One of the major questions in speech science has been how are language production and comprehension related Both processes require rapid and accurate retrieval of words and appropriate grammatical structures from the mental lexicon. Theories propose that word retrieval involves the selection of at least two levels of lexical information, semantic representations and word-forms. Spontaneous speech is characterized by various phonetic processes such as co-articulation, the variability of the phonetic form of words, and by various types of disfluency phenomena. As speakers we come across difficulties during speech planning while as listeners we have to cope with other peoples' speech difficulties resulting in erroneous utterances. Speech difluencies are generally defined as phenomena that interrupt the flow of speech and do not add propositional contents to an utterance. There are various forms of disfluencies that occur and might slightly differ across languages. This talk will discuss the results Of a series of experiments that have been carried out with Hungarian speakers/listeners focusing on speech disfluencies. A corpus containing more than 5 000 perceived disfluency items was analyzed while recorded spontaneous speech material of about 8 hours was analysed Several specific experiments were also carried out to investigate the problem with the participation of young people, children and elderly persons. The following questions were posed (i) How do the various types and occurrences of difluency highlight the difficulties the speaker faces during speaking? (ii) What kinds of interrelations exist between the phonetic and phonologocial operations resulting in errors during speech planning? (iii) Do pauses refer to specific operations in the mental lexicon predicting the phonetic output? The temporal analysis of word retrieval was carried out in a "tip-of-the-tongue" elicitation experiment while pauses signaling the speaker's word-finding difficulties were measured also in spontaneous speech. (iv) How does the speaker's self-monitoring work? What kinds of strategies lie behind the diverse outcomes of the self-monitoring process? There are disfluencies that the speaker fails to notice, on the one hand, and there are difluencies that are noticed by the speaker but still remain uncorrected, on the other. There are yet other disfluencies that are corrected during speaking without the action being noticed by the speaker. (v) Are there any differences among the types and occurrences of difluency depending on the age of the speaker? (vi) What kinds of effects do speakers' disfluencies have on listeners? Speech perception is an extremely fast process, given that while the mechanism interprets the incoming waveform as a series of linguistic segments and suprasegmentals, it is also continuously ready to receive and correct incoming erroneous messages. How are speakers able to monitor their own speech while listening at the same time to another's speech (using a shadowing technique) or while being disturbed by background speech noise (Lombard-effect)? What kind of strategy helps speakers to comprehend erroneous speech? Our findings enable a hypothesis to beformulated explaining the differences in speech disfluencies (both in production and perception) depending on their type, the speaker's age, the actual context and the speech planning level where they occur.
