Learning to understand cause and effect

One of the most difficult things to demonstrate scientifically is cause and effect. Often a study will show that two items—say, smoking and lung cancer—are associated with one another. But it's another thing entirely to suggest that smoking causes lung cancer in humans. Only after hundreds of studies have been conducted do we now accept that hypothesis as fact.

Now consider a small child, trying to make sense of her world. She might want to know what makes the lights come on in her room. She notices that the lights tend to come on whenever a grown-up walks in. But sometimes, in the middle of the night, a grown-up walks in the room and the light does not come one. Perhaps something else is the cause.

So how do children assign cause and effect to everyday actions? Maybe they believe they themselves have control—by crying, they can summon an adult to the room, and this will cause the lights to come on. Or maybe they assess the probability of an action occurring—since an adult coming into the room is usually associated with lights, it must be the cause.

Tamar Kushnir and Alison Gopnik designed an experiment to try to separate these two mechanisms for determining causality. They showed 4-year-olds a box that appeared to play music when a wooden toy building block is placed upon it. In fact, the box was controlled by the experimenter; this allowed them look at the child's reaction to different interactions between the building blocks and the musical box.

In every trial, one of two different blocks (Block A and Block B) was placed on the box three times—for a total of six (two blocks, three times each). In a given trial, Block A might result in the box playing music three times, while Block B might result in the box playing music only once. Then the child was asked which block should be placed on the box to "make it go" again. In cases like this, 68 percent of the time (significantly better than chance) the child selected Block A.

But what if Block A only caused the box to play music twice out of the three times? If Block B works just one out of three times, then Block A should still be the better bet, right? Kushnir and Gopnik found that in this case, it depended on how the process was presented. For example, consider the following scenario:

In this case, only 63 percent of kids selected Block A—not significantly better than chance. But a change in the procedure caused even more dramatic results. If the experimenter let the child place the blocks on the box for the last two tries, then despite the fact that Block A had corresponded to music twice and Block B had only once, the kids chose Block B a whopping 79 percent of the time: they themselves had placed Block B on the box and seen the music play. They were weighing the effects of their own actions more strongly than the probability of a particular block corresponding to music.

The sequence in which the blocks corresponded to music mattered, too. Consider this variation:

In this case, no matter whether experimenter placed all the blocks on the box herself or the child took over for the last two tries, the child picked Block A nearly all the time (95 percent when the child took over, and 84 percent when the experimenter handled all the blocks).

It appears that even these young children can respond both to probability and to the results of their own actions, using them to make predictions about cause. When probability and the results of their own actions conflict, the child will weigh his or her actions more strongly than the probabilistic choice. So from a very young age, children seem to believe that they have an important influence on events. Only when they see clearly that their own influence is minimal do they take into account other possibilities—even though they are well equipped to do so.

Kushmir, T., & Gopnik, A. (2005). Young children infer causal strength from probabilities and interventions. Psychological Science, 16(9), 678-683.