Many studies concerning social capital in academia have used social network theory and social network analysis as an approach. Social network analysis focuses on a boundary set of actors in a network and what it reveals as an outcome of social capital. However, social capital is also a precursor or catalyst for cooperative work. Here, we investigate researchers' perceptions of social capital based on the hypothesis that what academics do when they collaborate (as a habit), may not relate to what they actually prefer. We have piloted a questionnaire-survey to test this, focusing on the goal-seeking behavior of publishing new research. Data were collected from 1,092 academics, across 6 faculties at the University of Copenhagen: Health and Medical Science, Science, Social Sciences, Humanities, Law, and Theology. The survey of collaboration habits first revealed significant differences at the level of gender, academic position, years active in publishing, and faculty. Collaboration preferences, interpreted from social capital theory, were also measured according to three interrelated dimensions--i.e. "cognitive," "relational," and "structural." Survey respondents tended to prefer the "cognitive-relational" aspects of collaboration (i.e. "reciprocity and obligation"; "shared understandings"), over certain "structural" determinants (i.e. "linking to networks"). Few habits and preferences correlated and did not correlate strongly. Amongst the researchers who indicated a preference for "bridging networks" (i.e. working with people who have different types of expertise), few confirmed this as a collaboration habit, particularly with experts from organizations outside academia (i.e. the business/public sector).