Turning Washington's heuristics in favor of Vandiver's conjecture

A famous conjecture bearing the name of Vandiver states that $p\nmid h_p^{+}$ in the $p$ - cyclotomic extension of $\text{Q}$. Heuristics arguments of Washington, which have been briefly exposed in Lang (1978), p. 261 and Washington (1996), p. 158 suggest that the Vandiver conjecture should be false if certain conditions of statistical independence are fulfilled. In this note, we assume that Greenberg’s conjecture is true for the \nth{p} cyclotomic extensions and prove an elementary consequence of the assumption that Vandiver’s conjecture fails for a certain value of $p$: the result indicates that there are deep correlations between this fact and the defect ${\lambda }^{-}>i(p)$, where $i(p)$ is like usual the irregularity index of $p$, i.e. the number of Bernoulli numbers $B_{2k}\equiv 0p,1<k<(p-1)\mathrm{/}2$. As a consequence, this result could turn Washington’s heuristic arguments, in a certain sense into an argument in favor of Vandiver’s conjecture.