We present evidence that the splice sites in mammalian pre-mRNAs are brought together via a three dimensional diffusion mechanism. We tested two mechanisms for splice site pairing: a lateral diffusion (‘scanning’) model and the currently favored three dimensional diffusion (‘jumping’) model. Two lines of evidence that distinguish between these two models are presented. The first utilized bipartite splicing substrates tethered by double-stranded RNA stems predicted to provide either a moderate or severe block to splice site pairing via a scanning mechanism. Splice site pairing via a jumping mechanism was expected to be unaffected or affected minimally. The second approach utilized a flexible poly(ethylene glycol) moiety within the intron. This insertion was predicted to reduce scanning efficiency but not the efficiency of a three dimensional diffusion mechanism. The best explanation for the data with the bipartite RNAs is that splice site pairing occurs through three dimensional diffusion. Kinetic analysis of the poly(ethylene glycol) containing substrate showed that neither the lag phase nor the initial rates of mRNA production and spliceosome assembly were affected by this insertion. Therefore, both experimental approaches supported the three dimensional diffusion model of splice site pairing.