Parental Expertise, Trustworthiness, and Accessibility: Parent-Adolescent Communication and Adolescent Risk Behavior
Abstract
A communication framework of persuasion and attitude change was utilized to analyze parent-adolescent communication about adolescent risk behavior. Three parent dimensions were deemed important: (a) perceived expertise, (b) perceived trustworthiness, and (c) perceived accessibility. Data were collected in surveys from 668 mother-adolescent dyads in economically disadvantaged neighborhoods in New York City (N= 668). Results showed weak correspondence between how expert, trustworthy, and accessible mothers thought they were on the one hand and how their sons and daughters characterized them on the other. All dimensions were related to how often adolescents said they talked with their mothers about a risk behavior, which, in turn, was predictive of lower levels of adolescent risk behavior. Implications for future research are discussed.
Considerable research has been conducted on parent-adolescent communication as a factor that influences adolescent risk behavior. Whereas early research questioned a link between these constructs, more recent studies have been supportive (e.g., Hutchinson, Jemmott, Jemmott, Braverman, & Fong, 2003; Jaccard, Dodge, & Dittus, 2002; Miller, Forehand, & Kotchick, 1999; Simons-Morton, 2002). In addition to direct links, studies have found that parent-adolescent communication moderates relationships between other variables and adolescent risk behavior. For example, Whitaker and Miller (2000) found peer norms were associated more strongly with sexual activity for adolescents who had not discussed sex or condoms with a parent, as opposed to those who had engaged in such conversations. Although some studies have yielded contradictory results (e.g., Darling & Hicks, 1982; Widmer, 1997), the bulk of current research suggests that parent-adolescent communication serves as a protective factor for adolescent risk activity (Dittus, Jaccard, & Gordon, 1999; Litrownik et al., 2000).
Most research on parent-adolescent communication correlates measures of the frequency of communication about the risk behavior in question to measures of risk activity. The communication process obviously is more complex than this, and the field can benefit from richer theoretical frameworks that can organize and guide research on parent-adolescent communication. Jaccard et al. (2002) present such a framework. Their approach draws on perspectives from attitude change and persuasion research in social psychology. It is restricted to the situation where a parent is trying to persuade or convince a child to perform or not perform a behavior. Jaccard et al. (2002) recognize that parent-adolescent communication is reciprocal and dynamic in character but suggest that it also is useful to think about such communication from a top-down perspective.
The theoretical framework conceptualizes communication in terms of five facets: (a) the source of a communication (the parent), (b) the communication itself (often referred to as the message), (c) the medium or channel through which the message is transmitted (e.g., face-to-face, written materials), (d) the recipient of the communication (the adolescent), and (e) the context in which the communication occurs. Each of these components has subcomponents. For example, sources of a message differ in their age, gender, expertise, and trustworthiness. Recipients of communications differ in their motivational states, emotional states, past experiences, and expectations. The surrounding environment varies in terms of its temporal, physical, social, and cultural features. Variations in each of these five components represent independent variables that ultimately affect how an adolescent responds to parental communications.
The present research focuses primarily on one facet of the theory, namely, source characteristics of parents that influence adolescent risk behavior. The research delineates three key source dimensions that have rarely been studied in the parent-adolescent communication literature (namely, perceived expertise, trustworthiness, and availability) and examines how adolescents and parents differ on their perceptions of these dimensions, how the dimensions relate to communication frequency, and the extent to which communication frequency mediates the association of the dimensions with adolescent risk behavior.
Parental source characteristics
Although a multitude of source characteristics may influence the effectiveness of a communication (e.g., Jaccard et al., 2002; Lefkowitz, Romo, Corona, Au, & Sigman, 2000), research in social psychology suggests two core dimensions as being particularly relevant, perceived expertise and perceived trustworthiness. We define perceived expertise as the perception on the part of the adolescent that the parent gives good advice and is a credible source of information (i.e., the parent is knowledgeable). We define perceived trustworthiness as the perception on the part of the adolescent that the parent is honest, trustworthy, and looking out for the best interests of the child.
Much research has linked perceived expertise and trustworthiness to attitude change (see Albarracín, Johnson, & Zanna, 2004, for a review of this research). For example, in experimental studies, when exactly the same message is provided to two groups by sources who differ in perceived expertise, the more expert source typically produces more attitude change than the less expert source. Attributions of expertise have been shown to vary as a function of how likeable the source is, the gender of the source, the height of the source, how rapidly the source speaks, the attractiveness of the source, formal credentials of the source, and a host of other cues that the recipient takes into account at the time of message processing (Chaiken, 1987; Cialdini, 1993; Johnson, Maio, & Smith-McLallen, 2004). Similar results have been found for trustworthiness. In experimental studies, when the same message is delivered to two groups but by sources who differ in their trustworthiness, the more trustworthy source typically generates more attitude change (Cialdini, 1993; McGuire, 1985). Sources who are thought to have a vested interest in a topic are often seen as less trustworthy, as are individuals who tend to look out for themselves at the cost of others (McGuire, 1985). If a source is not seen as trustworthy, then even though the source may be perceived as an expert, the message delivered will result in little attitude change because the source cannot be trusted to be objective and honest about the underlying issues. The research on perceived expertise and trustworthiness clearly establishes the centrality of these two source dimensions in the communication process.
A third source variable that is important in the parenting context is the perceived availability or accessibility of the parent. If a parent is seen as being too busy and generally unavailable, then the adolescent will not seek out the parent’s advice (i.e., the probability of message exposure will be low). Similarly, if parents are too busy to spend time talking with their children, then message exposure will be low. Part of building a relationship with one’s adolescent is being there for the child in times of need (e.g., Laible & Carlo, 2004). Lower levels of perceived availability may thus lead to reduced levels of attitude change.
We suggest that these three source dimensions, perceived expertise, perceived trustworthiness, and perceived accessibility, represent core dimensions that affect message effectiveness. Other source characteristics may influence the effect of a message, but many of these effects will be mediated by these three core dimensions. For example, the gender of the parent may affect the influence a parent has on the adolescent in the context of communication. Such gender differences probably are traceable through perceived source differences in expertise, trustworthiness, or accessibility (or all) as a function of the gender of the parent. To be sure, parental gender also can influence the type of message the child receives and the ways in which that message is delivered, variables that fall in the conceptual category of message characteristics in the Jaccard et al. (2002) framework. Independent of such message differences and focusing strictly on source dimensions, however, gender effects also may be attributable to differences in perceived expertise, trustworthiness, and accessibility.
To our knowledge, no study has investigated adolescent perceptions of parental expertise, trustworthiness, and accessibility as predictors of parent-adolescent communication and adolescent risk behavior. Also, no research examined the interrelationships among these constructs. The present research does so.
