Testing trade-offs between gender and number indexing and other cues to A and P arguments: A corpus-based perspective