The problem of finding a graceful labelling of a graph, or proving that the graph is not graceful, has previously been modelled as a CSP. A new and much faster CSP model of the problem is presented, with several new results for graphs whose gracefulness was previously unknown. Several classes of graph that are conjectured to be graceful only for small instances are investigated: after a certain size, it appears that for some of these classes the search to prove that there is no graceful labelling is essentially the same for each successive instance. The possibility of constructing a proof of the conjecture based on the search is discussed.
Barbara M. Smith