Each V, D, and J gene segment is flanked by a recombination signal sequence (RSS), composed of a conserved heptamer and nonamer separated by a 12- or 23-bp spacer. Variations from consensus in the heptamer or nonamer at specific positions can dramatically affect recombination frequency, but until recently, it had been generally held that only the length of the spacer, but not its sequence, affects the efficacy of V(D)J recombination. In this study, we show several examples in which the spacer sequence can significantly affect recombination frequencies. We show that the difference in spacer sequence alone of two V(H)S107 genes affects recombination frequency in recombination substrates to a similar extent as the bias observed in vivo. We show that individual positions in the spacer can affect recombination frequency, and those positions can often be predicted by their frequency in a database of RSS. Importantly, we further show that a spacer sequence that has an infrequently observed nucleotide at each position is essentially unable to support recombination in an extrachromosmal substrate assay, despite being flanked by a consensus heptamer and nonamer. This infrequent spacer sequence RSS shows only a 2-fold reduction of binding of RAG proteins, but the in vitro cleavage of this RSS is approximately 9-fold reduced compared with a good RSS. These data demonstrate that the spacer sequence should be considered to play an important role in the recombination efficacy of an RSS, and that the effect of the spacer occurs primarily subsequent to RAG binding.