According to Mirolli and Parisi’s hypothesis (M. Mirolli and D. Parisi, New Ideas in Psychology, 2011, 29: 298-311), it could be possible to overcome a gap existing between natural human cognition and usual cognitive robotics allowing to model in computer only simple low-level human cognitive functions (sensory-motor coordination, perception, etc.). They suppose this possibility could be implemented if a language would be considered, like Vygotsky has done, not only as a communication system, but also as a cognitive tool. Thus, social interaction with the help of language is like an extended cognitive system, which is internalized by child, when she entertains with adults and peers. However, during such internalization through the phase of ‘egocentric speech’ of child, the intersubjective phenomena become the intrinsic to child, intrapersonal ones. In the process of internalization, the child can become a subject (agent) of skills mediated by language. It means that the child becomes to use this kind of speech in intentional manner, i.e. she has conscious intention (e.g. ‘I go to walk’) and the child use the language for her purposes (‘I draw a boy’). Thus, for Vygotsky, natural human cognition presupposes that a child become a subject of cognition, but it remains unclear how computer could be an intentional subject. This egocentric speech could be considered, as we suppose, like a script which helps to child to become an intentional agent (subject) of actual for child routine practices (Neeley et al., 2001: Early Childhood Education Journal, vol. 28 (4): 243-246), though we must realize script can’t create his agency. In this paper, we analyze semiotic properties of expressive interactional training, which is used for making people with intellectual disabilities to be a subject of their expressive movement in integrated theatre, and we show how non-verbal scripts are applied for this aim.https://youtu.be/ZFTKJkg1Eb4