Mechanisms of influence
At least two mechanisms may operate with respect to how perceived expertise, trustworthiness, and accessibility influence adolescent risk behavior. First, if adolescents think their parents give good advice, are trustworthy, and are not too busy for them, then adolescents may seek out parents for help and advice more often. This should result in increased opportunities for parents to talk about important issues and to convey certain points of view (i.e., messages) to adolescents. Similarly, if adolescents think parents are credible and trustworthy, they may be more open to talking and problem solving with parents when parents approach them to initiate conversations. The net result is that higher levels of perceived expertise, trustworthiness, and availability will increase the frequency of communication, and this increased communication, in turn, will result in the adolescent being less likely to engage in risk behaviors as they are exposed to parental perspectives.
A second mechanism is that higher levels of perceived expertise and trustworthiness will make the adolescent more apt to accept the advice given by parents, everything else being equal. Thus, two parents may discuss matters with the same frequency and say the same things, but the adolescent will be more receptive to and more influenced by parents who are perceived to have expertise and to be trustworthy.
Another purpose of the present study was to empirically evaluate if the effects of expertise, trustworthiness, and accessibility are mediated by the frequency of communication between parent and adolescent or if the dimensions also have independent effects on adolescent risk behavior. We hypothesize that adolescent perceptions of perceived expertise, trustworthiness, and accessibility will be predictive of the frequency of parent-adolescent communication about adolescent risk behavior, which, in turn, will be negatively associated with adolescent risk behavior per se. In addition, we hypothesize that adolescent perceptions of parental expertise, trustworthiness, and accessibility will be predictive of adolescent risk behavior over and above their association with communication frequency. The latter prediction derives from the fact that higher levels of perceived expertise and trustworthiness make parental messages more effective, everything else being equal.
Are adolescent and parent perceptions of source dimensions congruent?
Positive attributions of parental expertise, trustworthiness, and accessibility are important from an adolescent perspective because they reflect the adolescent’s perceptions of the parent-adolescent relationship and are likely to affect an adolescent’s motivation to communicate with parents and to behave in accordance with parents’ values. Adult perceptions also are important, however, as is the congruency of adolescent and parent attributions. For example, if parents think they are accessible and available to their adolescents but adolescents do not perceive them as such, then this is important to document. Similarly, parents who believe that they have gained the respect of their children in terms of trust and expertise but whose children fail to see them in such terms may be insensitive to the need to strengthen their relationship with their child for purposes of maximizing communication effectiveness. No research to date has addressed these issues. Another purpose of the present research is to explore the congruence between adolescent characterizations of parental expertise, trustworthiness, and accessibility and parent characterizations of these same dimensions.
Summary
Although parent-adolescent communication has received increased attention in the social science literature, it has tended to lack an organizing framework for understanding how parents influence their children in a communication context. Jaccard et al. (2002) provide such a framework. The present research (a) develops brief measures of perceived expertise, trustworthiness, and availability that can be used in applied research settings; (b) tests the relationship between these dimensions and adolescent risk behavior; (c) evaluates mechanisms by which the constructs are thought to influence behavior; and (d) evaluates the correspondence in adolescent and parent perceptions of expertise, trustworthiness, and availability. We know of no research that has formally explored these issues in the context of parent-adolescent communication. The present study does so using two different risk behaviors, smoking cigarettes and engaging in sexual intercourse, among a diverse population of middle school students in an economically disadvantaged, inner-city setting. The study is unique in that it identifies three core source dimensions that should be attended to when trying to understand the dynamics of parent-adolescent communication, drawing upon analyses of source effects that have a long and rich research tradition in the social psychology of persuasion and attitude change.
Method
Respondents
Respondents were 668 mother-adolescent dyads who were recruited from Grades 6, 7, and 8 from six middle schools in the South Bronx in New York City. The Bronx is New York City’s poorest borough, with a median household income significantly below the state average (U.S. Census Bureau, 2000). Approximately 20% of the total population has an income below the federal poverty level. For families with children, approximately 40% have an income below the federal poverty level. The Bronx has been greatly affected by the HIV/AIDS epidemic. As of 2002, the borough accounted for 26,771, or 21%, of New York City’s total AIDS cases (New York State Department of Health, Bureau of HIV/AIDS Epidemiology, 2002). The student population of the communities we sampled is approximately 75% Latino and 25% African American, with Dominicans and Puerto Ricans comprising the majority of the Latino population.
A school roster was obtained from each school, and student names were randomly selected from the rosters. A total of 820 families were selected. Families were contacted via telephone by bilingual callers who invited the resident mother and adolescent to attend a data collection activity at the adolescent’s school on a particular date. Resident mothers were defined as the woman residing within the adolescent’s home primarily responsible for the adolescent’s care. In some cases, telephone numbers were incorrect or disconnected (n= 146, 18%). For families whose numbers were inaccurate or nonexistent, the researchers attempted to obtain accurate phone numbers or home addresses from teachers and staff at the school and, in some cases, from students. We also contacted directory assistance and used reverse directory assistance services of the local telephone company. This enhanced phone-calling strategy allowed us to contact 120 (82%) of the 146 families that initially could not be contacted. All mothers who were contacted agreed to have an information packet about the study sent to them. The packet included a cover letter, consent forms, and supplemental materials about the project. Project staff followed up these contacts with a telephone call to the mother a few days later to answer questions and to confirm the mother and adolescent would participate in the study. For families whose telephone numbers could not be identified (e.g., a family had no home telephone), the information packet was sent to the mother at the most current address available or sent home from school with her adolescent child.
Overall, 668 (81%) mother-adolescent pairs were recruited into the study. Of those recruited, 623 families attended one of the in-school data collection activities. Forty seven percent (n= 312) of those attending did so after being scheduled to participate the first time, 28% (n= 188) attended after being scheduled a second time, and 18% (n= 123) attended after a third scheduling. We offered the remaining families a home visit. Seven percent of families (n= 45) were interviewed in their residences, with two interviewers separately interviewing the mother or the adolescent in different rooms.
In all, 152 of the original 820 randomly selected families could not be recruited. Forty of these families (5%) directly refused to participate in the study, and 112 families (14%) never attended sessions or availed themselves of the home interviews. A refusal bias survey was administered during the recruitment process. We were able to gather information on key demographic variables (e.g., ethnicity, receipt of public assistance, employment status, maternal educational level) from all but 15 families. Those who refused or did not attend an event but were scheduled and those who did participate were demographically similar.
Mothers signed participation consent forms for themselves and permission forms for their adolescents, and the adolescents signed assent forms. All forms were written in both English and Spanish and were reviewed with the mothers and adolescents by project staff. Compensation was offered for participation, $30 per mother and $20 per adolescent. Institutional Review Board approval was obtained at Columbia University and the Centers for Disease Control and Prevention.
Our research focused on mothers rather than fathers because interviewing both mothers and fathers in the inner-city raises substantial challenges and costs. Fathers are difficult to recruit, and there are nontrivial selection effects in those who agree to participate (Costigan & Cox, 2001). Statistical modeling is complicated by the presence of nonrandom missing data (because of single-mother households), and research tends to show that, in both the sexual domain and the tobacco-use domain, mothers are more influential in terms of being a source of information for and affecting the risk behavior of their adolescents (e.g., DiIorio, Kelley, & Hockenberry-Eaton, 1999; Hutchinson, 2002). Although it would be ideal to study both mothers and fathers, it was not economically feasible. Focusing on mothers initially seemed reasonable, with future research to be conducted on fathers being our next goal.
