Resumé

Political socialization research so far has focused primarily on the direction of attitudes among children and youngsters. The preconditions for the development of political attitudes among these age groups have been neglected. In theoretical discussions cognitive development has been offered as a major prerequisite. In the article it is posited that political involvement is a second major prerequisite.

The effects of political involvement are examined with respect to four aspects
of political attitudes among Danish school children: the ability to express attitudes,
the correlation between indicators of a particular attitude, the stability of
attitudes, and the correlation between different attitudes. Except for the stability
of attitudes, results are clear. Political involvement is related to the formation of
attitudes. The effect of political involvement is stronger than the effect of any of
the traditional socio-economic variables ordinarily considered in socialization research.

The proclaimed ambition of political socialization research has been to study political attitudes in a developmental perspective. Present attitudes are seen as a product of all the stimuli that a. person has experienced throughout his whole life. This perspective differentiates socialization research from the neighbouring fields of public opinion and electoral behaviour, where attitudes are primarily examined in terms of stimuli active at the time of observation.

Socialization research so far has focused on the attitudes of children and adolescents. The literature is rich in studies which deal with the effects of socio-economic variables like sex, age, education, and social status, and with the effects of the various agents of socialization, notably the family and the school.

One might have expected the study of pre-adult attitudes to deal primarily with the effects of the various factors on the development of attitudes. Actual studies, however, have overwhelmingly focused on the direction of attitudes. The assumption has been that children and adolescents

do in fact possess political attitudes. Thus, little attention has been given to the preconditions of attitudes. Instead researchers have hastened to examine the direction of attitudes. The uncritical examination of the direction of attitudes is also quite surprising considering the ample evidence of non-attitudes among adults. In this article we shall examine political involvement as a factor related to the development of political attitudes.¹

1. The Hypotheses

Lack of empirical research concerning the development of attitudes has not been an obstacle to speculation on this subject, resulting in a number of 'models' of the socialization process. The two most important models are the accumulation model and the model of cognitive development.

According to the accumulation model, children and adolescents acquire attitudes and identifications as the outcome of a learning process. The child is gradually filled with knowledge, information, and opinions. Political attitudes accumulate solely due to influences from the social environment. The model explicitly disregards the characteristics of the socialized person. Individual characteristics neither limit nor facilitate the learning process. Both the level of cognitive development and other psychological charateristics are disregarded (Hess & Torney 1967, pp. 19-22).

Compared to the accumulation model the model of cognitive development represents a higher level of sophistication, as it identifies important preconditions for the development of attitudes. The cognitive development model as presented by Merelman (1971) posits that the development of political attitudes in children and adolescents is contingent upon the level of cognitive development. The development of political attitudes therefore in gross terms should occur in the same sequence for everybody. Age as an indicator of cognitive development becomes the central explanatory variable.

The strength of the cognitive development model is its identification of cognitive limits to the development of political attitudes. The political thinking of a person, of course, cannot exceed his general cognitive potential. The model, however, has little to say about factors determining the degree to which the cognitive potential of the child or adolescent will actually be realized with respect to political thinking. Its heavy reliance on age as an explanatory variable suggests an implicit assumption that children and adolescents will develop political attitudes corresponding to their level of cognitive development.

This is obviously a tenacious assumption, as it runs counter to the many studies of adult opinion and behaviour which have repeatedly demonstrated that politics is a very peripheral phenomenon to most people and that people do not develop rooted attitudes toward objects which they do not care about.

Empirical research on public opinion and electoral behaviour have demonstrated that only a slight minority of the grown up population do have rooted political attitudes. The overall impression is the same when one looks at the 'number of don't knows', the correlation between issue items, the correlation between party identification and issue items, or the stability of opinions. Only a minority has rooted attitudes (see e.g. Campbell et al. 1960; Converse 1964a, 1964b, 1970).

