Abstract
In many languages, overt case marking hides ϕ‐features from agreement probes. In some languages, such as Hindi, this effect is absolute. But in Adıyaman Kurmanji, our focus here, the effect is partial: the number feature of third‐person oblique subjects can be registered on the verb. We account for the difference by saying that overtly case‐marked nominals are in fact agreed with in the syntax, but the actual transfer of ϕ‐feature values happens at PF, where it is sensitive to postsyntactic operations like Fusion. All and only heads that fuse with case (a K head) can feed valuation and thus be realized on the agreement probe. Adıyaman Kurmanji is a language where number fuses with K and this feeds agreement, resulting in partial agreement with overtly case‐marked nominals (in number but not person), whereas Hindi is not. The possibility of object agreement in these languages is, we claim, the result of Multiple Agree, which creates complex feature bundles on agreement probes that are only partially realized morphologically at PF. We further support our theory by applying it to Icelandic and Faroese, languages that are often thought not to have agreement with oblique subjects but that may allow agreement in very specific situations, which our theory elucidates.