The study compares EFL writers' processes in composing reading-based writing (RW) and graph-based writing (GW) tasks developed for a university English proficiency exam. Think-aloud protocols and interviews of ten university-level nonnative English-speaking writers were collected to explore writers' composing processes. The results revealed that both types of the tasks require global comprehension of source texts as well as integrative manipulation of available information for writing. Some differences, however, existed across tasks and writers of varying score levels, with the RW tasks eliciting a more interactive and facilitative process than the GW tasks for the higher scoring writers. These results suggested that these tasks might measure different aspects of academic writing ability. Several considerations of the task constructs should apply in properly determining their use in a language test. The findings could be used to provide insights into the nature of RW and GW tasks and contribute to the validity of source-based writing tasks.