Measures
Self-administered questionnaires were used. To encourage truthful responding, we assured adolescents and their mothers their responses were confidential and explained to them the difficulty in linking a name to a questionnaire. Our previous research with inner-city populations suggests respondents find frequent scale format changes to be confusing. Whenever possible, we structured our items so they used either a 5-point strongly agree to strongly disagree scale (1 =strongly agree, 2 =moderately agree, 3 =neither agree nor disagree, 4 =moderately disagree, 5 =strongly disagree) or a 5-point positive to negative scale (1 =very positive, 2 =moderately positive, 3 =neutral, 4 =moderately negative, 5 =very negative). Instructional sets were reviewed verbally, and respondents were given practice items to eliminate warm-up effects and to ensure scale understanding. The language and wording used in the items was based on linguistic analysis of previously conducted focus groups with the target population. Gender variant versions of the questionnaire were developed for male and female adolescents, for mothers of male adolescents, and for mothers of female adolescents. The questionnaire was forward and backward translated into Spanish utilizing methods described in Marín and Van Oss Marín (1991). The questionnaires were pilot tested for readability and comprehension. Participants were asked to identify their language preference for completing the questionnaire. The questionnaire assessed both the mother’s and adolescent’s perspective on communication about two risk behaviors: smoking cigarettes and engaging in sexual intercourse. In addition to the customary demographics, we included a social desirability response scale to assess social desirability response tendencies. In our description of measures, we use phrasing consistent with the female version of the instrument; appropriate wording changes were made for male adolescents.
Measures of expertise, trustworthiness, and accessibility The constructs of perceived expertise, trustworthiness, and accessibility were measured using three items for each construct for both adolescents and their mothers. Each item was rated on a 5-point agree to disagree scale. For adolescents, the items for expertise were as follows: (a) my mother gives me good advice, (b) the advice my mother gives me is helpful when we talk about important topics, and (c) when I need good advice about something important, I go to my mother for help. The items for trustworthiness were as follows: (a) I can trust my mother when we talk, (b) my mother keeps her promises to me, and (c) my mother is honest with me. The items for accessibility were as follows: (a) it is difficult for my mother and me to find time to talk, (b) my mother is too busy when I want to talk to her about things, and (c) my mother has trouble finding time to talk with me. The parent items paralleled the adolescent items but were phrased to assess how the mother thought her child viewed her. For example, the three expertise items were (a) my daughter thinks I give good advice to her, (b) my daughter finds my advice helpful when we talk about important topics, and (c) when my daughter needs good advice about something important, she comes to me for help. For the accessibility items, however, the parallel mother items were phrased to reflect how accessible the mother thought she was to the adolescent. The three items were (a) it is difficult for my daughter and me to find time to talk, (b) I find I am too busy when my daughter wants to talk with me about things, and (c) I have trouble finding time to talk with my daughter. The alpha coefficients for the expertise, trustworthiness, and accessibility items for adolescents were 0.72, 0.82, and 0.73, respectively, and for mothers they were 0.70, 0.68, and 0.70, respectively. Each item was scored from 1 =strongly agree to 5 =strongly disagree, with the accessibility items being reverse coded so higher scores implied higher levels of accessibility. All measures were correlated trivially with our measure of social desirability response tendencies.
Measures of communication Adolescents indicated how much they had talked with their mother about specific topics using a 4-point rating scale with response categories 1 =not at all, 2 =somewhat, 3 =a moderate amount, and 4 =a great deal. For communication about the consequences of engaging in sexual intercourse, conversations on three domains were assessed, each with three items. One domain focused on the physical and health risks associated with sexual behavior and included the following items: (a) my mother and I have talked about what might happen to me if I were to get pregnant; (b) my mother and I have talked about how if I had sexual intercourse at this time in my life, I might get a sexually transmitted disease (STD); and (c) my mother and I have talked about how if I had sexual intercourse at this time in my life, I might get HIV/AIDS. A second domain focused on the social consequences of engaging in sexual intercourse and included the following items: (a) my mother and I have talked about how I might get a bad reputation if I had sexual intercourse at this time in my life, (b) my mother and I have talked about how embarrassing it would be for me if I got pregnant now, and (c) my mother and I have talked about how my boyfriend might lose respect for me if we had sexual intercourse at this time in my life. The third domain focused on the moral consequences of engaging in sexual intercourse and included the following items: (a) my mother and I have talked about how having sexual intercourse at this time in my life would be morally wrong, (b) my mother and I have talked about how if I had sexual intercourse at this time in my life I might regret not waiting until I was married, and (c) my mother and I have talked about how I would feel guilty if I had sexual intercourse at this time in my life. Parallel items were asked of the mother. The alpha coefficients for the three domains were 0.88, 0.85, and 0.89 for the adolescents and 0.88, 0.86, and 0.89 for the mothers, respectively.
For smoking cigarettes, two domains of communication were assessed, each with three items. The first focused on the physical and health consequences of smoking cigarettes and included the following items: (a) my mother and I have talked about how I could die sooner if I were to smoke cigarettes at this time in my life, (b) my mother and I have talked about how my lungs would get black if I were to smoke cigarettes at this time in my life, and (c) my mother and I have talked about how it would be unhealthy for me to smoke cigarettes at this time in my life. The second domain focused on the social consequences of smoking cigarettes and included the following items: (a) my mother and I have talked about how smoking cigarettes is not a good way to get accepted by my friends, (b) my mother and I have talked about how smoking cigarettes is not a good way to become more popular, and (c) my mother and I have talked about how smoking cigarettes is not a good way to try to look cool. Parallel worded items also were asked of the mother. The alpha coefficients for the two domains were 0.88 and 0.89 for the adolescents and 0.90 and 0.90 for the mothers, respectively.
Adolescent risk behavior Adolescents were provided with a definition of vaginal sexual intercourse and then were asked if they had ever engaged in it and, if so, how many times. Responses were scored in six categories ranging from zero to five times or more. Risk behavior for smoking was assessed by asking adolescents if they had ever smoked a cigarette, even a few puffs. Adolescents who replied no were assigned a score of 0 on the behavioral measure. Those who responded yes were further queried about how many days during the past month they had smoked cigarettes. Responses were initially scored from 0 =never smoked cigarettes to 10 =smoked cigarettes 10 or more times based on the frequency they provided, and this score plus 1 was assigned to the behavioral measure. Thus, scores could range from 0 to 10, with a score of 0 =the adolescent had never smoked cigarettes, a score of 1 =the adolescent had smoked cigarettes but not in the last 30 days, a score of 2 =the adolescent had smoked cigarettes but only once in the past 30 days, and so on up to a score of 10 (so the top range was nine or more cigarettes smoked in the last 30 days). The upper end of the scores for both measures were winsorized (i.e., topped at 5 and 10, respectively) because there were few cases with scores greater than this, and they were likely to serve as too influential outliers in the analyses. We used two different variants of the measures, a simple dichotomy reflecting whether adolescents had engaged in sexual intercourse and a count variable reflecting the number of times they had engaged in sexual intercourse over the course of their lifetime. For smoking, the dichotomy was whether they reported having smoked at least once in the past 30 days.
