Philosophy & Ethics · Reality & existence
What does it mean for one thing to cause another?
Test a cause by asking what would have happened without it. Then meet the cases that strain the test, including a backup pump, two rock throwers and a Queen who never watered your flowers.

What you will learn
- Identify a counterfactual question about an event
- Apply a dependence test to a simple, fully described system
- Identify background assumptions in a causal comparison
- Explain why changing several conditions can obscure what difference one event made
Lesson 1 of 4 · 4 min
Testing a cause by asking what if it hadn't happened
In 1748 David Hume gave a test for causes in one line. One thing causes another where, "if the first object had not been, the second never had existed."
Try it on a lamp. Say a working lamp is wired to power through one switch. You turn the switch on and the lamp lights. Would it have lit if you'd left the switch off? In this setup, no.
That question is about something that didn't happen. Philosophers call a claim about what would have happened a counterfactual.
Put the two cases side by side. With the switch on, the lamp lights. With the switch off, it stays dark. Everything else stays fixed, including the working lamp and the power supply. So the lamp's lighting depends on the switch.
That's a stronger claim than saying the switch came first. David Lewis, whose 1973 theory is the best known of this kind, put it like this. "We think of a cause as something that makes a difference, and the difference it makes must be a difference from what would have happened without it."
Now add an unrelated event. A second before the lamp lights, someone outside closes a window. The window closing came first, yet nobody would call it the cause.
The test shows why. Leave the window open and the lamp still lights, because the window isn't part of the circuit. Plenty of things happen just before any event. The test picks out the ones the outcome depended on.
The lamp is easy because you know the whole circuit. Real cases are harder. Say you take a remedy and your cold clears up two days later. Would it have cleared up anyway?
The test tells you what to ask. Answering takes evidence about how colds and the remedy behave.
Try to recallWhy doesn't a window closing just before a lamp lights show that the window caused the light?
Coming first isn't enough. The window isn't part of the circuit, so the lamp would still have lit with the window left open. The lighting didn't depend on it.
To test whether one thing caused another, take it away in your head and keep everything else the same. If the outcome would have happened anyway, it didn't depend on that thing. The answer is only as good as your picture of the system.
Lessons in this topic
Testing a cause by asking what if it hadn't happened
One way to test a cause is to take it away in your head and ask whether the outcome would still have happened. Try the test on a lamp, a switch and a window.
What you hold fixed decides what counts as the cause
A what-if question only tests a cause if you change one thing and leave the past alone. See why two people can name different causes of the same fire and both be right.
Why a backup breaks the what-if test for causes
A backup that never runs makes the real cause look unnecessary. See how David Lewis patched his theory, and how two rock throwers broke the patch.
Can something that didn't happen be a cause?
A gardener who fails to water your flowers seems to have killed them. Follow that thought and the Queen killed them too. See two ways philosophers respond.
Three quick stories from this topic
Two rocks and one bottle
Two rocks thrown at one bottle helped sink David Lewis's first theory of causation
Suzy's rock shatters the bottle a moment before Billy's arrives. Her throw broke it. Yet the bottle would have broken without her, so a test that asks what would have happened otherwise can't credit her. Lewis presented a replacement theory in 1999.
The test it breaks
Lewis's 1973 theory tied causes to dependence. An effect depends on a cause if it wouldn't have happened without it. With Billy's rock right behind, the shattering doesn't depend on Suzy's throw.
What Lewis tried next
His new theory asks what happens if a throw is altered slightly. Alter Suzy's and the shattering changes. Alter Billy's and it doesn't. Critics doubt that this fully solves the problem.
Hume's two definitions
Hume defined cause twice in one sentence, and the two definitions don't match
In 1748 he first said a cause is something always followed by its effect. Then he wrote "or, in other words" and gave a second test, that without the cause the effect "never had existed." Philosophers now treat the two as rival theories.
The two halves
The first says a cause is "an object followed by another, and where all the objects, similar to the first, are followed by objects similar to the second." The second adds "where, if the first object had not been, the second never had existed." The Stanford Encyclopedia of Philosophy calls this the first explicit counterfactual definition of causation.
A road not taken
Hume never explored the counterfactual approach, and generations of empiricist philosophers followed him in preferring regularity. The counterfactual analysis became popular only after David Lewis's 1973 theory.
The Queen and your flowers
By a standard test for causes, the Queen killed your flowers by not watering them
The test asks whether the outcome would have happened otherwise. Had your gardener watered the flowers, they'd have lived, so his neglect killed them. The same is true of the Queen. The philosopher David Lewis accepted that and said most causes just aren't worth mentioning.
Why absences count
Lewis amended his theory so that a failure to act could be a cause. Without that change it couldn't say the gardener's neglect killed the flowers.
The other way out
Some philosophers add a rule that a cause must be abnormal. The gardener was supposed to water the flowers. The Queen never does, so her omission doesn't count.
Sources
- Counterfactual Theories of Causation, Stanford Encyclopedia of Philosophy