There has been a lot of discussion about the factors accounting for the low incidence of attitudes among mass publics. The discussion has centred on the relative importance of education and political involvement. Research carried out in USA in the aftermath of the upheavals of the late 1960's points heavily to political involvement as the key factor (Nie with Anderson 1974; Pomper 1972; Aberbach & Walker 1970). The higher the political involvement, the higher the tendency to develop political attitudes.

Against this background cognitive development cannot alone be considered sufficient to explain the development of political attitudes. We shall hypothesize that political involvement is a most important factor explaining the occurrence of political attitudes.

2. The Dimension of Political Involvement

Political involvement is a psychological variable. It refers to the degree to which a person is interested in politics, concerned about politics or dealing with political phenomena. Campbell et al. (1964, p. 101) have talked about 'the individual's psychological involvement in polities', Verba & Nie (1972, p. 83) about 'the general psychological involvement in polities'. Even if one can speak about 'general psychological involvement in politics' , it is well known that people are not necessarily involved in the same aspects of political life. Some people care about international politics, others care about various aspects of domestic politics. Nevertheless the aim of this section is - if possible - to identify a more general dimension of political involvement and to develop an empirical measure of this dimension.

Conceptually political involvement is closely related to saliency and

centrality. If political objects or political attitudes are salient or central to a person, the person undoubtedly must also be characterized as politically involved.² The close relationship between political involvement and the two other concepts is also revealed by the fact that all have been operationalized in more or less identical ways.

Political involvement has been operationalized in terms of 1) related psychological variables like interest or sense of political efficacy (Campbell et al. 1960, pp. 101-02; Nie & Verba 1972, pp. 367-69; Martinussen 1971, pp. 154-59); 2) activity variables like discussion of politics or use of mass media (Nie & Verba 1972, pp. 367-69); 3) political information (Martinussen 1971, pp. 154-59); and 4) the ability to express opinions, i.e. by answering a questionnaire (Martinussen 1971, pp. 154-59).³

As it is our intention to examine the relationship between political involvement and the development of political attitudes, it is of course impossible to use the ability to express opinions as an indicator of political involvement. The use of this indicator would introduce a tautological •element in the empirical analysis. To obtain the most valid operationalization we have looked for indicators from the three remaining categories. A special problem arieses with the activity variables. Very few opportunities for overt political activity are open to children and young adolescents. They can talk and read about politics, and that is about all they can do. For this reason our questionnaire also contained questions concerning the disposition to future political activity, like joining a political party when old enough. Some of these questions are used because of lack of indicators of real political activity.

We end up with 11 indicators for the measurement of political involvement. First, there is the ordinary question concerning 'interest in politics'. Closely related is a question about the inclination for the subjects of history and civics in school. Next, there are two questions dealing with discussion of politics with parents and friends. There are two questions concerning attention to politics in the mass media, like newspapers and TV. There are four questions related to future political activity: voting, joining political party, joining political youth organization and taking pait in a demonstration. Finally, a measure of political information is taken into consideration.⁴ The 11 indicators have been selected because they are assumed to cover important aspects of the dimension of political involvement. But the question still arises of whether one can meaningfully speak of one general dimension of political involvement or whether we just have indicators of a number of vaguely related aspects of involvement. To answer this question a factor analysis was performed. The result is given in

Table 1. The figures given are the factor-loadings, indicating the correlation between the various indicators and the factors which resulted from the analysis. Three factors resulted from the analysis. Factor 1 can be interpreted as the general dimension of political involvement. Factor 2 is related to membership of political parties. Factor 3 concerns the use of mass media.

