Purpose: The purpose of this study was to investigate the relationship between working memory (WM) capacity and speech recognition in noise in both a group of young adults and a group of older adults. Method: Thirty-three older adults with a mean age of 71.0 (range: 60.4-82.7) years and 27 young adults with a mean age of 21.7 (range: 19.1-25.0) years participated in the study. All participants had age-appropriate hearing and no history of central nervous system dysfunction. WM capacity was measured using the van den Noort version of the Reading Span Test, and recognition of sentences in the presence of a stationary speech-shaped noise was measured as the speech reception threshold for 50% correct identification by using the Leuven Intelligibility Sentence Test. Results: The older adults had significantly worse WM capacity scores, t(58) = 8.266, p < 0.001, and significantly more difficulty understanding sentences in noise than the younger adults, t(58) = -6.068, p < 0.001. In the group of older adults, a correlation was found (r = -0.488, n = 33, p = 0.004) between the results of the WM capacity test (Reading Span Test) and the results of the speech-recognition-in-noise test (Leuven Intelligibility Sentence Test), meaning that the higher the WM performance was, the better was the speech recognition in noise. This correlation cannot be found in young normal-hearing listeners. Conclusions: This study shows deleterious effects of age on both WM capacity and speech recognition in noise. Interestingly, only in the group of older adults was a significant relation found between WM capacity and speech recognition in noise. The current results caution against the assumption that WM necessarily supports speech-in-noise identification independently of the age and hearing status of the listener.