¹ Financial support for this project was provided by a National Science Foundation Grant for Doctoral Dissertation Research made to Donald J. McCrone (Grant #GS 34218), and by a research grant from the Graduate College of the University of Iowa. I would like to thank Professor Donald McCrone, for his sound substantive advice, and George Woodworth, Assistant Professor of Statistics, University of Iowa, for his advice on the general linear hypothesis. Thanks also to John C. Wahlke for a number of helpful suggestions, and to Susan Brantley and John Tatlock for many stimulating discussions relevant to this research.

² Campbell, Ernest, “Adolescent Socialization,” in Handbook of Socialization, ed. Goslin, David A. (Chicago: Rand McNally, 1969), p. 827Google Scholar.

³ Sigel, Roberta, Learning About Politics (New York: Random House, 1970), p. 104Google Scholar.

⁴ Hyman, Herbert, Political Socialization (Glencoe, Illinois: The Free Press, 1959), p. 72Google Scholar.

⁵ Connell, R. W., “Political Socialization and the American Family,” Public Opinion Quarterly, 36 (Fall, 1972), 326–327CrossRefGoogle Scholar.

⁶ Hess, Robert and Torney, Judith, The Development of Political Attitudes in Children (Chicago: Anchor Press, 1967), p. 113Google Scholar.

⁷ Jennings, M. Kent and Niemi, Richard, “The Transmission of Political Values from Parent to Child,” American Political Science Review, 62 (March, 1968), 172–176CrossRefGoogle Scholar.

⁸ The discussion here will be limited to the nature of attitude correspondence.

⁹ Connell discusses relative pair and group correspondence, but neglects the third possibility—absolute pair correspondence. Connell, pp. 324–325.

¹⁰ Connell, p. 325.

¹¹ For example, many investigators believe that strict child-rearing patterns lead to authoritarianism in child. This authoritarianism could, in turn, “structure” policy attitudes. For a discussion of indirect socialization see Dawson, Richard and Prewitt, Kenneth, Political Socialization (Boston: Little, Brown, 1969), pp. 65–73Google Scholar; for a critique of the structuring principle see Searing, Donald D., Schwartz, Joel J. and Lind, Alden, “The Structuring Principle: Political Socialization and Belief Systems,” American Political Science Review, 67 (June, 1973), 415–432CrossRefGoogle Scholar.

¹² A constant will not usually inflate correlations. One instance, however, in which it would inflate correlations is when the constant causes a large percentage of both parent and child scores to fall into one rank or interval. It is important, therefore, to check the distribution of the attitude scores for both parents and children.

¹³ A reduction due to partialing does not, of course, necessarily mean a spurious relationship exists. A reduction also occurs when a developmental sequence is present.

¹⁴ This methodology is discussed by Furstenberg, Frank F., “The Transmission of Attitudes in the Family” (Ph.D. Dissertation, Department of Sociology, Columbia University, 1967)Google Scholar, chapter 3.

¹⁵ Wieting, Stephen, “Family Factors and the Religious Belief Systems of Adolescents” (Ph.D. Dissertation, Department of Sociology, University of Minnesota, 1971), p. 87Google Scholar.

¹⁶ Greenstein, Fred, Children and Politics (New Haven: Yale University Press, 1965), pp. 71–72Google Scholar.

¹⁷ There are, however, differences between party identification and issues which are inherent in the attitude objects. For example, “independent” for party identification is not equivalent to “undecided” in the case of issue attitudes. This point will be discussed again later in the paper.

¹⁸ Richard Niemi reports that adolescents in the Michigan socialization study perceived themselves as attitudinally closer to parents by an average of .10 (tau-b) than they were in actuality. He suggests that this misperception serves an important psychological function for the youthful respondents by reducing whatever tension may arise from asymmetry in parentadolescent political attitudes. Niemi, Richard G., “Collecting Information about the Family: A Problem in Survey Methodology,” in Socialization to Politics, ed. Dennis, Jack (New York: John Wiley and Sons, 1973), p. 480Google Scholar.

¹⁹ Furstenberg, p. 172.

²⁰ See March, David, “Political Socialization: The Implicit Assumptions Questioned,” British Journal of Political Science, 1 (October, 1971), 460–464Google Scholar.

²¹ Interviewers consisted of eight male political science and sociology graduate students between 25 and 32 years old.

