Trust Matters: Distinction and Diversity in Undergraduate Science Education
by Robert K. Ream, James L. Lewis, Begoña Echeverria & Reba N. Page
Background: How do we account for the persistent difficulty the U.S. community of science has in educating larger numbers of talented and diverse undergraduates? We posit that the problem lies in the community’s unremitting focus on scientific subject matter knowledge and students’ ability to learn, to the neglect of interpersonal social relations—particularly trust. Our study focuses on trust in academic mentoring. This topic is particularly relevant in studies of the undergraduate segment of the pipeline into science that typically occurs in large, impersonal settings that often generate student alienation. Our study hypothesizes that trustworthy relations may be both a significant condition for academic mentoring as well as an important outcome.
Purpose: The study we report explores the promise of a new research direction in undergraduate science education, one focused on trust in academic mentoring. A focus on interpersonal trust is unusual, if not entirely overlooked, in most studies of undergraduate science education, and stands in sharp contrast to the situation in K–12 education where an emergent body of survey and field research measures trust and documents its educational importance. Our framework derives from social capital theory and cautions that a science curriculum that informs and motivates entails attention not only to cognition, as indicated by subject matter knowledge, but also to the educational significance of other complex forms of cognition that undergird social skill, relational awareness, and the development and maintenance of trust.
Research Design: Because the MARC-U*STAR training program funded by the National Institutes of Health sponsors only a select number of talented upper division science major students on each host campus, our correlational analyses were based on a unique and relatively small sample of data collected from undergraduate science major students (N=161) in 16 colleges and universities in the Pacific Southwest. We analyzed the data in two linked parts: (a) an analysis of the association between trust and motivation; and (b) an evaluation of whether the Program affects trust. We link the two parts by using the same longitudinal dataset in each and, further, by building the evaluation on the findings from models that compare trust and motivation among upper division science major students enrolled in the Program (n=49) with trust and motivation among upper division science major students not in the Program (n=112). Given the study’s exploratory purpose and design limits, we situate the results near the front end of the continuum from preliminary studies designed to suggest further research directions to the kinds of precise predictions we can expect from truly randomized controlled experiments.
Findings: Despite its necessarily preliminary findings regarding trust, this study contributes to the literature on academic mentoring in undergraduate science education in several ways. First, our study develops a novel conceptual perspective on the measure and utility of trust as a crucial form of social capital. Second, we used this perspective to develop a set of structural models that suggest interpersonal trust is producible in undergraduate science education; is pedagogically powerful, not a mere nicety; and is particularly important for students who are members of groups historically underrepresented in the sciences. Third, and surprising because the literature on trust does not prepare for them, are results suggesting that trust may work differently for different groups of students in different contexts.
Conclusions/Recommendations: Our findings are unusual and encouraging, and they bring several issues to light regarding (a) the utility of theory in research and practice in undergraduate science education, (b) inequitable representation in the sciences, and (c) contextual complexities that condition the development and maintenance of trust within institutional settings. First, our framework on trust derives from social capital theory and cautions that a science curriculum that informs and motivates entails attention not only to cognition but also to the educational significance of other complex forms of cognition that undergird social skill, relational awareness, and the development and maintenance of trust. In cognitively and socially consequential exchanges in undergraduate science education, trust and mistrust are important. Next, our most unusual finding—that trust in a mentor seems to matter more for the motivation and career expectations of the MARC-U*STAR Program students than for non-Program students—raises the question: How does trust in a university mentor matter for equitable undergraduate education in science? With our data, we cannot fully account for this unexpected finding. Drawing from studies that disentangle the concept of interpersonal trust from more generalized notions of trust in the social environment, we have worked at a tentative interpretation: Perhaps interpersonal trust is especially impactful in situations where the more general social environment poses perceptible threats. Last, when the literature on trust is examined concurrently with our findings, trust emerges not as an invariant entity, but as supremely contextualized. We conclude by calling for a more expansive, multidisciplinary, and multimethod research agenda focused on trust, which may contribute to a reorientation of undergraduate science education.
For more than half a century, federal attention in the United States has been directed at strengthening the scientific and technological labor force using reforms in STEM (science, technology, engineering, and mathematics) education. Despite significant infusions of public resources, however, efforts to increase the number, competitiveness, and diversity of undergraduates who major in STEM disciplines and pursue a career in science have continued to fall well short of stated goals (National Research Council, 2011a; National Science Foundation, 2009). Recent reports indicate that the science community continues to have difficulty fulfilling a Science for All agenda designed to recruit and educate talented and interested undergraduates from African American, Latina/o, and Native American groups historically marginalized in the sciences (Anderson & Kim, 2006; Carter, Mandell, & Maton, 2009; Lee, 1997; Malcom, Dowd, & Yu, 2010; Mutegi, 2011; NSF, 2006; Summers & Hrabowski, 2006).
One common explanation for this continuing difficulty is that many programs in undergraduate science education are poorly designed (Crisp, Nora, & Taggart, 2009). They are said to be based on unexamined assumptions about best practices rather than basic research grounded in theory (DePass & Chubin, 2009; Lewis, 2003; Mervis, 2006; U.S. Department of Education, 2007).
A different explanation, however, points to limits in the basic research itself (NRC, 2012). One limitation is that research about undergraduate science education has reproduced the commonplace that success in science depends principally on cognition, as indicated by mastery of time-honored subject matter knowledge (Nespor, 1987; Seymour & Hewitt, 1997), to the neglect of other forms of thinking, persuading, and decision making in meaningful interpersonal relations that can result in a students improved ability to learn (Heckman, 2008; NRC, 2010; Rose, 2013). A second and related limitation is that the basic research has often polarized social versus psychological accounts, with each position underemphasizing that, when college students contemplate and work toward careers in science, both social and psychological processes are entwined.1
The traditional standoff is shifting, however. Sociologists and scholars of higher education are increasingly interested in studying the interpersonal dynamics that influence undergraduate expectations and paths into science (Chang, Eagan, Lin, & Hurtado, 2011; Hurtado, Cabrera, Arellano, & Espinosa, 2009; Melguizo, 2011; Perna, 2006). The trend is toward putting a human face on broad social forces such as the authoritative science curricula, campus racial climate, stereotypes, and racial microaggressions (Chang, Denson, Saenz, & Misa, 2006; Cole, 2007; Locks, Hurtado, Bowman, & Oseguera, 2008; Pettigrew, 1998; Seymour & Hewitt, 1997; Solórzano, Ceja, & Yosso, 2000). At the same time, psychologists and social psychologists are increasingly skilled at measuring the socialization of dispositions, including science major students attitudes and expectations acquired in university settings (Elliot & Dweck, 2005; Enman & Lupart, 2000; Harackiewicz, Barron, & Elliot, 1998; Lent et al., 2003; Steele, 2010; Steele & Aronson, 1995). These moves toward a more comprehensive, interdisciplinary model of research bode well for the necessarily practical endeavor of science education, in which events do not come labeled sociological or psychological or, for that matter, academic. Moreover, while initiatives in undergraduate education typically include a mix of social and psychological componentsreflective perhaps of a commonsensical notion that pedagogy that includes a little of everything will workthey seldom make explicit a sociocognitive rationale that can give the assorted components coherence, direction, and purpose (Bandura, 1986; Jones & Nisbett, 1972; Steele, 1997).
THE CURRENT STUDY
Our study explores the promise of a new research direction in undergraduate science education, one focused on trust. Although the study employs longitudinal survey data, it differs in purpose and design from the strictly randomized experimental analyses that seek to furnish precise predictions and causal interpretations (Schneider, Carnoy, Kilpatrick, Schmidt, & Shavelson, 2007). We reason that an exploratory study that pays more attention to descriptive evidence than definitive proof is warranted insofar as it encourages the discovery of new social phenomena (Rozin, 2001) and promises to inform and, perhaps, help mitigate the persistent, limited success in the United States in achieving both an excellent and equitable community of science.
We design our study with four considerations in mind. First, we bring the sociological and psychological aspects together in the concept of trust. We investigate the effect of trustspecifically, trust between science major students and the mentors they designate as being most helpful in college lifeon the motivation and career expectations of juniors and seniors majoring in the biomedical sciences. A focus on interpersonal trust is unusual if not entirely overlooked in most studies of undergraduate science education (Ghosh, Whipple, & Bryan, 2001), and stands in sharp contrast to the situation in K12 education where an emergent body of survey and field research measures trust and documents its educational importance (Adams, Forsyth, & Mitchell, 2009; Bryk & Schneider, 2002; Goddard, 2003; Hoy, Gage, & Tarter, 2006; Louis, 2007).
Second, our study provides a concise definition of interpersonal trust, shows how trust can be operationalized in survey research, and explores its association with motivation and science outcomes. This design feature is important if trust is neglected in studies of undergraduate science because it seems impossible to define or measure objectively, much less produce in practice, particularly in contrast to more familiar curricular interventions such as remedial tutoring or hard criteria such as grade point average (Glaeser, Laibson, Scheinkman, & Soutter, 2000). Limiting research and practice to inputs and outputs that are relatively easy to count can distort the education deemed relevant to becoming a scientist and exacerbate the very difficulties the community of science has in attracting and retaining new recruits, particularly recruits from groups long outside the community of science.
Third, our study focuses on trust in academic mentoring. This topic is particularly relevant in studies of the undergraduate segment of the pipeline into science that, in contrast to graduate programs, typically occurs in large, impersonal settings that often generate student alienation (Hurtado et al., 2009; Lewis, 2003; Seymour & Hewitt, 1997). Moreover, while many policy documents, along with some research, attest to the educational importance of facultystudent mentoring in higher education, particularly in STEM programs (Carter, 2001; Jacobi, 1991; Kuh & Hu, 2001; Maton, Hrabowski, & Schmitt, 2000; Nettles, 1991), little is known about what it takes to make mentoring relationships successful, particularly for students who are members of historically underrepresented groups (DuBois & Karcher, 2005; NRC, 2005; Sánchez & Colón, 2009). At best, we know that the effects of mentoring are variable, and that they depend in part on the quality of the relationship between students and faculty (Anaya & Cole, 2001; Milem, 1998; Tinto, 1993). Therefore, our study hypothesizes that trustworthy relations may be both a significant condition for academic mentoring as well as an important outcome (Rhodes, 2002).
Finally, in contrast to cross-sectional studies in undergraduate science education, we use longitudinal survey data collected from science major students in 16 colleges and universities in the Pacific Southwest. Because all 16 campuses host a supplemental training program funded by the National Institutes of Health (NIH)the Minority Access to Research Careers - Undergraduate Student Training in Academic Research (MARC-U*STAR), hereafter the Programthe sample includes a subset of upper division science major students who are members of groups traditionally underrepresented in the sciences.
The aim of the MARC-U*STAR Program is to increase the number and competitiveness of talented, minority undergraduates who go on to graduate school and careers as research scientists. Although some variation in the Program across campuses is expected, MARC-U*STAR provides that all the selected students will engage, for a consecutive 24-month period during the final two years of their undergraduate training, in a curriculum focused on academic mentoring under the supervision of science faculty. The mentoring entails supervised laboratory research experience, a yearlong seminar devoted to presenting and critiquing research articles, and assistance in applying to graduate school. The curriculum also includes an annual stipend and travel funds to attend professional conferences (see http://www.nigms.nih.gov/Training/MARC/USTARAwards.htm).
We analyze and evaluate the longitudinal survey data in two linked parts: (a) an analysis of the association between trust (seen as a form of social capital) and motivation; and (b) an evaluation of whether the Program affects trust. We link the two parts by using the same longitudinal dataset in each and, further, by building the evaluation on the findings from structural analyses that compare trust and motivation among upper division science major students enrolled in the Program to trust and motivation among upper division science major students not in the Program. Social capital theory guides both parts of the study (Bourdieu, 1986; Coleman, 1990; Portes, 1998).2
Thus, we first investigate how trust and motivation are related in the academic mentoring of junior and senior science major students by the institutional agent each student identified as most helpful in college life. Structural models test the hypothesis that trust and motivation are social and psychological factors that work together rather than as discrete, unrelated educational inputs. These models also test the hypothesis that science major students motivations and behaviors may be different in different social domains, and that program status, i.e., participation in the MARC-U*STAR Program, moderates sociocognitive processes in undergraduate science education.
The second, evaluative part of our investigation is built on the findings from the first part. Using the same longitudinal, multisite dataset, it develops single-group autoregressive models (Loehlin, 1998) to measure the influence of the Program on trust. We reason that if trust and motivation are educationally important, then knowing whether a science education program can affect trust and motivation also merits consideration. We pursue the following research questions:
Do trustworthy mentors affect science major students motivation and their educational behaviors and expectations?
Does motivation mediate the effect of trust on the expectations of science major students?
Do links between trust, motivation, and education outcomes differ between science major students who are in the MARC-U*STAR Program and science major students who are not in the Program?
Can we infer the Programs influence by measuring its association with trust?
