A few years back, during the holidays, I was asked to help out one of my nieces with a small science project. She had to read a book and come up with an experiment based on something in the book, and I was a logical choice to guide her in this endeavor.

And then I was told that baking something would count as a science project, which elicited a mental face-palm. Baking is not science. I pointed out to some of the assembled friends and family that science is about quantifying things — making a measurement.

“You measure things when you bake. A cup of flour, a teaspoon of salt, etc.”

OK, let’s try again: science is about quantifying a prediction, and measuring a result. And that has to be a little more involved than “if I measure these things and follow the recipe, cupcakes will result.” I’m not sure if I got through.

I avoided further controversy by finding a passage in the book where someone observed that a barrel of water froze, and the ice expanded. Aha! That’s a prediction (albeit a general one) and we can measure the expansion of ice. We measured the volume of some water, froze it, and measure the increase in volume by immersing the ice cubes in some cold water in a measuring cup, so there was a discussion of Archimedes principle in there as well. And we got an answer of about 10% expansion. Good enough for grade-/middle-school science.

But things I read keep bringing me back to this: following a recipe is not science. Memorization is not science. They are useful tools for doing science, but individually they are just that, tools. If you want to use baking as an example of science, make a quantifiable prediction and test it. What happens if you change ingredients? Does it matter if you mix the flour and sugar together before adding the eggs and butter? Why? There might be some chemistry there — that gets you started down the path of scientific inquiry.

There’s a section of Surely You’re Joking, Mr. Feynman where Feynman tells about teaching while on sabbatical, and how the students had memorized everything. If you asked the right question, you got a great answer. But when you went off-script, you got nothing, because the students didn’t understand what the words meant and what the definitions implied. It was like memorizing a bit of text in a foreign language, and speaking it perfectly, but having no idea of the meaning of what you just said.

One practical reason that following a recipe or rote memorization does not qualify as science is this: in any reasonably complicated experiment, something will go wrong, and it’s the job of the scientist to figure out why and track down where the problem is. Because there is cause and effect, you can diagnose what signals (or a lack thereof) mean, and uniquely identify a problem. Sometime you build that into the apparatus (e.g. this warning light turns on when that signal gets too big or small, or a needle will go into the red-line area if a bird drops a baguette into your exhaust fan), but for a lot of table-top apparatus it’s a matter of twiddling knobs and isolating individual components. It’s frustrating to have to track down these problems, but ultimately rewarding (to me, anyway) to figure out the puzzle.

Here’s a recent, small example: I was setting up a spectroscopy layout; we had tried this before and couldn’t make it work well enough to suit us, but a newer paper had come out explaining a trick or two we hadn’t considered. I set it up according to the drawing in the paper, and got a funny (funny unexpected, not funny ha-ha) result. Up to that point, I was following a recipe — split up a beam of light with a polarizing beamsplitter, and send them through a vapor cell to do saturated absorption spectroscopy, and modulate one of the beams. But when I adjusted the angle of the half-wave plate (which should have controlled how much light goes into each polarization path), I couldn’t get one of the beams to go to zero intensity, as it should have, and that’s where the science comes in: there was a problem with the polarization. Was the light somehow elliptically polarized? That could be from a bad waveplate. Was it the laser itself, or the optical isolator at the output? Was it the beamsplitter or another piece of optics? That’s where the science came into play. It turns out that the beamsplitter was made for a different wavelength had been put into the wrong drawer, and was always reflecting some light, even when it should have had 100% transmission. Problem solved. By Science™!