Discussions of interaction in second language performance assessment h
ave generally been loosely psychological in orientation, forming part
of attempts to model the nature of communicative ability within the in
dividual. But in investigating the validity of performance assessments
involving interactions between individuals (for example between candi
date and interlocutor in speaking assessments), the intrinsically soci
al nature of performance needs to be recognized. What would be the con
sequences for language testing research if it adopted a social perspec
tive on the nature of interaction? The paper explores the necessity fo
r such a reorientation, and suggests areas that would feature in a con
sequent research agenda.