In what follows, we begin with a brief review of the literature on social capital, trust posed as a form of social capital, and motivation in postsecondary education. We follow with a more detailed description of our two-part research design. We then turn to the heart of the paper: the structural analyses linking trust, motivation, and science education outcomes, and the evaluation of trust in the MARC-U*STAR Program designed to supplement the standard undergraduate science curriculum for small cohorts of juniors and seniors who are members of groups long disadvantaged in university science. We conclude with a discussion of the implications and limitations of our findings for future research in undergraduate science education.
REVIEW OF THEORY AND RESEARCH
An extensive literature demonstrates the varying impact of social relationships on the ability of college students to succeed in higher education and in STEM fields (Deil-Amen & Rosenbaum, 2003; Hurtado et al., 2009; Kuh & Hu, 2001; Maldonado, Rhoads, & Buenavista, 2005; Pascarella & Terenzini, 2005; Patton, Morelon, Whitehead, & Hossler, 2006). Relatively few studies in higher education, however, employ a social capital frameworkalthough a handful utilize social capital theory in investigating transitions to college (OConnor, Hammack, & Scott, 2009; Perna & Thomas, 2008; Perna & Titus, 2005; Sandefur, Meier, & Campbell, 2006). Fewer still focus on the importance of trust, which is often acknowledged as the most widely recognized form of social capital (Bryk & Schneider, 2002; Coleman, 1990; Putnam, 2000). Instead, the theoretical bases in much research addressing social conditions in postsecondary education are social integration (Spady, 1971; Tinto, 1993), group involvement (Astin, 1993), social support (House, 1981), and social cognitive career theory (Bandura, 1986; Lent et al., 2003; Melguizo, 2011). We orient this study using an explicit sociocognitive framework that draws heavily from social capital theory.
Three features of social capital theory recommend it as a suitable frame for our research and, more broadly, as an important perspective for research in higher education and undergraduate science education. First, it emphasizes the value of trust in the efficient exchange of resources (Coleman, 1990; Putnam, 2000); second, it calls critical attention to mistrust and its effects on the exchange of information (Lin, 2001; Ream, 2003); and third, it points to a dialectical relation between the social world as constituted and the interested practices and internalized dispositions of people as social actors (Bourdieu, 1984; Brubaker, 1985).
Elaborating on the first of these three features, trust is working as social capital where there is an objective state of social actors voluntarily opening themselves to and cooperating with the actions of others (Coleman, 1990; Rousseau, Sitkin, Burt, & Camerer, 1998). Trust lubricates the inevitable frictions of social and economic life and facilitates the network exchange of resources (Coleman, 1988; Granovetter, 2002). Indeed, trust encourages risk-taking among interdependent actors who share a belief in keeping promises and in the duty of honoring ones declaration of will (Seligman, 1997). Such risk-taking situations entail an exchange of the vulnerability of one actor and the perceived ability and good will of another (Moorman, Zaltman, & Deshpande, 1992).
Questions about trust and what motivates cooperative behavior are generating increased interest within and across various disciplines, including sociology (Kramer, 2006), psychology (Dunn & Schweitzer, 2005), K12 education (Bryk & Schneider, 2002), organizational management (Schoorman, Mayer, & Davis, 2007), economics (Glaeser et al., 2000), and political science (Putnam, 2000).3 Survey research typically assumes that trust entails sweeping judgments, such as whether most people are trustworthy or whether trust is characteristic of human nature in general.4 Other scholars, however, draw on philosophical and empirical studies to distinguish between generalized trust in the context of the abstract social environment versus interpersonal trust between interdependent actors (Cook, 2001; Schoorman et al. 2007). We posit that trust is more validly measured on a case-by-case basis, depending on what social actors perceive about particular others and the specific stakes involved in a relationship (Russell Sage Foundation, 2000).
Oriented by this view of interpersonal trust, we measure the aggregate of science major students perceptions of the trustworthiness of institutional agents, specifically the mentors they identify as most helpful. The importance of trust is often acknowledged, but less often examined, in higher education, perhaps because interpersonal trust is quite difficult to model in survey research. Our measure builds upon recurring definitions of trust that assume that students estimate mentor trustworthiness based on three particular qualities in their mentorscompetence, benevolence, and integrityand risk making themselves vulnerable to a mentor on the basis of their perceptions of his or her qualities (Mayer, Davis, & Schoorman, 1995; Tschannen-Moran & Hoy, 2000). Thus, we conceptualize trust as one partys willingness to be vulnerable to another party based on the confidence that the latter party is (a) competent, (b) benevolent, and (c) a reliable person of integrity who acts transparently.
Competence concerns the perceived expertise of the trustee to perform actions of importance to those who trust (Oliver & Montgomery, 2001). In appraising a mentors role-specific competence, for instance, a science major student may judge the mentors expertise in unpacking a complex mathematical equation or demonstrating an important laboratory technique. Benevolence, perhaps the most common facet of trust, concerns the degree to which the trustee is perceived to care about and respect the one who trusts (Hardin, 2002). In appraising benevolence, a science major student may appreciate a mentors capacity for altruism and for holding the students best interest in mind because the mentor values the continuation of the relationship with the student. Integrity concerns the degree to which the trustee is perceived to act transparently and with fairness and predictability (Tschannen-Moran, 2004). In appraising a mentors integrity, a science major student may be aware of the mentors consistency in following through on doing what was promised.5
ACCOUNTING FOR MISTRUST
Mentoring is often assumed to be an irrevocably positive process, and it is widely advocated for STEM education, especially in interventions aimed at recruiting students from underrepresented groups (Cole & Espinoza, 2008; Cronan-Hillix, Gensheimer, Cronan-Hillix, & Davidson, 1986; Nettles & Millett, 2006). However, in relying on social capital theoryspecifically the second feature that recommends itwe are turned to a critical awareness of the downside of trust. We examine undergraduates appraisals of their mentors as reflective of the level and quality of trust in the relationship and assume that the effects of mentoring vary (Anaya & Cole, 2001; DuBois & Karcher, 2005). For example, a recent evaluation of NIH-sponsored minority training programs in science red-flagged mentoring, warning that some students perceptions of benign neglect by their mentors or, at best, a lack of encouragement threatened the already low numbers of minority trainees at this relatively advanced career stage (NRC, 2005, p. 8).
Negative relations are rarely integrated into social theory. Yet, critical perspectives on social capital emphasize that interpersonal relations are never free of wider institutional and historical contexts (Kao, 2004; Orr, 1999; Ream, 2003). Although trust often depends on the social class and/or ethnoracial characteristics of social actors (Tschannen-Moran & Hoy, 2000), it may also depend on the competitive structures and institutional playing fields of undergraduate science education where resources and information are exchanged or guarded. People tend to extend trust more readily to others they perceive as being similar to themselves; people also tend to trust institutions that they perceive as embodying their values (Zucker, 1986). Insofar as heightened awareness of racial microaggressions and stereotypes reinforce a historically embedded skepticism among minorities regarding how welcome they are in institutions of higher education (Allen, Epps, & Haniff, 1991; Ancis, Sedlacek, & Mohr, 2000; Cabrera & Nora, 1994; Solórzano et al., 2000; Steele & Aronson, 1995; Tierney, 1999), mistrust may protect some students from being taken advantage of by science. Mistrust may even enhance the self-preservation of people at risk of being exploited by science (Bogart & Bird, 2003; Gauchat, 2012). Thus, for historically underrepresented minority (URM) students, a willingness to place trust in science faculty may require the voluntary acceptance of a unique vulnerability to institutional authority (Gambetta, 1988; Guiffrida, 2005).
SOCIALIZING DISPOSITIONS: LINKING TRUST AND MOTIVATION
The third feature recommending social capital theory, apart from its important emphasis on trust and its critical recognition of mistrust, is that it (like social cognitive career theory) rejects siloing sociological versus psychological first principles. Instead, it calls attention to a dialectical relation between the determinative power of social groups and the undetermined motivations of free agents (Ortner, 2006) and thus preserves insights gained by important viewpoints across disciplines. Given this perspective, we seek to investigate whether and how trust and motivation work in relation to one another, rather than as discrete inputs.
In surveying the literature on motivation (Elliot & Dweck, 2005; Wentzel & Wigfield, 2009), we grappled with what is external to individuals versus what is in them. Much work on intrinsic motivation attends to the social-contextual conditions that impact cognitive regulation (Connell & Wellborn, 1991; Deci & Ryan, 1985; Weiner, 1991). Although this work does not explicitly link trust and motivation, it suggests that interpersonal relatedness bears mightily upon individual agency and will (Rosenhan, 1973; Rotter, 1980). Both Self-Determination Theory (SDT) and a subtheory of SDT labeled Cognitive Evaluation Theory specify social factors that explain variability in motivation (Deci & Ryan, 2000). By pinpointing competing needs for relatedness and autonomy, this research identified sociological processes (e.g., positive feedback in social settings) that fulfill innate psychological requirements for the development of motivation (Ryan & Deci, 2000).6
While we acknowledge that people may be motivated to engage in an activity for reasons that range on a continuum from inherent interests to external surveillance or even coercion (Ryan & Grolnick, 1986), we avoid re-inscribing the notion that motivation is somehow produced sui generis. A person may have a unique passion for science, but that passion can be accelerated (or extinguished) by experiences in the world, including science education. Using the word intrinsic implies that motivation is somehow self-produced and belongs to people by their very nature. Yet, initially eager science major students may elect not to pursue a career in science, or they may participate only on a pro forma basis if, for example, they perceive their mentors to be incompetent or unsupportive. Accordingly, social capital theory and SDT scholarship in psychology lead us to couple trust in a mentor with what we term internalized motivation. Our results show that making this link is warranted inasmuch as (a) motivation is measurably associated with relatedness to mentors students perceive as trustworthy and (b) science education outcomes depend significantly on internalized motivation.
We have pointed out that social capital theory provides an illuminating sociocognitive framework, one that directs attention to neglected variables such as trust and mistrust, as well as to familiar variables such as motivation, that deserve meaningful and ongoing clarification. We use social capital theory to frame an investigation of the relation between trust and motivation in academic mentoring in undergraduate science education. In the next section, we describe the design of our study, including its use of longitudinal survey data and structural equation models (SEM) that interrogate the relationship that trust in a mentor has to internalized motivation among upper division science major students.
Our investigation began with the design of a longitudinal survey instrument that integrated views on trust across social science disciplines (Mayer et al., 1995; Schoorman et al., 2007). The survey facilitates the parsimonious measure of trust between upper division college science major students and their mentors. We also used social capital theory to operationalize our analyses in two parts. In the first part, we tested structural models that can account for the hypothesized association between the trust that science major students have in mentors they deem as most helpful and the motivation of such students to succeed in science education. The models further investigate whether trust and motivation correlate with educational outcomes. We also tested whether enrollment in the federally supported MARC-U*STAR Program moderated associations between trust and motivation. The second part of the study evaluates the overall effect of the Program on trust over time. It is based on findings from the structural models that suggest trust is an important facet of science education, particularly when the students involved are from groups historically underrepresented in the sciences.
We collected the longitudinal data at 16 research and comprehensive universities (four-year colleges and universities) in the Pacific Southwest, all of which had the Program on their campuses. The base year sample (fall 2005) included juniors and seniors majoring in biology, chemistry, biochemistry, or related subdisciplines in the biological and physical sciences.7 Students who participated in the base year survey were re-contacted one year later in fall 2006 and again in spring 2007 to complete follow-up surveys that contained most of the base year items. The resulting panel data contained a total of 161 undergraduate science major students.8
Because one of our goals was to explore similarities and differences in thinking and behavior among different student populations and to arrive at comparative program-based analyses, we divided the panel data (N=161) into two groups: upper division science major students participating in the MARC-U*STAR Program (N=49) and upper division science major students not participating (N=112) in the Program.9 Although some information was missing for a portion of survey participantsthe degree of missing data on observed variables ranged from 1.0% to 6.2%, with an average of less than 4.0%we were able to retain missing cases by using imputation techniques so as to make the sample more plausibly representative of students who failed to answer all the questions in the survey or who failed to participate in each wave of measurement.10
DEPENDENT, BACKGROUND, AND CONTROL VARIABLES
African American, Latino/a, and Native American minorities are as likely as White and Asian students to begin college expecting to pursue careers in science (NRC, 2011b). Underrepresented minorities are less likely, however, to persist as their science expectations wane over time (Chang et al., 2011). This may partly explain why science educators are increasingly in pursuit of research that measures not only test scores and course performance outcomes but also dispositional outcomes, such as motivation and expectations, that are important in the process of educational attainment (Bohon, Johnson, & Gorman, 2006; Cabrera & LaNasa, 2000; NRC, 2012; Riegle-Crumb, Moore, & Ramos-Wada, 2011). Thus, in this study, we measured two dependent variables to ascertain whether trust and motivation affect science major students educational expectations: (a) plans to attend graduate school, and (b) plans to pursue a career as a research scientist.
