| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
a Department of Human Development, University of Maryland, College Park, MD
Correspondence: Gregory C. Smith, EdD, Department of Human Development, University of Maryland, College Park, MD 20742. E-mail: gs80{at}umail.umd.edu.
Decision Editor: Vernon L. Greene, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
, comparative fit 
, root mean square residual 
, standardized 
, root mean square error of 
) after a negative direct path from age to generalized beliefs in primary control was added to the a priori model.
Key Words: Adult development • Social support • Successful aging
Social support and personal control are widely recognized by clinicians and researchers as perhaps the two most important predictors of morbidity, mortality, and psychological well-being in adulthood (for reviews, see Lachman, Ziff, and Spiro 1994
; Rowe and Kahn 1987). Yet, the interrelationship between these two constructs remains largely unexplored despite numerous assertions that they are clearly related (Bandura 1997; Bisconti and Bergeman 1999; Hansson and Carpenter 1994; Pierce, Saranson, Saranson, Joseph, and Henderson 1996; Rowe and Kahn 1987; Ryan and Solky 1996; Syme 1990; Zautra, Reich, and Newsom 1995). As Rowe and Kahn recommended over a decade ago, however, the "factors that influence successful aging should be studied in interdependent combinations as well as singly" (p. 149).
In this article, we first describe several conceptual shortcomings that may have discouraged previous investigations of the association between personal control and social support. We then link these two constructs theoretically in a structural model where generalized or global beliefs in personal control, presumed to be the direct antecedent of psychological well-being, are derived from two distinct sources: (a) personal agency, or achieving desired outcomes on one's own behalf (e.g., through ability, choices, perseverance, or planning), and (b) interpersonal agency, or obtaining positive ends through interactions with others (e.g., by expressing needs or behaving cooperatively). We believe age affects both forms of agency indirectly as a result of age-related changes in physical health and social support. Not only does the model include a developmental perspective on how age-related changes in adulthood may influence beliefs in personal control, it also holds that the widely documented relationship between social support, specifically emotional support, and psychological well-being is mediated by interpersonal agency and generalized beliefs in personal control.
Bandura 1997 recently described two weaknesses in the personal control literature that may help explain why past researchers have so often failed to link this construct with social support. One problem is that most measures of control beliefs simply ask respondents whether or not they can attain desired outcomes, without regard for the means used to do so. Yet, as Bandura noted, "When people are asked, without suggesting any means, whether they can attain desired outcomes, they undoubtedly consider on their own the means they have at their disposal in making their judgments" (p. 28). An obvious mandate for researchers, then, is to explore the means used by individuals to achieve desired outcomes. In this context, the term agency refers to whether personal control is derived from one's own efforts or from sources outside of oneself (Shapiro 1994).
Bandura 1997 additionally noted that even when conceptions of agent causality are considered, investigators have typically emphasized individual agency while ignoring the fact that people often work together to produce desired results. In response to this issue, he further suggested two principal types of agency by which people exercise control to achieve desired ends: (a) direct personal control, where they mobilize skills and resources on their own, and (b) socially mediated proxy control, where they try to get those who wield influence and power to act on their behalf. In Bandura's words, "Although the alternative forms of control require different types of means, both are agentive" (p. 28).
Similar to Bandura 1997 concepts of direct personal control and proxy control, Schulz and Heckhausen 1996 introduced the "agentic" constructs of selective primary control (i.e., the focused investment of resources such as effort, time, and ability toward a chosen goal) and compensatory primary control (i.e., the use of external resources such as assistance from others or technical aids). Unfortunately, however, no published studies in the control literature have explicitly dealt with either the assessment of these various agentic constructs or their possible linkages with social support and psychological well-being.
Limitations also exist within the social support literature that may have discouraged past attempts to investigate the relationship between personal control and social support. The implicit model in social support research, for example, is that providers perceive a need for support on the part of a potential recipient and then decide to give support to that person or not. This view fails to acknowledge, however, that an individual may ponder seeking support from several potential providers before seeking support from one of them or that he or she may choose not to seek support from any of them (for discussion, see Pierce et al. 1996). In contrast, a view of individuals as active seekers of support is not only consistent with agentic constructs like proxy control and compensatory primary control, it also represents an important step toward the theoretical marriage of personal control with social support.
