This paper presents the design and development of the Hong Kong Sign Language-Sentence Repetition Test (HKSL-SRT). It will be argued that the test offers evidence of discriminability, reliability, as well as practicality and can serve as an effective global measurement of individuals' proficiency in HKSL. The full version of the test consists of 40 signed sentences of increasing length and complexity. Specifically, we will evaluate the manual and non-manual components of these sentences to find out whether and to what extent they can differentiate three groups of deaf signers, namely, native signers, early learners and late learners. Statistical analyses show that the test scores based on a correct repetition of the manual signs of each sentence bear a significant negative correlation with signers' age of acquisition. Including the correct repetition of non-manuals in the scoring scheme can result in higher reliability and separation index of the test in the Rasch model. This paper will also discuss how psychometric measures of Rasch analysis, including the concept of fit and the rankings of items/persons in the Wright map, have been applied to the original list of the 40 sentence items for the development of a shortened test.