In science classrooms across the country, children learn how to carry out experiments. Imagine two sixth graders looking over a second-story balcony (properly supervised, of course). Jane holds a notebook, pencil at the ready. Jim holds a golfball and a large marble at arm's length over the railing. He releases the two objects. Jane and Jim listen. The golfball and the marble strike almost simultaneously; they hear a sharp rat-tat on the pavement, the sounds nearly overlapping. They repeat the experiment a dozen times, with the same result.

Science! Jane and Jim have replicated Galileo's famous experiment of dropping two balls of different mass from the top of the Leaning Tower of Pisa. (He probably never did this, in reality, but it's a good story.) Two objects of different mass fall at the same rate. In the classroom, with the help of a teacher and some knowledge of algebra--Jane and Jim are advanced for their age--the children are able to connect their results to Newton's Second Law of Motion. The science class unit they're studying is titled "The Scientific Method."

Most of us adults, perhaps with a few reminders, can dredge up memories of how the scientific method works. It helps when when we read a newspaper story about new or continuing scientific findings: the Earth is warming; neutrinos may travel faster than the speed of light; a new gene that governs aging has been discovered. We trust these results because we know about the scientific method. It's typically described along these lines:

1. You start with some general question about the world--you're a curious person at heart.
2. You develop a hypothesis, a specific kind of what-if question. "If I were to make observations of such and such, what would I expect them to look like?" Your hypothesis is based on what you already know: your experience and your knowledge of others' experience.
3. You design an experiment to collect observations that will test your hypothesis.
4. You run your experiment.
5. You analyze your results. Are they in line with your hypothesis? Whether they are or not, you've learned something new.

This seems pretty straightforward. But it's not enough. A few years ago I was talking with a friend online, and she said something along these lines: "I can understand some of the science I read about, but other times it's so complicated that I can't follow it. Do I just need to trust the scientists who did the work?"

This is a poser, and my description of the scientific method doesn't hold an obvious answer. It may seem as though trusting a scientific result is a matter of faith in people. But that's because I've left out two important aspects of the way science is done in practice. These ideas are touched on in science education, but they're not emphasized, sometimes neglected entirely. The first idea is peer review, and the second is replication.

Almost every scientist spends time reviewing the research of others. You receive an email message from an editor of a journal in your area, asking you to review a newly arrived article. You read the abstract, sometimes the entire article, and judge whether you're qualified to evaluate the research. If you accept the request, you write a detailed review of the work, covering both strengths and weaknesses. Other reviewers are doing the same. You usually don't know who they are, nor they you. The reviewers often don't know who the authors of the article are, and the authors generally don't know who you are. There's anonymity all around, except for the editor managing the process, in order to avoid obvious biases. The focus is on the validity and importance of the research. Peer review is one safeguard against unjustified results entering the literature.

Replication provides another, more important safeguard. When scientists publish their findings, they include enough information about what they did so that others can replicate the results. For especially surprising results, the work will be viewed with some scientific skepticism until others have produced the same findings, perhaps by repeating the same experiment exactly, or by reaching the same results by a different path, or by testing the implications of the new theory behind the results. As more evidence accumulates, there's more and more reason to believe (or disbelieve) the initial findings.

This means that we don't need to understand every detail of all the research papers that have been published on a topic, and yet we can still accept some specific results. It is a matter of trust, but not of individual scientists. It's trust of the process. It's not reasonable to be skeptical of a process that's worked well for centuries. It's also unreasonable to cherry pick specific findings and claim that they undercut an entire field of study, as creationists and climate change deniers often do. I sometimes grow impatient with such folks, because they don't seem to understand how science works.

Take global climate change. You might wonder whether the evidence has been fixed, somehow, perhaps by some fishiness in the review process for results outside the current consensus. This is what Climategate was about: researchers were accused of subverting the review process in order to preserve claims about the existence of global warming. In the end, independent panels made recommendations about improving transparency but found no evidence that that the peer review process was subverted [1]. Things could have turned out differently, but they didn't. The safeguard was tested and found to be sound.

You might also wonder whether there's something wrong with the basic science. Despite the review process and the replication, perhaps climate scientists are going about their research in the wrong way. This was the thought behind the establishment of the Berkeley Earth Surface Temperature project [2], an effort to reanalyze climate data using new techniques. The group found that "global warming is real." Replication worked.

 One interesting aspect of the BEST project was that it involved outsiders to the field of climate research. It might be surprising to think that outsiders could tell scientific experts anything that they didn't already know, but this is a natural part of science--there are no closed scientific guilds. I've found this to be true myself, in a different field of study. Some years ago I became interested in animal behavior, as a path toward understanding how we might build better robots. My students and I read the literature and developed what we thought was a useful new perspective on how to think about animals that use tools in nature. Was it a worthwhile perspective? There was a good way to find out: We wrote a paper and submitted it to a reputable journal. Experts vetted the piece, and eventually, after several revisions, it was published [3]. Not bad for computer scientists. Anyone with the interest, time, and ability can do science. I should emphasize the "time" part: I remember talking about our initial ideas in 1999, and the article didn't appear until 2008. But if we hadn't gone through the process, we wouldn't have been able to tell, really, whether our ideas had much value.

Of course, not everyone has the time or motivation to effectively become a scientist, just to understand some bit of science. What then? As I hope I've illustrated, you can trust the process. Once you discover what the scientific consensus is, it's straightforward. Not to trust well-accepted scientific views, in a rational way, would mean coming up with some other explanation for why they're well-accepted. A global conspiracy? Hardly. Collusion among authors and reviewers with personal connections? It's possible to detect such problems, and peer review is just one safeguard; there are others. External incentives to produce bad results? Again, such problems can be identified and rooted out. In the end, not accepting strong scientific consensus is something like not believing, say, in the existence of the island of Madagascar. You've never been there, and what you've seen at second hand could be the result of a grand conspiracy between biologists, map makers, nature photographers, and the Disney Corporation. Fortunately, the consensus is that there's a simpler explanation.  

1. The Reviews into the University of East Anglia's Climatic Research Unit’s E-mails, The Science and Technology Committee, The House of Commons, UK.
2. http://berkeleyearth.org
3. Robert St. Amant and Thomas E. Horton, Revisiting the definition of tool use, in Animal Behaviour, Volume 75 (2008), Number 4, pp. 1199-1208.