The 2003 NIST Language Recognition Evaluation was very similar to the last such NIST evaluation in 1996. It was intended to establish a new baseline of current performance capability for language recognition of conversational telephone speech and to lay the groundwork for further research efforts in the field. The primary evaluation data consisted of excerpts from conversations in twelve languages from the CallFriend Corpus. These test segments had durations of approximately three, ten, or thirty seconds. Six sites from three continents participated in the evaluation. The best performance results were significantly improved from those of the previous evaluation.