Methods using the reactions of beneficial arthropods to stimuli of prey and habitats are becoming more and more important in the biological control of pests. Especially olfactory cues, which are measurable with so called ‘olfactometers’ are of great importance. In the present study, a modified four-armed airflow olfactometer is described by explaining these modifications. As an example, the results obtained for the predatory ladybeetle Coccinella septempunctata are explained in detail. Various odour sources caused evident differences in the reactions of the predator, especially comparing the reared beetles and the individuals collected in the field. The volatiles emitted by the odour source ‘bean plant + aphids’ were attractive to 28 % of the reared and to 38 % of the field-collected test individuals. These values were clearly higher than in the cases of the odour source ‘aphids’ with 18 % for both groups. About 15 % of the reared and 13 % of the field-collected coccinellids reacted to the olfactometer chamber without an odour source. On the other hand, it became obvious in test series with groups of adult beetles that group tests were not so suitable because the test individuals diverted each other so that the reactions to the odours were reduced. In the case of the group tests, 68 % of the reared and 60 % of the field-collected individuals did not react to one of the offered odour sources. In further surveys, the most favourable pre-experiment starvation period of the tested individuals and optimal duration for a single experiment were determined. For adults of C. septempunctata, the most favourable starvation time was 6 hours and the most suitable duration of a single experiment 30 minutes. In two further experiments, it was investigated whether there were differences between males and females or between beetles of different ages. There were slight differences between males and females reacting to distinct odour sources and also in the case of coccinellids of different ages but in both cases without clear tendencies.