An additional significant difference between the two groups was in birth order. 

 Members of the top performing group were likely to be first-born in their 

 families than were those in the bottom group (F = 4.23, £<.05) . This signif- 

 icant difference between the two groups is in the opposite direction from that 

 found in SEALAB II (where first and only-borns performed significantly less well 

 than later -borns) but was predicted. As mentioned in the introduction, the 

 stress level induced in the shallow, warm-water habitat was assumed on a priori 

 bases to be significantly lower than that created in the more cramped habitat 

 located at 200 feet in murky, cold water in SEALAB. Where stress levels are 

 low or moderate, first-borns typically outperform later-borns (see Altus, 1966, 

 for a review of such findings). The obtained result is, therefore, consistent 

 with previous birth order findings and confirms the assumption that extreme 

 levels of stress were not induced by the TEKTITE environment. 



Predictive Equations 



Heretofore we have been discussing predictive variables only in isolation. This 

 is inadequate because such variables interact. That is, predictive variables 

 may be correlated with each other to a greater or lesser degree. To obtain 

 maximal effectiveness in the prediction of a behavioral criterion, one must con- 

 sider the combined effects of the predictive measures employed. This is typic- 

 ally done by using the statistical technique of multiple regression. While a 

 discussion of multiple regression is beyond the scope of this report (see 

 McNemar, 1970, and Cohen, 1968), suffice it to say that multiple regression 

 results in a statistic, the multiple correlation coefficient (R) ; R squared is 

 the percentage of variance on a criterion measure which is accounted for by a 

 given set of predictors. The multiple regression procedure consists of the 

 calculation of weights for each predictive variable which produce the highest 

 possible correlation between the criterion measure and the group of predictive 

 variables used. 



The ability of Life History Questionnaire variables to account for observed 

 variance in behavior was tested by performing multiple regression analyses using 

 the three sets of conceptual variables independently. It must be emphasized 

 that these analyses are preliminary and that the development of final, predic- 

 tive, test batteries will be undertaken only after extensive cross-validation 

 studies of the LHQ in other populations have been completed. Despite this dis- 

 claimer, it can be seen that impressive multiple correlations with the criterion 

 can be obtained using conceptually discrete sets of predictors. 



The multiple correlation between nine Environmental LHQ variables and the cri- 

 terior was .79. Variables were selected from the E pool using the Wherry Test 

 Selection method (Wherry, 1946; Hutchins, 1970). The obtained multiple R means 

 that more than 627o of the criterion variance was accounted for by this group of 

 predictors. The variables used for prediction were: clothing quality, food 

 quantity and quality, family size, height, parental physical affection, parental 

 physical punishment, father's employment, hometown size, and size of school. 



A second multiple regression was computed using 1 LHQ variables as predictors. 

 Five I^ variables formed the equation, again using the Wherry technique. The 

 resultant multiple R was .69, accounting independently for 487o of the criterion 

 variance. The variables used in this prediction were: financial independence, 

 school performance, religious activity, work (during the school year) and work 

 (during summer vacations) . 



VIII- 57 



