of each vertical bar so that they would appear in each case one less than the figures given. With Fig. 249 as it is, the 1908 bar shows a great increase over 1907 and the reader is quite justified in wondering how it happened that a greatly increased number of telephones were installed during a panic year. From the chart as shown the reader is not-likely to realize that 1908 is getting credit for the telephones installed during 1907, which happened to be a very prosperous business year. Having the data recorded as of December 31 each year overcomes the difficulty and makes certain that no false impression can be obtained.
Fig. 250 shows the data of Fig. 249 plotted as a smooth curve. For a trained class of readers the curve presentation is preferable to the bar presentation, for it permits seeing the fluctuations which have occurred from year to year more easily than they can be seen by glancing from bar to bar in Fig. 249. Within a few years it is probable that curves will be so well understood that a report to stockholders could best be made using the method of Fig. 250 instead of the method of Fig. 249.
In order to show the different impressions which may be had if various proportions between the horizontal and vertical scales are used, Fig. 251 has been plotted from the same data as Fig. 249 and Fig. 250. For Fig. 251 an assumption was made that the chart would be printed on exactly the same size page as was used for Fig. 249. The scales for Fig. 250 were, however, arranged in the other direction on the page and the co-ordinate ruling was made so that some space would be allowed for extension of the curve in future years. As seen from Fig. 251 the growth in the telephone business does not appear nearly so rapid as would be thought from observing Fig. 250. Each of these charts is, however, plotted to exact scale and the difference in the impression obtained is caused only by the proportions of the vertical and the horizontal scales. The appearance of less rapid growth in Fig. 250 is assisted somewhat by the fact that the large-type title of the chart is arranged horizontally instead of in the form of a square as seen in Fig. 250. The heavy black type with much greater spread horizontally than vertically tends to overshadow the curve itself and causes for the curve a more distinctly horizontal impression than would otherwise be obtained. A person reading charts must take great care that he does not give too much weight to the actual appearance of the curve on the page, instead of basing his conclusions on the percentage increase or decrease as judged from the figures of the vertical scale. The proper choice of scales for curve plotting is largely a matter of judgment, and