Another obstacle within the social support literature concerns the tendency to equate autonomy with independence, which leads to the notion that the need for relatedness is in opposition to the need for autonomy and personal control (for discussion, see Ryan and Solky 1996). Although researchers in the field of child development have long recognized that supportive relationships are either autonomy enhancing or autonomy reducing, this distinction is illusive in adult development and aging research (Rowe and Kahn 1987). Instead, the tendency there is to equate older adults' seeking of help in areas of functional limitations with dependence on others, even when such aid is solicited to preserve their autonomy (Bandura 1997). Yet, as Ryan and Solky noted, "In ideal relationships one need not feel diminished by being dependent, but instead may be encouraged and vitalized by responsive, facilitative, and caring others" (p. 264). Moreover, no empirical evidence suggests that perceptions of the effectiveness of others operating on one's behalf implies relinquishment of personal control (Skinner 1995).
We believe that the conceptual problems discussed in the preceding paragraphs are countered in the proposed model (Fig. 1) where global beliefs in personal control are presumed to result from two distinct but related sources: (a) personal agency, which is similar to both direct personal control and selective primary control, and (b) interpersonal agency, which is akin to both proxy control and compensatory primary control. We use the term agency rather than control, however, to more clearly denote the causal sequence in which greater use of both personal and interpersonal efforts (or agency) are hypothesized to result in higher levels of generalized control beliefs. Despite differences in nomenclature, however, this model is consistent with Bandura 1997 assertion that "a control event consists of an agent using certain means to produce performance attainments that give rise to various outcomes" (p. 24).
|
more narrowly defined proxy control as attempts to get those who wield influence and power to effect desired changes through such tactics as persuasion and social coercion. The primary difference between interpersonal agency and compensatory primary control (Schulz and Heckhausen 1996), on the other hand, is that the latter includes the use of technical aids (e.g., wheelchairs and hearing aids) to achieve one's ends, whereas interpersonal agency solely encompasses human relationships.
A key feature of the proposed model is that we hypothesize global beliefs in primary control to exert a positive direct effect on psychological well-being. Although the concept of control has been examined from diverse research traditions (e.g., locus of control, self-efficacy, learned helplessness) and defined in numerous ways (for reviews, see Bandura 1997; Rodin 1990; Rodin, Timko, and Harris 1985; Thompson and Spacapan 1991), the literature has repeatedly shown that greater personal control is associated with better emotional health regardless of how it is operationalized (Lachman et al. 1994; Rodin 1987; Skinner 1995; Thompson and Spacapan 1991). Moreover, results of control-enhancing experimental studies have revealed that the chain of causality is from increased personal control to higher levels of psychological well-being (for review, see Rodin et al. 1985). Thus, the proposed direct effect of primary control beliefs on psychological well-being is well substantiated in the literature.
There are two chief reasons why we selected perceived global belief in primary control, which refers to beliefs that one can obtain desired outcomes through one's own actions and initiatives, as the particular focus of the proposed model. First, primary control (i.e., direct actions on the environment to be in control vs cognitive adjustments to have one's emotions under control) is believed to be the most crucial element of successful adaptation across the life span (Schulz and Heckhausen 1996). Second, the mere perception or belief that one has control is as important to psychological well-being as objective control (Bandura 1997; Rodin 1990; Shapiro 1994).
Another salient aspect of the proposed model involves the hypothesized causal sequence in which emotional support exerts indirect effects on both global beliefs in primary control and psychological well-being through its direct effect on interpersonal agency. Although it is well established that increased social support acts as a buffer against psychological distress and that support is linked with positive mental health outcomes (Chappell 1995; Rowe and Kahn 1987; Ryan and Solky 1996), the underlying process by which social support leads to such benefits is not well understood (Antonucci 1990). One possibility, as indicated in our model, is that having emotionally supportive relationships leads to a heightened sense of interpersonal agency. In turn, this contributes to generalized feelings of primary control, which then fosters psychological well-being.
