error bars of the 1-year group. The open circles represent the results of the first test and the closed circles those of the 1-year retest. Figure 18.9 shows the results of the 2-year group. In both sets of data, there was no significant change in the slope of the power function, nor was there a significant shift of the position of the curves on the vertical coordinate. Analysis of the numerical responses of the individual subjects in both groups also showed no significant change in response well beyond the .01 level of confidence. We must conclude that absolute magnitude estimation, at least for the judgment of apparent length, yields extremely stable results over time, both in terms of individual performance and for the averaged performance across individuals.
In summary, I think we have shown that the method of absolute magnitude estimation, its counterpart magnitude production, and the combination of the two methods called numerical magnitude balance can be used effectively to measure the growth of sensation magnitude of vibrotaction. These methods and the method of direct intensity matching produced remarkably similar results.