The conclusions reached in this study lead to some rather perplexing implications. The first relates to the whole matter of measuring human perceptions through the use of a questionnaire. It is possible that the individuals sampled in this study did not give completely honest answers to every question. If, in fact, this did occur, the reason may be that each of the individuals has had a considerable amount of safety related training throughout his federal service career as a part of his skill training. Thus, he may give the "proper" answer even though he may not implement the safety principle. Certainly, the questionnaire used in this study is not completely without fault. One can never be sure, for example, that rating scales are sufficiently balanced or that the questions are completely understood by the respondents. Thus, in any study employing questionnaires, there is always some error built in. However, the questionnaire was the same for all respondents, therefore, the error can be assumed to be equally distributed throughout the sample group.

A second implication is that the responses to a number of the questions contained a sufficiently large number of U answers that analyzing A and Not-A instead of D and Not-D would have had a significant affect on the outcome of the analysis. While A responses outnumbered Not-A responses by 319 to 289 (52.5 to 47.5 per cent, respectively) this difference is not significant. However, the outcome of the analysis of a number of questions would have changed and it may have been possible to make a stronger case for accepting Kerr's proposal if the U responses had been considered as Not-A responses.

A third implication deals with the matter of the interaction of factors in influencing the perceptions supervisors hold about safety. In this study, age and supervisory level were treated separately. However, it is possible that a thorough analysis of the interaction of these two variables taken from a larger sample size could prove significant. Unfortunately, in this study the number of respondents in high age, lower level and low age, higher level categories was too small to perform statistically significant analysis.

A final implication deals with difference within populations in regard to skill specialty. It is possible, for example, that vehicle mechanics hold differing perceptions of appropriate safety behavior than do aircraft mechanics. It would be interesting, and hopefully significant, to determine if differences in skill specialties have any effect on the perceptions of appropriate safety behavior.