This thesis investigates attitudes towards varieties of Arabic. A key characteristic of the sociolinguistics of the Arab world is that Standard Arabic, Classical Arabic, and Modern Standard Arabic co-exist in a diglossic relationship with vernacular Arabic varieties spoken in many different Arab countries. The spoken regional varieties are not always comprehensible to other speakers from different geographical regions. The linguistic study of Arabic has often focused on the relation between Fusha (Standard Arabic) and ammiyya (the vernacular spoken Arabic varieties). This thesis explores the relationship between Fusha and ammiyya varieties of Arabic, from several different perspectives. Firstly, the research employs a direct approach, using accent labels in a questionnaire and focusing on Jordanian participants’ attitudes towards their dialects and dialects of other Arabic varieties. This is referred to as Study 1. Secondly, with an indirect approach, it uses listening experiments with audio clips of Arabic speakers to explore attitudes towards Arabic and Arabic-accented English in two different speech styles (reading vs. speaking). As well examining listener attitudes along the dimensions of ‘status’ and ‘solidarity’, I also examine listeners’ ratings of ‘comprehensibility’ and ‘accentedness’. This is referred to as Study 2.

The results obtained in Study 1 show that Jordanian participants hold different attitudes towards standard and non-standard varieties, as expected. Participants are, in general, proud of their own dialects, but overall, the Jordanian Urban dialect is the most preferred. However, participants stated that the Urban dialect should not be used in educational domains. The Bedouin dialect is seen to be the ‘original dialect of Jordan society’, conveying a sense of historical prestige. MSA was rated the highest in terms of characteristics such as ‘power’, ‘understandability’, and ‘pleasantness’, but it was rated lower on ‘wealth’ and ‘toughness’. However, the results indicate possible change over time: younger participants rate MSA lower for standardness and prestige compared to older participants. When rating other dialects, Jordanian participants often ranked Jordanian dialects high for social characteristics such as ‘pleasantness’, but Jordanian dialects are not always ranked highest overall (Urban is ranked amongst the lowest for ‘toughness’). For dialects from outside Jordan, Moroccan Arabic consistently received low scores. Taken together, these results increase our understanding of the attitudes of Jordanian speakers to their own and other Arabic dialects.

The results obtained in Study 2 showed listeners are more likely to correctly identify the regional origin of the speakers when speaking Arabic than when speaking English, and

in Arabic when speaking rather than reading. Perhaps surprisingly, listeners can identify the regional origin of some speakers even when using MSA (i.e. in the reading style). The Egyptian speaker was the most correctly identified, almost at ceiling rate in Arabic, and very often even when speaking English. The correct identification of the Egyptian dialect is likely due at least in part of the high frequency in which Egyptian Arabic is broadcast on television and in other forms of media. The other varieties were more often correctly identified in Arabic than in English, and more often correctly identified in Arabic speaking style than in Arabic reading style. For example, Moroccan Arabic was correctly identified 99% of the time in spoken Arabic, 40% in reading Arabic, but only 8% and 6% of the time in English reading and speaking, respectively. This shows, for example, that although MSA is often viewed as an invariant target, in actual fact phonological variation in production provides listeners cues about the regional origin of the speaker. In general, the attitude results in Study 2 are congruent with those for Study 1. For example, Moroccan Arabic is rated low for ‘pleasantness’ and ‘educated’ in Study 2, as in Study 1. The results also show that attitude scores can change if the regional variety is correctly identified or not. For example, the Jordan Urban speaker scored higher for ‘standardness’ and ‘educated’ when correctly identified than when incorrectly identified in Arabic reading style (and English speaking style, for ‘standardness’). This suggests that when listeners believed they were listening to a Jordan Urban speaker, but they were not, their responses nevertheless matched beliefs about the Jordan Urban dialect (shown also in Study 1). Study 2 also examined the accentedness and comprehensibility of the speakers. In keeping with other results, Moroccan Arabic was rated the most accented. However, the Moroccan speaker in the English reading style was rated to be less accented when incorrectly identified than when correctly identified. This suggests again that listeners awarded ratings based on their beliefs about the regional origins of the speaker.

Taken together, these results advance our knowledge of language attitudes in the field of Arabic sociolinguistics, and show that dialect identification, attitude ratings, accentedness, and comprehensibility ratings should be studied in combination, to shed new light on their complex relationship.