An acyclic mapping from an $n$ element set into itself is a mapping $\varphi$ such that if $\varphi^k(x) = x$ for some $k$ and $x$, then $\varphi(x) = x$. Equivalently, $\varphi^\ell = \varphi^{\ell+1} = \ldots$ for $\ell$ sufficiently large. We investigate the behavior as $n \rightarrow \infty$ of a Markov chain on the collection of such mappings. At each step of the chain, a point in the $n$ element set is chosen uniformly at random and the current mapping is modified by replacing the current image of that point by a new one chosen independently and uniformly at random, conditional on the resulting mapping being again acyclic. We can represent an acyclic mapping as a directed graph (such a graph will be a collection of rooted trees) and think of these directed graphs as metric spaces with some extra structure. Heuristic calculations indicate that the metric space valued process associated with the Markov chain should, after an appropriate time and ``space'' rescaling, converge as $n \rightarrow \infty$ to a real tree ($\R$-tree) valued Markov process that is reversible with respect to a measure induced naturally by the standard reflected Brownian bridge. The limit process, which we construct using Dirichlet form methods, is a Hunt process with respect to a suitable Gromov-Hausdorff-like metric. This process is similar to one that appears in earlier work by Evans and Winter as the limit of chains involving the subtree prune and regraft tree (SPR) rearrangements from phylogenetics.