We use expectations as a concrete indicator of an undergraduates science ambitions. We conceptualize expectations as reflecting not what science major students hope for, but what they plan for and believe will actually occur (Bohon et al., 2006; Hanson, 1994). Specifically, a single Likert-type survey item asked each student to rate whether he or she plan(s) to attend graduate school on a 5-point scale where 1 = strongly disagree and 5 = strongly agree. Another single Likert-type survey item asked each student to rate whether she or he plan(s) to become a research scientist on a 5-point scale where 1 = strongly disagree and 5 = strongly agree.
We included students socioeconomic status (SES), race/ethnicity, and months remaining until graduation as background variables in each of our exploratory models. We also included controls at the base year for motivation and, importantly, for each of the dependent variables. Each students prior outcome score is included in each of our models. SES is widely regarded as the most powerful predictor of racial gaps in educational attainment, and a highly disproportionate number of minority students are from lower socioeconomic households (Kao & Thompson, 2003; Roscigno, 2000; Tienda & Jensen, 1988). Thus, we included an SES composite consisting of three observed variables: family income and each parents education. An 8-point scale measured family income: $2,999 or less corresponded to a score of 1, and family income of $250,000 or more corresponded to a score of 8. Similarly, we acquired data on each students mothers and fathers levels of education: 1 on a 7-point scale measured up to an eighth-grade education, and advanced degrees (MD, PhD, or JD) corresponded to a score of 7. Family income and the two parent education measures were then factor analyzed. Factor loadings were used to create a single composite socioeconomic status score for each participant by summing the product of each loading and its respective item (a = .68).
To measure ethnoracial characteristics, we asked students to report their primary racial/ethnic group(s) on the survey. Self-reported race/ethnicity selected participants into two groups: members of a historically underrepresented group (African American, Hispanic, and Native American; N= 58) or not (N=103). Lastly, because our survey included science major students who were juniors or seniors when the base year survey was administered in fall 2005, at each measurement wave we asked students to report the year and term in which they expected to graduate with a bachelors degree. Using this information in conjunction with the date the survey was completed, we calculated, for each student, the months remaining until graduation. This variable was also included as a control in all statistical models to reduce error variability and to equalize participants on a variable with potentially confounding effects.11
MEASURING TRUST AND MOTIVATION
We used the survey data to address our research questions by developing multi-item, latent measures of trust and motivation and employing them in the measurement component of the structural models.
Trust tends to be measured inconsistently in survey research (Rousseau et al., 1998). One common problem is the failure of researchers to specify the trusteei.e., the person to whom something of value is entrusted. Anticipating this problem, we tracked science major students perceptions of the competence, benevolence, and integrity of the mentor whom students identified as being most helpful during their college science careers.
Trustworthy Mentor is a three-item construct representing science major students views of the institutional agent (i.e., the Program mentor, an assigned faculty advisor, or other faculty member) they identified as most helpful to them in college. This latent construct shows strong inter-item consistency (α= .86, fall 2005) and reflects the degree to which students perceive this person as professionally competent, benevolent, and behaving with integrity.
Measuring internalized motivation as a latent construct is also complicated. In characterizing motivation on an extrinsic-to-intrinsic continuum (Deci & Ryan, 2000), we use the term internalized motivation rather than intrinsic motivation to call attention to the influence of accumulated and often situation-specific experiences on seemingly spontaneous or inherent inclinations toward interest, exploration, and mastery (Elliot & Dweck, 2005; Ryan & Grolnick, 1986; Wentzel & Wigfield, 2009). We constructed a latent measure to quantify students acquired, rather than inborn, dispositions toward learning and achievement in science.
Internalized Motivation is a three-item construct representing the degree to which science major students are motivated to learn and achieve in their field of study, i.e., the degree to which students pursue learning as its own reward, and not primarily for extrinsic rewards such as a job/money, recognition, or a formal diploma. (α = .61, fall 2006).
We began our analyses of the survey data by examining descriptive statistics and mean differences on study variables across groups. Our longitudinal analyses were conducted within the latent variable modeling framework and relied on the path analytic strengths of structural equation modeling.
We stacked the multiple-group models to test specifically for group differences according to program status (i.e., in the Program versus not in the Program) in the estimated associations among the two latent constructstrust and motivationand the educational outcomes. The inferences we drew from the comparisons depended, in part, on cross-group measurement invariance in which items tapped the same latent construct for the Program and non-Program students. To test for invariance, we constrained and then freed factor loadings across groups for items that corresponded to the latent Trustworthy Mentor and Internalized Motivation variables in 2006. Each construct was invariant across the two groups of science major students (χ²Δ (7) = 4.60, p > .15).
The magnitude and significance of each path coefficient was estimated using the expectation maximization (EM) algorithm and a robust full information maximum likelihood estimator (Asparouhov & Muthén, 2005; Yuan & Bentler, 2000). Below, we detail the multigroup process by which we investigated whether internalized motivation represents the mediator through which trust influences educational outcomes, and whether program status moderates the relationships among these variables (Baron & Kenny, 1986; Kenny, Kashy, & Bolger, 1998).12
We were careful to ensure that our multiple-group models were empirically identified, such that the number of estimated parameters in each model were fewer than the sample size of our smallest group (the Program students, N=49). Accordingly, we pre-estimated covariances for exogenous variables, factor loadings, and item residuals for our 2006 latent measure of trust (see Figures 68 in the Appendix), and we imputed them in subsequent models.13 We also pre-investigated the tenability of imposing group equality constraints on control paths by examining models wherein dependent variables were regressed on controls only. Where comparisons of formal models were conducted, we used χ² difference testing with correction for the robust maximum likelihood estimator (Satorra, 2000; Satorra & Bentler, 1999). All reported models were positive-definite and converged normally. We used the following indices to test the appropriateness of all statistical models: the Tucker-Lewis index (TLI), comparative-fit index (CFI), Akaikes information criterion (AIC), and root mean square error of approximation (RMSEA).14 For more detailed information regarding data analyses procedures, see the Appendix.
In the second part of our analyses, we used single-group autoregressive models (Loehlin, 1998) to investigate the Programs association with students perceptions of the trustworthiness of their mentors as measured in fall of 2006 and in spring of 2007.
Autoregressive models are especially useful for measuring program effects. This is particularly so where, as in our case, mean levels on the dependent variable, mentor trustworthiness, are unequal across levels of the primary independent variable (the Program status) at the initial point of measurement. Because participants in the survey included both juniors, who were relatively new to the Program, and seniors, who were not, some Program effects were already present in the base year data. We sought a statistical model that would allow us to equalize the Program and non-Program students on base year measures. Of course, the success of the Program might derive, in part, from strategic recruiting practices that skimmed off top-performing and perhaps more trusting minority students. That objectionthat trust is an important criterion for eligibility rather than outcome of the Programis why we accounted for differences in trust upon entry into the Program. In sum, we used autoregressive models to control for levels on the dependent variable, trust, at any previous measurement occasion, and to enable time-specific gains associated with the Program to be isolated from the Program students higher initial propensity to trust.
The results are presented in three sections. First, we describe and compare study variables across the MARC-U*STAR Program and non-Program students. Second, we model sociocognitive processes by examining (a) the association between mentor trustworthiness and internalized motivation, (b) whether motivation mediates the effect of trust on the outcome measures, and (c) whether the associations among these variables are moderated by program status. Third, we explore the influence of the Program on trust while accounting for the Program students higher initial propensity to trust.
Although the Program-sponsored science major students are less advantaged economically than those not in the Program, they report higher levels of motivation, academic expectations, and trust in their mentors in science. These advantages are measurable and appear to persist over time, as we describe below.
In terms of background characteristics, the majority of the Program students self-identify as African American, Latino/a, or Native Americanall underrepresented groups in the biomedical sciences (again, the Program targets talented historically underserved minorities). Non-Asian minorities constitute 72% of the Program science major students compared to 21% of non-Program science major students, per Table 1. In relation to socioeconomic status, the Program students average .66 standard deviations (SD) below the mean SES of non-Program science major students (see Mean Difference column). Yet the Program students are a motivated group with levels of motivation, as measured in the 2005 base year, that are higher than non-Program students motivation (.57 SD above non-Program students). The Program students also report higher grades and science expectations in fall, 2005. For example, on a scale of 1 to 5, their plans to pursue research science careers eclipse non-Program students plans by 1.47 points. By the second follow-up survey in spring 2007, the base year advantage in plans to pursue research science careers persists4.31 versus 2.83 for the Program and non-Program students, respectively. These comparisons between science major students in and out of the Program suggest that high levels of motivation among the Program students may offset some of the educational challenges that tend to coincide with economic disadvantage.
Table 1. Variable Descriptions and Comparisons Across Program Status
Background Characteristics Fall 2005
| || || || || || |
Underrepresented Minority Status
Categorical variable (1 = URM, 0=not URM)
Standardized three-item composite variable
Standardized three-item composite variable
I am genuinely interested in the material I study in college
I do what I need for the grade without caring about understanding the material
I go to college primarily to get the degree I need for a good job (reverse coded)
Grades in Science
(5 = Mostly As, 1 = Below D)
Plans to Attend Graduate School
Likert-Type Item (5 = Strongly Agree, 1= Strongly Disagree)
Plans to Become a Research Scientist
Likert-Type Item (5 = Strongly Agree, 1 = Strongly Disagree)
Months Remaining Until Graduation
Trustworthy Mentor Fall 2005
| || || || || || |
I respect this person as a teacher
This person respects and appreciates me
This person is fair in his/her dealings with me
Trustworthy Mentor Fall 2006
| || || || || || |
I respect this person as a teacher
This person respects and appreciates me
This person is fair in his/her dealings with me
Trustworthy Mentor Spring 2007
| || || || || || |
I respect this person as a teacher
This person respects and appreciates me
This person is fair in his/her dealings with me
Internalized Motivation Fall 2006
| || || || || || |
Standardized three-item composite variable
I am genuinely interested in the material I study in college
I do what I need for the grade without caring about understanding the material (reverse coded)
I go to college primarily to get the degree I need for a good job (reverse coded)
Educational Outcomes Spring 2007
| || || || || || |
Plans to Attend Graduate School
Likert Type Item (5 = Strongly Agree, 1= Strongly Disagree)
Plans to Become a Research Scientist
Likert Type Item (5 = Strongly Agree, 1 = Strongly Disagree)
Note. For categorical variables, we use Chi-Square Tests of Independence to perform significance tests. For composite variables and Likert-type items, we use Univariate ANOVAs. Fall 2005 N=161 (the Program n=49, non-Program n=112); Fall 2006 N=146 (the Program n=43, non-Program n=103); Spring 2007 N=131 (the Program n=47, non-Program n=84). All means and tests, however, are based on N=161 (the Program n=49, non-Program n=112) after imputing missing data using single imputation via the EM algorithm. *p ≤ .05, **p ≤ .01.
Furthermore, the Program students across all three survey waves report high levels of trust in the mentor or college professional they identified as being most helpful, as compared to science major students not in the Program.15 As Figure 1 displays, the Program students perceptions of mentor trustworthiness in fall 2005 (M = .61 SD) exceeds by nearly one full standard deviation the student average among non-Program students (M = -.27 SD). One year later, the advantage narrows slightly, and then it widens again to .90 SD in spring 2007.16
Figure 1. Science major students perceptions of mentor trustworthiness by program status
Note. Results of analysis of variance indicate the mean differences in science major students’ perceptions of mentor trustworthiness are statistically significant (p ≤ .01) between the MARC-U*STAR Program and non-Program students. Fall 2005 N = 161 (Program n = 49, Non-Program n = 112); Fall 2006 N = 146 (Program n = 43, Non-Program n = 103); Spring 2007 N = 131 (Program n = 47, Non-Program n = 84). All means and tests, however, are based on N=161 (Program n = 49, Non-Program n = 112) after imputing missing data using single imputation via the EM algorithm (Little & Rubin, 2002).
Of course, membership in the Program is associated with several factors (e.g., race/ethnicity, SES, and time until graduation) that may partly account for the ebb and flow of trust and the magnitude of the apparent grouplevel differences. Later, we control for these factors in autoregressive models designed to gauge the overall influence of the Program on students perceptions of mentor trustworthiness.
MODELING SOCIOCOGNITIVE PROCESSES
In the exploratory models that follow, we use multiple-groups SEM to systematically compare whether trust and motivation are related over time to the expectations for graduate school and careers in research science as expressed by the Program versus non-Program students. We find that, particularly among the Program students, trust in mentors is associated with motivation to achieve in science.
