While within-subjects designs often require fewer subjects, that in itself can lead to issues in terms of the data. For instance, subjects are being used in multiple trials, so they would typically be more tired of the study, especially if it's the same general experiment performed repeatedly over consecutive trials. This can lead to multiple confounds, such as the subject not participating as they normally would, or even dropping out of the study and therefore nullifying any of the data previously collected. Regardless of what happens, it can interfere with the interpretation of the results. It therefore seems that while it uses fewer subjects, the risk is that you may need to run a study again if you have enough subjects drop out. For instance, imagine you had a within-subjects design that only required 12 subjects. If half of them drop out, that may not be enough to run a complete analysis, so you may need to run another study. In the end, running repeated trials will cost more than running one larger trial. In this respect, do you think within-subjects designs pose more of an issue than the benefits they seem to provide?