Political interest has the highest loading on factor 1. Discussing politics with parents or friends follows next. The loadings for information, use of media, and intention to vote are somewhat lower, but still relatively high. The indicators mentioned cover various important aspects of the phenomenon we have called political involvement. The interpretation of factor 1 as a factor for general political involvement is based on the high loadings of these indicators. Factor 2 is primarily defined by membership of political parties and their youth organizations. The other indicators of future political activity only have small loadings on factor 2, and the two defining indicators only have small loadings on Factor 1. Accordingly, we interpret Factor 2 as a rather narrow factor for party membership. Factor 3 also has a rather narrow reference. It is mainly defined by the two indicators which have to do with the use of mass media. As already mentioned, the mass media indicators also load heavily on Factor 1, and thus support our interpretation of Factor 1 as representing the general dimension of political involvement. The operationalization of political involvement which will be used here onwards is a reconstruction of Factor 1. Technically the reconstruction has been performed by using factor-score coefficients. In this way indicators with relatively high loading contribute

more to the constructed variable than indicators with relatively low loadings
5

3. Political Involvement and the Development of Political Attitudes

The general hypothesis that political involvement is related to formation of political attitudes among children and adolescents will be tested with respect to three different attitudes: economic liberalism, political trust, and party preference.

The three attitudes have been selected because together they represent central parts of the universe of attitudes which have been dealt with in socialization research. Socialization research so far has studied both attitudes which relate to the conflictual aspects of politics and attitudes which relate to political processes and political institutions. Economic liberalism and party preference clearly relate to political conflict, whereas political trust relates to political institutions and political processes.

Economic liberalism, political trust, and party preference have also been chosen because they are central to Danish politics. Economic liberalism and political trust have been seen as the two main dimensions structuring the realignment of the Danish party system which occurred in the 1970's (see e.g. Borre et al. 1976).

In the questionnaire economic liberalism, political trust, and party preference were measured in different ways. The questionnaire contained a number of items to measure economic liberalism and political trust. Party preference was registered by a single question concerning which party the respondent would vote for if of voting age.⁶

The effect of political involvement on the formation of political attitudes will be examined with respect to four aspects of the crystallization of attitudes. First, there is simply the effect on the ability to express political attitudes. Next, there is the effect on the correlation between items intended to measure the same attitude. Thirdly, we shall look at the effect of political involvement on the stability of political attitude. The fourth aspect to be investigated is the effect on the relationship between party preference and each of the attitudes of economic liberalism and political trust. The choice of these four aspects of the crystallization of attitudes is inspired by the literature about belief systems and non-attitudes among adult mass publics in the USA. The bulk of this literature deals with one or more of the aspects to be examined here. None of them has so far been investigated thoroughly in socialization research.

The original hypothesis, then, is subdivided into four parts. They are: 1) the higher the political involvement, the higher the ability to express political attitudes; 2) the higher the political involvement, the higher the correlation between indicators of the same attitude; 3) the higher the political involvement, the more stable the attitudes; 4) the higher the political involvement, the greater the relationship between party preference and economic liberalism and political trust.

3.1. Political involvement and the ability to express political opinions

Analysis of the ability of children and adolescents to express political opinions is scarce in the literature on political socialization. The exception is Hess & Torney (1967, Ch. 2), who have dealt somewhat with the dimension. They have looked at the propensity of various groups to choose the response categories 'don't know' and 'no opinion'. It can be assumed that this operationalization is quite valid with children. They are probably little affected by social pressures to have political views.

Hess and Torney found that the ability to express opinions increases rapidly around the age of 10 or 11. They did not investigate the relationship between political involvement and the ability to express opinions. They did, however, demonstrate that 'concern over the election outcome' has considerable effect on the ability to express opinions, and 'concern over the election outcome' sometimes has been used as an indicator of political involvement (1967, pp. 199ff).

Our operationalization of ability to express opinions follows the guidelines from earlier research. With economic liberalism and political trust respondents who choose one of the alternatives 'agree', 'disagree', and 'neither agree nor disagree' are considered to express opinions,

whereas those choosing the alternative 'no opinion' are not. Concerning
the party preference the sheer mentioning of a preferred party indicates
ability to express opinions.