Our initial attempt at modeling links between trust and motivation examines the effect of mentor trustworthiness on internalized motivation, controlling for background variablesincluding motivation as measured at the base year in 2005. Based on preliminary analyses (see Figure A1 in the Appendix), all covariances are constrained to equality across groups as are all control paths, except the association between motivation in 2005 and 2006. This model shows a good fit to the data and suggests two important differences between the Program and non-Program students. First, the association between trust and motivation among the Program students is positive and significant (t = 2.29, p < .05), while the association for non-Program students approximates zero (t = .23, p = .82). Second, although the base year and first follow-up measures of motivation are positively associated, the magnitude of the association is more pronounced for non-Program students, as indicated again by competing models in which this path is freed and constrained across groups (χ²Δ (1) = 52.64, p < .001).
Building upon these preliminary analyses, we estimate the model in Figure 2.17 Results suggest that trust affects motivation among the Program students for whom a 1.0 unit increase in mentor trustworthiness is associated with a .50 unit (.37 SD) boost in internalized motivation (t = 2.29, p < .05, b = .50, β = .37). Trust is not measurably associated with motivation, however, among non-Program students.
On one hand, non-Program students start out with comparably low levels of motivation at the base year, and their motivation remains relatively stable through the first follow up in 2006. The Program students, on the other hand, demonstrate higher levels of motivation at the base year, and yet the association between base year motivation in 2005 and motivation in 2006 is much less pronounced (b = .25 for Program students versus b = .65 for non-Program students). Thus, even after persevering to the upper division ranks in undergraduate science education, the Program students motivation still seems amenable to the influence of trustworthy science mentors. The positive association between the Program students’ trust in mentors and internalized motivation (t = 2.29, p < .05, b = .50, β = .37) can be interpreted as a medium association according to the criterion set forth by Cohen (1988).18 In short, program status appears to moderate links between trust and motivation, per Figure 2.
Figure 2. Internalized motivation regressed on trustworthy mentor Note. Top and bottom coefficients are the MARC-U*STAR Program and non-Program students, respectively. The path weights are presented as unstandardized parameter estimates. All covariances among exogenous variables are constrained to equality across groups (see Figure A1 in the Appendix); whether constrained or freed, the models fit equally well (χ²Δ (9) = 9.25, p < .001). Individual items that contribute to the latent constructs (depicted as ellipses) and all error terms and covariances are excluded from the figure to improve its readability. Fit Indices: [χ² (91) = 56.54, p = .99; CFI = 1.00, TLI = 1.11, RMSEA = .00, AIC = 4192.689]
EXPLORING MOTIVATION AS A MEDIATOR
Next, we explore whether motivation mediates students science expectations as measured at the second follow up in 2007.19 Specifically, we test mediation against two primary criteria. First, the initial variable, Trustworthy Mentor, should demonstrate a measurable association with the mediator, Internalized Motivation. Second, the mediator should affect the dependent variable(s). To ascertain whether mediation is complete and consistent, we also examine whether the association between trust and each dependent variable is negligible when motivation (including its impact on the dependent variable) is included as a mediator in the model (Baron & Kenny, 1986; Mackinnon & Fairchild, 2009). We evaluate each dependent variable separately.
In Figure 3, we model as the dependent variable students plans to attend graduate school. In this model, we constrain all control paths directed at the dependent variable to equality across the two groups with one exception: Preliminary results comparing freed and constrained models (χ²Δ (1) = 12.93, p < .001) show the link between students’ plans to attend graduate school as measured in 2005, and then again at the second follow up in 2007, to be strongly dissimilar across groups (b = .02 versus b = .64 for the Program and non-Program students, respectively). This path remains freed.
Figure 3. Graduate school expectations regressed on trust and motivation
Note. Top and bottom coefficients are the MARC-U*STAR Program and non-Program students, respectively. The path weights are presented as unstandardized parameter estimates. All covariances among exogenous variables are constrained to equality across groups (see Figure A2 in the Appendix); whether constrained or freed, the models fit equally well. Individual items that contribute to the latent constructs (depicted as ellipses) and all error terms and covariances are excluded from the figure to improve its readability. Fit Indices: [χ² (135) = 85.68, p = .99; CFI = 1.00, TLI = 1.13, RMSEA = .00, AIC = 5082.321]
For non-Program students, neither trust nor motivation is significantly associated with the dependent variable (p > .60 in both cases). The Program students’ plans to attend graduate school, however, seem more amenable to the influence of motivation, which is positively associated with the dependent variable (t = 3.82, p < .001, b = .58, β = .65). This particular association is not subtle, surpassing Cohen’s (1988) criteria for qualifying as a large association.
Although the direct link between mentor trustworthiness and the dependent variable in Figure 3 is insignificant for both groups (p = .40), here again trust is positively associated with the Program students motivation (b = .50, p < .05), which, in turn, is positively associated with their plans to attend graduate school (the mediated effect of trust on the dependent variable is .29 units [.24 SD]). For the larger sample of non-Program students, however, none of these associations meet the threshold of formal statistical significance.
In Figure 4, we examine yet another measure of science major students expectations, namely, their plans to pursue careers in research science. Once again, for non-Program students, neither trust nor motivation is significantly associated with the outcome variable (p > .30 in both cases). By contrast, among the Program students, motivation affects research science career plans (t = 3.10, p < .01), although, here again, the direct link between trust and the outcome variable is not significant (p = .69). Thus, in Figure 4, we constrain to zero the direct link between trust and the outcome variable for both groups, while freeing the link between motivation and the outcome for the Program students (and constraining it to zero for non-Program students). Again, in this instance, mentor trustworthiness seems to bolster the Program students motivation, which is positively associated with the outcome variable (t= 3.20, p < .01, b = .36, β = .50). The mediated association between trust and the outcome (.18 SD) is complete and consistent. Thus, for the Program students, trust in a mentor is positively associated with expectations for a career in research science via the link between trust and motivation. There is no measurable association between trust and motivation for non-Program students, however, nor does trust appear to influence their plans to pursue a career in research.
Figure 4. Expectations to become a research scientist regressed on trust and motivation
Note. Top and bottom coefficients are the MARC-U*STAR Program and non-Program students, respectively. The path weights are presented as unstandardized parameter estimates. All covariances among exogenous variables are constrained to equality across groups; all control paths directed at the dependent variable are constrained equal across groups (see Figure A3 in the Appendix). Whether constrained or freed, the models fit equally well. Individual items that contribute to the latent constructs (depicted as ellipses) and all error terms and covariances are excluded from the figure to improve its readability. Fit Indices: [χ² (136) = 94.59, p = .99; CFI = 1.00, TLI = 1.14, RMSEA = .00, AIC = 5089.947]
To this point, our results suggest that for the MARC-U*STAR Program students, trust in a mentor is positively correlated with internalized motivation, which appears to mediate students expectations to pursue graduate school and careers in research science. By contrast, these associations turn out to be statistically inconsequential among the larger group of non-Program science major students. Thus, program status apparently moderates associations among trust, motivation, and science outcomes.
THE EFFECT OF THE PROGRAM ON TRUST
Lastly, we build upon the results in the previous sections by using autoregressive models to explore the hypothesized association between the Program and our focal variable, trust. Insofar as trust affects the Program students expectations via motivation, then investigating whether the Program is associated with science major students perceptions of mentor trustworthiness is also important.
Exploring the Program is also important because it is plausible that MARC-U*STAR does not produce the desired effects, but simply selects for them. After all, upon entry into the Program, the students in our sample are already motivated to achieve at a high level in science education and generally view their mentors as trustworthy. Apart from this ability to attract gifted science major students who have overcome extraordinary challenges, what is the value of the Program if the students can already make it on their own?
Anticipating these concerns, we explore whether the MARC-U*STAR Program affects the trajectory of trust in mentors over time, controlling for the Program students background characteristics and, in particular, students initial perceptions of mentor trustworthiness, per the single-group autoregressive model depicted in Figure 5.20
This model associates the Program with our latent measure of trust at three points in time. At the first data wave in 2005, we correlate trust and each exogenous variable, including Program status, underrepresented minority status (URM), SES, and Months to Graduation. The association is clearly present in 2005 (r = .46, p < .001), which is unsurprising since we have already established that science major students enter the Program with higher proclivities to trust institutional agents. However, one year later and controlling for base year perceptions of mentor trustworthiness, the Program students again show scores on trust that significantly exceed non-Program students scores (t = 2.27, p < .05, β = .20). Importantly, these scores are modeled as a function of the Program when accounting for the temporally previous score on trust, plus the impact of the control variables, plus error. Thus, above and beyond their initial advantage in trust, the Program students perceptions of the trustworthiness of their most helpful mentor again exceeds non-Program students by .20 SD in fall 2006 and increase even more rapidly vis-à-vis non-Program students between fall 2006 and spring 2007 (a difference of .30 SD [t = 2.99, p < .01, β = .30]).
By exploring trust and motivation in science education and then evaluating the MARC-U*STAR Program based on the results of our preliminary models, our findings suggest that the value of the Program is predicated on more than stringent selection and eligibility criteria and a supplementary curriculum aimed at strengthening and/or remediating cognitive gains. It appears also to entail the Programs capacity to bolster its upper division minority enrollees already optimistic perceptions of the trustworthiness of mentors in university settings.
Figure 5. Associating the program with trust
Note. The path weights are presented as standardized parameter estimates. Individual items that contribute to the latent constructs (depicted as ellipses) and all error terms are excluded from the figure. Interyear factor loadings for corresponding items were constrained to equality and correlations between interyear error variances for corresponding items were estimated. Source: Longitudinal panel data containing 161 undergraduate science major students, including undergraduate biomedical science major students who participated in the MARC-U*STAR Program (N=49) and those science major students who did not participate in the Program (N=112). Fit indices: χ² (46) = 65.90, p = .03; CFI = .96, TLI = .95, RMSEA = .05, AIC = 4813.049
This investigation presents several methodological challenges. As previously discussed, the study uses a unique and relatively small sample of longitudinal, multisite survey data to explore trust as a little-used construct in undergraduate science education. Given the studys purpose and design limits, we situate the results near the front end of the continuum from preliminary studies; it is designed to suggest further research directions to the kinds of precise predictions we can expect from truly randomized controlled experiments (Gutting, 2012; Rozin, 2001; Schneider et al., 2007). Before we discuss the results in more detail, we elaborate upon the methodological challenges posed by our study and how we dealt with them. In general, we regard design as a series of trade-offs. Researchers must weigh design decisions in light of the purpose(s) they pursue in a study.
One methodological challenge entailed striking a balance between costly data collection versus securing sufficient statistical power. We decided to collect a smaller sample that included multiple survey items measured at three points in time, rather than a larger sample with more stable parameter estimates but fewer waves of assessment. We reasoned that (a) few longitudinal survey research projects explore college populations of minority students in contemporary science education; (b) fewer still employ data in which the same undergraduate science major students are surveyed about the quality of their relationships with mentors over time; and (c) in a demanding, two-year curriculum in upper division science or specifically in MARC-U*STAR, interpersonal trust might well fluctuate over time rather than being a once-and-for-all achievement.
At the same time, however, we acknowledge that the larger the sample, the more stable the parameter estimates. Yet, quantitative researchers do not agree on what constitutes large; instead, they often operate according to rules of thumb. By contrast, the field of structural equation modeling (SEM) has established some guidelines, and most researchers agree that parameter estimates in SEM become stable beginning at five cases per estimated parameter (Bentler, 1995; Westland, 2010). When the parameter estimates in our study did not meet this guideline, we double-checked our results using a multiple regression framework. We examined all of our study hypotheses in the regression framework in which our latent variables were replaced with summed composites, and analyses were again conducted separately for Program and non-Program students. With this check, the pattern of significant and non-significant findings did not change and the general magnitude of effects remained quite similar. These results supported all of the inferences put forth in this paper.
A second methodological challenge is that variables omitted from the structural models may bias the results. For example, undergraduates may gravitate toward or be assigned to faculty mentors on the basis of unmeasured characteristics that condition the educational impact of trust. We sought to address this challenge by controlling for various background factorsin particular, each students prior scores on each dependent variable (Cooper, Robinson, & Patall, 2006; Kane, Rockoff, & Staiger, 2008). Without randomized assignment, we cannot know whether we assessed all possible sources of initial differences between individual science major students. But, we can claim that the findings are intriguing and robust enough to warrant further studies of trust, especially insofar as few empirical studies, whether with large or small samples, have documented that trust sparks motivation and that educational programs can make a difference regarding undergraduate distrust of university science, most particularly among minority students (Bogart & Bird, 2003; Gauchat, 2012).
A third methodological challenge entails the strain between examining each feature of an intervention for its impact on trust versus evaluating the overall impact of an intervention. Given the exploratory purpose of our research, we evaluated the impact of the MARC-U*STAR Program as a whole, and, accordingly, we used autoregressive models that are not designed to tease apart particular aspects of the Program that may be associated with increased trust. Thus, our study does not examine the differential impact of specific curricular components, such as laboratory research or an annual stipend, on students trust and science priorities. Nor does it consider whether these or other curricular components are robust when matched on additional covariates such as institutional selectivity, campus diversity, or types of financial aid (Brewer, Eide, & Ehrenberg, 1999; Malcom et al., 2010; Melguizo, 2010). Given important concerns such as these, which our study cannot address as well as its promising results, we recommend an agenda for additional research.
