Knudsen's formula given above applies equally well to this averaging instrument. Neglecting for the moment the quantity in the square brackets, we have I = c p where c = (c1 + c2)/2 but we also have I = h l where h is the mean height and l the length of the figure, therefore h l = c p; but in this instrument we make p = l. Hence h = c, or the mean height of the figure is equal to the mean distance between the dents. The quantity in the brackets is too great to be neglected, however. If we were always dealing with circles, the ratio (R/2p)² would be a constant, and numerically equal to 1/16 or 6.25 per cent. Then all we should have to do would be to use a scale 6.25 per cent. longer than the true scale. But with a long narrow figure such as an indicator diagram, this ratio is much smaller. The measurement of a large number of diagrams gave a mean value of 1/60 for diagrams 4 in. long. It is obvious that, if a diagram be shortened, this ratio will increase, for the value of R does not decrease as rapidly as p, and vice versa; hence this ratio varies approximately inversely as the length of the diagram. Taking the value of 1/60 for the 4 in. diagram, this is equivalent to saying that there is an error of 1 in 60, or 1.67 per cent., in the result, and from the formula it will be seen that the result is too great by this amount; hence, if we make the length, l, between the legs of the instrument 1.67 per cent. of 4 in., or 0.067 in. less than the length from the tracing point to the center of the hatchet, p, we shall compensate for the error on a diagram 4 in. long. But the ratio of this constant quantity 0.067 in. to the length of the diagram also varies inversely as the length in just the same manner as the ratio R/2p, hence this method of correcting the instrument is approximately right for all lengths of diagrams. It must be remembered that if this correction were entirely neglected, it would not exceed two per cent.; hence any inaccuracy in this correction is an exceedingly small quantity, well under 1 per cent. Whenever errors have been attributed to the instrument, on examination it has always been found that they were due to carelessness in setting the length to the diagram, or to the tracing leg having been grasped so tightly as to cause side slip. The accuracy of the instrument may be easily demonstrated by drawing a rectangle, say about 4 in. long and 2 in. high, and finding the mean height by the averager, then by doubling the paper over and comparing its height with the mean distance between the dents, it will be found that they agree if the instrument has been carefully used.