Carter, L. F. (1947). An experiment on the design of tables and graphs used for presenting numerical data. Journal of Applied Psychology, 31, 640-650.

The relative effectiveness of using tables and graphs for presenting functional relationships was investigated by having subjects use tables and graphs which presented the same equation. On the basis of these subjects' results and those of a previous experiment it is concluded that:

1. The speed and accuracy with which tables can be used vary considerably with the type of construction of the table. In general a table containing every point is to be preferred since for most problems it can be used as rapidly and more accurately than simpler tables.
2. The differences between results with different graphs are not systematic enough, nor are they large enough, to indicate that the frequency of coordinate rulings are important in determining the speed and accuracy with which a graph can be used.
3. Except for convention, it makes no difference whether a graph is entered on the x or y-axis.
4. Within the limits of the material investigated, a table in which every point is given is as rapid, and more accurate to use, than any other method of presenting data.
5. Whenever possible tables should be reduced to the very simplest form and entered with the nearest tabulated arguments. If such errors as would arise from this procedure cannot be tolerated, then a very complete table should be used.
6. Both AAF pilots and college students are very slow at, and make large errors in, interpolation.
7. Graphs should not be used as a technique for presenting data unless an interpretation of the shape of the curve presented is important; or unless the speed and accuracy with which the graph is used are relatively unimportant.