The bivariate relationship between political involvement and the ability to express opinions is given in Table 2. For this analysis the continuous variable of political involvement has been divided into four categories ranging from low to high. As we expect to find only a small minority of school children with a crystallized pattern of attitudes the category labelled 'high' only contains 10 per cent of the respondents. We shall pay special attention to this group of children. The three other categories on political involvement each contain 30 per cent of the respondents.

Table 2 gives the proportion of the respondents expressing an opinion. As expected, the ability to express political opinions increases with political involvement. Concerning party preference there is a large difference between the two highest categories on political involvement. On this particular attitude the most highly involved children clearly differ from other children.

Next we shall compare the effect of political involvement with the effect of sex, grade (corresponding roughly to age), educational line, and social status. We will also look at the interaction between political involvement and these socio-economic variables. They are all familiar from traditional socialization research. Educational streaming is the closest one can get to a variable indicating the level of education when analyzing school children. Socialization theory would lead us to expect political attitudes to be most highly developed among boys, children of the highest grades, chlldren

in the highest stream, and among children with a high status
background.

The simple correlation between the ability to express opinions and political involvement and the socio-economic variables are presented in Table 3.⁷ All socio-economic variables seem to have a certain effect on the ability to express opinions. The direction of the effect is as expected. Sex seems to be the strpngest, social status the weakest factor. The effect of political involvement is considerably higher than the effect of the socioeconomic factors.

One might suspect that political involvement is an intervening variable between the socio-economic variables and the variables measuring the development of attitudes. This interpretation is consistent with the partial correlation coefficient also reported in Table 3. When controlling for political involvement, the effect of grade, educational streaming, and social status diminishes. Some of the original effect of these variables is due to related differences in political involvement. Sex is an exception to this pattern. The effect of sex on the crystallization of attitudes seems to be independent of political involvement.

The picture of political involvement as an important factor is further supported by the statistics given in Table 4. Here the reverse control of the one just presented has been carried out. The correlation between political involvement and the ability to express political attitudes has been controlled for socio-economic variables. In general the correlation coefficients for political involvement are not affected by this control. Political involvement works independently of these factors.

To summarize: The analysis clearly depicts political involvement as a
potent predictor of the ability to express opinions. The hypothesis relating
to this aspect of the formation of attitudes is supported.

3.2. Political involvement and the correlation between indicators of political attitudes

The size of the correlations between items indicating the same attitudes can be taken as a measure of the crystallization of the attitude. The more crystallized an attitude, the higher the correlation that must be expected among indicators of the attitude. Accordingly, correlations between indicators should be expected to rise with increasing levels of political involvement.

This particular aspect of the development of political attitudes has only been dealt with in a few studies. Easton & Dennis (1967) have analysed variations in the correlation between indicators of political efficacy. They found that this attitude was fairly well crystallized with American children of 8 or 9 years. Merelman (1971, pp. 66 ff) has looked at correlations between indicators of various democratic norms. Democratic norms could

be traced with 12 and 13 year-old children, who were the youngest group in the study. Correlations between indicators of democratic norms were increasing with increasing age. Neither Easton and Dennis nor Merelman have dealt with the effect of political involvement on the correlation between indicators of political attitudes.

A simple measure for the correlation between the indicators for economic liberalism and political trust has been constructed. For each attitude we have correlated all indicators with each other, and then calculated the mean of the resulting correlations.⁸ For economic liberalism with 7 items in the questionnaire, it is a mean of 21 coefficients, for political trust with 8 items, it is a mean of 28 coefficients. The party preference, which was measured by only one question, has to be omitted from this part of the analysis.

In Table 5 the mean coefficients for economic liberalism and political trust have been calculated for each of the subgroups we get when grouping the children by political involvement and by each of the socio-economic variables of sex, grade, stream, and social status. As expected, average coefficients are increasing with increasing levels of political involvement. Socio-economic variables also have some effect on the correlation bet

ween indicators. As expected, average correlations are somewhat higher with pupils in the 9th grade, with children in the examination stream, and with high status children. Somewhat surprisingly, the average correlation for political trust is a little higher with girls than with boys.

