We present results of a two-semester study to gauge the impact of collaborative two-stage exams on student learning and attitudes in university-level introductory astronomy classes for non-science majors. In the collaborative two-stage exam setting, students first completed an exam individually, and then they reconsidered a subset of exam questions within their previously established groups, discussing the questions with their peers to arrive at a common answer.Students took three to four exams during the semester using this format. At mid-semester, we surveyed the students to gauge their attitudes about collaborative work and its perceived influence on their exam preparation and performance. At the end of the semester, students sat an individual-only final exam, which contained all previous collaborative-phase questions, as well as a subset of questions seen only on the individual portions of the exams. When we compare the normalized gain on final exam questions that were included in the collaborative portions to that on questions found in only the individual portions, we find higher normalized gains in general for questions encountered on the collaborative portions of the exams. These gains are accompanied by a statistically significant effect size (Cohen's d). We note, however, that this improved performance appears to be dependent upon several factors. Those factors might include diminished retention over time, the assessment of overly complex concepts, and concept saturation. Our mid-semester survey indicates that the collaborative experience appears have a positive influence on their overall attitudes and their study habits.