is observed: the B3 (larger C Q ) site dephases more slowly than the B4 site at all investigated MAS frequencies (5 to 20 kHz) at 14.1 T. Increasing the MAS frequency leads to markedly slower dephasing for the B3 site, while there is a much less evident effect for the B4 site. Considering samples at 5, 25, 80 (natural abundance) and 100 % 11 B isotopic 15 abundance, dephasing becomes faster for both sites as the 11 B isotopic abundance increases. The experimental behaviour is rationalised using density matrix simulations for two and three dipolarcoupled 11 B nuclei. The experimentally observed slower dephasing for the larger C Q (B3) site is reproduced in all simulations and is explained by the reintroduction of the dipolar coupling by the so-called "spontaneous quadrupolar-driven recoupling mechanism" having a different dependence 20 on the MAS frequency for different quadrupolar frequencies. Specifically, isolated spin-pair simulations show that the spontaneous quadrupolar-driven recoupling mechanism is most efficient when the quadrupolar frequency is equal to twice the MAS frequency. While for isolated spin-pair simulations, increasing the MAS frequency leads to faster dephasing, agreement with experiment is observed for three-spin simulations which additionally include the homogeneous nature of the 25 homonuclear dipolar coupling network. First-principles calculations, using the GIPAW approach, of the 2 J 11B-11B couplings in lithium diborate, metaborate and triborate are presented: a clear trend is revealed whereby the 2 J 11B-11B couplings increase with increasing B-O-B bond angle and B-B distance. However, the calculated 2 J 11B-11B couplings are small (0.95, 1.20 and 2.65 Hz in lithium diborate), thus explaining why no zero crossing due to J modulation is observed experimentally, 30 even for the sample at 25 % 11 B where significant spin-echo intensity remains out to durations of 200 ms.