Results
The results are organized into four sections. First, we describe preliminary analyses that document basic descriptive statistics of demographic variables and adolescent risk behavior and the extent to which outliers, missing data, and nonnormality are present. Second, we use confirmatory factor analysis to explore the relationship between adolescent attributions of maternal expertise, trustworthiness, and availability and the corresponding perceptions of the parent on these dimensions. Third, we document our measures of communication frequency and explore the relationship between adolescent characterizations of communication and parental characterizations of communication. The results set the stage for the final section, which relates the three core source dimensions to adolescent tobacco and sexual risk behavior.
Preliminary analyses
Descriptive statistics Table 1 presents a demographic profile of the sample, and Table 2 presents means, standard deviations, and indices of skewness and kurtosis for key variables used in the formal models we tested. Overall, 11.2% of adolescents reported having ever engaged in sexual intercourse, and 16.0% reported having ever smoked a cigarette. Among female adolescents, 5.0% had ever had sex and 15.8% had ever smoked a cigarette, whereas among male adolescents, 17.1% reported sexual activity and 16.2% reported ever smoking. Reports of sexual behavior and cigarette smoking were higher among African American than among Latino adolescents, with 19.9% and 9.5% reporting having ever engaged in sexual intercourse, respectively, and 20.0% and 15.6% reporting ever having smoked a cigarette, respectively.
| African American | Latino | |
|---|---|---|
| Sample size | 140 | 528 |
| A: M age | 13 | 13 |
| A: % malea | 54 | 51 |
| P: M age | 42 | 40 |
| P: % completed high school or moreb | 73 | 53 |
| P: % not living with partnerc | 59 | 45 |
| P: % workingd | 53 | 53 |
| P: % receiving public assistancee | 40 | 44 |
- Note: A = adolescent sample; P = parent sample.
- aDummy coded variable with 0 =female, 1 =male.bDummy coded variable with 0 =less than high school education, 1 =completed high school, attended some college, completed college, or attended post college.cDummy coded variable with 0 =married or living with a partner, 1 =being single and living with no partner, divorced or separated and living in different households, or widowed with no partner.dDummy coded variable with 0 =parent reported they currently are not working, 1 =parent reported they currently are working.eDummy coded variable with 0 =parent reported they are not receiving public assistance, 1 =parent reported they are receiving public assistance.
| Adolescents | Mothers | |||||||
|---|---|---|---|---|---|---|---|---|
| Variables | M | SD | Skewness | Kurtosis | M | SD | Skewness | Kurtosis |
| SE: gives good advice | 4.40 | 0.95 | −1.55 | 1.41 | 4.28 | 0.96 | −1.46 | 1.65 |
| SE: gives helpful advice | 4.13 | 1.22 | −1.24 | 0.31 | 4.05 | 1.08 | −1.05 | 0.22 |
| SE: when want advice, go to | 3.94 | 1.35 | −0.98 | −0.45 | 4.34 | 0.93 | −1.48 | 1.60 |
| ST: keeps promises | 3.72 | 1.35 | −0.66 | −0.93 | 4.42 | 0.82 | −1.54 | 2.14 |
| ST: can trust | 4.13 | 1.25 | −1.24 | 0.19 | 4.62 | 0.67 | −2.12 | 5.71 |
| ST: is honest | 4.23 | 1.12 | −1.31 | 0.55 | 4.65 | 0.67 | −2.35 | 6.68 |
| SA: difficult to find time | 2.25 | 1.39 | 0.85 | −0.66 | 2.30 | 1.48 | 0.77 | −0.94 |
| SA: too busy | 2.00 | 1.25 | 1.16 | 0.19 | 1.62 | 0.95 | 1.86 | 3.26 |
| SA: trouble finding time | 2.10 | 1.34 | 1.02 | −0.27 | 1.69 | 1.13 | 1.79 | 2.62 |
| CC: die from smoking | 3.10 | 1.13 | −0.81 | −0.87 | 3.17 | 1.03 | −0.90 | −0.52 |
| CC: black lungs | 3.18 | 1.09 | −1.01 | −0.47 | 3.20 | 1.03 | −0.98 | −0.38 |
| CC: unhealthy for you | 3.24 | 1.04 | −1.10 | −0.16 | 3.30 | 0.98 | −1.22 | 0.27 |
| CC: bad way to get accepted | 3.12 | 1.14 | −0.91 | −0.73 | 3.22 | 1.03 | −1.08 | −0.13 |
| CC: bad way to be popular | 3.08 | 1.15 | −0.84 | −0.85 | 3.10 | 1.10 | −0.86 | −0.71 |
| CC: bad way to look cool | 3.14 | 1.12 | −0.93 | −0.67 | 3.24 | 1.00 | −1.10 | −0.05 |
| CS: consequences of pregnancy | 2.84 | 1.19 | −0.43 | −1.37 | 2.77 | 1.20 | −0.33 | −1.46 |
| CS: possibility of STDs | 2.92 | 1.18 | −0.59 | −1.20 | 2.98 | 1.10 | −0.62 | −1.03 |
| CS: possibility of AIDS | 3.00 | 1.15 | −0.69 | −1.04 | 3.05 | 1.09 | −0.75 | −0.82 |
| CS: morally wrong | 2.91 | 1.18 | −0.57 | −1.23 | 2.75 | 1.22 | −0.31 | −1.50 |
| CS: regret not waiting | 2.70 | 1.28 | −0.25 | −1.63 | 2.55 | 1.23 | −0.06 | −1.60 |
| CS: feel guilty | 2.63 | 1.25 | −0.18 | −1.61 | 2.45 | 1.25 | 0.07 | −1.64 |
| CS: get bad reputation | 2.72 | 1.23 | −0.28 | −1.52 | 2.57 | 1.26 | −0.09 | −1.65 |
| CS: be embarrassing | 2.82 | 1.25 | −0.44 | −1.48 | 2.66 | 1.28 | −0.21 | −1.65 |
| CS: lose respect for me | 2.58 | 1.28 | −0.124 | −1.70 | 2.55 | 1.27 | −0.06 | −1.67 |
| BC: frequency of smoking | 2.00 | 5.15 | 3.58 | 13.80 | ||||
| BS: frequency of intercourse | 0.36 | 1.15 | 3.33 | 9.95 | ||||
- Note: N= 668. SE = source expertise item; ST = source trustworthiness item; SA = source accessibility item; CC = frequency of communication item about cigarettes; CS = frequency of communication item about sexual intercourse; STD = sexually transmitted disease; BC = behavior cigarettes; BS = behavior sexual intercourse.
Outliers, missing data, and nonnormality In all our modeling efforts, we pursued outlier analyses to identify potentially influential cases that might mask more basic trends in the data. Specifically, we calculated a leverage score for individuals based on their multivariate profile for the variables in a given model. The mean leverage score was computed, and an outlier was defined as anyone having a leverage score three times the value of the mean (Jaccard & Wan, 1995). In addition, for models that had endogenous variables, we examined standardized dfbetas using limited information estimation approaches based on the linear equations implied by the model. An outlier was defined as an individual who had a standardized dfbeta larger than 3.0. No outliers were evident in any of our analyses.
