A virtual human is a computer-generated, humanlike character. Different modelling techniques are used in order to produce this human-likeness in behaviour as well as appearance. It has been shown that people use similar attributions, evaluations, and emotional responses towards virtual humans, as towards real human beings. This has opened up for research on human behaviour, interaction and communication by means of virtual human technology. A main advantage is that a given variable of interest can be effectively isolated and systematically examined without confounds.  As an example consider the effect of body posture on perceptions of a speaker’s credibility. In a study with a real human speaker as stimuli one has to deal with several confounding cues, such as the speaker’s natural facial expressions, head movements, and gestures. Using a virtual human, however, these can all be controlled whereas slight changes in a program parameter can create precise body posture manipulations. Another advantage is that it is possible to study the effects of subtle or fine-grained changes that real humans cannot systematically produce in a precise manner. In CCL we are currently using virtual human technology to study details of speech-gesture coordination in human speech production and comprehension.