To get some idea of the interaction between political involvement and socio-economic variables average correlations between indicators have also been calculated for the subgroups produced by grouping children by both political involvement and a socio-economic variable. These correlations are given in Table 6.

In the preceding section we found some evidence of political involvement as an intervening variable between the socio-economic variables and the ability to express opinions. The same interpretation cannot be supported with respect to the correlations between indicators. Looking at the effect of socio-economic variables controlled for political involvement, no overall picture emerges. Sometimes the effect is smaller, and sometimes the effect is larger than on the whole data set.

The result is much clearer if we consult the table in order to see whether the effect of political involvement is dependent on the values of socioeconomic variables. The effect of political involvement in general is sustained within each category relating to socio-economic variables. The few exceptions are primarily instances of where the two lowest categories relating to.political involvement appear in the 'wrong' order. In general the effect of political involvement within each category is just as high as on the ungrouped data. Accordingly, the effect of political involvement seems to be independent of socio-economic factors. This result is parallel to the findings in the previous section.

Considering the clear result of the examination of the effect of political involvement within socio-economic groups the hypothesis that correlations between indicators rise with rising levels of political involvement is supported.

3.3. Political involvement and the stability of attitudes

The third aspect of the formation of attitudes to be analysed is the stability of attitudes. We assume that rooted political attitudes will be stable, at least for a shorter period of time. Non-attitudes, on the other hand, will be fickle and unstable. As we expect relatively more rooted attitudes among highly involved school children, we expect the stability of attitudes to be rising with increasing levels of political involvement.

The question of attitude stability has rarely been dealt with in the
literature on political socialization. The most important exception is an

article by Pauline Vaillancort (1973). The author, who is well aware that non-attitudes are probably widespread among children, makes an attempt to explain variation in stability by socio-economic factors. The result is not encouraging. There are no differences between age groups, between sexes, or between levels of intelligence. But there are clear racial variations and variations between status groups. The author hints that better results might be obtained by taking the dimension of political involvement into consideration.⁹

A test-retest procedure is the conventional way to assess the stability of attitudes. Respondents are twice presented with the same question. Identical answers are taken to indicate stability, whereas diverging answers are taken to represent instability. This procedure, of course, does not give us a clearcut distinction between attitudes and random answers. Some of the people answering by chance are bound to come out with identical answers in the two interviews. A much higher rate of identical answers, however, must be expected among holders of attitudes than among random answers.

396 of our respondents were interviewed twice with an interval of one
year. The stability of attitudes can only be assessed with respect to this
group of children.

Among the children twice interviewed there were many who gave one or two 'no opinion' answers. Identical answers, therefore, may be found both with respondents who did not express an attitude at either of the two interviews and with respondents expressing attitudes at both interviews.

The obvious solution may seem only to analyse respondents which gave two substantial answers. The consequence, however, would be a drastic reduction in the number of cases available for analysis. In the group lowest on political involvement only one person yielded a party preference at both test and retest. The reduction in the number of cases would be somewhat smaller at higher levels of political involvement. Thus, not only would we get a drastic reduction of cases, but also a reduction which is clearly related to the variable of political involvement. Accordingly, it would hardly be possible to analyse the effect of political involvement if we only look at respondents who gave two substantial answers.

The only possibility is to include respondents giving one or two nonsubstantial answers in the analysis. In consequence it might be more correct to say that our analysis will be dealing with 'response stability' rather than with 'attitude stability'.¹⁰

The statistical measure of stability used is the contingency coefficient.
The higher the coefficient between answers, the higher the stability. With
the party preference, where only one question was asked, coefficients

have been calculated on this single question. With economic liberalism
and political trust, where several indicators were used, our measure is the
mean of the coefficients calculated for each of the indicators.

