty on the mechanism can add decades of delay to admitting something is causing harm and acting to reduce that harm.
Chris (and Dr. Abell, who says "The obsession with depending on carefully controlled lab studies is non-sensical with real world problems."
- I'll yield to the expert opinion, found with the help of AI.
Biochemical evidence is crucial for epidemiological studies because it transforms observational data into precise, mechanistic, and validated findings, addressing the inherent limitations of questionnaire-based, observational research. It provides objective, measurable markers—often referred to as biomarkers—that significantly increase the sensitivity, specificity, and accuracy of exposure assessment and disease identification.
Key reasons for incorporating biochemical evidence include:
-Precise Exposure Assessment: Instead of relying on questionnaires, which can be subject to recall bias, biochemical markers can quantify the actual internal dose of a toxicant or pathogen in tissues or bodily fluids (e.g., blood, urine, or DNA adducts).
-Validation of Causal Mechanisms: Biochemical data provide evidence of biological plausibility, helping to strengthen causal claims by demonstrating the exact mechanism by which an exposure leads to a disease.
-Reduced Misclassification: It helps to reduce errors in classifying exposure or disease status,, thereby improving the validity and power of studies.
Identification of Susceptibility: Biochemical and genetic markers can identify specific, high-risk populations who may be more vulnerable to certain exposures due to genetic factors, thus providing insight into effect modification.
-Early Detection: Molecular markers can detect early biological effects or subclinical, pre-disease changes, which allows for earlier intervention and more effective prevention strategies.
Strengthening Risk Assessment: By providing more accurate data on the dose-response relationship, biochemical evidence is critical for formulating evidence-based policy, regulatory standards, and public health interventions.
I watched Martin's video and was underwhelmed. It was really a thinly veiled advert for the University of Limerick and for his own YouTube channel. I agree on the need to "act immediately". Hill's work was originally on workers health in industrial settings. The real focus re causation should be mechanism of harm - which is key to plausibility. And that can very much include "confounding variables". The obsession with depending on carefully controlled lab studies is non-sensical with real world problems.
Take, for example, the issue over vaccination that resulted in non-scientist Robert Kennedy Jr becoming Czar over regulatory agencies in the U.S. Is autism "caused" by vaccines? Is it caused by agricultural chemicals?
According to Samson et al, based on a massive time series correlation over multiple years between CDC data and Ag data on use of glyphosate (Round Up - Bayer/Monsanto) the correlation with autism is R = 0.989. (JOS_Volume-9_Number-2_Nov_2014-Swanson-et-al. I have the .pdf)
But Kennedy's crusade focuses on vaccines — and for a reason. Many incidents reported by parents appear to show vaccination as a trigger for rapid onset of autism. Is there a connection? Is there one or more confounding variables here? Yes. Is the mechanism clear? Yes. Is policy an issue. Yes. Is the never ending search for profit an issue, Yes.
Policy - earlier vaccination of younger populations because of workforce issues. (more day care with very young children whose immune system is still under development)
Profit - (1) "let's just combine multiple disease vaccines into one. Cheaper to administer. Let's say measles, mumps, and rubella. (Let's find a way to take over the bodies own cells and have THEM create the trojan horse (mRNA) with no real research on the long term impacts.)"
(2) "We need a cheap adjuvant - thimerisol or similar (mercury Hg) or Aluminum compounds. Mercury and aluminum are toxic and known to cause brain damage but the body (the liver specifically) screens them out of the blood. Worth the small risk ...."
Confounding variable - glyphosate, which owes its "effectiveness" to being a chelating agent.
Mechanism
(1) In Sri Lanka, an "epidemic" of chronic kidney disease in an agricultural area was identified by scientists as probably resulting from glyphosate chelating mutliple toxic metal ions and thereby having them bypass the livers screening mechanism.
(2) Samsel and Seneff have published multiple papers on glyphosate suppression of critical body and brain hormones.
So (1) make it easy for brain-toxic metals to make it to the centre of intelluctual development of young children while simultaneously (2) depriving the same brain of critical hormones.
All science starts with observation and perception of what is happening with the subject or effects being observed
Uncompromised observation of cause and effect leads to scientific rigour over time, however, vested interests want absolute proof as opposed to risk considerations with foolish arguments and marketing to protect profits an face saving: a stupid and idiotic way of dealing with health
Thank you Richard. I would add that this behavior is short-sighted and, at times, sociopathic. We shouldn’t expect large corporations to voluntarily change practices that increase profits, but we should expect government agencies to do a far better job of protecting the public.