Only a few cases had missing data on any variable. No coherent pattern to the missing data was identified. For those cases with missing data, values were imputed to conform to covariance estimates consistent with the application of the expectation-maximization approach to missing data. Given the small number of instances of missing data, concerns surrounding estimation with missing information are moot.
Most of our analyses used structural equation modeling (SEM). Traditional maximum likelihood methods of SEM assume that the continuous variables in the model are multivariately normally distributed. A necessary but not sufficient condition for multivariate normality is normality at the univariate level. Nonnormality was evident for several variables used in our models, as seen in Table 2. For these variables, we pursued parameter estimation using three different approaches, traditional maximum likelihood, bootstrapping, and the use of robust standard errors. For the bootstrap analyses, overall fit of the tested models was calculated using the Bollen-Stine bootstrap approach in place of the traditional chi-square statistic (Bollen & Stine, 1993). In general, conclusions were consistent across the three estimation approaches. All reported results are from the bootstrap analyses.
Confirmatory factor analyses of expertise, Trustworthiness, and Availability
Model and model fit To explore the factor structure of our core source measures, we conducted a confirmatory factor analysis with six latent variables and 18 observed variables. The unit of analysis was the mother-adolescent dyad. Three of the latent variables focused on adolescent perceptions of maternal expertise, trustworthiness, and accessibility, and the other three represented maternal perceptions of how expert and trustworthy they thought their adolescent saw them as being and of how available they perceived themselves to be. Each latent variable had three indicators, as discussed in the Method section. All latent variables were assumed to be correlated. The model yielded good fit to the data. The Bollen-Stine p value for the chi-square test was .51, the standardized RMR (root mean squared residual) was 0.026, the CFI (comparative fit index) was 0.99, the RMSEA (root mean squared error of approximation) was 0.015 with 90% confidence intervals of 0.00 and 0.025, and the p value for the test of close fit was greater than .99. Inspection of modification indices and standardized residuals suggested no signs of ill-fit at a more focused level or any theoretically coherent parameters that should be freed up to improve model fit. No offending estimates were observed. All the standardized path coefficients linking a latent variable to an observed indicator were reasonable in magnitude (they ranged from 0.56 to 0.83 with a median of 0.72). Taken together, these results affirm the use of the 18 items as indicators of the six latent constructs.
Estimated correlations Table 3 presents the estimated correlations among the six latent variables. Several noteworthy trends were evident in these correlations. First, for both adolescents and mothers, perceptions of expertise were highly correlated with perceptions of trustworthiness. For adolescents, the estimated correlation was 0.96 (p < .01), and for mothers, it was 0.71 (p < .01). This suggests adolescents who see their mothers as having expertise also tend to see their mothers as being trustworthy, and mothers who believe their adolescents see them as experts also think their adolescents see them as trustworthy. We discuss later a model that collapses these latent constructs into a single factor for adolescents and mothers, separately.
| Variables | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| 1. A-Trust | 0.96* | ||||
| 2. A-Access | 0.25* | 0.30* | |||
| 3. M-Expert | 0.13 | 0.02 | 0.10 | ||
| 4. M-Trust | 0.06 | 0.02 | 0.07 | 0.71* | |
| 5. M-Access | 0.04 | 0.07 | 0.10 | 0.34* | 0.38* |
- Note: A-Expert = adolescent ratings of expertise; A-Trust = adolescent ratings of trustworthiness; A-Access = adolescent ratings of accessibility; M-Expert = mother ratings of adolescent perceptions of expertise; M-Trust = mother ratings of adolescent perceptions of trustworthiness; M-Access = mother ratings of accessibility.
- * p < .01.
Second, the correlations between adolescent perceptions of their mother on a given dimension and how the mother thinks the adolescent perceives her on that dimension are generally low and nonsignificant (ns). For example, the correlation between how accessible the adolescent says the mother is and how accessible the mother says she is was only 0.10 (ns). The correlation between the quality of the advice that the adolescent says the mother gives and how expert the mother thinks she is in the eyes of the adolescent was 0.13 (ns). The corresponding correlation for trustworthiness was 0.02 (ns). These results suggest a disconnect between mothers and adolescents with respect to the fundamental dimensions of communication.
Correspondence of reports of frequency of communication
We next examined the correspondence between parent and adolescent reports of the frequency of communication using confirmatory factor analysis. Specifically, we fit a 10-factor model to the 30 communication measures. Five of the latent variables were adolescent characterizations of communication, and five were mother characterizations of communication. The five latent variables within a group corresponded to communication frequency for the two domains for smoking (health consequences and social consequences) and the three domains for sexual intercourse (physical/health consequences, social consequences, and moral consequences). Each latent variable had three indicators, as described in the Method section. All latent variables were allowed to be correlated, and all measurement errors were uncorrelated. Although a statistically significant Bollen-Stine index of fit was observed (p < .001), all other indices of fit were satisfactory. The standardized RMR was 0.029, the CFI was 0.98, the RMSEA was 0.04 with 90% confidence intervals of 0.03 and 0.05, and the p value for the test of close fit was greater than .99. Inspection of modification indices and standardized residuals suggested no significant signs of ill-fit at a more specific level. No offending estimates were observed.
For both mothers and adolescents considered separately, the estimated correlation between the two latent variables for talking about smoking cigarettes was highly correlated (r= 0.96 for adolescents and r= 0.97 for mothers). These results suggest that mothers who talk with their children about the physical and health consequences of smoking cigarettes also are likely to talk with their children about the social consequences of smoking cigarettes. These two latent variables were collapsed into a single latent variable in later analyses. Because the adolescent latent variables tended to be only weakly correlated with the mother latent variables (average r= 0.11), separate latent constructs were maintained for mothers and adolescents.
With respect to the topic of sexual intercourse, for both mothers and adolescents separately, the estimated correlations between the latent variables for discussions about the moral consequences and the latent variables for discussions about the social consequences of sexual intercourse tended to be highly correlated (r= 0.98 for adolescents and r= 0.96 for mothers), suggesting that mothers who talk with their children about the social consequences of sexual intercourse also tend to talk with them about the moral consequences. These variables were collapsed into a single latent variable in later analyses. In addition, both of these latent variables tended to correlate with communication about the health and physical consequences of sexual intercourse but to a lesser extent (average r= 0.59). These results suggest mothers who talk with their children about the moral and social consequences of sexual intercourse also are likely to talk with their children about the physical and health consequences of sexual intercourse. Because the latent variables tended to be weakly correlated across sources (average r= 0.18), separate latent constructs were maintained for mothers and adolescents. For reasons discussed later, we did not collapse the three communication factors for sexual intercourse into a single latent variable, but instead we maintained distinctions between talking about the physical and health consequences of sexual intercourse on the one hand and talking about the moral and social consequences of sexual intercourse on the other hand.