We included emotional social support rather than instrumental support in the proposed model as an antecedent of interpersonal agency because emotional support is thought to be the core element of supportive relationships (Krause 1987; Ryan and Solky 1996). Consistent with this view, Krause found that emotional support and integration influenced older adults' feelings of control, whereas tangible help and information from others did not. It is also believed that instrumental support is more likely than emotional support to jeopardize feelings of personal control (Grams and Albee 1995; Hansson and Carpenter 1994; Rowe and Kahn 1998). This is because tangible help from others may remind the recipient of personal inadequacies and make it difficult to take credit for desired outcomes (Hansson and Carpenter 1994).
The hypothesized relationship between emotional support and interpersonal agency has additional empirical and theoretical justification. Cicirelli 1980, for example, found that those older people who felt the closest to supportive family members most likely perceived that their own actions produced desired consequences in dealing with that family member. Pearlin 1985 maintained that effective support seeking is associated with relationships characterized by intimate exchange and communication that goes beyond superficial levels. In like manner, Ryan and Solky 1996 concluded from their extensive review of relevant research that an individual's willingness to rely on others for tangible support depends on whether or not potential providers are emotionally supportive. Thus, it is sensible to predict that those adults reporting greater levels of emotional support will be more inclined to interact with others to achieve their goals.
It is noteworthy that a direct effect of emotional support on psychological well-being is deliberately absent in the proposed model, despite the vast literature suggesting a strong positive relationship between these two variables (Rowe and Kahn 1987; Ryan and Solky 1996). This omission reflects our belief that the extent to which individuals feel that they can elicit support from others when desired is the critical determinant of whether social support contributes to psychological well-being. This is reflected in the proposed model where we hypothesized that interpersonal agency and global beliefs in primary control would mediate the relationship between emotional support and psychological well-being.
The findings of two prior studies substantiate the logic of this hypothesis. Bohm 1983, for example, found that other salient features of older adults' supportive relationships (e.g., frequency and scope of contacts, as well as such qualitative characteristics as reciprocity, intimacy, and stability) did not make significant additional contributions to predicting psychological well-being in a hierarchical regression analysis once perceptions of interpersonal efficacy and generalized control beliefs emerged as significant predictors. More recently, Bisconti and Bergeman 1999 likewise found that perceived control over the elicitation of social support mediated the relationship between diverse measures of social support and the outcomes of psychological well-being and physical health. Not only was the rival hypothesis of perceived social control as a moderator of the support–outcome relationship ruled out by these investigators, the same mediational pattern was observed across two independent samples of older adults. Bisconti and Bergeman concluded that "increasing a person's sense of control in social relationships may actually be the underlying factor that produces better psychological and physical outcomes in old age" (p. 101).
Despite their important contributions toward elucidating the relationship between social support, beliefs in personal control, and overall well-being, neither Bohm 1983 nor Bisconti and Bergeman 1999 distinguished perceived control (or agency) within social situations from control (or agency) without the involvement of others. This distinction is important because of the possibility that perceived social control may have been a proxy for more generalized beliefs in personal control in these studies. In contrast, our model explicitly addresses this issue by designating personal and interpersonal agency as distinct constructs that arise from separate antecedent variables.
Another unique feature of the proposed model is that it addresses the impact of normal age-related changes on generalized control beliefs and psychological well-being. Specifically, we hypothesized that age would exert negative indirect effects on both personal agency and interpersonal agency as a result of age-related changes in self-reported health and emotional support. The causal sequence involving age, self-reported health, and personal agency, for instance, is based on the fact that normal age-related changes in physical health become increasingly prevalent toward the end of middle age and that these changes yield constraints that adversely affect one's ability to exercise direct control (Heckhausen and Schulz 1995; Rodin et al. 1985; Zautra et al. 1995). We additionally hypothesized a positive direct effect of self-reported health on psychological well-being, given the vast body of research showing that physical health is a major predictor of subjective well-being in adulthood (for review, see Larson 1978).
The hypothesized causal sequence involving age, emotional support, and interpersonal agency is based on the belief that a steady decline in social activity and contacts starts to occur in middle age (Carstensen 1991; Pearlin and Skaff 1995), which then affects the availability of support from others. Pearlin 1985, for example, maintained that the diverse needs experienced by an individual require specialized kinds of support from specialized sources to the extent that "the greater level of attachment to and interaction with membership groups, the greater is the likelihood that they will provide the most fertile harvest of supports of various kinds" (p. 45). Conversely, we hypothesized that as social contacts shrink in the later years, so too does the availability of specialized sources of social support. In turn, we expected that as these various forms of support become less accessible to call on with advanced age, the ability to use interpersonal agency would diminish as well.
