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.