²² The sample consists of 183 adolescents (54 per cent male, 46 per cent female) and 322 parents (45 per cent male, 55 per cent female). Data on both parents were obtained for 139 of the adolescents. Twenty-six adolescents came from one-parent homes. In the case of the remaining eighteen adolescents, the second parent was not home at the time of the scheduled interview or call-backs.

I shall not, in this report, distinguish between mothers and fathers. Therefore, to make maximum use of the sample size each adolescent is paired with both the mother and the father (when both are present). This procedure renders tests of statistical significance based on F-ratio's inappropriate. See footnote 43 for a further discussion.

²³ Wilier, David, Scientific Sociology (Englewood Cliffs, N.J.: Prentice-Hall, 1967)Google Scholar, chapter 6.

²⁴ Bohrnstedt, George W., “Reliability and Validity Assessment in Attitude Measurement,” in Attitude Measurement, ed. Summers, Gene (Chicago: RandMcNally, 1970), pp. 87–88Google Scholar.

²⁵ Kerlinger, Fred N., Foundations of Behavioral Research (New York: Holt, Rinehart and Winston, 1964), pp. 453–455Google Scholar.

²⁶ The common factor model was used with communalities estimated by squared multiple correlation coefficients. The rotation was oblique. The item falling below .50 was the first racial integration item (see appendix) which loaded at .38.

²⁷ The factor scores were estimated by multiple regression following the procedures outlined in Harmon, Harry, Modern Factor Analysis (Chicago: University of Chicago Press, 1968), pp. 350–354Google Scholar.

²⁸ Jennings and Niemi use the product-moment coefficient for party identification, and tau-b for the issues.

²⁹ Positive scores mean adolescents are more liberal than parents, negative scores mean the reverse. The distribution of these scores approximates a normal curve. The standard deviations for the race issue, China policy and marijuana laws are 4.1 3.7, and 4.3, respectively.

³⁰ The most plausible hypothesis, alternative to direct parental influence, is that parents and children have similar attitudes because each individual family member is exposed to the same social influences. The example used earlier was social class. Variables of this type may determine the position of the parent and child within their respective groups thus producing a spurious correlation. One test of the spurious relationship hypothesis is controlling for a third variable. Controls were introduced for education, social class, and religion; at no point, for any issue or party identification, was the bivariate correlation reduced by more than .02.

³¹ Connell, pp. 329–330, Jennings and Niemi, p. 169.

³² Perceptual accuracy was collapsed into three categories to insure that each category had an adequate sample size. The linear patterns was, however, even more pronounced in the five-category case, although the N's were small. In the five-category breakdown, the correlations in the category exhibiting the greatest perceptual accuracy for the race issue, China policy, and the marijuana issue were .66, .72, and .61, respectively (based on 30–35 cases).

³³ The question is: “How often do you discuss politics with your son or daughter?” (1) never, (2) a few times a year, (3) several times a month, (4) several times a week.

³⁴ It is possible that adolescents are influencing parents. But if this were the case, correlations should not be so strongly affected by adolescent perceptual uncertainty, unless there is a functional relationship between adolescent uncertainty over parental attitudes and parental uncertainty over adolescent attitudes. Any conclusive test would require longitudinal data. Another nonrejectable alternative is the possibility that parents and peers have attitudes which are strongly related. If this were the case, it could be argued that the adolescent's attitude is due to peer influence.

³⁵ Jennings and Niemi, p. 181, and Middleton, Russell and Putney, Snell, “Political Expression of Adolescent Rebellion,” American Journal of Sociology, 68 (March, 1963), 532–533CrossRefGoogle Scholar.

³⁶ Nogee, Philip and Levin, Martin, “Some Determinants of Political Attitudes among College Voters,” Public Opinion Quarterly, 22 (Winter, 1958–1959), 460CrossRefGoogle Scholar.

³⁷ Ernhart, Claire B. and Loevinger, Jane, Authoritarian Family Ideology: A Measure, Its Correlates, and Its Robustness (Austin, Texas: Society for Multivariate Experimental Psychology, 1969)Google Scholar.

³⁸ A four- as opposed to three-category breakdown is reported because any curvilinear pattern is more likely to be evident in the former. The results with three categories reflect the same pattern.

³⁹ Jennings and Niemi, p. 181, and Middleton and Putney, pp. 532–533.

⁴⁰ From this point forward, the term “issue salience” will be used to refer to partisan salience as well as issue salience.