In summary, we hypothesized that greater amounts of perceived primary control and higher levels of self-reported health would exert positive direct effects on psychological well-being. In turn, we thought that global beliefs in primary control stem from two related but distinct means: (a) interpersonal agency, which is associated with increased emotional support, and (b) personal agency, which is enhanced by better physical health. Finally, we hypothesized that increased age would lead to both less support and reduced levels of self-reported health. The proposed model not only delineates a causal relationship between emotional support and beliefs in personal control, but also encompasses a developmental perspective in which normative age-related changes are regarded as antecedents of change in perceived control.
|
Methods |
|---|
|
TOPAbstractMethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
years) from the Greater Baltimore–Washington Metropolitan area volunteered to participate in this study. Participants were recruited in a convenience sample from a wide variety of settings (e.g., senior citizen centers, senior apartments, colleges and universities, social service agencies, clubs, organizations, shopping malls, and religious groups) in order to obtain a broad representation of people from diverse backgrounds and age groups. Table 1 presents the resulting demographic characteristics of this sample by age group. It can be seen that the majority of participants were female (73.2%) and White (73.4%) and possessed high levels of educational attainment. Nevertheless, the sample was quite diverse with respect to age group and marital status. Although a high percentage of the entire sample (38.4%) indicated that they had never been married, the vast majority of these individuals (37.3%) were under age 35. All participants resided in the community at large or in retirement communities of a noninstitutional nature (e.g., Leisure World). The small number of participants in the "other" category of living arrangements (4.1%) consisted of college students in dormitories and older adults from senior citizen apartments. Although a total of 538 individuals actually participated, 56 of these were dropped from the study because they resided in institutional settings.
|
Measures
Table 2 shows the zero-order correlation matrix, means, standard deviations, observed ranges, and averaged reliability estimates for the indicator variables used to test the model.
|
), which consists of the following three items: "How have you been feeling in general in the last month?" "How happy, satisfied, or pleased have you been with your personal life in the past month?" and "Has your daily life been full of things that were interesting to you in the past month?" Respondents rated each item on a 6-point Likert-type scale. For example, the first item was rated on a scale ranging from 1 (in very low spirits) to 6 (in excellent spirits). Responses to all three items were then summed, and a mean score was calculated for each respondent.
McDowell and Newell 1996 reported extensive evidence of this instrument's validity and reliability, including multitrait and factor analyses of the entire PGWB, which confirmed the existence of the Positive Well-Being subscale. Internal consistency coefficients for the six PGWB subscales were reported by these authors to range from .72 to .88, and a 1-week test–retest reliability estimate for the Positive Well-Being subscale was found to be .74. Cronbach's alpha (.79) and split-half (.74) reliability values obtained with the present sample were acceptable.
Global belief in primary control was assessed by two items measuring control over positive outcomes from the Pearlin and Schooler 1978 Mastery Scale (i.e., "I can do just about anything that I set my mind to do" and "What happens to me in the future mostly depends on me"). Both items were rated along a Likert-type scale ranging from 1 (strongly disagree) to 4 (strongly agree), and a mean score on the two items was then calculated for each respondent. The low Cronbach's alpha (.46) and split-half reliability values (.46) observed with the present sample are likely attributable to the brevity of this measure.
Although we originally intended to use the entire seven-item version of the Mastery Scale as an indicator of global belief primary control, work by Zautra and colleagues 1995 eventually led us to use the two-item version instead. Their confirmatory factor analysis of the entire Mastery Scale with a sample of 269 older adults revealed a two-factor solution. One factor, labeled Control Over Positive Outcomes, consisted of the two items used in the present study. The second factor, labeled Fatalism, included the remaining five items of the Mastery Scale related to control over negative life events (e.g., "There is really no way that I can solve some of the problems I have").
