We prospectively examined 128 patients with acute first-ever stroke to determine the prevalence of swallowing disorders, the diagnostic accuracy of our clinical assessment of swallowing function compared with videofluoroscopy, and interobserver agreement for the clinical and videofluoroscopic diagnosis of swallowing disorders and aspiration. We found clinical and videofluoroscopic evidence of a swallowing disorder in 51% [95% confidence interval (CI) 42–60%] and 64% (95% CI 55–72%) of patients, respectively, and aspiration in 49% (95% CI 40–58%) and 22% (95% CI 15–29%) of patients, respectively. The optimal clinical criteria for detecting videofluoroscopic evidence of a swallowing disorder and aspiration were any clinical evidence of a swallowing disorder (sensitivity 73%, 95% CI 62–82%; specificity 89%, 95% CI 76–96%), and any clinical evidence of aspiration (sensitivity 93%, 95% CI 76–99%; specificity 63%, 95% CI 53–72%). The interobserver agreement between two speech pathologists for the clinical diagnosis of a swallowing disorder (κ: 0.82 ± 0.09) and aspiration (κ: 0.75 ± 0.09) was good, and between a speech pathologist and radiologist for the videofluoroscopic diagnosis of a swallowing disorder (κ: 0.75 ± 0.09) and aspiration (κ: 0.41 ± 0.09), it was good and fair, respectively. Although clinical bedside examination underestimates the frequency of swallowing abnormalities and overestimates the frequency of aspiration compared with videofluoroscopy, it may still offer valuable information for the diagnosis of swallowing impairment. Long-term follow-up studies are required to determine the independent functional significance of the findings of the bedside and videofluoroscopic examinations in predicting the occurrence of important outcome events such as aspiration pneumonia.