The errors that can enter into any kind of measurement, and so limit its accuracy, may be classified into two main types.

First, systematic errors arise from inaccuracies in calibration or the general performance of experimental instruments. The errors are systematic when they are consistent and reproducible. For example, in measuring the length of an object using a ruler whose own length markings are, say 1% too close together, a consistent overestimate by 1% will be obtained. The measurement may be repeated many times using the same ruler, but will give the same error. Of course, in the complicated electronic and mechanical systems used in sophisticated diagnostics, many more complicated possibilities exist for systematic errors to arise. Nevertheless, the principle remains that these errors cannot be revealed by repeated measurement with the same instrument. There is very little to be said in the way of general analysis of systematic errors except that the best way to reveal them is to compare measurements of the same quantity using different instruments (or even different techniques). In the absence of such a check, the experimenter must attempt to estimate the systematic uncertainties from a fundamental knowledge of how an instrument works and what potential flaws there are, or else from his own experience.

The second type of error is random or statistical. There are many possible sources of such errors. In our length measurement example they may arise from slight misreadings of the scale due to misalignment or parallax, for example.