Even though close inspection of the two Control Over Positive Outcome items suggests that they may be tapping positive, negative, or neutral outcomes, correlational analyses by Zautra and colleagues 1995 involving both factors on the Mastery Scale revealed "that control over positive events appears to have a distinctive effect in the preservation of health and well-being among older adults, not accounted for by assessments of control over negative life experiences" (p. 162). These findings suggest that people have different conceptualizations of their control of positive versus negative life experiences. Thus, to avoid confounding these two aspects of perceived control, we used only the two items from the Mastery Scale purported by Zautra and colleagues to measure positive control.
Agency Beliefs
We measured personal and interpersonal agency beliefs using the items shown in the Appendixthat were derived from a larger pool of items that we developed by (a) reviewing extant theories and measures of personal control, (b) eliciting input from relevant professionals and lay adults, and (c) pilot testing. In prior analyses of these items, we (Smith et al. 1999) first performed an exploratory factor analysis (EFA) on the initial pool of items that yielded the expected two-factor solution. We then performed confirmatory factor analyses (CFA) that (a) confirmed the initial EFA findings, (b) revealed that the correlation between the two factors was small (i.e., 
), and (c) demonstrated factorial invariance across gender and age. Three of the items purported to tap interpersonal agency, however, cross-loaded with the personal agency factor in the CFA and were thus dropped. The resulting items on the Interpersonal Agency and Personal Agency scales are shown in the Appendix.
In the present study, respondents were asked to rate each of the eight items tapping personal agency and the five items tapping interpersonal agency along a Likert-type scale ranging from 1 (never) to 4 (often). Mean scores on both the personal agency and interpersonal agency measures were then calculated for each respondent. Cronbach's alpha and split-half reliability values, respectively, were acceptable for both the Interpersonal Agency (.76 and .70) and Personal Agency (.78 and .73) scales.
It should be noted that these two measures of agency beliefs are similar to the measure of mastery used in this study to the extent that all three measures tap aspects of active control beliefs. Yet, there are important conceptual differences between the agency and the mastery constructs that are reflected in the precise wording of items on these respective measures.
Mastery has been defined as a self-reported individual difference variable that refers to the generalized belief that people have regarding their ability to exercise control over important circumstances of their lives (Pearlin and Schooler 1978). In this sense, mastery (e.g., "What happens to me in the future mostly depends on me") is equivalent to having an internal locus of control because both constructs reflect the individual's global belief that what happens to him or her is largely the result of one's own behaviors (Smits, Deeg, and Bosscher 1995).
In contrast, agency refers to specific acts that are performed intentionally to achieve desired ends (Bandura 1997). We defined personal agency in this study as the use of one's own efforts and abilities to achieve desired consequences (e.g., "Careful planning enables me to get what I want or need") and interpersonal agency as interacting with others for this purpose (e.g., "I achieve my goals by knowing when to ask others for help"). Thus, mastery encompasses the general sense of having control over events in one's life, whereas agency involves specific strategies used to achieve such control.
The pattern of zero-order correlations that we observed between the individual items on the three scales suggests that each scale taps the underlying construct it is purported to measure, and none are tainted by the similarity in measuring active control beliefs. Specifically, the bivariate relationships revealed that correlations involving within-scale items were greater in magnitude than correlations involving between-scale items. Further substantiating the distinctiveness of these constructs is the fact that the respective zero-order correlations between overall scale scores (see Table 2 ) were also low in magnitude.
Emotional support was assessed by seven items from the Expressive Support Scale developed by Pearlin, Mullan, Semple, and Skaff 1990, which taps the perceived availability of a person who is caring, trustworthy, uplifting, and a confidant. Respondents rated each item (e.g., "I have a friend or relative in whose opinions I have confidence") along a Likert-type scale ranging from 1 (strongly disagree) to 4 (strongly agree). Ratings were reversed for one negatively worded item on this scale and then a mean score was calculated for each respondent. The Cronbach's alpha (.81) and split-half reliability (.81) estimates obtained with the present sample were acceptable.
Self-Reported Health.
Respondents rated their physical health status on a single-item Likert-type scale ranging from 1 (poor) to 5 (excellent). Numerous studies have shown subjective measures such as this to be robustly correlated with objective indices of physical health (for discussion, see George 1996).
|
Statistical Analyses |
|---|
|
TOPAbstractMethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
).
