Abstract
This paper focuses on non-standard spelling patterns in Facebook messages. We analyse German Facebook messages from around 100 South Tyrolean Facebook users collected for the DiDi Corpus (www.eurac.edu/didi) with respect to the variants used for unstressed word-final (pronounced as [ɐ]) in polysyllabic words in Standard German. The DiDi Corpus contains a considerable portion of non-standard and dialectal messages; in addition, the corpus provides a wide range of socio-demographic information about the writers (such as age, gender, geographic origin and other) which is available as metadata. The corpus thus is suitable to investigate spelling variation from a sociolinguistic point of view taking into account extra-linguistic features of the Facebook users. We will provide examples of regional features and analyse the distribution of such features. Results show that the standard is the most preferred variant in the corpus; however, age is a strong predictor for variation as young writers more often use an alternative spelling than older writers do. In addition, the non-standard variants in the corpus reflect features of spoken varieties and differences within the investigated territory. Thus, the geographic origin of writers is a good predictor for the way of variation.