The refitted model with collapsed factors still provided good fit to the data. Although a statistically significant Bollen-Stine index of fit was observed (p < .001), all other indices of fit were satisfactory. The standardized RMR was 0.029, the CFI of 0.97, the RMSEA of 0.04 with 90% confidence intervals of 0.03 and 0.05, and the p value for the test of close fit was greater than .98. Inspection of modification indices and standardized residuals suggested no signs of ill-fit. The one exception was a large modification index (49.2) for correlated error focused on mother reports of communication about sexual intercourse for the moral/social factor item about getting a bad reputation and a second item loading on the same factor about losing the respect of one’s boyfriend or girlfriend. Because this was an isolated result among a large number of modification indices, we chose to view it as a chance effect and we did not free up the parameter. All conclusions we make are unaffected by whether this parameter is freed.
The most notable result in the refitted model was the generally weak correlations between adolescent reports of the frequency with which they had talked with their mother about a topic and mother reports of the frequency with which they had talked about those topics. For smoking cigarettes, the estimated correlation between frequency of talking was 0.13 (p < .01). For sexual intercourse, the estimated correlation for the physical/health consequences was 0.23 (p < .01), and for the social/moral consequences, it was 0.17 (p < .01). These results are consistent with a large corpus of past research that has observed weak correlations between adolescent and parent reports of the frequency of communication about topics related to risk behavior (Jaccard, Dittus, & Gordon, 1998; Miller, Kotchick, Dorsey, Forehand, & Ham, 1998). The standardized path coefficients from a latent construct to its observed indicators all were reasonable, ranging from 0.74 to 0.92.
The prediction of adolescent risk behavior
We next tested models that related the source dimensions to the frequency of communication and adolescent risk behavior. Figure 1 presents the structural model that guided the analysis for sexual behavior. To simplify matters, we estimated models separately for sexual intercourse and smoking cigarettes. The smoking model was identical in form to the sexual intercourse model, except that frequency of communication within a source was represented by a single latent variable (because of the high correlation between communicating about the health consequences of smoking and communicating about the social consequences of smoking, as noted above). Only structural relationships among the latent variables are shown in Figure 1. The measurement model is omitted, as are the correlations between latent exogenous variables. The expertise/trustworthy latent variable had six indicators (three for expertise and three for trustworthiness), reflecting the collapsing of the expertise and trustworthiness factors into a single latent variable. The availability/accessibility latent factor had three indicators. In the sexual intercourse model, the frequency of communication about pregnancy/AIDS latent variable had three indicators, and the frequency of communication about social and moral consequences had six indicators. For the smoking model, the frequency of communication latent variable had six indicators. The outcome variable in both models was a dichotomous variable, reflecting whether the adolescent had engaged in the behavior of interest. We also analyzed behavioral counts using Poisson regression analogs with continuous latent predictors.
Structural Relationships Among Source Variables, Communication Variables, and Adolescent Risk Behavior
Note: Correlations between latent variables are not shown but were estimated. The measurement model is not shown.
The model in Figure 1 assumes that the primary mechanism by which expertise, trustworthiness, and availability influence adolescent risk behavior is by increasing the frequency of communication about relevant topics between parent and child. If expertise and trustworthiness influence risk behavior independent of this mediator (e.g., by making messages more believable or acceptable), then statistically significant paths from the source dimensions to risk behavior should be evident, when holding frequency of communication constant. We tested this prediction in our model evaluations.
We estimated the models using the latent variable–modeling framework implemented in MPlus (Muthén, 2002). This framework permits the analysis of categorical and count outcomes in SEM. For the dichotomous outcome, estimation used a logit function to relate the latent continuous variables to the outcome. When the outcome was a count, estimation took the form of a zero-inflated Poisson regression model (see Long, 1997, for a conceptual discussion of such models). The analyses used a robust maximum likelihood criterion with numerical integration (see the MPlus documentation at http://www.statmodel.com). In all cases, model fit was good as indexed by the Akaike information criterion, the Bayesian information criterion (BIC), and a sample size–adjusted BIC relative to the independence model. In addition, there was an absence of offending estimates, and the parameter estimates were generally well behaved. We discuss each outcome, in turn.
Analysis of smoking cigarettes None of the parent-based latent variables were predictive of smoking behavior. These were dropped from the model, and analyses using only the adolescent measures were pursued. As expected, the logistic coefficient for predicting smoking behavior from the frequency of parent-adolescent communication was statistically significant (coefficient =−0.25, p < .05, exponent of coefficient = 0.78). The more that adolescents said they talked with their parents about the consequences of smoking, the less likely it was that the adolescent had smoked cigarettes. One of the source dimensions predicted smoking behavior over and above communication frequency. Specifically, the latent variable of perceived expertise/trustworthiness yielded a statistically significant path when it was added to the model in Figure 1. The logistic coefficient was −0.52 (p < .01, exponent of coefficient = 0.59). This indicates that, holding constant communication frequency, the more expert and trustworthy that adolescents perceived their parents as being, the lower were the odds that the adolescent smoked cigarettes in the past 30 days. This latter result is consistent with the idea that expertise/trustworthiness influences behavior independent of its effects on communication frequency.
Both perceived expertise/trustworthiness and perceived accessibility were predictive of the frequency of communication. The unstandardized regression coefficient for perceived expertise/trustworthiness was 0.72 (p < .01), and for perceived accessibility, it was 0.09 (p < .05). The metric of the latent predictors and latent criterion was a 1 – 5 strongly agree to strongly disagree scale. In general, higher levels of perceived expertise/trustworthiness and higher levels of perceived accessibility were associated with higher levels of communication frequency. When the behavioral outcome was a count representing the number of days in the month the adolescent reported smoking cigarettes, the results were comparable to those with the binary outcome, except the communication frequency measure was only a marginally significant predictor of risk behavior.
Analysis of sexual behavior The same dynamics described above for smoking cigarettes were evident for sexual intercourse, with some qualifications. As in the previous analysis, none of the parent latent variables were predictive of communication frequency or sexual risk behavior, and they were therefore dropped from the model. Recall that two dimensions of communication frequency were identified: one focused on discussions of pregnancy/AIDS and the other focused on discussions about the social/moral consequences of sexual activity. Only the latter was predictive of the odds of engaging in sexual intercourse (logistic coefficient =−0.58, p < .01, exponent of coefficient = 0.56). None of the source dimensions were predictive of sexual behavior independent of communication frequency.
Attributions of expertise/trustworthiness to the parent were significantly related to the frequency of communication about social/moral consequences (path coefficient = 0.61, p < .01). Perceived availability was not significantly related to the frequency of communication. When the behavioral outcome was a count focused on the number of times the adolescent reported having sexual intercourse, the results were comparable to those with the binary outcome.
In sum, for both risk domains, the frequency of communication between parent and child about the risk behavior was predictive of risk activity. The perceived expertise/trustworthiness of the parent predicted the frequency of such communication, and in the case of smoking cigarettes, it had independent effects on risk behavior over and above communication frequency. Perceived accessibility was related to the frequency of communication for cigarette smoking but not for sexual intercourse.
