Crossroads research revealed that in addition to patterns diffused over wide areas, local configurations play a decisive role. Watson (2018) showed that in the lexicon, convergences are explainable through shared social practices, for instance joint agricultural labour, and that differences in basic lexicon are upheld in Crossroads language despite close historical connectedness.
Cobbinah (2017) discovered that agreement patterns are not influenced by contact languages but by the prominence of particular agreement types in a particular language ecology, i.e. in the majority of languages in a particular sociolinguistic space. Cobbinah (forthcoming d) investigated the realisation of direct objects in multilingual speech in a quantitative study based on data collection with the Pear Story stimulus. The results show that object realisation or ellipsis is not uniform across languages but correlated with communities of practice, for instance inhabitants of a household or neighbourhood who share a pattern across all their languages, and does not correspond to the patterns described in available grammatical descriptions of these languages. This entails that, in the absence of a standard variety, which is often used in language contact study, there is no stable reference point from which to calibrate variation.
Building on this observation, Watson (2019) developed an empirical model that captures how speakers, based on their repertoires, categorise named languages, and with which linguistic features this prototypical cognitive reification is associated, so that the parameters they use to categorise speech can be revealed.
Krajcik (forthcoming 2019) conducted a quantitative study of speech and gesture in motion event description with speakers at the Crossroads. Counter cross-linguistic predictions, gesture and speech are not aligned; rather, study participants use gestures encoding the manner of placement or shape of the placed object even when in their accompanying speech, manner of placement is not expressed, independently of language used. This unexpected attention to manner and shape is likely to be caused by the preponderance of languages with noun class systems that encode size and shape of objects in their repertoires so that they are primed to encode these configurational aspects in their gesture, irrespective of the information encoded in the predicate, demonstrating, like Cobbinah’s and Watson’s research, the importance of local language ecologies and their linguistic facets.