Despite its limitations, this study contributes to the literature on academic mentoring in undergraduate science education in several ways. First, our study develops a novel conceptual perspective on the measure and utility of trust as a crucial form of social capital. Second, we used this perspective to develop a set of structural models that suggest interpersonal trust is producible in undergraduate science education; is pedagogically powerful, not a mere nicety; and is particularly important for students who are members of groups historically underrepresented in the sciences. Third, and surprising because the literature on trust does not prepare for them, are results suggesting that trust may work differently for different groups of students in different contexts. For instance, while we know that, in general, being non-White is negatively associated with trust in people and institutions in the United States (Brehm & Rahn, 1997; Weaver, 2006), we found that improving the perceptions of the MARC-U*STAR Program students, most of whom were minorities, regarding the trustworthiness of their mentors is positively associated with Program students motivation, which in turn mediates their expectations for graduate study and a career as a research scientist. Yet, for the larger group of non-Program science major students in our sample, most of whom were not minorities, none of these links were measurably significant. We discuss these three research contributions regarding the utility of theory, inequitable representation, and contextual complexities in turn, and conclude with a recommendation for a more expansive research agenda focused on trust, which may contribute to a reorientation of undergraduate science education.
With its focus on trust, our study demonstrates the value of theory development for undergraduate science education (Braxton, 2000; Crisp et al., 2009).21 For instance, it suggests that it is not surprising that science educators are unaware of trust, including its implications for motivation, if they lack a framework such as the one we introduce, which directs them to consciously look for and develop trust. Our framework is derived from social capital theory and cautions that a science curriculum that informs and motivates entails attention, not only to cognition as indicated by subject matter knowledge, but also to the educational significance of other complex forms of cognition that undergird social skill, relational awareness, and the development and maintenance of trust. Adherence to pedagogical theory focused on subject matter cognition is necessary, yet insufficient for increasing the number, competitiveness, and diversity of undergraduates who major in STEM and pursue research careers in the sciences (Handelsman et al., 2004). The sociocognitive framework we develop using social capital theory emphasizes that exchanges of information and resources entail both trust and mistrust, hypothesizes the association between trust and motivation, and posits that sociological and psychological factors work together rather than separately in mentorstudent relations. From this perspective, undergraduates in science education are not only learning about science but, simultaneously, are coming to know themselves and others and their places in particular school and social orders (Page, 1991). In such cognitively and socially consequential exchanges, trust and mistrust are important.
TRUST AND EQUITY
Our most unusual findingthat trust in a mentor seems to matter more for the motivation and career expectations of the MARC-U*STAR Program students than for non-Program studentsraises the question, how does trust in a university mentor matter for equitable undergraduate education in science? The question is important because, with limited empirical evidence about science mentorings inputs, quality, or effects (Jacobi, 1991; Nettles & Millett, 2006; NRC, 2009), it is nevertheless touted as among the most efficient means of tempering the alienating effects of off-putting faculty members and large, impersonal, undergraduate courses in the sciences (Lewis, 2003; Seymour & Hewitt, 1997).
With our data, we cannot fully account for this unexpected finding. Yet, drawing from studies that disentangle the concept of interpersonal trust from more generalized notions of trust in the social environment (Cook, 2001; Schoorman et al., 2007), we have worked at a tentative interpretation, which we hope other scholars may penetrate in future research: Perhaps interpersonal trust is especially impactful in situations where the more general social environment poses perceptible threats. Especially within institutions that threaten failure, does the trust that develops in meaningful interactions between individuals matter more than it might otherwise within institutions that are perceived to encourage success? And, does interpersonal trust matter in ways that can mitigate institutional shortcomings?
For instance, consider that all students who declare a science major may have reason to mistrust just how genuinely welcoming the community of science is, insofar as science faculty wink about weeding out students (Seymour & Hewitt, 1997) and roughly half of the freshmen who declare a science major upon entering are gone by the end of their sophomore year (Center for Institutional Data Exchange Analysis, 2000). However, note too that trust in mentors proved less measurably consequential for the non-Program undergraduates in science than for the Program students.
This surprising difference may be partly understandable given that the non-Program students, the great majority of whom are members of groups well-represented in the university and science, are likely to enter the university with fewer prior experiences that would lead them to mistrust the university or science, whereas the MARC-U*STAR Program students, almost all of whom are members of groups underrepresented in the university and science, may have more tangible reasons to maintain mistrust of the university and the community of science. The latter group may continue to wrestle with heightened awareness of racial stereotypes and micro-aggressions perpetuated within the university environment (Solórzano et al., 2000; Steele, 2010). They may also have specific reasons to perceive their interests as being at odds with those of the community of science (Chang et al., 2011; Major & OBrien, 2005). For example, studies surveying the recent history of American science, most infamously, the 40-year Tuskegee Institute/U.S. Public Health Service experiment with untreated syphilis in Black males, document why historically disadvantaged groups might mistrust the community of science (Bogart & Bird, 2003; Gamble, 1997; Gauchat, 2008).
However, for the MARC-U*STAR studentsfrom whom we might expect deep and lingering mistrusttrust in mentors appears to have deepened during the Program, affecting students motivation and subsequent aspirations for graduate school and careers as research scientists. Hence, our speculation that for the Program students, the interpersonal trust fostered in the Program moderates the students more generalized institutional mistrust.
To be clear, our study did not measure generalized trust in the institution of science or undergraduates overall perceptions of the culture of science (Brint, Cantwell, & Hanneman, 2008; Hurtado et al., 2009), nor did it establish whether trust in the general social environment of the university varied across our two groups of science major students. Therefore, further research with a much larger sample than ours is needed to probe the differential impact of trust on different groups of students across institutional settings, and the research will require a design and multiple forms of data able to account for trust at two levels simultaneouslythe individual level and the structural level.22
THE INFLUENCE OF CONTEXT
This study was designed to form a bridge between a conceptualization of trust and empirical studies of college science education. However, as previously mentioned, the study sheds little light on the particular components of the Program to ascertain their differential impact on trust. We know little about the shifting conditions of practice in which protégés and mentors in the community of science struggle over relations of trust. In relation to both points, trust emerges, not as an invariant entity, but as supremely contextualized. Accordingly, our call for a wideranging, multidisciplinary, and multimethod research agenda reflects our reasoning that comparative considerations of trust in diverse contexts are an excellent means of recognizing the complexity of interpersonal and generalized trust in schools, including universities.
For purposes of illustration, we note three contextual issues that arose in our study whose delineation requires a broader, more comparative approach than we pursued. The first is relations of trust in face-to-face contexts, such as science classrooms and laboratories. Our survey data does not account for the likelihood that the profiles, or identities, of faculty mentors differ markedly from their mentees profiles. Therefore, additional research is needed that will provide information about interpersonal trust between science mentors and undergraduates, particularly whether people extend trust more readily to others whom they perceive as similar to themselves.23 And what of the general, more abstract characteristics of a university itself? How does the hierarchical and sometimes alienating culture of science (Hurtado et al., 2009), one that tends to prioritize the instrumentality of quantitative skills and job-market prospects over ideas, values, and interpersonal relationships (Brint et al., 2008), influence relations of trust between faculty and students in undergraduate science classrooms?
The second contextual issue centers on higher education policy. New research should account for how trust and mistrust interact with myriad other reforms in undergraduate science education, a topic our study did not explore. For example, market- and accountability-based reforms are arriving at the university. As in K12 education, the reforms attempt to ascertain and influence faculty effectiveness in teaching by measuring it against growth in student scores on standardized tests (Hannaway & Woodroffe, 2003). But, will such reforms undermine other reforms in undergraduate education that seek more vital teaching and learning as a consequence of boosting relations of trust between faculty and students?
Third, and last, are the ecological domains that shape human developmentfamily, peer, school, and community (Bronfenbrenner, 1979). The effort to account for trust in undergraduate science education should occur not only within the university, but across these domains. For instance, inasmuch as familism is of particular importance among Latinos in higher education (Desmond & Lopez Turley, 2009), then understanding whether trust in parents impacts Latina/o motivation in science may be important for understanding the trajectories of Latino youth in undergraduate science (Alon, Domina, & Tienda, 2010). Similarly, investigating how trust among same-age peers in undergraduate science differs by student race and/or ethnicity seems important in light of often-cited research that suggests that if minorities will only work more collaboratively with their peers, then racial gaps in STEM attainment would diminish (Treisman, 1992). Because it is nearly impossible to isolate these spheres in order to measure the influence of each separately from the others, a more comprehensive and multidisciplinary model of research is needed (Ream, Ryan, & Espinoza, 2012).
As a social institution that is both competitive and collaborative, the U.S. community of science is faced with challenges building and maintaining relationships of trust between undergraduates and their university mentors. These challenges may be especially pressing for historically underserved college students for whom the experience of racial tensions and a deep-seated mistrust of science institutions can serve as obstacles to taking a chance with powerfully positioned institutional agents such as well-established scientists (Gambetta, 1988; Guiffrida, 2005). Yet, there can be no trust without risk taking (Rousseau et al., 1998). In effect, science major students must take a chance on mentors, trusting that they will exercise authority benignly, just as mentors must risk losing the approval of university colleagues if they devote too much time and energy to students. Such leaps of faith involve trusting in the good intentions of individuals in positions of institutional authority, including trusting that ones identity will be positively affirmed by those in authority, and that, by complying with the exercise of authority, ones own interests will be advanced (Erickson, 1993; Louis, 2007, Mitchell & Spady, 1983). Yet, while trust holds promise, we cannot ignore the possibility that feelings of mistrust may protect some students from being taken advantage of or exploited by science (Bogart & Bird, 2003; Gauchat, 2012). Mistrust may inoculate the healthy skeptic from the ill effects of trust misplaced. Perhaps the important concluding question, then, is not so much whether trust matters to historically underrepresented undergraduates, but whether minorities can be persuaded to view it as rational to take the risks and seek the help that make trust in mentorsnot blind trust in powerful institutional actors, but rationally calculative trust in authentic agents of authoritya possibility, even within mainstream institutions they have perceived to be unworthy of trust.
This study is one part of a larger research project supported by the National Institutes of Health (Grant Number 5R01GM72059-2). The data analyses and interpretations expressed in this article are the authors and do not necessarily reflect those of the granting agency. The authors thank other members of the Research Group for their assistance and support in all aspects of the study, including Natalie Becker, Marie-France Orillion, Virginia Panish, Eva Ritter, Maritza Rodriguez, and Sarah Ryan. The anonymous reviewers are also acknowledged for their useful comments.
1. Thus, many sociological studies have concentrated on the influence on undergraduate trajectories of large, faceless social forces such as chilly campus climates, institutional racism, and leaks in the pipeline into science, while psychological studies have focused on explanations of the achievement goals, personalities, and/or work ethics of individual students (Bensimon, 2005; Enman & Lupart, 2000; Hilton & Lee, 1988; Hurtado, 1994; Hurtado, Carter, and Spuler, 1996).
2. Although there is disagreement over its precise definition, social capital refers generally to individuals capacity to gain access to scarce resources by virtue of their membership in groups and participation in broader structures of society (Bourdieu, 1986; Coleman, 1988)resources that include knowledge and information (Becker, 1964), cultured dispositions (Lamont & Lareau, 1988), and/or economic capital and employment opportunities (Granovetter, 2002). By social capital, then, we mean the aggregate of actual and potential resources embedded in social networks that may be converted, via social exchange, into other manifestations of resources for use by individuals and groups (Portes, 1998; Ream & Palardy, 2008).
3. Economic sociologists show that trust, as social capital, greases the wheels of economic exchange (Gambetta, 1988; Granovetter, 2002); political scientists encourage the perception of trust as social capital by demonstrating its contributions to the health of civil society (Fukuyama, 1995; Putnam, 2000); and scholars in K12 education employ a social capital framework to show that kids learn more with trust in schools (Bryk & Schneider, 2002; Dika & Singh, 2002).
4. The General Social Survey (GSS), for example, contains three questions that researchers have consistently used to measure trust: (a) Generally speaking, would you say that most people can be trusted or that you can't be too careful? (b) Do you think most people would try to take advantage of you if they got a chance or would they try to be fair? (c) Would you say that most of the time people try to be helpful, or that they are mostly just looking out for themselves? (See http://www.norc.org/projects/General+Social+Survey.htm.)
5. Of course, student reports of their perceptions may consist of projections of their own values onto others. This bias has been targeted as a weakness of research on social dynamics (Kandel, 1996). From a different conceptual orientation, however, tracking college science major students perceptions of mentor trustworthiness is justified because people often act based on perceptions or other information they think is true, even when it is not (Rosenthal & Jacobson, 1968). Succinctly, meaning is causal for human action (Erickson, 1986).
