English is the most important lingua franca on the European continent, with countless Europeans using it daily for cross-cultural communication in their private and/or professional lives. Such English as a lingua franca (ELF) encounters are naturally characterized by considerable phonetic-phonological variation due to the presence of speakers from various L1 backgrounds. The question of how to maintain phonological intelligibility in ELF communication has thus concerned researchers for quite some time now. Previous research in this respect has focused primarily on the contribution of particular pronunciation features to phonological (un)intelligibility (e.g. Jenkins 2000, Deterding 2013), yet research on the impact of other factors, notably that of linguistic co-text and extralinguistic context, has been comparatively scarce. This is at least partly due to the influence of a study by Jenkins (2000), which suggested that most non-native ELF users do not profit much from co-textual and contextual cues in spoken word recognition. Moreover, most studies in this respect have been carried out on the basis of qualitative examinations of interactive speech data (e.g. Jenkins 2000, Deterding 2013, Kennedy 2012), where a thorough quantification of the contribution of various within- and between-subject factors regarding international pronunciation intelligibility has not been possible.

This paper fills this research gap by modelling the effects of co-text, context and several relevant listener variables (e.g. language attitudes or familiarity with a particular type of speech or a particular accent - cf. Gass & Varonis 1984, Rubin 1992) on phonological intelligibility in ELF communication. In this large-scale investigation (n=508), co-textual and contextual effects were operationalized in the form of four experimental conditions. Listeners were hearing
(1) isolated words (i.e. no co(n)text),
(2) words embedded in semantically ‘neutral’ syntactic co-text that only enabled the listener to identify the PoS of the target word: e.g. “It's quite _ .” (flat)
(3) words embedded in syntactic co-text containing a semantic prime: e.g. “His favourite colour is.” (purple)
(4) words embedded in syntactic co-text with a schematic prime: e.g. At the airport: “I need to pick up my__.” (bag)

All listeners were exposed to all four conditions in a randomized order. Phonological intelligibility was operationalized as the number of target words correctly transcribed by listeners. Logistic mixed regression revealed significant effects on phonological intelligibility of the variables condition, self-assessed listening proficiency in English, familiarity with the Austrian and similar accents in English, familiarity with various types of spoken English, as well as attitudes towards the speaker. Notably, among all variables examined, condition was the one that had the strongest effect on phonological intelligibility, though its effect was regulated by the variable listening proficiency. These results stand in stark contrast to Jenkins’ (2000) findings, since they strongly suggest the importance of co-textual and contextual effects regarding international pronunciation intelligibility. Alternative explanations for the weaker effect of some of the listener variables examined in the present study are discussed as well.