In recent decades, with an increasing application of computer technology to the delivery of oral language proficiency assessment, there have been renewed debates over the appropriateness of two different testing modes, namely, (a) face-to-face, or direct, testing, and (b) person-to-machine, or semi-direct, testing. Previous research conducted in various geographical regions has produced mixed results as to which format is more acceptable to test takers. The present investigation, carried out in the context of a university setting in Hong Kong, compared the popularity of two testing modes by analyzing reactions and perceptions of a group of test takers who had just sat for a direct test as well as a semi-direct test. The results indicate that, although a large proportion of the participants had no particular preference in terms of the testing mode, the number of participants who strongly favored direct testing far exceeded the number strongly favoring semi-direct testing. The participants' main reason cited for disliking semi-direct testing was its inability for the examiner and examinee to interact during the test, which appears to have created a psychological barrier for the test taker.[Copyright of Language Assessment Quarterly is the property of Routledge. Full article may be available at the publisher's website: http://dx.doi.org/10.1080/15434300902800059]