A Tri-modal Risk Attitude Indicator for Aviation Personnel: Design and Diagnostic Validity

Marian Popa
Ph.D., National Institute of Aeronautical Medicine, Bucharest, Romania
Bucharest University, Romania
Cezarina Rotaru
National Institute of Aeronautical Medicine, Bucharest, Romania
Bucharest University, Romania
Ioana Oprescu
National Institute of Aeronautical Medicine, Bucharest, Romania
Bucharest University, Romania


In accordance with Atkinson’s risk preference theory, we see the conduct within risk situations as a result of opposite tendencies between risk and prudence orientation. In order to asses risk attitude, we designed a system made up of three specific indicators, based on a genuine interpretation of AC-REF (ACceptance-REFusal) values inventory (Zörgö, 1977) . Risk significance (as a physical threat, danger for life or health) of the 272 AC-REF items has been subjected to an evaluation on a 1 to 10 scale, by a sample of 96 subjects. Following this procedure, 43 items with risk significance and 36 items with prudence significance were chosen. A sample of 1168 subjects has been evaluated with a computerized version of AC-REF inventory. Risk\prudence orientation modes have been calculated as a pondered sum of the transformed values of the items having risk\prudence significance. A composed indicator, meaning risk\prudence balance attitude, has been computed as difference between the two primary indicators. The three indicators, as three specific modes of risk attitude (risk preference, prudence preference, and risk \prudence balance), show very good psychometric features. The diagnostic (construct) validity of these indicators was tested against some major questionnaires such ALAPS (Armstrong Laboratory Aviation Personality Survey), Cattell 16PF, and Eysenck Personality Inventory. First of all, we got consistent and significant correlations between risk attitude mode indicators and some of the ALAPS scales, especially for “risk taking”. The other questionnaires also show significant association with these indicators. The three modal indicators show significant variations as far as psychosomatic type is concerned, but also related to age and gender.


There are very few human behaviors that do not imply a certain level of risk, meaning by that, on the one hand, a degree of uncertainty regarding target achievement, and, on the other hand, some related potential losses. The specific means of referring to a risk situation is approached through the concept of “attitude toward risk”. This refers to an individual feature endowed with a certain degree of constancy, which tends to actualize in a relatively constant manner in various risk situations.

Risk behavior has been seen as closely related to decision-making situations (Ceauşu, 1972, 1976, Plous, 1993), and also, as an important component of the usual repertoire of the pilot’s professional behavior. As a result of this, the assessment of attitude towards risk is especially important for the behavioral and adjustment prediction of the air personnel. The main goal of our study is to conceive a system made up of indicators relevant for the risk attitude, which may be used within the process of psychological assessment of aviation personnel. At this moment we intend mainly to develop a procedure of outlining these indicators and evaluating psychometrical and differential diagnostic qualities.


Two major models explain behavior in risk situations: cognitive and motivational. The cognitive model is well represented by the “risk homeostasis theory”, promoted by Wilde (1982). Mainly, this theory states that in any activity the individual tends to accept a certain level of risk regarding their safety and health, in exchange for certain benefits they hope to gain from that activity. Due to this fact, the perceived risk level is compared with the accepted risk threshold, aiming to maintain the distance between the two at a minimum. Thus, if the subjectively estimated risk is smaller than the accepted risk threshold, individuals tend to increase the accepted risk level. Conversely, when the perceived risk exceeds the level of the accepted risk, a more cautious behavior is adopted (Wilde, 1994, p.7). Simonet & Wilde (1997) found experimental evidence in the laboratory experiment based on a computer program to support the “risk homeostasis theory”.

The motivational model based on McClelland’s research (1953, 1985) and on that of his collaborators regarding self-actualization needs refers to each individual’s attempt to achieve the highest level possible of performance. A high level of achievement need has been associated with a tendency towards a moderate risk level and with perseverance in a chosen behavior. From this, Atkinson (1957, cf. Potkay & Allen, 1996, p.352) derived the well-known model of risk preference, according to which people are motivated simultaneously both by the attraction to success, and by fear of failure. The result of these two opposite tendencies is directly materialized in the accepted risk level. The more intense either motivation is compared to the other, the greater the accepted risk level.

Our study follows the path of Atkinson’s model . We predicted that the potential loss significance (intensity of the threat) is always reflected mainly in the emotional area of personality. In addition to that, constancy of the emotional pattern of the human conduct makes this an extremely useful target for psychological assessment.

