86 



DESIGN AND USE OF INSTRUMENTS 



schemes worth considering are graduations 

 by units, twos, fives, or tens. Vernon (79) 

 found it not difficult to show that scales 

 graduated in other ways were read Tvith 

 more frequent errors. 



Studies of the accuracy of reading scales 

 graduated by tens but using different scale 

 lengths per division are in agreement in indi- 

 cating that the accuracy of reading to tenths 

 of divisions increases as the distance between 

 graduation marks increases to about 0.75 

 inches (Grether and Williams, 36; Kappauf 



02 04 06 08 10 12 14 



ARC LENGTH EQUAL TO ONE SCALE UNIT- IN INCHES 



Fig. 6. Median scale reading error for dials 

 graduated by tens, plotted as a function of the 

 arc length representing unit scale distance. Data 

 from Grether and Williams (36) for 80 subjects, 

 each of whom made 20 readings on each of 4 scales 

 (from among 16 used). All scales were graduated 

 from to 50 and were read under speeded condi- 

 tions. 



Note: Median values have been reinter- 

 preted and are given in this plot as 0.5 scale 

 units less than those given in the original 

 report. 



and Smith, 41). The agreement among 

 these results is the more encouraging in view 

 of the differences in test conditions and the 

 variation in stimulus materials which were 

 used in the experiments. 



Grether and Williams used a speeded read- 

 ing situation where response times to the 

 presentation of single instruments were 

 measured. All scales ranged from to 50 

 but occupied different amounts of the cir- 

 cumference of four differently sized dials. 

 Kappauf and Smith used a serial reading 



task, 12 dials all ahke in design being read in 

 sequence under accuracy instructions. The 

 scales occupied the full circumference of 

 every dial, some scales ranging from to 100, 

 others to values of 200, 400, and 600. The 

 results of these two experiments are shown in 

 Fig. 6 and in Part A of Fig. 7. Note that 

 the curves reach low points, indicating small 

 average errors for Grether and Williams or 

 infrequent errors for Kappauf and Smith, as 

 arc length equal to one scale unit approaches 

 0.05 to 0.1 inches, or as mark separation 

 approaches the size 0.5 to 1.0 inches. 



The accuracy of reading scales which re- 

 quire interpolation to fifths of intervals has 

 been studied by Kappauf and Smith (41). 

 Here it appears that when subjects are read- 

 ing under accuracy instructions there is little 

 reduction in the frequency of their reading 

 errors when the arc length equal to one scale 

 unit is increased beyond 0.02 or 0.04 inches. 

 See Part B of Fig. 7. 



The foregoing paragraphs can be sum- 

 marized by saying that the evidence avail- 

 able at the present time suggests that for 

 greatest accuracy in reading scales to units, 

 the distance allocated to each scale unit 

 should be between 0.05 and 0.1 inches on 

 scales graduated by tens, and need not ex- 

 ceed 0.04 for scales graduated by fives. 

 Comparable data for the reading of scales 

 graduated by twos or units are not available. 



On the converse problem of the accuracy 

 of scale reading when unit scale distance 

 remains fixed and graduation values change, 

 the following data may be cited: For unit 

 scale distances exceeding O.Olj inches, scales 

 graduated by fives are read ^\ith fewer errors 

 than scales graduated by tens (40, 41) . This 

 can be seen by comparing Parts A and B of 

 Fig. 7. For the case of unit scale distance 

 equal to 0.12 inches, graduation by ones is 

 superior to graduation by fives for clock 

 reading (Grether, 29). Information on the 

 effectiveness of graduation by twos is inade- 

 quate at present. 



Discussion in the preceding paragraphs 

 has treated scale reading accuracy as de- 