Table 7 shows the contingency coefficients for each of the three attitudes. No clear picture emerges. There is some difference between the three attitudes, but it seems hard to explain variation in stability. Political involvement comes out as the best explanatory factor. In all instances the highest contingency coefficients are found with the group scoring highest on political involvement. Thus some support for our hypothesis can be found. But the relationship is weaker than expected.

The discouraging result may to a certain degree be attributed to the
'disturbance' caused by stable non-respondents. Stable non-respondents
are concentrated among the children of little political involvement.11 The

partly negative finding should, however, also give us a pause to make a
critical examination of our hypothesis.

So far we have assumed instability of attitudes to be related to the
phenomenon of 'non-attitudes'. Two alternative interpretations of instability
are feasible.

The first alternative interpretation relates to attitude change. We cannot,
of course, rule out the possibility that attitudes have simply changed
between test and retest. In our case a whole year separates the two tests.12

Opinion research has shown that attitudinal change is much more frequent in late adolescence and early adulthood than later on (see e.g. Butler & Stokes 1969, Ch. 3). Furthermore, one can assume that the highest volatility will be found among highly involved youngsters (cf. Converse 1962). They are most likely to receive information and other stimuli which may trigger attitudinal change. Instability in the form of change of attitudes therefore is bound to work contrary to the confirmation of our hypothesis.

The second alternative interpretation relates directly to the process of socialization. A major feature of socialization during adolescence is the gradual acquisition of attitudes. Accordingly some of our respondents may not have had an attitude at the first interview, but may have acquired one between the two interviews. There is no reason to assume that those acquiring opinions between interviews will be distributed the same way as those holding non-attitudes. Instability in the form of acquisition of attitudes therefore may also distort our analysis.

It is hard to reach any definite conclusion concerning the effect of political involvement on the stability of attitudes. It is somewhat difficult to reach a satisfactory operationalization of attitude stability, empirical results have been inconclusive, and it is possible to challenge the theoretical reasoning behind this particular aspect of our general hypothesis on the importance of political involvement.

3.4. Political involvement and the correlations between attitudes

It is a well-known fact about electoral behaviour in Denmark that party preference is related both to economic liberalism and to political trust. Supporters of socialist parties are in favour of economic equality, supporters of bourgeois parties against. Supporters of the Progress Party and other protest parties show considerable less political trust than supporters of the more established parties.

We expect correlations between party preference and economic

liberalism and political trust to rise with rising levels of political involvement.

Individuals strive to create correspondence between attitudes (e.g. Festinger 1968). The impulse to do so, however, is a function of the importance of attitudes. The more salient the attitudes, the more pertinent the need for consistent views. Consequently we should expect politically involved persons to have a great need for consistent political views, whereas the need for consistent political attitudes should be less with people more indifferent to politics.

Merelman (1971, p. 140) has shown that the correlation between party preference and democratic creed increased with age. The explanation offered is that the intellectual ability to establish the ideological link between attitudes increases with age. This explanation stresses the cognitive element in the formation of attitudes and corresponds to Merelman's general preference for the cognitive development framework as an approach to political socialization. Our alternative to his explanation would be to attribute the increase in the correlation between attitudes to the increase in political involvement which goes with rising age.

To simplify the analysis our party variable has been dichotomized.But we have had to work with two different dichotomies. For correlation with economic liberalism, the two categories are preference for a socialist party and preference for a bourgeois party. All children preferring the Progress Party are excluded for this analysis. For the correlation with political trust, the two categories are preference for the Progress Party and preference for other parties.¹³

The correlations between party preference and attitudes are presented
in Table 8. As earlier, calculations have been carried out within subgroups.

The effect of political involvement is as expected. The correlation between party preference and each attitude is increasing with increasing levels of political involvement. The increase is somewhat smaller with respect to political trust than with respect to economic liberalism. Relatively high coefficients occur with the 10 per cent group scoring highest on political involvement.

Among the socio-economic variables only social status produces the
expected variation. The pattern found for the other socio-economic variables
is hard to interpret.