In spite of the popularity of the Atkinson’s model, and of numerous studies generated by it, the concept of risk behavior depiction retained a narrow attention, in which the risk preference and risk avoidance are perceived as extreme values on a unique scale. We argued that there are two opposite tendencies, relatively independent but complementary to each other, that describe the risk attitude as a result of risk preference (the need for success, actualization) and of the caution preference (failure avoidance). So, we aimed to discover a three way indicator made up of specific indicators of the risk and prudence preference, and a final indicator derived from the difference between these opposite tendencies, thus building on the Atkinson’s model of risk preference.

Method and Sample

We chose the AC-REF Values Inventory conceived by Zörgö (1976) to be the basis of our measure. Within this measure the AC-REF allows participants to answer personally on a scale on a 1 to 100, which measured the acceptance or refusal of 272 items (words or expressions) referring to facts, objects or usual every day situations. Applied within of PSYCOMP computerized system (Popa, Popescu, 1994) and interpreted according to our special designed method (“in-subject standardization”), AC-REF proved to have a remarkable psychodiagnostical potential as far as professional adjustment for aviators is concerned (Popa, Draghici, 1994). In our opinion the main advantage of this instrument is its semi-structured nature, which allows the subject to express the intensity of his\her attitude without feeling constrained by the item structure, or by the rigid frame of a forced responses panel. When carrying out a lab examination with participants subjected to an assessment procedure with decisional purpose, this method of investigation may overcome the drawbacks specific to that of questionnaires (Friedenberg, 1995, Anastasi, Urbina, 1997).

The study followed two steps: (1) First, we asked a sample of “experts” (N=81) to assess the risk significance (as danger for health or for physical integrity), giving grades from 1 to 10 to all of the 272 AC-REF items. After this, we ordered the item list according to the median value of the grades given by the “experts” . Thus we conceived a list of 43 risk significance items and another list with 36 prudence significance items. (2) Secondly, a sample of 1168 aviation subjects were assessed with the AC-REF. For each subject, the risk and prudence preference indicator was calculated as the sum of row scores with the median value of each item. In order to emphasize more obviously the individual significance, they have been transformed into standardized scores (T scores). Finally, the compound indicator of the risk-prudence balance has been computed as the difference between the two primary indicators.

Results and Discussion

Below are the most significant results that emerge from the analysis of the three indicators. First we shall analyze their psychometrical features in order to see if they have the formal features required by the psychological diagnosis. Secondly, we shall assess the sensitivity of these indicators with assessments of certain personality traits and biographical data (as age and gender) which naturally show variations associated to the attitude towards risk.

•  Psychometric features

All the three computed indicators show adequate psychometric features, the distribution being noteworthy, as it is close to the normal curve.

Table 1: Correlations within the three risk attitude indicators (N=1168)
Risk preference Prudence preference Risk\Prudence balance
Risk preference 1 -0.374 ** +0.829**
Prudence preference 1 -0.829 **
Risk\Prudence balance 1
* p<0.05, ** p<0.01

Among the correlation of the three indicators shown in the Table 1, we find particularly significant the one between Risk and Prudence preference. The significant value still moderate as far as intensity is concerned, between the risk preference and prudence preference, is able to confirm our hypothesis, which states that, even related, this indicators have also an important degree of independence.

•  External construct validity

The “diagnostic”, or in other words, “construct” validity, refers to the extent to which the three indicators we have computed may be said to actually measure a construct related to risk attitude behavior. To do this, the risk attitude indicators were tested against personality traits and some biographical data, which have a proven to be in relation with risk behavior in previous research.

2.1) Risk attitude indicators (RAI) and original AC-REF indicators

In the first instance the relation between RAI and some genuine AC-REF indicators was questioned. From a relatively long list we have chose only five indicators related to psychosomatic type and social or performance behavior.

Table 2: Correlation between risk attitude indicators and AC-REF indicators (N=1168)
Risk preference Prudence preference Risk\Prudence balance
SOCIALNESS - 0.322** -0.231**
PERFORMANCE 0.357** - 0.208**
SOMATOTONIC 0.412** - 0.239**
VISCEROTONIC - 0.233** -0.180**
CEREBROTONIC -0.153** 0.632** -0.472**
* p<0.05; ** p<0.01 (only the significant correlations are shown)

As it can be seen there are some consistent correlations between the RAI and AC-REF indicators. The preference for risk is positively related to the performance orientation and somatotype indicator, but negatively with cerebrotype. Also, the prudence preference shows significant relation with sociability, viscerotype, and especially with cerebrotype. As far the risk\prudence balance is concerned, we can see a complete list of significant correlations. The degree of specificity of each indicator must be underlined as revealed by “independent” relation with each construct criterion.

