Toward a Vector Model of Intentions

Olli T. Ahtola, University of Florida
ABSTRACT - This paper takes a critical look at Fishbein's model of behavioral intentions. It argues that the model as conceptualized and operationalized by Fishbein and his associates contains some inconsistencies, unacceptable assumptions, and unexplained components. A new model is developed which proposes a solution to these weaknesses.
[ to cite ]:
Olli T. Ahtola (1976) ,"Toward a Vector Model of Intentions", in NA - Advances in Consumer Research Volume 03, eds. Beverlee B. Anderson, Cincinnati, OH : Association for Consumer Research, Pages: 481-484.

Advances in Consumer Research Volume 3, 1976      Pages 481-484

TOWARD A VECTOR MODEL OF INTENTIONS

Olli T. Ahtola, University of Florida

ABSTRACT -

This paper takes a critical look at Fishbein's model of behavioral intentions. It argues that the model as conceptualized and operationalized by Fishbein and his associates contains some inconsistencies, unacceptable assumptions, and unexplained components. A new model is developed which proposes a solution to these weaknesses.

FISHBEIN'S MODEL OF INTENTIONS

The so called "extended Fishbein model" was developed to predict and explain a specific behavioral act in a well defined situation. In a given situation, a person is assumed to form a specific behavioral intention which directly determines his subsequent overt behavior. According to the theory, there are two major factors that determine behavioral intentions: a personal or "attitudinal" factor and a social or "normative" factor. These two factors in the theory are given empirical weights. Symbolically, the central equation in the theory can be presented as follows (Fishbein and Ajzen, 1975, p. 301):

B - I = (AB) w1 + (SN) w2   (1)

where:

B = the behavior (in a specified situation)

I = the intention to perform behavior B

"B = the attitude toward performing behavior B

SN = the subjective norm, i.e., the influence of the social environment on behavior B

w1 and w2 are empirically determined weights

EQUATION   (2)

where:

"B = the attitude toward performing behavior B

bi = the strength of belief that performing behavior B leads to consequence or outcome i

ei = the evaluation of outcome i

n = number of slaient beliefs

Furthermore, according to the theory, the subjective norm (SN) is determined by the sum of the perceived expectations of specific referent individuals and/or groups weighted by the individual's "motivation to comply" with those expectations.

Algebraically (Fishbein and Ajzen, 1975, p. 302):

EQUATION   (3)

where:

SN = the subjective norm

bi = the normative belief (i.e., the person's belief that referent group or individual i thinks he should or should not perform behavior B)

mi = the motivation to comply with referent i

n = number of relevant referents

Empirical Weights

The behavioral intention is proposed to be a weighted sum of the two factors - "attitudinal" and "normative". Standardized multiple regression coefficients have been used to estimate the weights. Empirical evidence shows that the weights vary with the behavior, with the situation, and with individual differences. These weights are proposed to represent the individual's relative importance of the two factors (Fishbein and Ajzen, 1975). Why these importance weights have never been tried to measure directly and independently is not explained. Also, how is it possible to least squares estimate these weights if they vary with the behavior and with individual differences? If we want to least squares estimate the weights, at the minimum, there should be a theory which stipulates what kinds of individuals do not vary in the weights. Then the weights could be properly estimated for these individuals. Only constants (parameters), not variables, can be estimated through least squares fitting.

Furthermore, it is a well-known fact that if the predictor variables are correlated the precision of estimation falls so that it becomes very difficult, if not impossible, to distangle the relative influences of the predictor variables. Also, estimates of coefficients become very sensitive to particular sets of sample data, and the addition of a few more observations can sometimes produce drastic shifts in the coefficient. Empirical evidence shows that AB and SN (or Sbimi) tend to be correlated and, consequently, the above problems are likely. Given the nature of the model and the assumptions in multiple regression, it seems that the regression analysis is quite improper method to estimate the weights, especially when our purpose is to explain the intentions and not just to predict them, i.e., we are interested in the values of the weights and our sole purpose is not just to maximize R2.

Personal Attitude

The first factor, i.e., the "attitudinal" factor, is a conceptually clear measure of how much the individual likes or dislikes to perform the behavior. Often, however, this factor has been measured in terms of the individual's attitude toward the behavior (e.g., Ajzen and Fishbein, 1969) and not in terms of the individual's attitude toward he himself performing the behavior. The latter, of course, is the conceptually correct procedure. Fishbein makes a distinction between attitude toward an object and attitude toward a particular behavior (with respect to the object). It would seem logical that he made a distinction also between attitude toward the behavior in general, and attitude toward a particular actor (like the subject himself) Performing the behavior.