Reverse causality and analysis of methodological artifacts The above structural models assume that the direction of causality is such that increased communication leads to lowered adolescent risk behavior. It could be argued, however, that the causal direction is the reverse. That is, if parents learn or suspect that their adolescent child is engaging in sex or smoking, then this might motivate parents to talk with their child more about the issues. This logic predicts a positive association between frequency of communication and risk behavior, whereas our framework predicts a negative association. The results showed a negative association, which is consistent with a model that assumes communication is a protective rather than reactive factor.
Another possibility for reverse causality is when parents suspect that their children are engaging in risk activity, which can create conflict between parent and child. This conflict, in turn, can lead to lowered trust and perceived expertise by the adolescent. The negative association between expertise/trustworthiness and risk activity could be because of the influence of conflict resulting from parents learning of adolescent risk activity on adolescent perceptions of trustworthiness and expertise. We conducted two analyses to pursue this hypothesis. First, we asked parents if they thought their child had engaged in sexual intercourse (or smoked cigarettes) and then reran the models using only parents who did not think their child had done so. These parents were unlikely to be fighting with their child because of perceived involvement of the child with sex/smoking. The result of these tests replicated our original analyses with the total sample. Second, we estimated the models for the total sample but included an index of parent-adolescent conflict as a covariate in the models, thereby holding constant parent-adolescent conflict. The measure of conflict was the sum of ratings by mothers on a five-point strongly agree to strongly disagree scale in response to the following two items: (a) my daughter/son and I end up fighting when we talk and (b) no matter what I say, my daughter/son and I always seem to end up arguing. In these analyses as well as those described later, the covariate was partialled out by including a direct path from it to the two endogenous communication frequency variables as well as to the behavioral outcome. If the effects are a by-product of parent-adolescent conflict, then they should dissipate when conflict is statistically held constant. They did not. All conclusions derived from our initial model tests held up in these reanalyses.
Like many past studies, we found that adolescent characterizations of parent-adolescent communication were more predictive of risk behavior than parent characterizations of communication. Because the index of behavior was a self-report by the adolescent, it is natural to attribute the association to shared method variance. One source of shared method variance might be social desirability response tendencies that cause some adolescents to simultaneously inflate their characterizations of the frequency of communication with their parents and minimize their reports of sexual activity. We measured social desirability response tendencies using a scale based on the work of Paulhus (1991) and introduced the measure into our models as a covariate for predicting risk behavior from the communication items. All the statistically significant effects remained statistically significant, arguing against this interpretation.
Finally, we evaluated whether the fundamental relationships observed in our models were maintained when gender and age were included as covariates. This was indeed the case. We also tested if the effects would persist, holding adolescent relationship satisfaction with their mother constant as a covariate. Past research has shown that relationship satisfaction is an important predictor of adolescent risk behavior, and it is possible that our communication indices simply represent a proxy for it. All the paths that were statistically significant when adolescent relationship satisfaction was excluded from the analysis were also statistically significant when it was included as a covariate. The measure of relationship satisfaction was a single item responded to on a five-point strongly agree to strongly disagree statement “I like my relationship with my mother.” This single-item measure has been found to be highly correlated with more complex, multi-item measures of relationship satisfaction (Dittus & Jaccard, 2000).
Discussion
The present research relied on the general theoretical framework suggested by Jaccard et al. (2002) as a basis for analyzing communications between parents and adolescent children. The strength of the framework is that it brings to bear perspectives on attitude change and persuasion from social psychology and makes available a large corpus of knowledge from existing social psychological research to the analysis of parent-adolescent communication about adolescent risk behavior. The present research focused on source characteristics and explored three core dimensions of a source that are likely to influence behavior: (a) the perceived expertise of the source, (b) the perceived trustworthiness of the source, and (c) the accessibility or availability of the source. Numerous interesting results were observed that have both theoretical and practical implications.
Distinctions between expertise and trustworthiness
One interesting result was the finding that adolescents’ attributions of expertise on the part of their mothers were highly correlated with attributions of trustworthiness. If a mother was seen as being trustworthy, she also was seen as having expertise. This correlation raises questions about the need to distinguish the two constructs. On a conceptual level, the constructs are clearly different. One can know a great deal about a topic yet still not be honest, objective, and unselfish when laying out arguments with respect to one’s position. Research in social psychology suggests that the determinants of perceived expertise may not be the same as the determinants of perceived trustworthiness and that expertise and trustworthiness can, under certain conditions, affect message effectiveness differentially. This said, it appears to be the case that parents who are good at establishing credibility with their adolescent children also are good at establishing a trusting relationship. Although it may be unnecessary to distinguish the constructs in certain research contexts, we believe it is important to maintain distinctions at the conceptual and applied levels. For example, when developing intervention programs aimed at parents, it will be useful to discuss with parents not only strategies for giving high-quality advice but also strategies for developing trusting relationships. To assume that because the relevant skills co-occur in natural settings, one can teach one type of skill and ignore the other would, in our opinion, be a mistake.
Parent-adolescence correspondence in perceptions
Another important finding in the present research was that parent perceptions of how much credibility, trust, and accessibility they think they have established with their adolescents bear only a weak relationship to adolescent characterizations of parent credibility, trust, and accessibility. The findings suggest that parents who need to work on establishing these desirable facets may think they have already done so when this is not the case. Parents who think they are accessible and available to their adolescents but whose children do not agree may need help either in structuring their time for their children more effectively or, alternatively, in letting their children know about their availability.
It is possible that parents provide good advice to their adolescents but that the adolescents do not like what they hear from their parents because it is counter to their own wishes and desires. The result may be some loss in credibility on the part of the parent in the eyes of the adolescent. This might explain the tendency for adolescents to rate parents lower than parents rate themselves on perceived expertise and trustworthiness. We believe, however, if parents present their advice in a respectful and empathic way that engenders trust and understanding on the part of the adolescent, then losses in credibility as a result of providing counterdesired advice and information can be minimized.
Source dimensions as predictors of adolescent risk behavior
We posited two mechanisms by which perceived expertise, trustworthiness, and availability may predict adolescent behavior. Our study found support for both mechanisms, namely, (a) that higher levels of perceived expertise, trustworthiness, and availability are associated with increases in the frequency of communication between parent and child and (b) that higher levels of perceived expertise and trustworthiness have independent associations with risk behavior holding frequency of communication constant, perhaps because such expertise/trustworthiness enhances the likelihood that the adolescent will accept the messages conveyed by parents. Both mechanisms tended to be evident for smoking behavior, whereas only the first was evident for sexual behavior. The analysis of sexual activity included one more mediator than smoking behavior (namely, communication about moral consequences), and perhaps this is a source of the differential pattern of results. Future research needs to replicate this result and, if replicated, explore why such topic differences exist.