We used the maximum likelihood procedure of EQS 5.4 (Bentler 1996), which is fairly robust with respect to violations of non-normality due to skewness, to analyze the variance–covariance matrix and to estimate the parameters and goodness of fit. Because the chi-square statistic is sensitive to both the assumption of normality and sample size, we followed Bollen and Long 1993 recommendation to consider additional goodness-of-fit indices such as the goodness-of-fit index (GFI), the comparative fit index (CFI), the standardized root mean square residual (RMR), and the root mean square error of approximation (RMSEA).
Given the complexity of the proposed model (with five of the seven latent constructs measured by multiple indicators), the procedure recommended by Liang, Lawrence, Bennett, and Whitelaw 1990 to move from multiple indicators to composites while still incorporating error theory into model evaluation was undertaken as follows. First, we calculated coefficient alpha and split-half reliability estimates for each indicator variable. Second, we calculated the average of these two estimates to derive the proportion of observed score variance representing the unreliability of that measure. (For the single-item measures of self-reported health, where reliability estimates were unattainable, the average of the reliability estimates for all other variables [.71] was used instead. Perfect reliability [1.00] was assumed for chronological age. Third, we then derived the measurement error variance for each latent variable by multiplying the variance of its indicator times (1 - its average reliability estimate). Fourth, we used values obtained with the above formula to fix each latent construct's error variance rather than allowing them to be freely estimated by the EQS 5.4 program.
Although the reliability estimates for the two-item Mastery scale were low, using a structural equation model in which mastery is treated as an error-free latent construct is designed to overcome this limitation. Having relatively low reliability in a measured variable implies that its relations with other measured variables will be attenuated. Structural equation modeling, however, corrects for that attenuation and yields paths among theoretically error-free constructs that are unencumbered by the measurement error reflected in the scale's low reliability (Acock and Schumm 1993). Had the reliability of the Mastery scale been increased by adding more comparable quality items, the observed correlations with other measured variables would also increase because the mitigated measurement error would cause less attenuation. Still, the underlying latent relationships among constructs in the model would be unaffected in principle. Despite the benefits of structural equation modeling in accounting for measurement error, the stability of a construct from sample to sample is enhanced by having more items and higher reliability.
Although instances of missing data were rare, we used the recommended procedure of replacing missing values with means in the above analyses (Schumacker and Lomax 1996).
|
Results |
|---|
|
TOPAbstractMethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
is considered small (Cohen 1988), suggesting that they are related but distinct constructs. Third, chronological age correlated significantly with all other variables except for personal agency, but the magnitude of these correlations was small (i.e., 
to -.23).
An initial test of the a priori model with EQS 5.4 yielded a less-than-satisfactory fit to the data. The LaGrange multiplier tests results, however, suggested that the addition of one theoretically reasonable path (i.e., from age to perceived primary control) could better account for the data. Fig. 2 shows the results of the revised model, including the standardized path coefficients. Although the chi-square value remained high, 
p < .05, it showed a statistically significant drop (p < .001) from that observed for the a priori model. Three of the more stringent indices of model fit were good: The GFI (.98) and CFI (.94) met or exceeded the benchmark criteria of .90. Both the RMR (.02) and standardized RMR (.05) were at or below the acceptable value (.05). The RMSEA (.06) was within the acceptable range (
.06). The standardized path coefficients for each of the hypothesized direct effects in the revised model were statistically significant (p < .05). The revised model accounted for 44% of the variance in psychological well-being and 53% of the variance in perceived primary control.
|
|
|
Discussion |
|---|
|
TOPAbstractMethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
; Lachman et al. 1994; Rodin et al. 1985; Rowe and Kahn 1987; Schulz and Heckhausen 1996; Skinner 1995; Thompson and Spacapan 1991). We extend past findings, however, by identifying personal agency and interpersonal agency as two distinct means by which generalized control beliefs develop. This study also represents a formidable step away from the primarily descriptive approach in previous studies of identifying normative patterns of age-related change in perceived control (for discussion, see Lachman 1986) to one that accounts for individual differences by specifying the antecedents and consequences of such change.
This study obtained support for the construct validity of the scales used to measure personal and interpersonal agency. Not only did these constructs exert separate direct effects on subjective primary control as hypothesized, but they were also affected by the distinct antecedent variables specified in the model. Thus, future research is warranted to (a) develop more refined measures of these constructs, (b) identify their possible antecedents not included in the present model, and (c) compare them to related constructs such as compensatory primary control (Schulz and Heckhausen 1996) and proxy control (Bandura 1997).