Even though AB variable in itself is conceptually clear and easy to operationalize, the model of its determinants (2) has been argued to have several problems (Ahtola, 1973, 1975). However, because these problems have been discussed in detail elsewhere (Ahtola, 1973, 1975) they will not be discussed in this paper.

Normative Influence

"The second or normative component of the theory, SN, deals with the influence of the social environment on behavior. The subjective norm is the person's perception that most people who are important to him think he should or should not perform the behavior in question.'' (Fishbein and Ajzen, 1975, p. 302)

The above statement indicates that Fishbein believes that there exists in an individual's cognitive structure a unidimensional concept, subjective norm, which is structurally similar to attitude. It should be noted that this unidimensional variable was not proposed in his earlier writings (e.g., Fishbein, 1972). In his earlier writings Fishbein used normative beliefs and motivation to comply as first order determinants of intention without any intermediary variable, such as the subjective norm.

The question arises whether people really perceive any collective "important others" in normal circumstances. Important others are usually such things as spouse, father, mother, employer, friends, etc. It seems obvious that in many cases, perhaps in most cases, the opinions of these referent individuals or groups conflict or at least vary in magnitude. If this is the case, do referents, such as spouse, employer, and friends, get cognitively combined into a concept "important others"? It would Seem plausible that several different persons in a group, such as a social club, where members hold relatively homogenous norms about expected behavior of their members, will be cognitively combined into a collective group, such as "the members of my Rotary Club". But it seems very unlikely that the wife's opinion would be cognitively combined with the club members' opinion to form a generalized opinion of "important others". It is argued here that very seldom does there exist such a thing as generalized subjective norm in people's cognitive structure, except perhaps when the individual is contemplating some abnormal or asocial behavior which is uniformly sanctioned by the whole society.

Furthermore, if Fishbein's model of the determinants of SN (3) is correct it seems inconsistent to conceptualize (and operationalize) the SN as "the person's perception that most people who are important to him think he should or should not perform the behavior in question" (Fishbein and Ajzen, 1975, p. 302). This arises from the fact that the motivation to comply can be negative as well as positive, i.e., the individual may be motivated not to do what the referent wants. Let's take a hypothetical example in which each relevant referent of a person thinks he should perform the behavior but he is strongly motivated not to comply with their wishes. Under those circumstances Sbimi gives a highly negative score. If, however, the SN is measured directly as suggested (Fishbein and Ajzen, 1975, p. 314), i.e., in terms of whether "most people who are important to me think I should --- I should not perform behavior X", one definitely would expect a highly positive response. These are totally opposite results. This inconsistency results from the fact that the "subjective norm" measure does not seem to incorporate the "motivation to comply" influence. The SN measure is probably some kind of average bi score over the relevant referents, which, as indicated earlier, most likely has very little, if any, cognitive meaning to the individual.

Turning to the determinants of the social norm, one finds additional problems. These problems, however, are such that they do not conflict with the above argument about the questionable conceptual meaning of the social norm, i.e., while evaluating the determinants of SN it will not be assumed that SN is correctly conceptualized. In other words, there are internal problems in this submodel.

The normative belief, bi, is conceptualized as the person's belief that a relevant reference group or individual thinks he should or should not perform the behavior. There are clearly two conceptually independent variables in this conceptualization. One is the strength of the subject's belief and the other is the direction and polarity of the referent's opinion. Fishbein, however, measures bi unidimensionally. Perhaps the operationalization of bi indicates which variable is the one conceptualized relevant to the model.

At least three different operationalizations can be found in the literature. These are:

1. (Ajzen and Fishbein, 1969)

Referent X thinks I should perform the behavior

probable_____:_____:_____:_____:_____:_____:_____:_____:improbable

2. (Fishbein and Ajzen, 1975)

Referent X thinks

I should_____:_____:_____:_____:_____:_____:_____:_____:should not perform the behavior

3. (Ajzen and Fishbein,1969)

Referent X expects me to perform the behavior

probable_____:_____:_____:_____:_____:_____:_____:_____:improbable

The above three measures are not conceptually parallel. The first scale measures the strength of the subject's belief. If the subject marks anything but the extreme left category, he is indicating that there is a possibility that the referent is indifferent or against the behavior. The second scale, on the other hand, (especially if the standard Semantic Differentials Scale instructions are given to the subject) seems to ask the subject's perception of the content (direction and polarity) of the referent's opinion, i.e., how much the referent is for or against the behavior. One must draw conclusion that the model is not conceptualized to discriminate between the perceived polarity (i.e., the content) and the uncertainty (i.e., the strength) of the belief. These two aspects of the normative belief are logically different and it seems very plausible to argue that they are also "psycho-logically" different. This is because they need not even co-vary. For exam-le, it is quite conceivable that a person may think that referent X is most likely very much against the behavior but the person is somewhat uncertain about it. At the same time, the person may think that referent Y is slightly unfavorable and the person is certain about it (e.g., because the referent Y has told him what he feels, while the person may be just guessing the referent X's opinion).

