Embodied conversational agents, ECAs, are virtual entities with human-like communicative capabilities. They are used as a new form of human–machine interface. They often have a human-like appearance (see Figure 9.1) and are able to sustain a conversation with users or virtual agents (Cassell et al., 2000; Kopp et al., 2003; Granström et al., 2002; Gratch et al., 2007; Marsella & Gratch, 2009). ECAs communicate through verbal and nonverbal means such as facial expressions, hand and arm movements, body posture, and prosody. They can indicate a point in space (Martin et al., 2009), describe with their hands the shape of an object (Bergmann & Kopp, 2009), punctuate their speech with beats and eyebrow movements (Cassell, 2001; Granström et al., 2002), and display emotional expressions (Niewiadomski et al., 2009a, 2009b) or social signals (Bickmore & Picard, 2005). They are able to manage speaking turn exchange (Thórisson, 2002) and even show some ecologically valid listening behavior (Bevacqua et al., 2010).

ECAs have basically dual functions. On one hand they can serve various functions in a human–machine interface, and on the other they are tools to study human communication. Examples of the first application are Web agents that serve provided information, or pedagogical agents that provide support to students in their learning process (Graesser et al., 2007; Johnson et al., 2000; Moreno & Flowerday, 2006). Lately research has been undertaken to create ECAs as companions for young kids or for the elderly (Bickmore et al., 2008; Tartaro & Cassell, 2008). ECAs are fully parameterized entities. By controlling each of their parameters, one by one or in combination, one can study their role and impact. Visual prosody has been investigated by looking at the tight temporal synchrony of voice intonation and eyebrow movement (Cavé et al., 1996; Krahmer & Swerts, 2004). The role of body posture in turn taking (Cassell et al., 1999), eye movements in deictic function (Raidt et al., 2007), and importance