Using "error bars" on graphs is a good way to help students see that, within the inherent uncertainty of the measurements due to the instruments used for measurement, the data points do, in fact, lie along the line that represents the linear relationship. In this article, the author explains why connecting the dots on graphs of collected data is not enough. Whether it is because of uncertainty due to measurement or the lack of consideration of other variables, it is unrealistic to expect collected data to give a graph that describes an exact mathematical relationship--that all the points will fall exactly on a certain line or curve that represents a particular mathematical formula. Rather, what everyone should look for are general trends in the data. With enough experimental data and repetition, it is possible that these general trends can suggest or confirm exact mathematical relationships. The bottom line, though, is that the trends that lead to more exact relationships are not going to be obvious if all one does is connect the dots on graphs of collected data. One might end up with a drawing that looks vaguely like a bird or a snake, but not something of scientific value. (Contains 5 figures.)