6. Beyond its social antecedents, the educational importance of motivation has also been widely studied (Benware & Deci, 1984; Connell & Wellborn, 1991; Dweck, 1986; Harackiewicz et al., 2002; McDonald, Ing, & Marcoulides, 2010; Wentzel & Wigfield, 2009). Spence and Helmreich (1983) found that, even controlling for prior SAT scores, motivation predicts grade point averages among college undergraduates. Others demonstrated its association with improved science outcomes for students of color in higher education (Brown, 2002; Carlone & Johnson, 2007; Russell & Atwater, 2005). Engaging in an activity for its inherent satisfactionsthe aspect of cognitive regulation that psychologists often refer to as intrinsic motivationcaptures particular attention for its many educational advantages, including more behavioral effectiveness, enhanced self-esteem, and better assimilation of individuals within their social groups (Gottfried & Gottfried, 1996; Ryan & Deci, 2000; Schunk & Zimmerman, 2008).
7. For the purpose of survey distribution, the Program directors at each campus were asked if they would be willing to distribute survey packets to their Program-sponsored students (and/or allow members of the research team to do so). Program directors were also asked if they would be willing to facilitate access to one or more high-enrollment classes required for all or most upper division juniors and seniors majoring in the biomedical sciences.
8. Of the 161 undergraduate science major students, 119 completed all three waves of the survey, while 30 had data for the base year and first follow up, and 12 had data for the base year and second follow up. To determine the extent of similarity on study variables between students who did and did not respond to all three survey waves, analyses of variance, controlling for program status, were performed for each continuous observed measure in the base year sample. These analyses revealed no significant differences between students who did and students who did not complete both follow-up surveys on any of the items in the survey (p > .05).
9. Judging the adequacy of the Program sample size (N=49) involves recognizing that the MARC-U*STAR Program sponsors only a select number of talented upper division science major students on each campus. Among the 16 campuses in our survey, the modal number of Program enrollees was quite smalljust 6 MARC-U*STAR students per campus (with a range from 3 to 16 Program-sponsored students).
10. The software employed in the analyses confronts missing data with estimation by full information maximum likelihood (FIML) instead of relying on ad hoc methods such as listwise or pairwise deletion, or mean imputation. Unlike many other imputation methods, FIML estimation uses all the information from the observed data, estimating a coefficient for the relationship between variables (missing data are built directly into the estimation method), as opposed to imputing a value for an otherwise observed variable. For details, see Arbuckle (1996).
11. Data from the spring 2007 wave include responses from recent graduates who were prompted to reflect back on the final term of their undergraduate program when answering survey questions.
12. We examined whether motivation mediates the impact of trust on science outcomes using the four-step process described by Baron and Kenny (1986) and others (e.g., Kenny et al., 1998; Mackinnon & Fairchild, 2009). According to this framework, mediation requires that at least two of four criteria (the second and third criteria below) are met. First, the initial variable (Trustworthy Mentor) should significantly impact the dependent variables (expectations to attend graduate school and become a research scientist). Second, the initial variable should significantly impact the mediator (Internalized Motivation). Third, the mediator should significantly impact the dependent variables. Finally, complete and consistent mediation require that the impact of the initial variable on the dependent variables is clearly zero when the mediator (its impact on the dependent variable) is included in the model.
13. Because control variables and our latent measure of trust were always exogenous, covariances among these variables and factor loadings for trust were always pre-estimated and imputed in subsequent models that contained endogenous variables of interest. After examining the association between trust and motivation (see Figure 2), we noted the values of all the paths and the factor loadings for Internalized Motivation and imputed them in subsequent models that contained additional endogenous/dependent variables.
14. TLI and CFI are practical fit indices designed to address sample size issues; values of .9 and above indicate reasonable model fit. RMSEA is sensitive to the number of estimated parameters, acting as a barometer in estimating model parsimony. RMSEA values of .08 or less indicate reasonable model fit (Browne & Cudeck 1993). AIC is best used to compare competing models, where smaller values indicate a superior model fit.
15. The vast majority of students in the Program (87%) identified their formal MARC-U*STAR Program mentor, an assigned faculty advisor, or other faculty member as the most influential and helpful institutional agent, while 10% indicated other college personnel. One Program student identified a college guidance counselor; one other Program student indicated a graduate student was most helpful.
16. Although there are no set standards for interpreting differences between group means, Cohen (1988, pp. 2527) suggested that standardized differences of .20 are small, .50 are medium, and .80 are large.
17. In Figure 2, the trust-to-motivation path is freed for the Program students, but constrained to zero for non-Program students, while the path linking internalized motivation in 2005 and 2006 is free to vary across groups. Difference testing supports the tenability of this single-additional constraint (χ²Δ (1) = .25, p > .15).
18. In multivariate analyses, Cohen (1988, pp. 413–414) suggested that standardized parameter estimates of .13 are small, .36 are medium, and .51 are large.
19. In each of these models, we impute, based on coefficients obtained from previous models, (a) covariances among exogenous variables, (b) loadings and item residuals for trust, (c) loadings for internalized motivation as measured in 2006, and (d) all path coefficients directed at internalized motivation. We also constrain or free control paths across groups according to findings from previous analyses. All other coefficients are free. See the Appendix.
20. Time-ordered evaluation of the Program depended, in part, on whether the trust items tapped the same latent construct at each measurement wave. Thus, we constrained and then freed interyear factor loadings for items that corresponded to trust and found, in support of temporal invariance, that the models fit equally well (χ²Δ (6) = 5.40, p > .15).
21. To overlook theory as extraneous or even an impediment to STEM education policy could hamper research in science education, the evaluation of government-sponsored programs in the sciences, and the enormous potential of teaching science to undergraduates (Braxton, 2000; Crisp et al., 2009).
22. Such research might furnish particularly vital new information about how pedagogical practices can affect the relation of trust and mistrust to equity in undergraduate science. Recall, for instance, that our evaluation of the MARC-U*STAR Program showed it had boosted students interpersonal trust in mentors. However, Web-based and published descriptions of the Program do not mention trust, and mentoring is specified as academic and is most often described as monitoring student grades, requirements for graduation, and participation in Program events. Therefore, some benefits of the Programstudent motivation and career expectations spurred by the generation of trust between undergraduates and their faculty mentorsare produced without anyone in the Program explicitly planning or recognizing them. They are unintended consequences. Yet, without self-conscious knowledge of what the Program is trying to produce, faculty cannot fully monitor or moderate their activities, recognize their accomplishments, or disseminate the Program as a practical, efficacious model for the wider community of undergraduate science education.
23. For instance, do particular facets of faculty identities have to align with or even match undergraduates identities for trustworthy, pedagogical relations to develop in science classrooms? Do mentors ethnoracial characteristics have to match their mentees, especially given research that documents a general, negative association in the United States between being non-White and trusting others and institutions (Brehm & Rahn, 1997; Weaver, 2006)?
Adams, C. M., Forsyth, P. B., & Mitchell, R. M. (2009). The formation of parentschool trust: A multilevel analysis. Educational Administration Quarterly, 45, 433.
Allen, W. R., Epps, E., & Haniff, N. (Eds.). (1991). College in Black and White: African American students in predominantly White and in historically Black public universities. Albany, NY: State University of New York Press.
Alon, S., Domina, T., & Tienda, M. (2010). Stymied mobility or temporary lull? The puzzle of lagging Hispanic college degree attainment. Social Forces, 88, 18071832.
Anaya, G., & Cole, D. (2001). Latina/o student achievement: Exploring the influence of studentfaculty interaction on college grades. Journal of College Student Development, 42, 314.
Ancis, J. R., Sedlacek, W., & Mohr, J. (2000). Student perceptions of campus cultural climate by race. Journal of Counseling and Development, 78, 180185.
Anderson, E., & Kim, D. (2006). Increasing the success of minority students in science and technology. Washington, DC: American Council on Education.
Arbuckle, J. L. (1996). Full information estimation in the presence of incomplete data. In G. Marcoulides & R. Schumacker (Eds.), Advanced structural equation modeling: Issues and technique (243278). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Asparouhov, T., & Muthén, B. (2005). Multivariate statistical testing with survey data. Proceedings of the Federal Committee on Statistical Methodology (FCSM) Research Conference. Coyne, MD.
Astin, A. (1993). What matters in college: Four critical years revisited. San Francisco, CA: Jossey-Bass.
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall.
Baron, R., & Kenny, D. (1986). The moderatormediator variable distinction in social psychological research: Conceptual, strategic, and statistical consideration. Journal of Personality and Social Psychology, 51, 11731182.
Becker, G. (1964). Human capital. New York, NY: National Bureau of Economic Research.
Bensimon, E. (2005). Closing the achievement gap in higher education: An organizational learning perspective. New Directions for Higher Education, 131, 99111.
Bentler, P. M. (1995). EQS structural equations program manual. Encino, CA: Multivariate Software, Inc.
Benware, C., & Deci, E. (1984). Quality of learning with an active versus passive motivational set. American Educational Research Journal, 21, 755765.
Bogart L., & Bird, S. (2003). Exploring the relationship of conspiracy beliefs about HIV/AIDS to sexual behaviors and attitudes among African American adults. Journal of the National Medical Association, 95, 10571065.
Bohon, S. A., Johnson, M. K., & Gorman, B. K. (2006). College aspirations and expectations among Latino adolescents in the United States. Social Problems, 53, 207225.
Bourdieu, P. (1984). Distinction: A social critique of the judgment of taste. Cambridge, MA: Harvard University Press.
Bourdieu, P. (1986). The forms of capital. In J. Richardson (Ed.), Handbook of theory and research on the sociology of education (pp. 96111). New York, NY: Greenwood Press.
Braxton, J. M. (Ed.). (2000). Reworking the student departure puzzle. Nashville, TN: Vanderbilt University Press.
Brehm, J., & Rahn, W. (1997). Individual-level evidence for the causes and consequences of social capital. American Journal of Political Science, 41, 9991023.
Brewer, D., Eide, E., & Ehrenberg, R. (1999). Does it pay to attend an elite private college? Cross-cohort evidence on the effects of college type on earnings. Journal of Human Resources, 34, 104123.
Brint, S., Cantwell, A., & Hanneman, R. (2008). The two cultures of undergraduate academic engagement. Research in Higher Education, 49, 383402.
Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press.
Brown, S. W. (2002). Hispanic students majoring in science or engineering: What happened in their educational journeys? Journal of Women and Minorities in Science and Engineering, 8(2), 123148.
Browne, M., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136159). Thousand Oaks, CA: Sage Publications.
Brubaker, R. (1985). Rethinking classical theory: The sociological vision of Pierre Bourdieu. Theory and Society, 14, 745775.
Bryk, A., & Schneider, B. (2002). Trust in schools: A core resource for improvement. New York, NY: Russell Sage Foundation.
Cabrera, A. F., & LaNasa, S. M. (2000). Overcoming the tasks on the path to college for America's disadvantaged. New Directions for Institutional Research, 107, 3144.
Cabrera, A. F., & Nora, A. (1994). College students perceptions of prejudice and discrimination and their feelings of alienation: A construct validation approach. Review of Education/Pedagogy/Cultural Studies, 16, 387409.
Carlone, H. B., & Johnson, A. (2007). Understanding the science experiences of successful women of color: Science identity as an analytic lens. Journal of Research in Science Teaching, 44, 11871218.
Carter, D. F. (2001). A dream deferred?: Examining the degree aspirations of African American and White college students. New York, NY: Routledge Flamer Press.
Carter, D. F., Mandell, M., & Maton, K. I. (2009). The influence of on-campus, academic year undergraduate research on STEM Ph.D. outcomes: Evidence from the Meyerhoff Scholarship Program. Educational Evaluation and Policy Analysis, 31, 441462.
Center for Institutional Data Exchange Analysis. (2000). 19992000 SMET retention report. Norman, OK: University of Oklahoma.
Chang, M. J., Denson, N., Saenz, V., & Misa, K. (2006). The educational benefits of sustaining cross-racial interaction among undergraduates. The Journal of Higher Education, 77, 430455.
Chang, M. J., Eagan, M., Lin, M., & Hurtado, S. (2011). Considering the impact of racial stigmas and science identity: Persistence among biomedical and behavioral science aspirants. The Journal of Higher Education 82, 564596.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Cole, D. (2007). Do interracial interactions matter? An examination of studentfaculty contact and intellectual self-concept. The Journal of Higher Education, 78(3), 249281.
Cole, D., & Espinoza, A. (2008). Examining the academic success of Latino students in science, technology, engineering, and mathematics (STEM) majors. Journal of College Student Development, 49, 285300.
Coleman, J. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95S120.
Coleman, J. (1990). Foundations of social theory. Cambridge, MA: Harvard University Press.
Connell, J., & Wellborn, J. (1991). Competence, autonomy and relatedness: A motivational analysis of self-system processes. In M. R. Gunnar & L. A. Sroufe (Eds.), Minnesota symposium on child psychology (Vol. 22), Hillsdale, NJ: Erlbaum.