It seems very likely that both the uncertainty about the belief and the content of the belief (especially the polarity) affect behavioral intentions, and consequently both of them should be incorporated in the model.

The problem with the third scale is that it may not measure at all what the referent is perceived to think the subject should do. It may measure what the subject thinks the referent thinks he will do. This is the basic denotative meaning of the scale. The subject may know that his wife expects him to get drunk next Saturday night because that is what he always does. She still probably does not think that he should get drunk.

The second component in the normative influence sub-model, i.e., the "motivation to comply", is suggested to be measured in terms of how much the subject, in general, wants or wants not to do what the referent thinks he should do. There seems to be considerable uncertainty about the exact meaning of this component. Should it be independent of the referent's particular demands, or should it be specific to the particular behavior or behavioral domain under consideration, or should it be defined as the subject's general motivation to comply with the referent? The last conceptualization is advocated by Fishbein (Fishbein and Ajzen, 1975). The theoretical grounds for this choice are not referred to, but the guess of this author is that the grounds are to make the "motivation to comply" component independent of the "normative belief" component. This would be very convenient for predictive purposes; but is it conceptually sound? The subject might want to comply with his wife's strong wishes but could care less about her less polarized wishes. On the other hand, the subject may form a reactance to his father's strong demands, so that he does not comply if he feels high pressure, while if the pressure is lower he is happy to comply. What is argued here is that the "motivation to comply" is not independent of the content of the normative belief.

French and Raven (1959) and Kelman (1961) have convincingly argued that social influence of a referent increases with that referent's power to reward or punish the person (compliance), with the person's liking of, or desire to, identify with the referent (identification), with the referent's perceived expertise (internalization), and with the extent to which it is legitimate for the referent to make demands of the person. If the purpose of the motivation of comply component is to tap all the above influences, its operationalization as a unidimensional single measure seems utterly optimistic. As mentioned earlier, Fishbein suggests that the component be operationalized in terms of how much the subject wants or wants not to go along with the referent's opinion. This scale probably measures the identification influence quite well but it may not get at the compliance influence ("I don't want to do what my father thinks I should do, but I am still going to do as he wants because otherwise he is going to spank me"), or the expertise influence ("I really dislike my doctor and his advice, but I still do what he tells me to do because he is a real expert") or legitimate right influence ("I really dislike to send money to my no-good son at the college, but I'll do it anyway because I got money from my father and that is a tradition and responsibility in our family").

TOWARD A VECTOR MODEL OF INTENTIONS

The purpose here is to suggest what properties and what kinds of functional relationships the model of behavioral intentions should possess if the above criticism of the Fishbein model is accepted.

To understand and predict a person's intention to perform a specified act, we must know how much he would like to perform the act, i.e., what his personal attitude toward his performing of the act is. However, before deciding whether to perform the act or not he may want to consider what others would like him to do. That is, we must know what he perceives each of the relevant referents' attitude toward him performing the act to be. In other words, the subject's perception as to how much each referent wants him or wants him not to perform the act. In many cases, there may be uncertainty in the subject's mind about the polarity or perhaps even the direction of some of the referents' opinions.

Properties of Perception and Belief

People seldom perceive things along continuous dimensions but instead locate them in terms of categories on these dimensions. For example, some attitudes may be perceived "very favorable" while some others only "fairly favorable" or "slightly favorable". That is, the attitude dimension has been partitioned into categories, and the attitudes are described in terms of these categories, not in terms of some point on a continuous dimension. The number of categories perceived or utilized seems to differ somewhat among individuals (Osgood, Suci, and Tannenbaum, 1957). Seven categories have been found to be very common. Let's assume that the subject is using seven attitude categories, which he has labeled "very favorable", "fairly favorable", "slightly favorable'', "indifferent", "slightly unfavorable", "fairly unfavorable", and "very unfavorable".

Each category is a concept which together form the population of attitude concepts which the subject may associate with the referent (i.e., perceive the referent to have). Each referent elicits one or more of these attitude concepts (i.e., attitude categories). It is quite conceivable that under certain circumstances, the same referent elicits two or more of these attitude categories. It seems reasonable to assume that the most likely attitude category to the referent is elicited first (Fishbein, 1967). Then the other possible categories to the referent are elicited in order of their likelihoods. This takes place when the subject is not sure what the referent's attitude is.

For example, the subject has the belief that his mother most likely is "very unfavorable" toward him performing the act, but he is not quite sure and he thinks that his mother may be "fairly unfavorable" or perhaps only "slightly unfavorable". He knows, that his mother is not neutral or favorable.

