Game theoretic models of clear versus plain speech

Clear speech is a speaking style intended to improve the comprehension of the hearer, which is usually due to the external noise, less ideal listening conditions, or the speaker is intended to be more intelligible. Clear speech, which exhibits increased duration, pitch, amplitude, and more exaggerated articulation, consumes more energy in order to improve the likelihood of accurate communication. To strike a balance between the cost of clear speech and the improvement it brings, we use game theory to model the phenomenon of clear speech. The conventions that speakers and hearers use to communicate are considered as equilibria in the communication game, and we need to make predictions of how the equilibria changes under the different circumstances. How our models correspond to what is experimentally observed, and what predictions are made for experimental results are discussed in the thesis. In the basic model, we study the case where the speaker has to send one of two messages equally likely in one-dimensional acoustic space. Next, we make a further discussion of the basic model in a priori probability of the sent message, the number of messages, and the conflicts between clearness and comprehensibility. The third contribution of this thesis is to extend the one-dimensional acoustic space to two dimensions, by introducing uncontrastive and contrastive features.