There is no ideal method for detecting Helicobacter pylori. The 'standard' 13Carbon urea breath test (13C-UBT), which involves collecting eight to 15 breath samples and subsequent costly analysis, was modified by pooling 21 samples of expired breath taken at five minute intervals for 40 minutes into a collecting bag, from which a single 20 ml aliquot was taken and analysed by mass spectrometry. This test was evaluated on 50 patients after routine upper gastrointestinal endoscopy, and results were compared with those from the standard 13C-UBT, bacteriology, ELISA serology, and histology--the latter being taken as the gold standard. H pylori were seen in 34 of 50 (68%) patients (in three it was detected in biopsy specimens from the corpus alone). The modified 13C-UBT was positive (pooled excretion delta 13CO2 greater than 5 per mil) in 31 patients and negative in 19 (three false negative results), specificity was 100% (standard 13C-UBT 94%) and sensitivity 92% (standard 13C-UBT 93%). The modified 13C-UBT had a coefficient of variation within subjects of 3.7%. For the ELISA serology and culture the specificities were both 100%, but the sensitivities were 82% and 68% respectively. The 13C-UBT results correlated with the grade of histological gastritis. The modified 13C-UBT is simpler, cheaper, more reproducible, and provides an easy non-invasive method for the detection of H pylori.