We fit the fundamental plane, FP, to over 400 early-type galaxies in 20 nearby clusters (cz ∼ 4000–11000 km s-1), using our own photometry and spectroscopy, as well as measurements culled from the literature. We find that the quality of fit, rms(log σ), to the average fundamental plane, ⟨FP⟩, varies substantially among these clusters. A statistically significant gap in the distribution of rms(log σ) separates the clusters that fit ⟨FP⟩ well from those that do not. These two subgroups are of roughly equal size (11 low-rms clusters and nine high-rms clusters). The mean redshifts of the two subsamples are 6700 and 8000 km s-1, respectively, and both samples show a mean peculiar velocity of zero. However, the nine high-rms(log σ) clusters exhibit a scatter about the mean peculiar velocity of 785 ± 200 km s-1, similar in amplitude to large-scale motions reported by other observers. In contrast, the other 11 clusters exhibit significantly lower scatter, equal to 245 ± 55 km s-1. Assuming early-type galaxies in clusters are drawn from a single underlying population, there should be no correlation between the scatter in peculiar velocity and rms(log σ). While the source of the high values of rms(log σ) exhibited in this subset of clusters is not entirely known, there is some indication that it is an intrinsic cluster property, possibly related to the presence of significant substructure. For example, we find that all X-ray–bright clusters in our sample fit the ⟨FP⟩ well, i.e., they exhibit low values for rms(log σ). The evidence suggests that all-sky surveys that mix together both classes of clusters, as characterized by rms(log σ), may artificially inject signal into the observed peculiar velocity field. We therefore suggest that a prefilter such as the X-ray luminosity of clusters or FP scatter be used to select clusters to derive relative distances via the FP. When such a prefilter is applied to our sample a relatively quiescent Hubble flow is revealed.