The present findings also reinforce the growing belief that social support is an important contributor to feelings of personal autonomy and control across the life span (Hansson and Carpenter 1994; Krause 1987; Pierce et al. 1996; Ryan and Solky 1996; Syme 1990; Zautra et al. 1995). They are also consistent with the earlier studies by Bohm 1983 and Bisconti and Bergeman 1999, showing that beliefs in personal control mediate the relationship between social support and various indices of well-being. Thus, it appears that a research emphasis on discerning the factors that contribute to perceptions of control within interpersonal relationships will be more productive than identifying the specific features of relationships (e.g., satisfaction, intimacy, formal vs informal) that are correlated with well-being. Our findings also support Bisconti and Bergeman's conclusion that interventions designed to increase feelings of control in interpersonal relationships are likely to be more beneficial than efforts to increase the size or structure of one's social network. Several questions emerge, however, regarding the precise conditions under which supportive others might impart the highest levels of interpersonal agency.
One such question concerns the controversial issue of domain-specific versus generalized control beliefs that permeates the personal control literature. Although our model focused exclusively on generalized control beliefs, under the assumption that domain-specific beliefs predict fewer behaviors outside of that domain (Skinner 1995), it may be argued that the magnitude of correlations between model variables would be higher if domain-specific measures were used instead. Not only would it be meaningful in the future to test the model within the specific domains of functioning that researchers have already found are associated with age differences in personal control beliefs (e.g., health, cognition, finances), but it would also be important to examine how specific providers of social support (e.g., spouses, friends, adult children) may differentially affect interpersonal agency (Martire, Stephens, and Townsend 1998).
Another key question has to do with where the boundaries of control-enhancing interpersonal relationships begin and end. Stated differently, is there a threshold point beyond which continued support leads to dependence, enmeshment, and decreased feelings of agency (Krause 1987)? The complexity of this issue is revealed by the fact that even someone who relinquishes control to powerful others may perceive enough regularity in the actions of such people that he or she begins to believe that reinforcements are obtained through his or her purposeful action (Bandura 1997; Levenson 1981). The present findings also underscore an important distinction, especially in old age, between dependence on others and the seeking of help to preserve one's autonomy (Bandura 1997; Bohm 1983; Rodin et al. 1985).
A third relevant question involves the directionality of the relationship between emotional support and interpersonal agency as specified in the proposed model. In contrast to our conclusion that emotional support enables interpersonal agency, other researchers have argued just the opposite: that greater feelings of personal control facilitate the seeking of support (Eckenrode 1983; Lowenstein and Rosen 1995). It is also possible that a reciprocal relationship exists between these two variables (Bandura 1997; Pearlin and Skaff 1995). Yet, because all studies to date have involved cross-sectional designs, inferences regarding causality are inconclusive in the absence of longitudinal research.
It should be noted that feelings of efficacy in mobilizing support from others is a theme in our measure of interpersonal agency (e.g., "I get what I want or need by seeking the advice of others") that also exists in both Bohm 1983 measure of perceived efficacy in supportive relationships (e.g., "How satisfied are you with being able to find someone to shop for you if you were ill?") and Bisconti and Bergeman 1999 measure of perceived social control (e.g., "I find that if I ask my family (or friends) to visit me, they come"). This common focus on perceived efficacy is important because beliefs of personal efficacy are thought to be the key elements of human agency, to the extent that people will not attempt to make things happen if they feel that they do not have the power to produce the results they desire. It is also quite likely that having feelings of control over social support without corresponding feelings of efficacy may adversely affect one's psychological well-being. As Bandura 1997 noted, "Beliefs that outcomes are determined by one's own behavior can be either demoralizing or empowering, depending on whether or not one believes one can produce the required behavior" (p. 20). In this sense, demoralization occurs when people consider a desired outcome to be controllable, yet they cannot execute the means to achieve that end.