2.2) RAI and personality traits

When referring to the interaction between risk attitude and personality traits we may discuss two different trends. From the cognitive point of view, Wilde (cit. before, pp.180) describes a personality type oriented towards excessive risk taking. Its main features are: optimistic face, non-realistic assessment of his chances, small imagination, apparent preoccupation with physical threats, reduced awareness of personal motivations, “cyborg complex” (identifying oneself with the tool or machine that sustains the self-assertion), and belief in one’s “lucky star”. We often find examples of such personality type among the aviators or people addicted to extreme sports.

In a study regarding the personality and attitude involvement in traffic accident risk, West and Hall, (1997) concluded that the tendency towards norm and rule breaking is significantly associated with taking excessive risks. We can agree with Vasilescu (1995) who states that the “purely rational” decider is only fiction, the risk conduct being inevitably influenced by emotional and motivational variables, in their turn associated to personality constants.

Based on these assumptions we started to investigate the relation between the three modes of risk attitude and the results from other assessment carried out by using some personality inventory ordinarily used in aviation personnel assessment: Armstrong Laboratory Aviation Personnel Survey (ALAPS), Eysenck Personality Inventory (EPI-form A), and Cattell 16PF.

2.2.1) RAI and ALAPS

ALAPS is a 240 item questionnaire, recently developed by Retzlaff, King, McGlohn and Callister (1996), and especially designed for aviation personnel assessment. ALAPS has 15 scales divided into three major categories: personality scales (confidence, sociality, aggressiveness, orderliness, negativity), psychopathology scales (emotional instability, anxiety, depression, alcohol abuse), and crew interaction scales (dogmatism, deference, team oriented, organization, impulsivity, risk taking).

ALAPS was adapted for the Romanian population and experimentally run into PSYCOMP system on 53 young pilot students who had a previous AC-REF assessment. Table 3 provides all the obtained correlations between RAI and ALAPS scales.

Table 3: Correlations between RAI and ALAPS scales (N=53)

Risk preference Prudence preference Risk\Prud. Balance
Personality Confidence 0,228 -0,108 0,212




Aggressiveness 0,352** -0,162 0,325**
Orderliness -0,243 -0,073 -0,113
Negativity 0,060 -0,035 0,060
Psychopathology Emotional instability -0,188 -0,067 -0,082
Anxiety -0,082 0,097 -0,111
Depression -0,051 0,062 -0,070
Alcohol abuse -0,101 -0,074 -0,021
Crew Interaction Dogmatism 0,177 -0,048 0,143
Deference 0,041 0,042 0,001
Team oriented -0,427** 0,166 -0,376**
Organization -0,043 -0,123 0,045
Impulsivity 0,167 -0,142 0,193
Risk taking 0,478** -0,410** 0,554**
* p<0.05; ** p<0.01

As it can be seen, there are some significant correlations between RAI and ALAPS scales. First of all, the consistent correlations with “risk taking” of all risk attitude modes must be highlighted, including prudence preference, risk preference, and, especially, risk\prudence balance. This supports the prediction that our RAI accurately measures risk attitude. Based on the above correlations, it can be assumed that risk attitude is characterized by a combination of aggressiveness and self oriented attitude.

2.2.2) RAI and EPI (A) scales

The table below shows the correlations between risk attitude modes and the three traits measured by EPI: extraversion, neuroticism, and instability .

Table 4: Correlations between RAI and EPI-A (N=366)
Risk preference Prudence preference Risk\Prud. Balance
Extraversion (E) 0,126* -0,215** 0,204**
Neuroticism (N) -0,221** 0,166** -0,231**
Instability (E+N) -0,120* - -
* p <0.05; ** p<0.01 (only the significant correlations are shown)

The output values show moderate but significant correlations between RAI and all EPI scales. Risk preference seems to be related with extraversion (r=,126) but less intensely than prudence preference is related with introversion (r=-,215). Also, risk preference attitude is related with neuroticism tendency less than prudence preference is related with “emotional stability”. Risk\Prudence balance correlates in a positive manner with extraversion and in a negative one with neuroticism.

