The inflammatory component of asthma is usually assessed indirectly by symptoms and spirometry, these may be inaccurate. It can now be assessed directly and reliably by the examination of sputum cell counts. There is no information on how clinical assessment of the presence and type of airway inflammation compares with actual measurements. In this single-centre observational study, sputum was collected from 76 consecutive adults with asthma attending a tertiary chest clinic after their physicians had recorded the expected cell counts in sputum. The authors examined the extent of agreement between clinical judgement of sputum cell counts and actual counts in asthmatic patients (Cohen's Kappa) and the possible predictors of agreement (multiple logistic regression). Sixty-seven of the 76 sputum samples were suitable for analysis. Agreement between expected and actual cell counts occurred in 30/67 patients. The overall agreement for the different cell types was poor (estimated K=0.14, 95% confidence interval (CI)=0.02, 0.26). The experience of the physician in using sputum cell counts in clinical practice, steroid requirement at the time of assessment, and control of asthma as assessed by the physician or by the patient could not predict the chances of agreement or disagreement. Unaware of the sputum results, the physicians often changed treatment in a way that seemed inappropriate for the cell counts present. There is poor agreement between clinical judgement of the presence and type of airway inflammation in asthmatic patients and sputum cell counts. The impact of sputum examination on the outcomes of anti-inflammatory treatment now needs investigation.