Metamemory research so far has mainly focused on people’s ability to predict their item memory, for example what was said. However, it is also a useful skill to predict one’s source memory, for example the speaker. The aim of this study is to assess the relative accuracy of Judgements of Source (JOS) using multinomial processing tree (MPT) modeling, which provides a measure of source memory independent of item memory and corrected for the influences of guessing, allowing a better assessment of the relationship between source memory and JOS. We will report results from three experiments conducted in parallel (with random assignment). A total of 90 participants (n = 30 per experiment) study words and sources in the three classical source monitoring tasks. Participants either hear the words read out by two speakers (external), imagine themselves or a famous person saying the words (internal) or the combination – listening to one speaker or imagining themselves saying the word (reality-monitoring). Items will be vincentized into bins of different JOS level and aggregated over participants. The two-high-threshold model of source monitoring (2HTSM) will be used to estimate source memory parameters for each JOS level. We hypothesize that source memory d increases with increasing JOS level.