Recent articles in this journal (Foot 1999; Saville and Hargreaves 1999) have focused on the advantages and disadvantages of the paired format of the Cambridge Speaking Tests. This article aims to contribute to the debate by considering how the pairing of candidates may impact upon the language sample produced and could affect the assessment process. Data from the Speaking Tests are presented which suggest that pairing potentially affects linguistic performance if one candidate has higher linguistic ability than the other, or if candidates know each other. Pairing also seems to affect the amount of talk produced, and depending on the gender make-up of the pair there appear to be qualitative differences in the respective participation in the tests of each partner.