The interaction of proteins with DNA recognition motifs regulates a number of fundamental biological processes, including transcription. To understand these processes, we need to know which motifs are present in a sequence and which factors bind to them. We describe a method to screen a set of DNA sequences against a precompiled library of motifs, and assess which, if any, of the motifs are statistically over- or under-represented in the sequences. Over-represented motifs are good candidates for playing a functional role in the sequences, while under-representation hints that if the motif were present, it would have a harmful dysregulatory effect. We apply our method (implemented as a computer program called Clover) to dopamine-responsive promoters, sequences flanking binding sites for the transcription factor LSF, sequences that direct transcription in muscle and liver, and Drosophila segmentation enhancers. In each case Clover successfully detects motifs known to function in the sequences, and intriguing and testable hypotheses are made concerning additional motifs. Clover compares favorably with an ab initio motif discovery algorithm based on sequence alignment, when the motif library includes only a homolog of the factor that actually regulates the sequences. It also demonstrates superior performance over two contingency table based over-representation methods. In conclusion, Clover has the potential to greatly accelerate characterization of signals that regulate transcription.