The transcending theme of perceived self-efficacy in these three studies also means that evidence for the role of perceived control as a mediator of the relationship between social support and emotional well-being is presently limited to this self-efficacious aspect of personal control. It might be, however, that other elements of control over social support such as choice or predictability also mediate this relationship. In fact, these latter types of control over social support may even be easier to facilitate across adulthood than feelings of efficacy or agency in the mobilization of support. Examples of this range from college students being able to choose academic advisors to nursing home residents selecting the nursing aide who will attend to their needs. Thus, future studies are needed to examine the relative importance of these different aspects of social control in mediating the support–well-being relationship.
Our findings also have important developmental implications. That chronological age had both direct and indirect effects on generalized perceptions of primary control is intriguing given the repeated finding in the literature that global measures of perceived control fail to reveal age differences where more sensitive domain-linked ones do (for discussion, see Bandura 1997; Lachman et al. 1994; Rodin et al. 1985; Schulz, Heckhausen, and Locher 1991). However, given the assertion by Rodin and colleagues that these past findings may be explained by the fact that age-related changes rather than age per se influence generalized control beliefs, we presumed chronological age to have only indirect effects on subjective primary control in our a priori model.
In retrospect, however, the direct pathway from chronological age to perceived primary control included in the revised model seems justified by the fact that so few studies have involved samples with enough variance to test hypotheses regarding age differences in global control beliefs rigorously (Schulz et al. 1991). Simply put, the extensive age variance within the present sample may have allowed the strong relation between age and perceived control to emerge that was absent in past studies. Nevertheless, the prevailing view of age as a surrogate variable cautions that additional mediators of the relationship between age and perceived primary control need to be specified in our model. In any event, the present findings reinforce the conclusion of Rodin and colleagues 1985 that "the elderly may be psychologically and physically more vulnerable to the negative effects of uncontrollability because of environmental and physiological changes that commonly occur in old age" (p. 14).
One caveat regarding the effects of age in the proposed model, however, concerns the hypothesized pathway from chronological age to emotional support. Although the rationale for including this pathway is that a decline in social activity and contact beginning in middle age (Carstensen 1991; Pearlin and Skaff 1995) diminishes the likelihood or availability of receiving diverse kinds of social support as interaction with membership groups declines (Pearlin 1985), the latent construct of network contact was not included in the model. Thus, a more thorough test of these assumptions would involve a revised model in which this construct presumably mediates the relationship between age and emotional support.
Additional shortcomings of the present study that restrict the generalizability of the findings should be noted as well. First, although the sample was large and included adults of diverse ages, it was non–randomly selected and skewed toward individuals who were White, female, and highly educated. Second, despite our use of structural equation modeling, inferences regarding causality were severely limited by cross-sectional data. The findings were also restricted by the use of only one indicator to represent each construct included in the proposed model. In particular, use of the two-item Mastery scale to measure global beliefs in primary control resulted in low reliability, which constrains generalizabilty across samples. Third, evidence for the construct validity of our measures of personal agency and interpersonal agency is limited to the findings of this study. Finally, all measures were obtained through self-reported surveys in the absence of more objective indicators of the latent constructs. Self-reports of agency and subjective control are not so problematic, however, as actual control may not be nearly as important as a person's perception of controllability (Bandura 1997; Rodin 1990). Despite these limitations, the present study sheds important new light on the potential causal relationships that exist between chronological age, social support, personal control, and psychological well-being in adulthood.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication May 19, 1999. Accepted for publication January 28, 2000.
|
Appendix ENDIX |
|---|
|
TOPAbstractMethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
Interpersonal Agency Scale Items
|
References |
|---|
|
TOPAbstractMethodsStatistical AnalysesResultsDiscussionAppendix ENDIXReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. D. Lincoln Financial Strain, Negative Interactions, and Mastery: Pathways to Mental Health Among Older African Americans Journal of Black Psychology, November 1, 2007; 33(4): 439 - 462. [Abstract] [PDF]
|
|
|
|
 |
|
 D. S. Bailis and J. G. Chipperfield Compensating for Losses in Perceived Personal Control Over Health: A Role for Collective Self-Esteem in Healthy Aging J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2002; 57(6): P531 - 539. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. S. Fry Predictors of Health-Related Quality of Life Perspectives, Self-Esteem, and Life Satisfactions of Older Adults Following Spousal Loss: An 18-Month Follow-up Study of Widows and Widowers Gerontologist, December 1, 2001; 41(6): 787 - 798. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
