In a clinical setting the chest radiograph is the reference standard in establishing the diagnosis of community-acquired pneumonia (CAP). This study aimed to assess interobserver reliability (IR) of radiographic findings and the relationship to different causative pathogens in CAP. Chest radiographs of 192 patients with pneumonia, obtained from a database, were reviewed by 2 radiologists and 1 respiratory physician without specific clinical information. Main pattern of infiltrate, extent of pneumonia, presence of pleural fluid, thickened bronchial walls, lymphadenopathy and air bronchogram were scored. Also, the involved lobes were identified. Sputum cultures, blood cultures and serological tests were performed to identify the causative pathogen. IR was poor (kappa <0.4) for determining the main pattern of infiltrate and presence of air bronchogram, lymphadenopathy and thickening of bronchial walls. IR was fair to good (kappa 0.4-0.7) or even excellent (kappa>0.7) for determining the presence of pleural effusion, the extent of pneumonia and for identifying the lobes involved. Mycoplasma pneumoniae was associated more often with patchy alveolar opacities than Streptococcus pneumoniae ( P = 0.0 5). Chlamydia spp. were associated with unilobar involvement (86%), especially when compared to M. Pneumoniae ( P = 0.0 3) and S. pneumoniae ( P = 0.0 0 4). In conclusion, simple features such as presence of pleural fluid, the extent of pneumonia and identifying the involved lobes show fair to excellent IR. Other features such as main pattern of infiltrate are difficult to assess and show poor IR. Hardly any relation between different pathogens and radiological features was found. Therefore, chest radiographs are of limited value in predicting the causative pathogen, but are of good use to determine the extent of pneumonia and to detect complications such as parapneumonic effusion. © 2005 Elsevier Ltd. All rights reserved.