In physics we have a uniquely high standard for evidence and proof. We tend to talk about things like standard deviations, “5-sigma” uncertainty, and error propagation. When we talk about data and evidence, we’re not talking about anecdotes or qualitative reasoning. When we talk data, we mean numerical measurements with associated measurement uncertainties. Here’s an example taken from the Journal of Applied Physics to illustrate this point.
This group of scientists has made a plot of thermal conductivity of a material vs. temperature of the material. The blue curve is the predicted behavior, and the red dots are the measured values. Now the critical point: the vertical lines represent the uncertainty in each measurement. For individual data, this is usually represented as
y = x +/- Dx
, where x is the measured value and Dx is the uncertainty in the measurement. The “error analysis,” as we call it, is rigorous and fool-proof, and with observations like this plot we can see that the theoretical predicted curve (red) falls within the range of each individual measurement and its associated uncertainty. This kind of rigorous data analysis and criterion for what constitutes proof is a defining characteristic of physics discourse.
Another takeaway from this brief discussion is the essential use of pictures, which brings us to the topic of information in physics discourse. The easiest way to see the validation/invalidation of a theory is to make on of these graphs with theoretical predictions and measurements. One can actually see how good the model is. This amounts to a powerful argument. If an institution or physics community is able to produce good evidence, information, and arguments, then we qualify that institution as expert in the field. This amounts to saying that this institution is a reliable producer of knowledge in the field.
Charles Golden 