The major reason why results are not clearer probably has to do with a
drastic reduction in the number of cases. The analysis can only be carried
out with respect to children with a party preference and a position concerning

economic liberalism or political trust. The greatest reduction in the number of cases is caused by the party variable. One may assume most of the children with a party preference to be highly involved in politics. Political involvement, therefore, may interact with socioeconomic variables in a way confounding the correlations. Unfortunately the small number of cases prevents a control for political involvement.

In spite of these reservations we did after all find a rising correlation between attitudes with rising levels of political involvement. The examination of this aspect of the formation of attitudes therefore further supports our general thesis concerning the importance of political involvement.

4. Conclusion

The thesis we have tested in this article is that political involvement is an

important factor explaining variation in the occurrence of political attitudes. We have tried to examine the effect of political involvement with respect to four aspects of political attitudes: the ability to express attitudes, the correlation between indicators of a attitudes, the stability of attitudes, and the correlation between attitudes. Except for the somewhat tricky question of attitude stability the results have been clear. Political involvement comes out as a better predictor of the occurrence of attitudes among school children than any of the traditional variables of socialization research.

Our findings may necessitate a partial revision of the well-known model of cognitive development. The major implication is that cognitive development may still be considered a necessary, but no longer a sufficient condition for the formation of attitudes. A relatively high level of political involvement is a further precondition for the formation of political attitudes.

If so, two important tasks for future research may be spelled out. Political involvement seems to be a critical psychological variable in the socialization process. A top priority in future research must be to look for the most important determinants of political involvement. Besides traditional socio-economic variables, obvious variables to be examined are the political involvement of parents and the political or quasi-political activities of the child itself. If political involvement is a key factor in the socialization process, one might assume socialization processes to follow differing patterns depending on the level of political involvement. Accordingly, another top priority for future research must be to study the socialization process at various levels of political involvement.

---

NOTES

1 The analysis was carried out on 1896 Danish school children of 13-17 years, coming from 9 schools in Aarhus and 1 school in Morsø. The 10 schools were chosen to represent different social contexts, i.e. working, middle and upper class contexts in Aarhus and a stagnating rural area in Morsø. The questionnaire was administered in 1973. The pupils from Morsø - 3% in number - were interviewed once more after a lapse of one year. The study was made possible due to a grant from the Danish Social Science Research Council.

2 The conceptual difference is first, that 'political involvement' is a property of a person, whereas 'centrality' and 'saliency' relate to attitudes or objects of attitudes. Secondly, 'centrality' and 'saliency' often refer to more limited phenomena like the use of nuclear power or membership of the EEC. They are seldom used to characterize as farreaching a phenomenon as politics in general.

3 For a discussion of problems in operationalizing 'saliency', see Redder & Tonsgaard 1971.

4 The exact wording of the questions can be found in Jarlov, Svensson & Togeby 1977.

5 For the technical details of the index construction see Nie et al. 1975, pp. 487ff.

---

6 The items for economic liberalism were: 'wage differentials are much too high in this country', 'everybody should have the same wage no matter his job', 'people with high incomes should pay higher taxes than they do today', 'people with higher education deserve a high income', 'the state should confiscate major fortunes to the benefit of the whole population', 'it is only fair that some people make twice as much money as others', 'the income gap between rich and poor should be diminished'. The items for political trust were: 'politicians are too careless with taxpayers' money', 'politicians make agreat effort to the benefit of others', 'we should be more grateful to politicians for the great amount of work they carry out', 'there are a few bosses deciding everything', 'politicians generally care too little about the opinion of voters', 'the few people running everything don't care what people think', 'politicians do work to the benefit of the whole society, and not for any special group', 'politicians always care what ordinary people think'.

7 The operationalization of social status is a modified version of the code used by the Danish National Institute of Social Research.

