146 



DESIGN OF CONTROLS 



bility to practical situations is not yet 

 known. 



1. Range effect. It has long been known 

 that judgments of the magnitude of a vari- 

 able are distributed over a narrower range 

 than the input magnitudes. That is, large 

 inputs are underestimated, small inputs over- 

 estimated, mean inputs unaffected. This 

 phenomenon has recently been observed in 

 experimental investigations of responses to 

 sudden displacements of the "target." In 

 the Naval Research Laboratory experiment 

 (20), the equipment was similar to a com- 

 pensatory tracking system. A bright dot on 

 an oscilloscope, centered on a hairline, was 

 suddenly displaced and the operator was re- 

 quired to make a translatory movement with 

 a lever type device to bring the dot back to 

 the hairline as soon as possible. In the 

 Indiana University study (11) the "dot" 

 formed by a vertical line moving behind a 

 narrow horizontal slit was suddenly dispaced 

 and the operator was required to realign his 

 sliding pointer with the target "dot" as 

 quickly as possible. In both of these studies 

 large displacements were undershot on the 

 average and small displacements were over- 

 shot on the average. Results from another 

 study at Indiana University (12) show a sig- 

 nificant difference between the mean ampli- 

 tude of response to a one-inch displacement 

 when that displacement is the largest of a 

 sequence of variable displacements and when 

 it is the smallest of a sequence. The former 

 two studies were designed to investigate the 

 response to a step-function input in continu- 

 ous tracking (Type C) . However, it may be 

 that these responses are more accurately de- 

 scribed as Type B or "setting" adjustments. 

 2. Predictability. This variable of the 

 total input sequence is still in the status of a 

 reasonable hjrpothesis. The results of the 

 Indiana University studies seem to indicate 

 that regularities in the input sequence allow 

 the operator to predict certain aspects of fu- 

 ture inputs. Such prediction adversely af- 

 fects the proportionality betw^een the input 

 and the output of the system. The impor- 



tance of the relationship will be discussed 

 later. 



Operator Variables 



We have noted above that the correlation 

 between input and output in the case of the 

 operator can not be "built in." We have 

 also noted that "the operator" is an abstrac- 

 tion of the transmission functions of a human 

 being in a particular situation. The degree 

 to which the designer's abstraction fits the 

 individual operator in any given instance is 

 affected by the variability between operators 

 and the variability within one operator as a 

 function of time. A few brief statements are 

 therefore pertinent before considering the 

 system variables related to the operator. 



Variability of operator characteristics in 

 respect to a particular system design involves 

 changes due to a change of operators, 

 changes accompanying practice, and changes 

 due to fatigue. By means of selection and 

 training the range of individual differences 

 may be reduced considerably. By designing 

 the system in terms of the "average opera- 

 tor" the effects of these differences may be 

 further reduced. The effects of practice 

 may be controlled to a satisfactory extent by 

 utilizing the characteristic form of the 

 "learning curve." The development of 

 standardized motor responses follows a de- 

 celerating course. By training operators to 

 a point where little if any change results from 

 additional practice and by maintaining them 

 at that level, learning effects may be mini- 

 mized. (It should be noted that this prin- 

 ciple carries implications for the designer and 

 the research worker relative to the differences 

 between characteristics of trained and un- 

 trained operators.) Finally, the effects of 

 fatigue may be reduced by control of the 

 operating environment, by attention to the 

 difficulty and duration of the task, and by 

 the conditioning and adaptation which ac- 

 companies the regular performance of a task. 

 In control systems the outputs of the op- 

 erator which are standardized for a given 

 class in inputs consist of movement patterns. 

 These patterns may vary in terms of dura- 



