The return rates of 26% for students and 30% for the internal medicine faculty members in survey 3, as well as a response rate of 28.8% in survey 1 can be considered as representative and are comparable to data on return rates for email surveys for university students at 21% as well as for surveys by regular mail at 32% [7, 8]. The overall return rate in survey 2 of 10.8% seems quite low. However, results showed to be reproducible in four consecutive semesters. We therefore assume that these results are reliable.
The ratio of time spent on self study and quantity of instruction (TSS/QI) is an important measure for effective instruction as proposed by Gijselaers and Schmidt . This concept dates back to the time-on-task hypothesis by Carol, which claims that it is time dedicated to studying that determines the learning outcome . Given that medical curricula require processing a big quantity of information, it does not seem counterintuitive that it is crucial to provide students with enough time for self study that can be used for individual learning activities and memorization techniques. With regard to a future curriculum reform, TSS/QI would be a parameter that can be easily modified by reducing QI. Compared to other curricular measures seeking to improve learning outcomes and to increase effectiveness of instruction, like the reduction of group size [10, 11] or a long-term mentorship between students and teachers , an adjustment of TSS/QI could be swiftly implemented and translated into improvement of the curriculum.
Our results show that TSS/QI is unacceptably low in our current curriculum (see Figure 1). This finding is particularly unfavorable when considering that, according to Gijselaers and Schmidt, TSS/QI is positively correlated to academic achievement. Additionally, Gijselaers and Schmidt found that students are not willing to spend more than 37 hours total per week on education. Given the low TSS/QI in our present curriculum, this suggests that our current instruction time is too high. Moreover, in our survey a majority of students wishes for more time for self-study and criticizes that under the present curriculum mandatory courses are not distributed evenly among the internal medicine and surgery semesters. Given the fact that some weeks of the present surgery semester leave almost no free time for self study this criticism appears justified. Therefore it can be concluded that reducing instruction time is necessary for future curriculum changes.
However, the relationship between mandatory lessons and TSS is not strictly linear, in other words we cannot simplify the relationship to the statement "the less we teach the more students study". Van der Drift et al. could show that the relationship is rather curvilinear, meaning that up to a certain extent scheduled lessons enhance TSS while mandatory courses beyond that point rather impede TSS . In their study the ideal QI was found to lie between 8 and 10 hours per week, a value by far below what our current curriculum provides within most weeks. The German law exactly defines the minimum of mandatory instruction time required for sufficient guidance and study stimuli. These requirements are clearly below the quantity of instruction in our current curriculum. Accordingly it seems reasonable to redesign students' schedules so that QI does not to exceed German legal requirements and does not fluctuate significantly between and within semesters.
Apart from time available for self study, and necessary guidance to efficiently use this time, time actually spent on self-study also seems to be influenced by test frequency. Our data suggest that students study mainly prior to exams and that teachers and students consider multiple smaller exams during the semester preferable to one big exam in the end of the semester. The peak of TSS/QI ratio and TSS in week 8 of the internal medicine semester can be explained by a mid-semester exam. In the surgical semester, no such additional exam exists. A mid-semester peak of TSS/QI and TSS can therefore not be detected (Figures 1 and 2). In both semesters a clear increase of TSS/QI over 1 can be shown prior to the respective final exam. These results demonstrate the correlation between time spent on self-study and exam frequency. Findings in cognitive psychology studies in general and in medical education studies in particular further support our conclusion that test frequency should be increased when implementing a new curriculum. Testing is not merely able to serve as an assessment tool but has been shown to provide mnemonic benefits. These beneficial effects of testing on long term retention of information can mainly be attributed to two effects, a direct and an indirect one . The direct effect results from students' awareness that they will be tested on the topics they are reading about or listening to in lectures. According to Larsen et al. this is an even stronger boost to later retention than repetition of the respective topic . Additionally, the retrieval of knowledge itself strengthens long-term memory . The indirect effects of testing result from increased study time and improvement of study strategies with regard to an up-coming exam. This phenomenon can be further promoted by more frequent testing (e.g. monthly) as students are prompted to adopt constant learning habits with an even distribution of study time over the academic year or semester. Studies demonstrate that evenly spaced study sessions promote memory performance . Testing can serve as a tool to influence students learning habits likewise. Accordingly a meta-analysis of forty studies by Bangert-Drowns, Kulik and Kulik has shown, that frequent testing before a final exam results in better performance compared to less frequent (or no) testing prior to the final exam . The effect on performance in exams which do not take place immediately after a course, but two to four years after a respective instruction unit (like the German medical licensing examination) might be more compelling and remains to be investigated. One further beneficial effect of exams is inherent in the process of test taking itself. Studies show that taking a test leads to better retention of information than restudying the material for an equivalent amount of time [17, 18]. Apart from students' performance, there is evidence that more frequent testing improves students' evaluation of faculty [19, 20].
Although our results are comprehensive and conclusions for further curricular changes can be drawn, there are several limitations to this study. One constraint is the relatively low response rate, especially in surveys 1 and 2 that might result in a selection bias, as highly motivated students could be over-represented. However, as those students are also likely to show a high willingness to dedicate time for self-study despite many compulsory lessons, this potential bias would presumably result in an overestimation of TSS/QI. As we claim that TSS/QI is too low in our current curriculum, this is biasing our results towards a more "conservative" estimate rather than towards incorrectly low TSS/QI values.
A second point of caution pertains to the outcome variables of our study. While we could show, that increased test frequency results in raised TSS/QI and TSS, that is increased quantity of learning, this study did not demonstrate the impact of TSS on learning outcome. Previous studies, even on a multicenter level  could already provide evidence for the positive relation between TSS and learning outcome [3, 6]. However, future studies are recommended to correlate test frequency or TSS directly to exam results, levels of competence or graduation rates in order to verify that this evidence based assumption holds true.
It is also to note that this study exclusively focused on quantity of instruction, not on quality. It is well possible that exceptionally high quality mandatory instruction has a positive impact on learning that exceeds the negative impact of its reducing TSS. We also did not differentiate between different teaching formats, although it is likely that their impact on learning outcome is different. Schmidt e al. report that the strong inverted relation between lecture time and graduation rate was found to be weaker for other scheduled activities like practical training and tutorials . In summary, when reducing quantity of instruction, it is important to carefully choose which formats to keep as guidance and stimulants for TSS and which formats to cut down on because they rather impede effective TSS.