Perceptual decisions are often modeled as successively accumulating evidence. The theoretical basis for this are sequential probability ratio tests, which propose that the total evidence is the sum of multiple individual pieces of evidence collected throughout the perceptual process. In contrast to this notion, existing performance measures for perceptual decisions, such as accuracy and sensitivity d’, are not additive. Here, we suggest computing a performance measure based on participants’ decisions and confidence ratings that we dub accumulated evidence E. This measure quantifies evidence additively: Evidence accumulated during a first stimulus presentation interval, E_{[0, a]}, adds up with that during a second interval, E_{[a, b]}, to the total evidence accumulated during the whole presentation E_{[0, a]} + E_{[a, b]} = E_{[0, b]}. This additivity allows measuring the otherwise not directly accessible evidence that is exclusively accumulated during the second presentation interval, E_{[a, b]} = E_{[0, b]} - E_{[0, a]}. In a first experiment, we provide direct support for previous findings suggesting a deceleration in evidence accumulation (E_{[0, a]} > E_{[a, b]}). In a second experiment, we quantify how much evidence accumulation in later stages (E_{[a, b]}) depends on that of earlier stages (E_{[0, a]}). Limitations arise from the reliance on confidence ratings, which are biased by various factors so that careful interpretations are needed. Overall, this approach allows formulating more precise hypotheses about evidence accumulation during perception.