Lab Skills: Uncertainty and Error Bars

One measurement, however carefully taken, could be a fluke. Maybe your hand slipped on the stopwatch, or the ruler was tilted just once. Five or ten measurements, scattered a little around some central value, tell you something a single number never can: how repeatable your result is. That scatter becomes an uncertainty, and on a graph it becomes an error bar, a short line through each point showing the range the true value probably sits in.

Error bars turn curve fitting into a real test. When you draw a best fit line through a set of points, it should pass through or near most of the error bars, not just hit the dots themselves. If the line threads through nearly every bar, the data supports that model. If it keeps missing bars, either the model is wrong or something in the setup is behaving strangely, and that is worth chasing down rather than ignoring.

Error bars also separate two very different problems. Random scatter shows up as points jumping above and below the fit line in no pattern, with error bars overlapping the line roughly half the time on each side. A systematic offset looks different: every point sits consistently above or below the expected line, often with small, tight error bars, meaning your precision is fine but something about the method itself is skewed, like a miscalibrated sensor or a timer that starts late. Spotting which pattern you have tells you whether to take more data or fix your equipment.