I believe we SHOULD expect large corporation to be honest and to work for the good of both workers and customers. That was part of the tenants of Keynesian economics. Speaking of "causal" you can trace the increasing wealth inequity back to Milton Friedman economics that arose in the 1960's.
"But when epidemiologists document higher risks of cancer, autism, or cardiovascular disease linked to toxic chemicals or pollution, skepticism hardens into paralysis. Why?"
Well said. Noncorrelational human studies are an impossible barrier to breach when studying if something is toxic. They would be unethical. You can’t randomize people to lead exposure vs control for example. This point is left unsaid by critics of Epidemiology. Correlation is all we will ever have, so we need to accept these findings when replicated.
Another great article. And I love that the header image seems to echo the "See no evil, hear no evil, speak no evil" cartoons. That's a good summary of the denialism about environmental health harm evidence that vested interests have promoted for years. Was that an intentional reference or just accidental?
I am surprised you did not mention Hill's criteria. This is a systematic approach leading from correlation to causation that should and can almost always be applied. The Seneff studies on glyphosate were time series correlations over a long period (20 years) for example. As for sociopathic - clearly! Andrew Harvey, in his book "The Hope", told of a meeting initiated by the head of a major agribiz company in the mid '90s. The CEO said "My company knows the devastation we are causing to thousands of lives. I don't care" I quoted Harvey (with permission) in my book on wealth inequality, "Gush Up".
You missed the other half of industry’s ‘correlation is not causation’ : ‘animals are not humans’
Tony: You’re right—I covered that in an earlier essay. Here, I’m focusing on how journalists, geneticists, and other scientists often reinforce the reflexive dismissal of observational research.
"Correlation may not prove causation, but it is a good hint!"
Observations which involve noting correlations is the beginning of the scientific process. It is hair-pulling lunacy to ignore correlation. It's like my former MIL telling me that the fact that I got violently ill every time after eating certain foods didn't mean anything and I should continue to eat those foods. Yeah, she was a bitch and so is anyone who tells you to continue to consume poison because it isn't "proved" yet.
Maggie: I agree—scientific insight often begins with an anecdote, a careful case study, or an observed correlation. Those early signals shouldn’t be dismissed simply because they are preliminary or incomplete. They are how hypotheses are born. At the same time, we distinguish between signals that generate questions and evidence that settles them. Some correlations—like the population-level patterns that revealed herd immunity—become powerful tools when they are consistent, biologically plausible, and repeatedly confirmed. Others remain clues that require deeper investigation.
The was an interesting Cochrane Review article several years ago on sources of lead poisoning in children. They concluded paint was not a source of lead exposure to children because they excluded any studies that were not controlled. All of the studies of children's blood lead levels going down after lead paint abatement didn't have a control group where there was a group of children with no intervention. Of course it would have been unethical to have such a control group.
Holly: Do you mean the 2020 Cochrane review which found that environmental, educational, or combined interventions, such as cleaning up dust or removing paint, often fail to significantly reduce lead levels in children's blood? They didn't dismiss that lead paint was a potential source, but they did conclude the evidence supporting specific household interventions to reduce lead exposure was weak. We conducted a RCT of lead hazard reduction and found some evidence that blood lead levels declined in the more heavily exposed children, but lead exposure was relatively low. You can read the trial here: https://jamanetwork.com/journals/jamapediatrics/fullarticle/2696978
Thank you, this is such a crisp diagnosis of a deadly double standard.
I love the Jebel Irhoud contrast because it exposes the epistemic asymmetry in one move: we’ll revise human origins on fragments of bone, yet when bodies, at population scale, show consistent, biologically coherent signals of harm from involuntary exposures, we suddenly demand an impossible bar (randomized trials of pollution, absolute certainty, “no confounding ever”). That isn’t scientific rigor; it’s selective skepticism with predictable winners and predictable casualties. 
Your point about the “convenient confusion” around correlation is also exactly right. Good epidemiology doesn’t stop at association, but it stress-tests it: timing, dose–response, replication across settings, and mechanistic plausibility. Meanwhile, other correlational designs (twin studies, anecdotes, fossil fragments) are routinely treated as insight-generating and actionable. 
