A spoken dialogue or discourse can be globally organized into coherent discourse segments in which local salience and coherence properties apply. This paper looks at the regions between these discourse segments in American English spontaneous speech dialogue. These bridge regions do not form a coherent discourse segment themselves. When fluent, their utterances belong to neither the previous nor subsequent discourse segment. Often the utterances are disfluent. Dialogue turns in these bridge regions are characteristically shorter, on average, both in duration and number of words. Moreover, there are fewer pronouns, on average, in bridge turns, compared to the entire dialogue.