⁴¹ The general form of model is:

Y = a_o + a_i + b_oX_j + b_jX_j + e,

where a_o represents the grand mean, a_i represents the deviation of the grand mean from the kth category, b_oX_j is the regression of Y on the interval independent variable which is common to all categories, b_jX_j represents the deviation of the optimum regression function in the kth category from the common slope. The question of interest is whether or not the common regression is parallel to the within-category regression. If they are parallel, it means that categorizing the parent attitude (by, for example, issue salience) has no effect. If the common regression slope and the slope computed within categories are not parallel, it means the regression within categories provides the line of best fit. The sums of squares due to fitting the line within the categories will thus be greater than the sums of squares due to the common regression. The explained variance and the correlation coefficient will increase because of this nonparallelism. See Overall, John E., “Multiple Covariance Analysis by the General Least Squares Regression Method,” Behavioral Science, 17 (May, 1972), 313–320CrossRefGoogle Scholar.

⁴² Most readers are no doubt aware that two-way analysis of variance computed by standard methods requires equal cell frequencies. Unbiased estimates are usually obtained, when cell frequencies are not equal, by using the weighted squares of means approximation. The multiple covariance linear model is the mathematical equivalent of this approximation. Fennessey, James, “The General Linear Model: A New Perspective on Some Familiar Topics,” American Journal of Sociology, 74 (July, 1968), 4CrossRefGoogle Scholar.

⁴³ Tests of statistical significance based on F-ratio's are not appropriate for these data because the least squares assumption of “independence” in independent random samples is not met (183 adolescents are paired with 322 parents). This means that the degrees of freedom for the residual sums of squares cannot be estimated in an unbiased manner. Consequently, the mean square residual (the denominator in the F-ratio) cannot be computed. The sums of squares, the crossproducts, and the degrees of freedom for the numerator are, however, unaffected.

⁴⁴ If, however, parental influence were indirect (for example, overprotected adolescents being conservative on China policy) it is possible that this indirect influence could lead to adolescent perceptual accuracy. Adolescents, because they are influenced, could be motivated to learn the parent's attitude. But if parental influence is direct, as is argued for the attitudes studied here, it is very unlikely that parental influence could be antecedent to adolescent perceptual accuracy.

⁴⁵ In these equations the a's are the intercepts and the b's are the regressions. The a's are necessary for unbiased estimates of the within-category slopes (their effects on Y are partialed out). The numerals above the a's and b's are the number of categories used in constructing the dummies and the first-order interaction terms (although perceptual accuracy for party identification has two categories rather than three).

⁴⁶ The information presented in Table 8 will be used to test hypotheses 6, 7, and 8, and to partition the explained variance between issue salience and perceptual accuracy. Total explained variance for the partitioning analysis is that variance explained by issue salience and perceptual accuracy. It is important to note that by the terms “issue salience” and “perceptual accuracy” is meant the parental attitude categorized by these two variables, or the effect of the two variables on the parent-adolescent regression (parental influence).

⁴⁷ Using the race issue as an example, entering perceptual accuracy immediately after the parent attitude increases the correlation coefficient from .32 to .44. Entering perceptual accuracy after the parent attitude and issue salience (issue salience controlled) increases the correlation from .38 to .48.

⁴⁸ In the case of the race issue, entering issue salience immediately after the parent attitude increases the correlation from .32 to .38. Entering issue salience after the parent attitude and perceptual accuracy (perceptual accuracy controlled) raises the correlation from .44 to .48.

⁴⁹ The variables are entered in the order called for by the equation for hypothesis seven.

⁵⁰ Blalock, Hubert M. Jr., “Causal Inferences, Closed Populations, and Measures of Association,” American Political Science Review, 60 (March, 1967), 130–136CrossRefGoogle Scholar.

⁵¹ If the variable “attractiveness” is added to the equations for either hypothesis 7 or hypothesis 8, the results will be statistically identical.

⁵² The correlations between attractiveness and perceptual accuracy for the attitudes toward racial integration, China policy, marijuana laws and party identification are .08, .09, —05 and .05, respectively; for attractiveness and issue salience they are .01, .00, .12, and —05, respectively.

⁵³ It should be noted from Table 6 that there is a reversal for the within-category progression for party identification. The statistical model will make no adjustment for this fact. If we assume conceptually that the progression must be linear, then the model overestimates the impact of attractiveness on the transmission of party identification.

⁵⁴ It should be kept in mind that what is being discussed is the independent effect of attractiveness on successful transmission after the effects of issue salience and perceptual accuracy have been partialed out.