PREDICTION OF IN-HABITAT BEHAVIOR 



Robert Helmreich, John Wilhelm and Roger Bakeman 

 The University of Texas at Austin 



Roland Radloff 

 National Science Foundation 



The development of predictive tests has been a major preoccupation of 

 psychology. Countless hours have been spent on the design of paper and pencil 

 measures to predict job performance, psychological states, and a variety of 

 other behaviors. One practical need for successful prediction is obvious. 

 When behavior can be explained as a function of antecedent conditions, indi- 

 viduals can be selected who have the maximum chance of success in any given 

 role. A good example of a situation demanding precise prediction is the selec- 

 tion of astronauts. Training costs are high in terms of money and time, mission 

 costs are enormous, and the psychological or physical failure of one man could 

 not only abort a mission but also imperil the lives of teammates. Although 

 selection of pilot-astronauts has been successful to date, the overall state of 

 the art in psychological selection leaves much room for improvement. 



As we mentioned previously, demographic variables appear to be most successful 

 in prediction of behavior but have generally not been used or studied system- 

 atically. We have described the development of the Life History Questionnaire 

 as an instrument for demographic investigation and prediction. In this section 

 we will discuss the effectiveness of pre-mission predictive variables in account- 

 ing for the actual behavior recorded during missions. 



Criteria for Prediction 



One of the most persistent problems in the development of predictive instru- 

 ments is obtaining valid, quantifiable criteria of performance. For example, 

 attempts to predict adult mental health as a function of childhood experiences 

 face the problem of defining mental health in an objective, quantifiable manner. 



One solution has been to employ dichotomous criteria such as passing or failing 

 a course, completion or noncompletion of pilot training, etc. Considerable 

 success has been obtained in predicting such dichotomous criterion variables. 

 However, this type of prediction has serious limitations because it forces a 

 broad spectrum of behavior into a limited number of arbitrary categories. It 

 does not, for example, provide information on the difference between individuals 

 who barely meet the criterion and those whose performance is outstanding. This 

 is particularly important when selecting from a population of highly qualified 

 candidates, all of whom can meet the criterion. The task in this case is to 

 discriminate between those whose performance will be superb and those whose will 

 be merely excellent. 



We have pointed out that one of the reasons for our enthusiasm about TEKTITE 

 was the availability of continuous, objective behavioral criteria. With high 

 reliability, a large number of observations (2866 to 4770 records on each cri- 

 terion per aquanaut, depending on mission length), and large individual differ- 

 ences, the project provided a magnificent research setting for the investigation 

 of predictive measures. For the purposes of this discussion, we have chosen 



VII I -50 



