Gaze patterns and fillers: empirical data on the difference between Dutch ‘euh’ and ‘euhm’

In natural conversations, people often use fillers such as ‘euh’ and ‘euhm’. These fillers do not always stem from a cognitive production difficulty but can also serve a communicative goal (Lake et al., 2011; Walker et al., 2014). Different interactional functions have been attributed to fillers, such as the marking of discourse boundaries (Swerts 1998), warning for a delay (Clark, 1994), a turn holding, taking or yielding function (Beattie, 1983; Clark & Fox Tree, 2002; Cook & Lalljee, 1972; Maclay & Osgood, 1959). This study explores the correlation between fillers and gaze aversion by a speaker. Just as fillers, gaze aversion can serve different communicative goals, which seem to be related to those of the fillers. For example, people look away when they want to continue their turn (Hirvenkari et al., 2013) and when they start a new turn (Oertel et al., 2012). More specifically, speakers’ gaze patterns when producing two different manifestations of the filler (‘euh’ vs. ‘euhm’ in Dutch) are compared. The distribution of gaze just before, during and after ‘euh’ and ‘euhm’ was studied in three triadic conversations between acquainted students. In each triad, the participants had a free conversation of approximately 15 minutes. The triads were recorded in an experimental setting, using mobile eye tracking glasses (Gullberg & Kita 2009; Jokinen 2010; Brône & Oben 2015). After synchronizing the videos made by the mobile eye trackers and the static camera, the conversations were transcribed and annotated for gaze target and fillers. We expect a difference in gaze pattern between ‘euh’ and ‘euhm’, as the two realizations of the filler also differ in other aspects. As already observed before, the formal realization of a filler affects its communicative function. The vocal ‘uh’/’euh’ seems to warn the interlocutor for a short interruption, whereas the vocal-nasalic ‘um’/’euhm’ announces a longer delay (Clark, 1994). This functional difference correlates with differences on the formal level. For example, ‘euhm’ occurs more frequently at major discourse boundaries and is more likely to be surrounded by silent pauses than ‘euh’ (De Leeuw, 2007). The analysis of four triadic conversations reveals that also in gaze distribution, ‘euh’ and ‘euhm’ show some interesting differences. In the data, 81 occurrences of ‘euh(m)’ were found. More than ‘euh’, ‘euhm’ is accompanied by a gaze aversion away from the previous target or gaze fixation to the background. During ‘euh’, on the contrary, speakers’ gaze stays more frequently fixed on an interlocutor. In addition, other interactional differences could be found: ‘euhm’ is surrounded by more silent pauses and occurs more often in turn-final and turn-initial position. The multimodal analysis we present in this study thus supports previous findings of an interactional difference in vocal and vocal-nasalic fillers.status: publishe