Cook, K. S. (Ed.). (2001). Trust in society. New York, NY: Russell Sage Foundation.
Cooper, H., Robinson, J., & Patall, E. (2006). Does homework improve academic achievement? A synthesis of research, 19872003. Review of Educational Research, 76, 162.
Crisp, G., Nora, A., & Taggart, A. (2009). Student characteristics, pre-college, college, and environmental factors as predictors of majoring in and earning a STEM degree: An analysis of students attending a Hispanic serving institution. American Educational Research Journal, 46, 924942.
Cronan-Hillix, T., Gensheimer, L., Cronan-Hillix, W., & Davidson, W. (1986). Students' views of mentors in psychology graduate training. Teaching of Psychology, 13, 123127.
Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. New York, NY: Plenum Press.
Deci, E. L., & Ryan, R. M. (2000). The what and why of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11, 227268.
Deil-Amen, R., & Rosenbaum, J. (2003). The social prerequisites of success. Can college structure reduce the need for social know-how? The Annals of the American Academy of Political and Social Science, 586,120143.
DePass, A., & Chubin, D. (Eds.). (2009). Understanding interventions that encourage minorities to pursue research careers: Building a community of research and practice. Bethesda, MD: American Society for Cell Biology.
Desmond, M., & Lopez Turley, R. (2009). The role of familism in explaining the Hispanic-White college application gap. Social Problems, 56, 311334.
Dika, S., & Singh, K. (2002). Applications of social capital in educational literature: A critical synthesis. Review of Educational Research, 72, 3160.
DuBois, D., & Karcher, M. (Eds.). (2005). Handbook of youth mentoring. Thousand Oaks, CA: SAGE Publications.
Dunn, J. R., & Schweitzer, M. E. (2005). Feeling and believing: The influence of emotion on trust. Journal of Applied Psychology, 88, 736748.
Dweck, C. S. (1986). Motivational processes affecting learning. American Psychologist, 41(10), 1040-1048.
Elliot, A., & Dweck, C. (Eds.). (2005). Handbook of competence and motivation. New York, NY: Guilford Press.
Enman, M., & Lupart, J. (2000). Talented female students' resistance to science: An exploratory study of post-secondary achievement motivation, persistence, and epistemological characteristics. High Ability Studies, 11, 161178.
Erickson, F. (1986). Culture difference in science education. The Urban Review, 18, 117124.
Erickson, F. (1993). Transformation and school success: The politics and culture of educational achievement. In C. Jordan & E. Jacob (Eds.), Minority education: Anthropological perspectives (pp. 335356). Norwood, NJ: Ablex Publishing Corporation.
Fukuyama, F. (1995). Trust: The social virtues and the creation of prosperity. New York, NY: Free Press.
Gambetta, D. (Ed.). (1988). Trust. Oxford, UK: Blackwell.
Gamble, V. (1997). Under the shadow of Tuskegee: African Americans and health care. American Journal of Public Health, 87, 17731778.
Gauchat, G. (2008). A test of three theories of anti-science attitudes. Sociological Focus, 41, 337357.
Gauchat, G. (2012). Politicization of science in the public sphere: A study of public trust in the United States, 1974 to 2010. American Sociological Review, 77, 167187.
Ghosh, A., Whipple, T. W., & Bryan, G. A. (2001). Student trust and its antecedents in higher education The Journal of Higher Education, 72, 322340.
Glaeser, E. L., Laibson, D. I., Scheinkman, J. A., & Soutter, C. L. (2000). Measuring trust. Quarterly Journal of Economics, 115, 811846.
Goddard, R. D. (2003). Relational networks, social trust, and norms: A social capital perspective on students chances of academic success. Educational Evaluation and Policy Analysis, 25, 5974.
Gottfried, A. E., & Gottfried, A. W. (1996). A longitudinal study of academic intrinsic motivation in intellectually gifted children: Childhood through early adolescence. Gifted Child Quarterly, 40, 179183.
Granovetter, M. (2002). A theoretical agenda for economic sociology. In M. F. Guillen, R. Collins, P. England, & M. Meyer (Eds.), Economic sociology at the millennium (pp. 3560). New York, NY: Russell Sage Foundation.
Guiffrida, D. (2005). Othermothering as a framework for understanding African American students definitions of student-centered faculty. The Journal of Higher Education, 76, 701723.
Gutting, G. (2012, May 17). How reliable are the social sciences? [Web log post]. Retrieved from http://opinionator.blogs.nytimes.com/2012/05/17/how-reliable-are-the-social-sciences/
Handelsman, J., et al. (2004). Scientific teaching. Science, 304, 521-522.
Hannaway, J., & Woodroffe, N. (2003). Policy instruments in education. Review of Research in Education, 27, 124.
Hanson, S. L. (1994). Lost talent: Unrealized educational aspirations and expectations among U.S. youths. Sociology of Education, 67, 159183.
Harackiewicz, J. M., Barron, K., & Elliot, A. (1998). Rethinking achievement goals: When are they adaptive for college students and why? Educational Psychologist, 33, 121.
Harackiewicz, J. M., Barron, K. E., Tauer, J. M., & Elliot, A. J. (2002). Predicting success in college: A longitudinal study of achievement goals and ability measures as predictors of interest and performance from freshman year through graduation. Journal of Educational Psychology, 94, 562-575
Hardin, R. (2002). Trust and trustworthiness. New York, NY: Russell Sage Foundation.
Heckman, J. (2008). Schools, skills, and synapses. Economic Inquiry 46, 289324.
Hilton, T., & Lee, V. (1988). Student interest and persistence in science: Changes in the educational pipeline in the last decade. Journal of Higher Education, 59, 510526.
House, J. S. (1981). Work stress and social support. Reading, MA: Addison-Wesley.
Hoy, W., Gage, C., & Tarter, C. (2006). School mindfulness and faculty trust: Necessary conditions for each other? Educational Administration Quarterly, 42, 236255.
Hurtado, S. N. (1994). The institutional climate for talented Latino students. Research in Higher Education, 35, 2141.
Hurtado, S. N., Cabrera, M., Arellano, L., & Espinosa, L. (2009). Diversifying science: Underrepresented student experiences in structured research programs. Research in Higher Education, 50, 189214.
Hurtado, S. N., Carter, D. F., & Spuler, A. (1996). Latino student transition to college: Assessing difficulties and factors in successful college adjustment. Research in Higher Education, 37, 135157.
Jacobi, M. (1991). Mentoring and undergraduate academic success: A literature review. Review of Educational Research, 61, 505532.
Jones, E., & Nisbett, R. (1972). The actor and the observer: Divergent perceptions of the causes of behavior. In E. Jones, D. Kanouse, H. Kelley, R. Nisbett, S. Valins, & B. Weiner (Eds.), Attribution: Perceiving the causes of behavior (pp. 7994). Morristown, NJ: General Learning.
Kandel, D. (1996). The parental and peer contexts of adolescent deviance: An algebra of interpersonal influences. Journal of Drug Issues, Special Issue, 26, 289315.
Kane, T. J., Rockoff, J., & Staiger, D. (2008). What does certification tell us about teacher effectiveness? Evidence from New York City. Economics of Education Review, 27, 615631.
Kao, G. (2004). Social capital and its relevance to minority and immigrant populations. Sociology of Education, 77, 172175.
Kao, G., & Thompson, J. (2003). Racial and ethnic stratification in educational achievement and attainment. Annual Review of Sociology, 29, 417442.
Kenny, D., Kashy, D., & Bolger, N. (1998). Data analysis in social psychology. In D. Gilbert, S. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (pp. 233265). Boston, MA: McGraw-Hill.
Kramer, R. M. (Ed.). (2006). Organizational trust. New York, NY: Oxford University Press.
Kuh, G. D., & Hu, S. (2001). The effects of student-faculty interaction in the 1990s. The Review of Higher Education, 24, 309332.
Lamont, M., & Lareau, A. (1988). Cultural capital: Allusions, gaps and glissandos in recent theoretical developments. Sociological Theory, 6, 153168.
Lee, O. (1997). Scientific literacy for all: What is it and how can we achieve it? Journal of Research in Science Teaching, 34, 219222.
Lent, R. W., Brown, S., Schmidt, J., Brenner, B., Lyons, H., & Treistman, D. (2003). Relation of contextual supports and barriers to choice behavior in engineering majors: Test of alternative social cognitive models. Journal of Counseling Psychology, 50, 458465.
Lewis, B. (2003). A critique of literature on the underrepresentation of African Americans in science: Directions for future research. Journal of Women and Minorities in Science and Engineering, 9, 361373.
Lin, N. (2001). Social capital: A theory of social structure and action. Cambridge, UK: Cambridge University Press.
Little, R., & Rubin, D. (2002). Statistical analysis with missing data (2nd ed.). New York, NY: Wiley.
Locks, A., Hurtado, S., Bowman, N., & Oseguera, L. (2008). Extending notions of campus climate and diversity to students transition to college. The Review of Higher Education, 31, 257285.
Loehlin, J. C. (1998). Latent variable models: An introduction to factor, path, and structural analysis. Mahwah, NJ: Lawrence Erlbaum Associates.
Louis, K. S. (2007). Trust and improvement in schools. Journal of Educational Change, 8, 124.
Mackinnon, D. P., & Fairchild, A. J. (2009). Current directions in mediation analysis. Current Directions in Psychological Science, 18, 1620.
Major, B., & OBrien, L. (2005). The social psychology of stigma. Annual Review of Psychology, 56, 393421.
Malcom, L. E., Dowd, A. C., & Yu, T. (2010). Tapping HSI-STEM funds to improve Latina and Latino access to the STEM professions. Los Angeles, CA: University of Southern California.
Maldonado, D. E. Z., Rhoads, R., & Buenavista, T. L. (2005). The student-initiated retention project: Theoretical contributions and the role of self-empowerment. American Education Research Journal, 42, 605638.
Maton, K., Hrabowski, F., & Schmitt, C. (2000). African American college students excelling in the sciences: College and post-college outcomes of the Meyerhoff Scholars Program. Journal of Research in Science Teaching, 37, 629654.
Mayer, R., Davis, J., & Schoorman, F. (1995). An integrative model of organizational trust. Academy of Management Review, 20, 438458.
McDonald, S., Ing, M., & Marcoulides, G. (2010). An investigation of early parental motivational strategies on mathematics achievement by ethnicity: A latent curve model approach. Educational Research and Evaluation, 16, 401419.
Melguizo, T. (2010). Are students of color more likely to graduate from college if they attend more selective institutions? Evidence from a cohort of recipients and nonrecipients of the Gates Millennium Scholarship Program. Educational Evaluation and Policy Analysis, 32, 230248.
Melguizo, T. (2011). A review of the theories developed to describe the process of college persistence and attainment. In J. C. Smart & M. B. Paulses (Eds.), Higher education: Handbook of theory and research (pp. 395424). Netherlands: Springer Science Business Media B.V.
Mervis, J. (2006, April). NIH told to get serious about giving minorities a hand. Science, 311, 328329. Retrieved from http://www.sciencemag.org/cgi/content/full/310/5752/1263b?eaf
Milem, J. (1998). Attitude change in college students: Examining the effect of college peer groups and faculty normative groups. The Journal of Higher Education, 69, 117140.
Mitchell, D., & Spady, W. (1983). Perspective: Authority, power and the legitimation of social control. Educational Administration Quarterly 19, 533.
Moorman, C., Zaltman, G., & Deshpande, R. (1992). Relationships between providers and users of market research: The dynamics of trust within and between organizations. Journal of Marketing Research, 29, 314328.
Mutegi, J. W. (2011). The inadequacies of Science for All and the necessity and nature of a socially transformative curriculum approach for African American science education. Journal of Research in Science Teaching, 248, 301316.
National Research Council. (2005). Assessment of NIH minority research and training programs: Phase 3. Oversight Committee for the Assessment of NIH Minority Research Training Programs. Washington, DC: The National Academies Press.
National Research Council. (2009). Rising above the gathering storm two years later: Accelerating progress toward a brighter economic future. Washington, DC: The National Academies Press.
National Research Council. (2010). Exploring the intersection of science education and 21st century skills: A workshop summary. Washington, DC: The National Academies Press.
National Research Council. (2011a). Assessing 21st Century skills: Summary of a workshop. Washington, DC: The National Academies Press.
National Research Council. (2011b). Expanding underrepresented minority participation: Americas science and technology talent at the crossroads. Washington, DC: The National Academies Press.
National Research Council. (2012). Discipline-based education research: Understanding and improving learning in undergraduate science and engineering. Washington, DC: The National Academies Press.
National Science Foundation. (2006). Science and engineering degrees, by race/ethnicity of recipients: 19952004. Arlington, VA: National Science Foundation.
National Science Foundation. (2009). Women, minorities, and persons with disabilities in science and engineering: 2009 (NSF 09305). Arlington, VA: National Science Foundation.
Nespor, J. (1987). Knowledge in motion: Space, time, and curriculum in undergraduate physics and science. London, UK: Falmer Press.