And the policy takeaway is the part that should be stapled to every chemical regulatory meeting: when exposure is widespread, persistent, and non-consensual, waiting for “perfect proof” isn’t neutrality, but it’s a decision to keep running the experiment on the public while industry externalizes the costs. Lead, PFAS, tobacco, asbestos—same script, same delay, same downstream harm.
You wrote: "Nowhere is this blind spot more consequential than with pollution and chemicals in commerce."
Fluoride is a major industrial pollutant. During the period after World War 2 there were more lawsuits over fluoride air pollution than all over air pollutants combined.
Fluorides are also vital industrial chemicals. We need them for our advanced civilization. So we have to accept some fluoride exposure, but we don't have to accept it being added to our tap water.
Science has consistently found harm from fluoride exposure starting before World War 2, during the 20th century and ever earlier. This explains why there was a lot of resistance to the first proposal to add fluoride to tap water in 1939. And this explains why most of the world has never adopted this practice or stopped doing it entirely.
With the now admitted (by the EPA) neurotoxic effect of fluoride ingestion, we see pro-fluoridation studies that look at the IQ of adults long after they had their fluoride exposure when young. They claim there is no negative effect from early fluoride exposure in adults , but we really don't know how much each individual ingested in these studies.
It seems these studies are being done because they can't produce science showing children do not have negative brain effects from fluoride exposure.
I had always though that if the correlations met Bradford Hill's criteria, causation is highly likely and governments should act immediately to limit exposure of the population to the toxicant. But Greg Martin says more needs to be done. https://www.youtube.com/watch?v=0JV1orh1QGE In my opinion, proving plausibility is crucial, requiring biochemical evidence of damage. Epidemiologists often ignore the biochemical evidence because they do not have enough experience in understanding the in vitro, animal and human biochemical studies.
Hardy: I agree that government agencies should act once there is sufficient evidence. I indeed, I am not convinced they should wait until all of Bradford Hill’s criteria are met. I also agree that identifying biochemical effects or biomarkers—whether in laboratory or human studies—is valuable. Observational findings are harder to dismiss when there is supporting evidence for a specific mechanism. That said, plausibility does not require a defined mechanism. Plausibility asks whether an observed association is consistent with existing biological knowledge—whether it makes biological sense given what we know at the time. Mechanistic evidence goes further: it provides direct evidence of how an exposure produces an effect by identifying specific pathways, targets, or intermediate processes.
Hardy, we've discussed this issue before. I disagree that "proving plausibility is crucial". And what does it mean to "prove" plausibility? Plausibility is not much different than "possibility" and does not confer certainty. Also, it is too much to require proof of biochemical damage. In many cases, the biological mechanism for harm of environmental agents is not well known by the time it is obvious there is harm being caused by the agent. Lead and IQ loss is a good example. Observational epidemiology and animal toxicology studies can be black boxes as far as the biological mechanism. Sure, it can strengthen the case for causality when you also have evidence of a biological mechanism that explains how the harm is caused, but it is not necessary, and waiting to have any certainty on the mechanism can add decades of delay to admitting something is causing harm and acting to reduce that harm.
Chris (and Dr. Abell, who says "The obsession with depending on carefully controlled lab studies is non-sensical with real world problems."
- I'll yield to the expert opinion, found with the help of AI.
Biochemical evidence is crucial for epidemiological studies because it transforms observational data into precise, mechanistic, and validated findings, addressing the inherent limitations of questionnaire-based, observational research. It provides objective, measurable markers—often referred to as biomarkers—that significantly increase the sensitivity, specificity, and accuracy of exposure assessment and disease identification.
Key reasons for incorporating biochemical evidence include:
-Precise Exposure Assessment: Instead of relying on questionnaires, which can be subject to recall bias, biochemical markers can quantify the actual internal dose of a toxicant or pathogen in tissues or bodily fluids (e.g., blood, urine, or DNA adducts).
-Validation of Causal Mechanisms: Biochemical data provide evidence of biological plausibility, helping to strengthen causal claims by demonstrating the exact mechanism by which an exposure leads to a disease.
-Reduced Misclassification: It helps to reduce errors in classifying exposure or disease status,, thereby improving the validity and power of studies.
Identification of Susceptibility: Biochemical and genetic markers can identify specific, high-risk populations who may be more vulnerable to certain exposures due to genetic factors, thus providing insight into effect modification.
-Early Detection: Molecular markers can detect early biological effects or subclinical, pre-disease changes, which allows for earlier intervention and more effective prevention strategies.
