The power of random assignment  (Dan Reisberg) (ch.4)

Is it hazardous to talk on a cell phone while driving? Many people believe it is, and point to evidence showing that people who use the phone while driving are more likely to be involved in accidents, compared to people who do not use the phone while driving. But is this evidence persuasive?

Actually, this evidence is ambiguous-open to more than one interpretation. Being alert to this ambiguity is crucial for science, because, if data can be interpreted in more than one way, then we can draw no conclusions from them. What is the ambiguity in this case? Perhaps talking on a phone while driving is distracting, and increases the likelihood of an accident. But, as an alternative, perhaps the drivers who use cell phones while on the road are the drivers who were, from the start, less cautious, or more prone to take risks. This lack of caution is why these people talk on the phone while driving, and it's also the reason why they're more often involved in accidents. Thus, cell-phone use and accidents go together, but not because either one causes the other.

In technical terms, the problem here is that the cell-phone users are a self-selected group. In other words, they decided for themselves whether they'd be in our "experimental group" (the cell-phone users) or our "control" group (people not using cell phones). Presumably, they put themselves in one group or the other for some reason-because of some tendencies or attributes. And it might be those tendencies or attributes that cause the observed outcome-in this case, the increased accident rate.

If we really want to examine the effects of phone use on driving, therefore, we need to make sure that our "phone group" and our "no-phone group" are equivalent to begin with, before cell phones enter the scene. If we then discover that cell phones are associated with more accidents, we'd know that the cell phones are indeed at fault, and not some preexisting difference between the groups.

Psychologists usually achieve this matching of groups by means of random assignment. Rather than allowing participants to sort themselves into a group of phone users and a group of nonusers, we, the experimenters, assign them to one group or the other on some random basis (perhaps a coin toss). This won't change the fact that some drivers are careful and others are not, or that some are more attentive than others. But our coin toss ensures that careless drivers have an equal chance of ending up in the phone or no-phone groups, and likewise for careful drivers, or risky ones. As a result, our two groups will end up matched to each other, with each group containing the same mix of different driver types.

Random assignment is one of the most important tools in a psychologist's research kit-ensuring that groups are matched before an experiment begins. That way, if the groups differ at the end of the experiment, we can be sure it's because of our experimental manipulation, and not because of some preexisting difference.

And with all of this said, what about cell-phone use? The evidence suggests that talking on a cell phone while driving is dangerous, because of the distraction. Hence there is an important message in these data-but it's a message we can offer only because of the power of random assignment.