In this paper I present cross-linguistic data of NRNR, TP RNR, and composed plurality to argue for the multi-value asymmetry between the N and the T domain. I propose that the asymmetry can be accounted for by assuming that T heads have multiple unvalued number features while N heads have only one. The proposal is compatible with the unified Agree analysis of concord and agreement and it is further supported by the mismatch cases of multi-valuation.