This paper reports on the on-going evaluation of a
computer-assisted system (CEOTS) for the assessing of
spoken English skills among Chinese university students.
This system is being developed to deal with the negative
backwash effects of the present system of assessment of
speaking skills which is only available to a tiny minority.
We present data from a survey of students at the
developing institution (USTC), with follow-up interviews
and further interviews with English language teachers, to
gauge the reactions to the test and its impact on language
learning. We identify the key issue as being one of
validity, with a tension existing between construct and
consequential validities of the existing system and of
CEOTS. We argue that a computer-based system seems to
offer the only solution to the negative backwash problem
but the development of the technology required to meet
current construct validity demands makes this a very long
term prospect. We suggest that a compromise between the
competing forms of validity must therefore be accepted,
probably well before a computer-based system can deliver
the level of interaction with the examinees that would
emulate the present face-to-face mode.