The project "Mechanisms and representation in comprehending speech" (MaRCS) started in 2009. The project investigates how we map the acoustic information in spoken utterances onto stored lexical knowledge, and hence constructing interpretations of speakers' intentions. The project evolved out of the projects "Decoding continuous speech" and "Phonological learning for speech perception".

These projects (2001-2008) investigated the stable system and the developing system, respectively. The MaRCS project shares with its predecessors the focus on the processing of acoustic-phonetic information but focuses more sharply on theoretical questions, such as

- How are different sources of information sources (segmental, suprasegmental, lexical, visual, ...) integrated?

- How does episodic learning affect speech comprehension?

- To what extent is speech perception special?

- How does spoken-word recognition relate to semantic and conceptual processing and to speech production?

The project's research is divided in two complementary subprojects. In one subproject, we focus on every detail of the speech-comprehension process to elucidate the inner workings of speech comprehension. This project leads to small-scale models of how speech comprehension might work. In the other subproject, we put these models to the test in various large-scale domains, such as spontaneous conversation or speech comprehension in the elderly.

There are close links with the projects on "Adaptive listening" and "Categories across language and cognition". With the former, there is a common interest in the carrier of communication, the speech signal, with the latter there is a shared interest in basic theoretical issues concerning the acquisition and flexibility of cognitive categories.