Background: The manifestation of impaired production of closed-class morphemes in Chinese aphasia has been examined in recent years (Lu, 1994; Packard, 1990, 1993; Tzeng, Chen, & Hung, 1991; Yiu & Worrall, 1996). However, these studies employed tasks that either allowed the patients to use avoidance strategies to conceal their impairment or modelled the target morphemes and thus provided cues to the patient. Hence, a contextually highly constrained task was used in this study. Aims: This study aimed at confirming previous observations regarding continuity across clinical types in grammatical morpheme production, dissociable performances among markers of the same morpheme category and among homophonous functors serving different functions, and to identify factors determining the accessibility of grammatical morphemes. Methods & Procedures: The Cloze test without modelling of target morphemes was employed. A wide range of functors were examined including: aspect markers; negative markers; classifiers with sub-types of verbal, sortal, container, and collective classifiers; pronouns encompassing personal, predicative, and adverbial pronouns; coverbs; structural particles including the nominative, adjectival, and adverbial particles, and one modal particle; and structural suffixes including one associated with descriptive complements and one with resultative complements. Six fluent and four non-fluent Cantonese-speaking aphasic patients were involved. They suffered either a head trauma or a unilateral left CVA at least 6 months prior to the first test session. Outcomes & Results: The main findings were largely consistent with previous reports: (1) Pronouns, coverbs, structural and modal particles, and structural suffixes presented greater difficulty to the patients than aspect markers, negative markers and classifiers. (2) Dissociations in performance were found among different members of the same category and between homophonous morphemes serving different functions. (3) Performances of the fluent and non-fluent groups differed quantitatively rather than qualitatively, with better performance on all categories by the fluent group. (4) Neither substitutions nor omissions dominated the error patterns of either aphasia group. Conclusions: The results confirm previous observations that the difference between fluent and non-fluent aphasic patients with regard to the production of grammatical morphemes is non-categorical, and that morphemes of the same category and homophonous markers of different functions may present different degrees of difficulty to the patients. Furthermore, factors including global vs local relationships, conceptual complexity, semantic value, frequency of occurrence, and optionality of occurrence may help determine the relative accessibility of functors.