The online map-based accent-recognition task: a new pathway for perceptual dialectology

We present and evaluate the map-based accent-recognition task, in which listeners place speech fragments of regionally accented Netherlandic Standard Dutch on a map in a browser-based experiment. On the basis of 15,780 accent placements (from 1,578 participants), we demonstrate that this task successfully elicits listener associations between speech and space, which are mediated by listeners’ sociolinguistic representations. We subsequently assess the task’s validity by comparing the listeners’ map-based associations with objective dialectological distances: results show a significant correlation between production and perception and hence establish convergent validity. Finally, we use a hierarchical cluster analysis to map out listeners’ perceptual isoglosses. These reveal that the accent placements largely clustered along the province boundaries, but with dialectologically motivated deviations and subcategorizations. We discuss methodological implications for perceptual dialectology and sociolinguistic implications for Dutch.