Interestingly, only adolescent as opposed to parental attributions of expertise/trustworthiness and accessibility were predictive of adolescent risk behavior. These results are consistent with a large number of studies in other domains that show how adolescents perceive their environment is more important in influencing their behavior than the structure of that environment per se or how their parents perceive that environment (e.g., Dittus & Jaccard, 2000; Jaccard et al., 1998). Although some studies suggest independent effects of environmental and parental orientations (Jaccard et al., 1998), the subjective world of the adolescent has been a consistent predictor of adolescent risk behavior. Supplementary analyses we conducted as well as those from past research suggest that these results cannot be attributed to method variance in adolescent self-reports as reflected by social desirability response tendencies.
Parent-adolescent correspondence in communication frequency
Our research replicated previous studies that have found generally low correlations between adolescent and parent characterizations of the frequency of communication about risk behaviors (Jaccard et al., 1998; Miller et al., 1998). In our opinion, one report is not necessarily more accurate than the other in documenting the true levels of communication that have transpired. Parent reports probably reflect all communication attempts that have been made by the parent, whereas adolescent reports probably reflect only those communication attempts that have actually registered with the adolescent. Retrieval of communication episodes from memory are likely to be different for parents and adolescents, given the diverse ways in which the episodes are stored and encoded in long-term memory, the available cues for memory retrieval, and the different motives and desires that can influence the retrieval process. In addition, how parents define a discussion may differ from how an adolescent defines a discussion. A fairly consistent finding in this area, however, as well as the current study, is that adolescent characterizations of the frequency of communication are predictive of risk behavior relative to parent characterizations.
An interesting applied result in the current study was the finding that parent-adolescent discussions about pregnancy consequences and STDs were not significantly associated with adolescent risk behavior. Instead, parent-adolescent discussions about the moral and social consequences of sexual intercourse were predictive. It is possible that adolescents already are well exposed to and knowledgeable about the consequences of pregnancy and HIV/AIDS, given the widespread attention these topics receive in the media and in sex education classes. By contrast, the social and moral consequences of early sexual activity receive less attention, so that parents who discuss these topics are providing their adolescents with a broader base of motivations for postponing sex. This, in turn, leads to lessened risk behavior. Future research should explore these issues.
Additional research directions
Additional theories The present research used the general attitude change and persuasion framework described by Jaccard et al. (2002) as drawn from social psychology to analyze parent-adolescent communication. More fine-grained models of persuasion exist that also can be applied, including the elaboration likelihood model (Petty & Cacioppo, 1981, 1986; Petty & Wegener, 1999) and the heuristic-systematic model (Chaiken, Liberman, & Eagly, 1989; Chen & Chaiken, 1999). The elaboration likelihood model emphasizes the cognitive processing of message content and the motivational factors that are likely to influence such processing. When recipients are highly motivated to process message content, then the characteristics of the message itself (e.g., the quality of the arguments) should affect attitude change. When the motivation to process message content is low, however, then more peripheral cues become relevant, such as the expertise and trustworthiness of the source and the recipient’s mood states (Petty & Wegener, 1998). The elaboration likelihood model also describes factors that can produce bias in the processing of message content per se. The heuristic-systematic model is similar but emphasizes the use of simple heuristics in the processing of messages and posits a dual-processing approach to information processing. Future research on parent-adolescent communication that applies the framework of Jaccard et al. (2002), as well as these more fine-grained theories (see also Johnson et al., 2004), will probably yield important insights into how communications from parents influence adolescent risk behavior.
General versus topic-specific expertise and trustworthiness The present study measured perceived expertise and trustworthiness on a general level without regard to topic area. It is possible that adolescents will perceive different levels of expertise and trustworthiness depending on the content domain in question. A parent who is seen as having useful advice and perspectives about sexual intercourse may not be seen as having much to offer for issues about drug use. Even within a content domain, differential levels of expertise may be perceived, such as with respect to the topics of condoms versus abstinence in the sexual domain. Considerable literature in social psychology suggests predictors that are more directly tied to a behavior will be more predictive of that behavior than general measures that focus on global attitudes or global cognitions (e.g., Ajzen & Fishbein, 1980; Jaccard, King, & Pomazal, 1977). An alternative approach that might prove fruitful is to measure perceptions of the three core source domains at a topic-specific level. We are currently developing such measures. Topic-specific measures will permit researchers to explore the extent to which attributions of expertise, trustworthiness, and availability in one domain generalize across other domains. We suspect some generality but domain-specific variance as well.
Subdimensions of expertise, trustworthiness, and availability A secondary goal of the present research was to develop brief measures of expertise, trustworthiness, and availability that could be used in applied research settings where interview time is limited. It should be possible to develop more elaborate measures that explore in greater depth different subdimensions of each source facet. For example, for trustworthiness, one can make conceptual distinctions between (a) being honest, (b) respecting the word and promises of the adolescent, and (c) having the best interests of the adolescent in mind. A multi-item scale exploring each of these facets of trustworthiness could be developed. In addition, one can examine each subdimension from four different perspectives: (a) how parents perceive themselves on the dimensions, (b) how parents think adolescents perceive them on the dimensions, (c) how adolescents perceive their parents on the dimensions, and (d) how adolescents think their parents perceive them on the dimensions. Such measures could explore the relevance of attributional dynamics in much greater depth than was done in the present research.
Caveats
Like any study, the present research must be interpreted within the context of its methodological limitations. The indices of adolescent risk behavior relied on self-reports and may be subject to some degree of bias and measurement error. The presence of measurement error, if not modeled correctly, can bias parameter estimates, thereby requiring caution in interpretation. Our results were correlational in nature and, of course, do not permit unambiguous causal attributions. We were able to conduct analyses that called into question issues of reverse causation, and this increases our confidence in the final models we settled upon. Specification errors can bias parameter estimates and must be taken into account. The study focused on an inner-city population of Latinos and African American middle school youth. Generalizations to other populations should be made with care and restraint. This is also true of the parent figure in the study, namely, mothers. The results should not be generalized to fathers and father-adolescent communication. For sexual activity, we focused on only a single type of sexual behavior, namely, vaginal sexual intercourse. Because of the sensitivity of this topic, the age of the adolescents, and the wishes of school administrators, we had to minimize the number of questions we asked about actual sexual behavior. Most other relevant sexual behaviors (e.g., number of partners, anal sex, condom use) have base rate problems in this population, so studying them is not feasible. This is not necessarily the case for oral sex, however. One should not, therefore, generalize our results to include oral sex because parental communication may relate differently to oral sex than to vaginal intercourse as a risk behavior.
Despite these limitations, we believe the current research provides insights to understand better the dynamics of parent-adolescent communication and sets the stage for more theoretically rich research in this domain. We have identified three core dimensions of source characteristics (perceived expertise, trustworthiness, and availability) and linked them to communication frequency and adolescent risk behavior in two different content domains. We found that parents’ characterizations of themselves on these three dimensions were only modestly related to adolescent characterizations of their parents, and we developed some of the applied implications of such disparities. Finally, we found that adolescent perceptions rather than parental perceptions of the dimensions were most critical in predicting risk behavior.
Note
This research was supported by funding from the Centers for Disease Control and Prevention Cooperative Agreement U87/CCU220155-3-0. The findings and conclusions in this paper are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
References
Citing Literature
- SupportingSupporting5
- MentioningMentioning127
- ContrastingContrasting1