2.2.3) Correlations between RAI and Cattell 16PF factors

The Cattell 16PF is one of the most frequently used questionnaires in the aviation community. For that reason, we evaluated the RAI against this classical instrument for personality assessment.

Table 5: Correlations between RAI and Cattell 16PF factors (N=471)
Risk preference Prudence preference Risk\Prud. balance
(A) outgoing vs. reserved -0,049 -0,097 0,031
(C) emotional stability vs. instability 0,079 0,018 0,036
(E) dominance vs. submissiveness 0,157** -0,112* 0,166**
(F) optimistic vs. pessimistic 0,082 -0,163** 0,152**
(G) strengthens vs. weakness superego -0,219** 0,079 -0,183**
(H) adventurous vs. timid 0,045 -0,089 0,084
(I) tender- vs. tough-minded -0,187** 0,408** -0,371**
(L) suspicious vs. trusting 0,085 -0,046 0,081
(M) imaginative vs. practical -0,034 0,153** -0,117*
(N) shrewdness vs. unpretentious -0,048 0,121** -0,106*
(O) insecure vs. self-assured -0,064 -0,011 -0,032
(Q1) radicalism vs. conservatism 0,124** -0,076 0,123**
(Q2) self-sufficiency vs. group dependence -0,035 0,155** -0,119*
(Q3) controlled-casual -0,042 0,089 -0,081
(Q4) tensed vs. relaxed -0,004 0,015 -0,012
* p <0.05; ** p<0.01

Although the correlations are relatively small, many of them are significant. It must be underlined again the relative independence between RAI and criterion indicators. And, especially, the fact that risk\prudence balance indicator has more correlations with criteria than the primary indicators. Overall, it must be observed that all the risk attitude modes are related with some of Cattell 16PF factors in a consistent way with other research regarding risk behavior.

The results presented allow us to conclude that:

( i ) RAI show a significant relation with a wide range a of personality traits assessed with laboratory instruments,

( ii ) there is a certain particularity of each risk attitude mode indicator, suggested by the correlation independence with personality criterions. This fact sustains the hypothesis that risk attitude must be assessed on a multidimensional scale instead of on a unidimensional one.

2.3) Risk attitude mode indicators and biodata (gender and age)

The existence of a difference between women and men is constantly revealed in the risk behavior researches (Goma-Freixanet, 1997, West &Hall, 1997 ). Generally, women are believed to have a greater predisposition to accept rules (norms) and, thus, a smaller involvement in risky situations. That is why we found necessary to emphasize the variation of the RAI according to gender variable.

Table 6: ANOVA statistic for RAI and gender
Risk preference Prudence preference Risk\Prud. balance
Males Mean(m) 52,86 47,34 53,32
N 546 546 546
Females Mean(f) 43,47 57,70 41,40
N 108 108 108
Mean(m)-Mean(f) 9,39 -10,36 12,02
t=80,614 t=99,598 t=133,810
Sig.=0.000 Sig.=0.000 Sig.=0.000

As it can be seen in the table above, the risk attitude mode indicators have a significant variation among gender criterion. It also must be underlined that the largest difference between males and females is the risk\prudence balance indicator.

Another biographical variable traditionally associated with risk attitude is age. In aviation we talk about the “conservatism” progressively displayed by the pilots as they are aging, which results in less frequent involvement in high-risk situations. The scientists from US Naval Safety Center (Eyraud & Borowski, 1985) carried out research regarding the relation between flight accidents and age of pilots by analyzing the flight accidents during 1977-1982. The results concluded that there is an association between the age of the pilots and the type of accident they are involved in. Along with other relevant aspects, they emphasized a tendency to exaggeratedly take risks in pilots under 30-33, also associated with a more “relaxed” attitude towards the regulation constraints.

In order to emphasize the “sensitivity” of the risk attitude mode indicators with regard to age, we computed the ANOVA statistic along 10-years of age intervals .

Table 7: ANOVA statistic for RAI and age (male subjects)
Risk preference Prudence preference Risk\Prud. balance
-20 Mean 53,4821 46,6148 54,1444
N 428 428 428
21-30 Mean 50,4794 49,5906 50,5364
N 249 249 249
31-40 Mean 47,7592 51,3296 47,8452
N 230 230 230
41-50 Mean 43,7897 53,4658 44,1605
N 43 43 43
F=29,199 F=18,402 F=36,113
Sig.=0.000 Sig.=0.000 Sig.=0.000

As it can be observed, there is an obvious variation of RAI related to age, showing a systematic decrement of risk preference and increment of prudence preference as far as age is growing up.