Strengthening Risk Assessment: By providing more accurate data on the dose-response relationship, biochemical evidence is critical for formulating evidence-based policy, regulatory standards, and public health interventions.
I watched Martin's video and was underwhelmed. It was really a thinly veiled advert for the University of Limerick and for his own YouTube channel. I agree on the need to "act immediately". Hill's work was originally on workers health in industrial settings. The real focus re causation should be mechanism of harm - which is key to plausibility. And that can very much include "confounding variables". The obsession with depending on carefully controlled lab studies is non-sensical with real world problems.
Take, for example, the issue over vaccination that resulted in non-scientist Robert Kennedy Jr becoming Czar over regulatory agencies in the U.S. Is autism "caused" by vaccines? Is it caused by agricultural chemicals?
According to Samson et al, based on a massive time series correlation over multiple years between CDC data and Ag data on use of glyphosate (Round Up - Bayer/Monsanto) the correlation with autism is R = 0.989. (JOS_Volume-9_Number-2_Nov_2014-Swanson-et-al. I have the .pdf)
But Kennedy's crusade focuses on vaccines — and for a reason. Many incidents reported by parents appear to show vaccination as a trigger for rapid onset of autism. Is there a connection? Is there one or more confounding variables here? Yes. Is the mechanism clear? Yes. Is policy an issue. Yes. Is the never ending search for profit an issue, Yes.
Policy - earlier vaccination of younger populations because of workforce issues. (more day care with very young children whose immune system is still under development)
Profit - (1) "let's just combine multiple disease vaccines into one. Cheaper to administer. Let's say measles, mumps, and rubella. (Let's find a way to take over the bodies own cells and have THEM create the trojan horse (mRNA) with no real research on the long term impacts.)"
(2) "We need a cheap adjuvant - thimerisol or similar (mercury Hg) or Aluminum compounds. Mercury and aluminum are toxic and known to cause brain damage but the body (the liver specifically) screens them out of the blood. Worth the small risk ...."
Confounding variable - glyphosate, which owes its "effectiveness" to being a chelating agent.
Mechanism
(1) In Sri Lanka, an "epidemic" of chronic kidney disease in an agricultural area was identified by scientists as probably resulting from glyphosate chelating mutliple toxic metal ions and thereby having them bypass the livers screening mechanism.
(2) Samsel and Seneff have published multiple papers on glyphosate suppression of critical body and brain hormones.
So (1) make it easy for brain-toxic metals to make it to the centre of intelluctual development of young children while simultaneously (2) depriving the same brain of critical hormones.
All science starts with observation and perception of what is happening with the subject or effects being observed
Uncompromised observation of cause and effect leads to scientific rigour over time, however, vested interests want absolute proof as opposed to risk considerations with foolish arguments and marketing to protect profits an face saving: a stupid and idiotic way of dealing with health
Thank you Richard. I would add that this behavior is short-sighted and, at times, sociopathic. We shouldn’t expect large corporations to voluntarily change practices that increase profits, but we should expect government agencies to do a far better job of protecting the public.
I believe we SHOULD expect large corporation to be honest and to work for the good of both workers and customers. That was part of the tenants of Keynesian economics. Speaking of "causal" you can trace the increasing wealth inequity back to Milton Friedman economics that arose in the 1960's.
"But when epidemiologists document higher risks of cancer, autism, or cardiovascular disease linked to toxic chemicals or pollution, skepticism hardens into paralysis. Why?"
Thank you Ken. Cheers from Bowen Island.
Well said. Noncorrelational human studies are an impossible barrier to breach when studying if something is toxic. They would be unethical. You can’t randomize people to lead exposure vs control for example. This point is left unsaid by critics of Epidemiology. Correlation is all we will ever have, so we need to accept these findings when replicated.
Another great article. And I love that the header image seems to echo the "See no evil, hear no evil, speak no evil" cartoons. That's a good summary of the denialism about environmental health harm evidence that vested interests have promoted for years. Was that an intentional reference or just accidental?
I am surprised you did not mention Hill's criteria. This is a systematic approach leading from correlation to causation that should and can almost always be applied. The Seneff studies on glyphosate were time series correlations over a long period (20 years) for example. As for sociopathic - clearly! Andrew Harvey, in his book "The Hope", told of a meeting initiated by the head of a major agribiz company in the mid '90s. The CEO said "My company knows the devastation we are causing to thousands of lives. I don't care" I quoted Harvey (with permission) in my book on wealth inequality, "Gush Up".