Segment deletion in the Kiel Corpus of Spontaneous Speech *

In this paper we examine the characteristics of symbolic con sonant deletions in the Kiel Corpus of Spontaneous Speech . This survey comprises the 31 phonetically labelled dialogues of CD-ROM#2 (IPDS 1995) together with a further 43 dialogues which form a subset of the data on CD-ROMs#3–4 (IPDS 1996, 1997). This da ta base is identical to that used in Helgason and Kohler (1996). The primary focus is on deletions that have been labelled with an MAmarker, which denotes that some phonetic correlates of the deleted segment can be detected in the signal (see Helg ason and Kohler 1996). Such cases, we believe, give valuable insights into the arti culatory processes involved in reduction. Symbolic consonant deletions refer to the deletion of conso nantal segments with reference to canonical lexical citation forms (as laid out i n Helgason and Kohler 1996). In the present discussion, a distinction is made between sym bolic deletions that have no adjacent deleted segment, and deletions that occur in con junction with another deletion, i.e. when a preceding and/or following vowel or conson ant is also deleted. Henceforth the former are referred to as single deletion and the la tt r s multiple deletion, and treated in separate sections. This refers only to deleti ons within a lexical item, so that, e.g., the consecutive deletion of two segments that oc curs across a word boundary is counted as two single deletions. The data used for this survey contain 3018 single deletions ( f vowels and consonants) and 295 multiple deletions 1. TheMAmarker has been applied in 427 instances