Nettles, M. (1991). Racial similarities and differences in the predictors of college student achievement. In W. R. Allen, E. Epps, & N. Haniff (Eds.), College in Black and White: African American students in predominantly White and in historically Black public universities (pp. 7591). Albany, NY: State University of New York Press.
Nettles, M., & Millett, C. (2006). Three magic letters: Getting the Ph.D. Baltimore, MD: Johns Hopkins University Press.
OConnor, N., Hammack, F., & Scott, M. (2009). Social capital, financial knowledge, and Hispanic student college choices. Research in Higher Education, 51, 195219.
Oliver, A., & Montgomery, K. (2001). A system cybernetic approach to the dynamics of individual- and organizational-level trust. Human Relations, 54, 10451063.
Orr, M. (1999). Black social capital: The politics of school reform in Baltimore, 19861998. Lawrence, KS: University Press of Kansas.
Ortner, S. (2006). Anthropology and social theory: Culture, power, and the acting subject. Durham, NC: Duke University Press.
Page, R. (1991). Lower-track classrooms: A curricular and cultural perspective. New York, NY: Teachers College Press.
Pascarella, E., & Terenzini, P. (2005). How college affects students: A third decade of research (Vol. 2). San Francisco, CA: Jossey-Bass.
Patton, L., Morelon, C., Whitehead, D., & Hossler, D. (2006), Campus-based retention initiatives: Does the emperor have clothes? New Directions for Institutional Research, 2006, 924.
Perna, L. (2006). Studying college access and choice: A proposed conceptual model. Higher Education, 21, 91157.
Perna, L., & Thomas, S. (2008). Theoretical perspectives on student success: Understanding the contributions of the disciplines. Hoboken, NJ: Wiley.
Perna, L., & Titus, M. (2005). The relationship between parental involvement as social capital and college enrollment: An examination of racial/ethnic group differences. Journal of Higher Education, 76, 485518.
Pettigrew, T. (1998). Intergroup contact theory. Annual Review of Psychology, 49, 6585.
Portes, A. (1998). Social capital: Its origin and applications in modern sociology. Annual Review of Sociology, 24, 124.
Putnam, R. (2000). Bowling alone: The collapse and revival of American community. New York, NY: Simon and Schuster.
Ream, R. (2003). Counterfeit social capital and Mexican American achievement. Educational Evaluation and Policy Analysis 25, 237262.
Ream, R., & Palardy, G. (2008). Re-examining social class differences in the availability and the educational utility of parental social capital. American Educational Research Journal 45, 238273.
Ream, R., Ryan, S., & Espinoza, J. (2012). Reframing the ecology of opportunity and achievement gaps: Why no excuses reforms have failed to narrow student group differences in educational outcomes. In T. Timar & J. Maxwell-Jolly (Eds.), Narrowing the achievement gap: Perspectives and strategies for challenging times (pp.3556). Cambridge, MA: Harvard Education Press.
Rhodes, J. (2002). Stand by me: The risks and rewards of mentoring todays youth. Cambridge, MA: Harvard University Press.
Riegle-Crumb, C., Moore, C., & Ramos-Wada, A. (2011). Who wants to have a career in science or math? Exploring adolescents' future aspirations by gender and race/ethnicity. Science Education, 95, 458476.
Roscigno, V. (2000). Family/school inequality and African-American/Hispanic achievement. Social Problems, 47, 266290.
Rose, M. (2013, January 16). Commentary: Giving cognition a bad name. Education Week, 32.
Rosenhan, D. (1973). On being sane in insane places. Science, 179, 250258.
Rosenthal, R., & Jacobson, L. (1968). Pygmalion in the classroom. The Urban Review, 3, 1620.
Rotter, J. B. (1980). Interpersonal trust, trustworthiness, and gullibility. American Psychologist,35, 17.
Rousseau, D., Sitkin, S., Burt, R., & Camerer, C. (1998). Introduction to special topic forum: Not so different after all: A cross-discipline view of trust. The Academy of Management Review, 23, 393404.
Rozin, P. (2001). Social psychology and science: Some lessons from Solomon Asch. Personality and Social Psychology Review, 5, 214.
Russell, M. L., & Atwater, M. M. (2005). Traveling the road to success: A discourse on persistence throughout the science pipeline with African American students at a predominantly White institution. Journal of Research in Science Teaching, 42, 691715.
Russell Sage Foundation. (2000). Examining the role of trust in society. Russell Sage Foundation NEWS, 6, 287290.
Ryan, R., & Deci, E. (2000). Self-Determination Theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55, 6878.
Ryan, R., & Grolnick, W. (1986). Origins and pawns in the classroom: Self-report and projective assessments of individual differences in childrens perceptions. Journal of Personality and Social Psychology, 50, 550558.
Sánchez, B., & Colón, Y. (2009). Race, ethnicity, and culture in mentoring relationships. In D. DuBois & M. Karcher (Eds.), Handbook of youth mentoring (pp. 191204). Thousand Oaks, CA: SAGE Publications.
Sandefur, G. D., Meier, A. M., & Campbell, M. E. (2006). Family resources, social capital, and college attendance. Social Science Research, 35, 525553.
Satorra, A. (2000). Scaled and adjusted restricted tests in multisample analysis of moment structures. In R. Heijmans, D. Pollock, & A. Satorra (Eds.), Innovations in multivariate statistical analysis: A festschrift for Heinz Neudecker (pp. 233247). London, UK: Kluwer Academic Publishers.
Satorra, A., & Bentler, P. (1999). A scaled difference chi-square test statistic for moment structure analysis (UCLA Statistics Series #260). Los Angeles, CA: University of California.
Schneider, B., Carnoy, M., Kilpatrick, J., Schmidt, W. H., & Shavelson, R. J. (2007). Estimating causal effects using experimental and observational designs (report from the Governing Board of the American Educational Research Association Grants Program). Washington, DC: American Educational Research Association.
Schoorman, D., Mayer, R., & Davis, J. (2007). An integrative model of organizational trust: Past, present, and future. Academy of Management Review, 32, 344354.
Schunk, D. H., & Zimmerman, B. J. (Eds.). (2008). Motivation and self-regulated learning: Theory, research, and applications. New York, NY: Lawrence Erlbaum Associates.
Seligman, A. B. (1997). The problem of trust. Princeton, NJ: Princeton University Press.
Seymour, E., & Hewitt, N. (1997). Talking about leaving: Why undergraduates leave the sciences. Boulder, CO: Westview Press.
Solórzano, D., Ceja, M., & Yosso, T. (2000). Knocking at freedom's door: Race, equity, and affirmative action in U.S. higher education. The Journal of Negro Education, 69, 6073.
Spady, W. G. (1971). Dropouts from higher education: Toward an empirical model. Interchange, 2, 3862.
Spence, J. T., & Helmreich, R. L. (1983). Achievement-related motives and behaviors. In J. Spence (Ed.), Achievement and achievement motives: Psychological and sociological approaches (pp. 1074). San Francisco, CA: Freeman.
Steele, C. (1997). A threat in the air: How stereotypes shape intellectual identity and performance. American Psychologist, 52, 613629.
Steele, C. (2010). Whistling Vivaldi: And other clues to how stereotypes affect us. New York, NY: W. W. Norton & Company, Inc.
Steele, C., & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology, 69, 797811.
Summers, M. F., & Hrabowski, F. A. (2006). Preparing minority scientists and engineers. Science, 311, 18701871.
Tienda, M., & Jensen, L. (1988). Poverty and minorities: A quarter century profile of color and socioeconomic disadvantage. In G. Sandefur & M. Tienda (Eds.), Divided opportunities: Minorities, poverty, and social policy (pp.2361). New York, NY: Plenum Publishers.
Tierney, W. G. (1999). Models of minority college-going and retention: Cultural integrity versus cultural suicide. The Journal of Negro Education, 68, 8091.
Tinto, V. (1993). Leaving college: Rethinking the causes and cures of student attrition. Chicago, IL: University of Chicago Press.
Treisman, U. (1992). Studying students studying calculus: A look at the lives of minority mathematics students in college. The College Mathematics Journal, 23, 362372.
Tschannen-Moran, M. (2004). Trust matters: Leadership for successful schools. San Francisco, CA: Jossey-Bass.
Tschannen-Moran, M., & Hoy, W. (2000). A multidisciplinary analysis of the nature, meaning, and measurement of trust. Review of Educational Research, 70, 547593.
U.S. Department of Education. (2007). Report of the academic competitiveness council. Washington, DC: National Center for Education Statistics, Institute of Education Sciences.
Weaver, C. N. (2006). Trust in people among Hispanic Americans. Journal of Applied Social Psychology, 36, 11601172.
Weiner, B. (1991). Metaphors in motivation and attribution. American Psychologist, 46, 921930.
Wentzel, K., & Wigfield, A. (Eds.). (2009). Handbook of motivation at school. New York, NY: Routledge.
Westland, J. (2010). Lower bounds on sample size in structural equation modeling. Electronic Commerce Research and Applications, 9, 476487.
Yuan, K-H., & Bentler, P. M. (2000). Three likelihood-based methods for mean and covariance structure analysis with non-normal missing data. Sociological Methodology, 30, 167202.
Zucker, L. G. (1986). The production of trust: Institutional sources of economic structure, 18401920. In B. Staw & L. Cummings (Eds.), Research in organizational behavior (Vol. 8, pp. 55111), Greenwich, CT: JAI Press.
COVARIANCES AMONG EXOGENOUS VARIABLES
We began by estimating covariances among exogenous variables including URM, SES, MONTHS, 2006 Trustworthy Mentor, and 2005 Internalized Motivation across groups. The model in which all covariances were constrained to equality across groups (χ²(34) = 20.79) and the model in which all were freed across groups (χ²(25) = 12.334) fit equally well (χ²Δ (9) = 9.25, p > .15). This suggested that covariances among these exogenous variables were equal across groups. Similarly, covariances were equal across groups in the model that included URM, SES, MONTHS, 2006 TR, 2005 IM, and 2005 Graduate School (χ²Δ (5) = 6.06, p > .15). Covariances were also equal across groups in the model that included URM, SES, MONTHS, 2006 TR, 2005 IM, and 2005 Research Scientist (χ²Δ (5) = 1.55, p > .15). Thus, all covariances were constrained to equality across groups and imputed to match the values obtained here in subsequent analyses.
We began by regressing 2006 IM on controls including URM, SES, MONTHS, and 2005 IM. The model in which all the paths were equal (χ²(37) = 35.64) and the model in which all paths were freed across groups (χ² (33) = 24.10) had a significantly different fit (χ²Δ (4) = 11.39, p < .01). To identify which control variable(s) had a significantly different impact on 2006 IM, we tested competing models in which individual paths were freed and compared to the model in which all paths were constrained. Models in which URM, SES, and MONTHS were individually freed did not differ from the fully constrained model. However, the model in which 2005 IM was freed fit significantly better than the model in which all paths were constrained (χ²Δ (1) = 31.81, p < .001), suggesting that this path was different across groups. As such, this path was freed to vary across groups in subsequent models where dependent and independent variables of interest were included.
Next, we regressed 2007 Graduate School on controls including URM, SES, MONTHS, and 2005 GS. The model in which all the paths were equal (χ²(16) = 17.90) and the model in which all paths were freed across groups (χ² (12) = 7.55) had a significantly different fit (χ²Δ (4) = 11.20, p < .01). To identify which control variable(s) had a significantly different impact on 2007 GS, we tested competing models in which individual paths were freed and compared to the model in which all paths were constrained. Models in which URM, SES, and MONTHS were individually freed did not differ from the fully constrained model. However, the model in which 2005 GS was freed fit significantly better than the model in which all paths were constrained (χ²Δ (1) = 9.53, p < .01), suggesting that this path was different across groups. As such, this path was freed to vary across groups in subsequent models where dependent and independent variables of interest were included.
Last, we regressed 2007 Research Scientist on controls including URM, SES, MONTHS, and 2005 RS. The model in which all the paths were equal (χ²(16) = 10.21) and the model in which all paths were freed across groups (χ² (12) = 6.60) did not differ significantly (χ²Δ (4) = 3.74, p > .15), suggesting that the paths were equal across groups. As such, these paths were constrained to equality across groups in subsequent models where dependent and independent variables of interest were included. See Figures 6–8.
Appendix Figure A1. Covariances among exogenous variables across groups.
Note: Coefficients are unstandardized. All loadings and covariances are equal across groups.
Appendix Figure A2. Covariances among exogenous variables across groups.
Note. Coefficients are unstandardized. All loadings and covariances are equal across groups.
Appendix Figure A3. Covariances among exogenous variables across groups.
Note. Coefficients are unstandardized. All loadings and covariances are equal across groups.
Cite This Article as: Teachers College Record, 2014, p. 1-50
http://www.tcrecord.org ID Number: 17428, Date Accessed: 10/19/2018 3:03:36 PM