<p>Previous studies have shown that semantic radical transparency supports vocabulary learning. However, whether this effect affects the reading of novel words (pseudowords) during incidental vocabulary learning is unclear. Accordingly, in order to investigate the effect of semantic radical transparency on Chinese novel word learning and how this effect changes with the gradual accumulation of novel word representations, the present study integrated the incidental vocabulary learning paradigm with a vocabulary knowledge test. We manipulated whether the semantic radicals of pseudowords were related to the meaning imparted in the informative learning context. Each pseudoword was described as a novel concept of a specific semantic category by 15 informative contexts, and these sentences were divided into five learning phases. We recorded the eye movements of the study participants during the learning phase and their responses during the test phase. A within-subjects experimental design of 2 (semantic radical transparency: transparent, opaque) × 5 (learning phases: 1st, 2nd, 3rd, 4th, 5th) was used. The results of the learning phase revealed that, in the first learning phase, readers had shorter reading times and made fewer regressions back to novel words containing transparent semantic radicals. The post-test outcomes showed that the accuracy of the choice of semantic category task for semantic transparent novel words was significantly higher than those for opaque ones. The results of this study indicate that proficient Chinese readers spontaneously use the semantic information of sublexical-level semantic radical units to facilitate incidental vocabulary learning.</p>