In summary, each belief statement has the content and the strength. The content in the above example is the direction and the polarity of the referent's perceived attitude and the strength is the likelihood or probability the subject assigns to the referent having the attitude with that content.

Motivation To Be Influenced

To understand and predict the effect of the perceived opinion of a referent we must know what the subject's motivation to be influenced by the referent's attitude is.

The subject may, and most likely does, have a different motivation to be influenced to differently polarized attitudes of the referent. The motivation may or may not be monotonically related to the polarity of the referent's perceived attitude. The subject may be motivated not to be influenced if the referent's attitude is very polarized (reactance), while he may he influenced positively by the less polarized wishes. The subject, at the same time, may have zero motivation to another referent's mild attitudes, while, when the attitudes are perceived to be polarized, he is motivated to go along. Father and wife of the subject might be such referents. The subject may be irritated by his father's extreme wishes and shows his independence by not complying, while he complies with his father's less extreme wishes because he accepts them as legitimate. On the other hand, the subject may not care at all about his wife's small wishes and takes his wife's wishes into consideration only when he perceives them to be really strong.

In operationalizing the motivation to be influenced, at least two scales seem necessary. These are the measures of 1. how much the subject wants or wants not to do as the referent wants him to do given the polarization of the referent's attitude, and 2. how much the subject feels he should or should not do as the referent wants him to do given the polarization of the referent's attitude. These two influences should be summative.

Influence of a Referent

The motivations to be influenced are viewed as summa-rive. Furthermore, the amount of behavioral response tendency available for summation is an absolute function of the strength of belief. It seems reasonable to argue that if the subject is sure what the referent's wish is, all the corresponding motivation is available, i.e., the corresponding motivation to be influenced has the weight 1. On the other hand, if the subject perceives no chance that a given attitude category represents the referent's opinion, none of the corresponding motivation to be influenced is available, i.e., the weight is 0. Various levels of uncertainties have weights between 0 and 1. The more likely the category for the referent the higher the weight. Because every referent must have some attitude (note, indifference is an attitude) the sum of the subjective probabilities assigned to different attitude categories for each referent should add up to 1.

Algebraically, this may be expressed as follows:

EQUATION   (4)

where:

Mr = motivation to be influenced by referent r (with respect to behavior B)

brj = the strength of belief that referent r has attitude j

mrj = motivation to be influenced by referent r's attitude j

g = number of attitude categories

or in vector notations:

Mr = br'mr   (5)

where:

br' = row vector of probabilities of referent r's association with the attitude categories (as perceived by the subject)

mr = column vector of the subject's motivation to be influenced by referent r's attitude categories

Intention

The influences of the various relevant referents are viewed as summative. The influence of the personal attitude and the influences of the various relevant referents are viewed as summative. Because the personal attitude and the influences of the referents are measured along somewhat different dimensions and using different scales, a scale conversion constant (which does not have any cognitive meaning) is needed.

In summary, the proposed model can be represented symbolically as follows:

EQUATION   (6)

where:

IB = an individual's intention to perform behavior B

"B = his personal attitude toward performing B

B = scale conversion constant

n = number of relevant referents

br' and mr are the same as in (5)

REFERENCES

Olli T. Ahtola, "An Investigation of Cognitive Structure within Expectancy-Value Response Models," Unpublished doctoral dissertation (University of Illinois, 1973).

Olli T. Ahtola, "The Vector Model of Preferences: An Alternative to the Fishbein Model," Journal of Marketing Research, 12 (February, 1975), 52-59.

Icek Ajzen and Martin Fishbein, "The Prediction of Behavioral Intentions in a Choice Situation," Journal of Experimental Social Psychology, 5 (1969), 400-416.

Martin Fishbein, "A Behavior Theory Approach to the Relations between Beliefs about an Object and the Attitude toward the Object," In M. Fishbein (ed.). Readings in Attitude Theory and Measurement (New York: Wiley, 1967), pp. 389-400.

Martin Fishbein, "The Search for Attitudinal-Behavioral Consistency," In J. Cohen (ed.). Behavioral Science Foundations of Consumer Behavior (New York: The Free Press, 1972), pp. 245-252.

Martin Fishbein and Icek Ajzen, Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research (Reading: Addison-Wesley, 1975).

John P. R. French, Jr. and Bertram H. Raven, "The Bases of Social Power," In D. Cartwright (ed.). Studies in Social Power (Ann Arbor: University of Michigan Press, 1959), pp. 150-167.

Herbert C. Kelman, "Processes of Opinion Change," Public Opinion Quarterly, 25 (1961), 57-78.

Charles E. Osgood, George J. Suci, and Percy H. Tannenbaum, The Measurement of Meaning (Urbana: University of Illinois Press, 1957).

---------------------------------------