One of our goals is to assess whether FactBank’s seven-category scheme is the right one for the task. To this end, we also evaluated whether a five-tag version would increase agreement and perhaps provide a better match with readers’ intuitions. Logically, PR − is equivalent to PS+, and PS − to PR+, so it seemed natural to try to collapse them into a two-way division between “probable” and “possible”. We thus ran the MTurk experiment again with the five-point scheme in Table 3.