8 In the calculations, persons answering 'no opinion' have been omitted.

9 The problem of stability is also dealt with in Hess & Torney 1967, pp. 238-40. Stability is increasing with age.

10 The problem we are discussing here is due only to our efforts to avoid answers reflecting 'non-attitudes'. When all respondents are forced to give substantial answers one luckily avoids stable non-respondents.

11 In the group scoring low on political involvement we find 87% of respondents were stable with no party preference, 1% stable with a party preference, and 12% changing between expressing and not expressing a party preference. In the group high on political involvement we find 10% of respondents were stable without a party preference, 45% stable with a party preference, 12% shifting between parties, and 31% changing between expressing and not expressing a party preference.

12 The questionnaire was administered in early summer 1973 and 1974. In between was the election of December 1973 which brought the Progress Party into the Folketing.

13 Political trust and economic liberalism are measured by means of an index constructed by using factor-score coefficients. Respondents are only assigned a value on the index in case they have given substantial answers to at least 70 per cent of the items. Respondents answering 'no opinion' to less than 30 per cent of the items are assigned the mean value on these items.

REFERENCES

Aberbach, J. D. and Walker, J. L. 1970. 'The Meaning of Black Power: A Comparison of
White and Black Interpretations of a Political Slogan', American Political Science Review
64, 367-88.

Borre, 0., Nielsen, H. J., Sauerberg, S. and Worre, T. 1976. Vælgere i 70'erne, København:
Akademisk Forlag.

Butler, D. and Stokes, D. E. 1969. Political Change in Britain. London: Macmillan.

Campbell, A., Converse, P. E., Miller, W. E. and Stokes, D. E. 1960. The American Voter.
New York: Wiley.

Converse, P. E. 1962. 'Information Flow and the Stability of Partisan Attitudes', Public
Opinion Quarterly, 26, 278-99.

Converse, P. E. 1964a. 'The Nature of Belief Systems in Mass Publics', pp. 206-61 in David
E. Apter (ed.), Ideology and Discontent. New York: Free Press.

Converse, P. E. 1964b. 'New Dimensions of Meaning for Cross-Section Sample Surveys in
Polities', International Social Science Journal 16, 19-34.

Converse, P. E. 1970. 'Attitudes and Non-attitudes: Continuation of a Dialogue' in Edward
E. Tufte (ed.), The Quantitative Analysis of Social Problems. Reading, Mass.: Addison

Easton, D. and Dennis, j. 1967, 'The Child's Acquisition of Regime Norms: Political
Efficacy', American Political Science Review 61, 25-38.

Festinger, L. 1968. A Theory of Cognitive Dissonance. Stanford: Stanford U.P.

Hess, R. D. and Torney, J. V. 1967. The Development of Political Attitudes in Children.
Chicago: Aldine.

Jarlov, C, Svensson, P. and Togeby, L. 1977. Kodebog for socialiseringsundersøgelsen.
Århus: Institut for Statskundskab.

Martinussen, W. 1972. 'Political Awareness and the Distribution of Other Social Resources:
The Case of Norway', Scandinavian Political Studies 6, 154-59.

Merelman, R. M. 1971, Political Socialization and Educational Climates, New York:
Rinehart and Winston.

Nie, N. H. with Anderson, K. 1974. 'Mass Belief Systems Revisited: Political Change and
Attitude Structure', Journal of Politics 36, 540-91.

Nie.N. H.etal. 1975. Statistical Package for the Social Sciences. New York: McGrawHill.

Pomper, G. 1972. 'From Confusion to Clarity: Issues and American Voters', American
Political Science Review 66, 415-28.

Redder, K. W. and Tonsgaard, O. 1971, 'Saliency and Political Attitudes', Scandinavian
Political Studies 6, 183-91.

Vaillancourt, P. M. 1973, 'Stability of Children's Survey Responses', Public Opinion Quarterly
37, 373-87.

Verba, S. and Nie, N. H. 1972. Participation in America. Political Democracy and Social
Equality. New York: Harper and Row.