Background. The relevance of speech-language-hearing therapy (ST) duration to language impairment remains unclear. Objective. To determine the effect of ST duration on improvement in language impairment as a stroke sequela and to compare the findings with those for occupational therapy (OT) and physical therapy (PT). Methods. Data regarding patients with stroke sequelae who were registered in the Japanese Association of Rehabilitation Medicine database were analyzed. Propensity scores for ST, OT, and PT duration were calculated using logistic regression, followed by inverse probability weighting in generalized estimating equations to examine the odds ratio for improvement in the Functional Independence Measures scores for comprehension, expression, and memory. Analyses stratified by age and dementia severity were also conducted. Results. Compared with short-duration ST, long-duration ST was significantly associated with improved scores for comprehension and expression in the overall study population and in some groups, with higher benefit especially for younger participants (<64 years) and those with more severe dementia. A significant but less pronounced effect was also observed for OT and PT. Conclusion. Long-duration ST is more effective than long-duration OT or PT for improving language impairment occurring as stroke sequela. However, these effects are limited by age and severity of dementia.