A causal relationship exists when one event directly produces another, not simply when two things happen to occur together. If you’re trying to figure out which scenario among several options represents a true causal link, the key is checking whether the supposed cause actually precedes the effect, whether a clear mechanism connects them, and whether a hidden third factor could explain the pattern instead. Understanding these criteria makes it straightforward to separate genuine cause-and-effect from coincidence.
Three Requirements for Causation
The classical framework for establishing causation, originally outlined by philosopher John Stuart Mill, requires three things. First, temporal precedence: the cause has to come before the effect. Second, covariance: the cause and effect must be statistically related. Third, no alternative explanation: you need to rule out the possibility that some unmeasured third variable is actually driving both.
A randomized controlled trial is considered the gold standard for proving causation precisely because it satisfies all three criteria. By randomly assigning people to a treatment group or a control group, researchers ensure that the groups are balanced at the start. Any difference in outcomes can then be attributed to the treatment itself, not to some pre-existing difference between the groups. Observational studies, by contrast, can show that two things are related and even that one comes before the other, but they struggle with that third requirement. There’s always the possibility of a confounding variable lurking in the background.
What Causal Relationships Look Like
The most famous example of a proven causal relationship in public health is smoking and lung cancer. As early as 1939, a German researcher named Franz Hermann Müller showed that people with lung cancer were far more likely to have been smokers. By 1954, British researchers Richard Doll and Austin Bradford Hill had demonstrated that people who smoked 35 or more cigarettes per day were 40.6 times more likely to die from lung cancer than nonsmokers. That same year, a large American Cancer Society study concluded the link was proven “beyond a reasonable doubt.”
This relationship checks every box: smoking precedes the cancer by years or decades, heavier smoking correlates with higher cancer rates (a dose-response relationship), there’s a clear biological mechanism (inhaled carcinogens damaging lung tissue), and no third variable has ever been identified that explains the pattern away. When you’re evaluating whether something is causal, this is the template to compare against.
How Correlation Tricks You
The classic example of a non-causal correlation is ice cream sales and drowning deaths. These two statistics rise and fall in near-perfect sync. But ice cream doesn’t cause drowning. The hidden third variable is summer weather. Hot temperatures drive more people to buy ice cream and also drive more people to swim, which increases the number of drownings. The two events correlate by chance through a shared cause, not because one produces the other.
Spotting these spurious correlations comes down to asking a simple question: is there a plausible mechanism by which A directly produces B, or could something else be causing both? If you can easily identify a confounding variable that explains the entire pattern, you’re probably looking at correlation, not causation.
Reverse Causality: When the Direction Is Wrong
Sometimes two things genuinely are connected, but people get the direction backwards. This is called reverse causality, and it’s a common trap in health research. A well-known example involves alcohol and heart disease. For years, studies suggested that moderate drinkers were healthier than non-drinkers, implying that alcohol had a protective effect. But researchers eventually noticed a problem: many of the “non-drinkers” in these studies were people who had quit drinking because they were already sick. The illness came first, then the decision to stop drinking. This “sick quitter” effect made abstainers look less healthy than they actually were, creating the illusion that moderate drinking was beneficial.
A similar pattern shows up with blood pressure. In older populations, blood pressure tends to decline in the years before death. If you only look at a snapshot of the data, it can appear as though lower blood pressure is associated with dying sooner. In reality, the approaching death is causing the blood pressure drop, not the other way around.
How to Evaluate a Causal Claim
Epidemiologists use a checklist of nine considerations, known as the Bradford Hill criteria, to evaluate whether an observed association is likely causal. You don’t need to memorize all nine, but the most useful ones for everyday thinking are:
- Strength: A larger effect size makes causation more likely. Smoking increasing lung cancer risk by a factor of 40 is hard to explain away as a fluke.
- Consistency: The relationship shows up across different populations, time periods, and study designs.
- Temporality: The cause always precedes the effect. This is the one non-negotiable criterion.
- Dose-response: More exposure leads to more of the effect. More cigarettes means higher cancer risk.
- Biological plausibility: There’s a known or reasonable mechanism explaining how A could produce B.
Biological plausibility is particularly important because it connects statistical patterns to actual biology. An association between two variables is much more convincing when scientists can explain the pathway by which one influences the other. Without that mechanism, even a strong statistical relationship could be coincidental.
Applying This to Multiple-Choice Scenarios
When you’re presented with several relationships and asked which is most likely causal, run through a quick mental checklist. Does one event clearly precede the other in time? Is there a direct, plausible mechanism connecting them? Could a third variable explain the pattern instead?
Relationships that are almost certainly causal tend to involve a direct physical or biological mechanism: taking a medication and experiencing a known side effect, exposure to a pathogen and developing an infection, or increasing exercise and seeing improved cardiovascular fitness. Relationships that are probably just correlations tend to involve two outcomes that rise and fall together without any logical connection, or where an obvious confounding variable (like age, income, season, or geography) could explain both.
If one option describes a scenario where A directly acts on the body or environment to produce B, and the other options describe patterns that could easily be explained by a shared third factor, the first option is your causal relationship. The presence of a clear mechanism, combined with the correct time sequence, is what separates cause from coincidence.