Phrase-based statistical machine translation systems depend heavily on the knowledge represented in their phrase translation tables. However, the phrase pairs included in these tables are typically selected using simple heuristics that potentially leave much room for improvement. In this paper, we present a technique for selecting the phrase pairs to include in phrase translation tables based on their estimated quality according to a translation model. This method not only reduces the size of the phrase translation table, but also improves translation quality as measured by the BLEU metric.