Markedness and voicing gaps in stop and fricative inventories
This study investigates the hypothesis that marked sounds are more likely to be gaps in a sound inventory. A gap is defined as an absence of an [α voice] stop or fricative when the [−α voice] counterpart exists. Different formulations of markedness are tested and evaluated on whether they label the gaps as more marked than attested sounds. Results show an overall success of markedness based on typological attestedness of sounds in labeling gaps as more marked. However, the success of markedness based on aerodynamics and cross-linguistic phonological processes is limited to stops and fricatives, respectively. Analyses also show that gaps in attested inventories are more likely to be marked than gaps from randomized artificial inventories. This discrepancy between attested and artificial inventories shows how markedness, feature economy, and symmetry interact in shaping sound systems of human languages.
- 3.1Inventories and feature systems
- 3.1.1Attested inventories
- 3.1.2Artificial inventories
- 3.1.3Feature systems
- 3.2.1Identifying gaps
- 3.2.2Generating gap-foil pairs
- 3.2.3Statistics of gaps and gap-foil pairs
- 3.3Modeling the identity of gaps through markedness
- 3.3.3Typological frequency
- 3.3.4Voicing likelihood
- 4.1Model comparison in attested inventories
- 4.2Are models’ success unique for attested inventories?
- 4.3Coverage of correct predictions
- 5.Discussion and conclusion