A new experimental approach to the famous problem of the anomalously slow Gamow-Teller (GT) transitions in the beta decay of the A=14 multiplet is presented. The GT strength distributions to excited states in 14C and 14O were studied in high-resolution (d,2He) and (3He,t) charge-exchange reactions on 14N. No-core shell-model calculations capable of reproducing the suppression of the beta decays predict a selective excitation of Jpi=2+ states. The experimental confirmation represents a validation of the assumptions about the underlying structure of the 14N ground state wave function. However, the fragmentation of the GT strength over three 2+ final states remains a fundamental issue not explained by the present no-core shell model using a 6homega model space, suggesting possibly the need to include cluster structure in these light nuclei in a consistent way.