Our study has to answer two main questions: (1) Are the three risk attitude mode indicators specific enough to justify their use within a unique assessment model of personal behavior in risky situations? (2) Is there a significant variation they would register according to individual variables whose relation with the risk behavior has already been proven?

The results of our study allow us to give a positive answer to both these questions. Firstly, independent associations with different personality variables were found. Secondly, RAI were found to have significant variation with regard to gender and age, which are well known for their relations with risk attitude and behavior.

Therefore, we believe this model of output interpretation with the AC-REF Values Inventory may be included in the psychological assessment process. Moreover, we also believe that other types of indicators must be developed in future via the same procedure.


1 The AC-REF inventory consists of 272 common words or expressions, of which refusal or acceptance must be evaluated with a number between 0 to 100.
2 In fact, this paper focuses on just one of the parts of a risk behavior assessment system, which will include three specific tests based on different theoretical backgrounds.
3 It must be said that the “expert” subjects did not take part in the following step of the study, as they where explicitly motivated as experts, in order to avoid the distortion by way of defense mechanisms.
4 We have called “instability” the computed scale (E+N), used by Eysenck himself in “crime proneness” scale in judiciary studies.
5 It must be stipulated that the results are quite similar for 5-year of age intervals.


Anastasi., A., Urbina, S., 1997, Psychological testing (7th ed.), Prentice-Hall, Inc.
Bruno, F.J., 1993, Psychological symptoms, John Wiley&Sons
Ceauşu V., 1972, De la incertitudine la decizie, Editura militară, Bucureşti
Ceauşu V., 1976, Psihologia zborului, Editura militară, Bucureşti
Eyraud M., Borovski B., 1985, Age and Pilot Performance, Aviation Space & Envirnmental Medicine, 56
Friedenberg, L., 1995, Psychological testing; design, analysis, and use, Allyn&Bacon
Goma-Freixanet M. 1997, Personality and risk taking behaviour in women: Some preliminary results, ECP (abstract), http://allserv.rug.ac.be/~ivanmerv/ecp8250.html
McClelland, D.C., 1985, Human Motivation, Glenview, Ill; Scot, Foresman
McClelland, D.C., Atkinson, J.V., Clark, R.W., Lowell, E.L., (1953), The Achievement Motive, New York : Appleton-Century, Crofts
Popa M., Drăghici M., 1994, “Structuri de orientare valorică specifice la aviatori”, Comunicare la Conferinţa Naţională de Psihologie, Bucureşti, 27-29 mai
Popa M., Popescu C., 1994, “Laborator psihologic informatizat în regim L.A.N.”, Comunicare la Conferinţa Naţională de Psihologie, Bucureşti, 27-29 mai
Potkay Ch. R., Allen P.B., 1986, Personality: Theory, Research, and Applications, Brooks\Cole Publishing
Retzlaff, P.D., King, R.E., McGlohn, S., Callister, J.D., 1996, The development of the Armstrong Laboratory Aviation Personality Survey (ALAPS), Interim Technical Report, AL\AO-TR-1996-0108
Plous, S., The psychology of judgement and decision making, McGraw-Hill, Inc.
Simonet, S., Wilde G.J.S., 1997, “Risk: Perception, Acceptance, and Homeostazis”, Applied Psychology: An International Review, 46, 235-252
Vasilescu I.P., 1995, “Eficienţa deciziilor, flexibilitatea comportamentală şi stabilitatea internă: investigaţie computerizată”, Conexiuni, coord. S. Chelcea, L. Mitrănescu, Editura I.N.I., Bucureşti
West R., Hall J., 1997, “The Role of Personality and Attitudes in Traffic Accident Risk”, Applied Psychology: An International Review, 46, p.253-264
Wilde, G.J.S., 1982, “The theory of risk homeostazis: Implications for safety and health”, Risk analysis, 2, p.249-258
Wilde, G.J.S., 1994, Target Risk, www http://pavlov.psyc.queensu.ca/target/index.html
Zörgö, B., 1976, “Inventarul de valenţe AC-REF”, Îndrumător psihodiagnostic, vol 3, Cluj, Universitatea Cluj-Napoca