Rationale for choices of stimuli set:

1.	We want to keep the number of elicited tokens per factor/category approximately the same. So, up to 3 pairs per factor is reasonable and sufficient for our purposes.

2.	We want to keep the tones as similar as possible, so the closer the tone levels were, the better the pair was for stimuli. If we were unable to find matching tones, then we favored level tone pairs over dynamic ones.

3.	We excluded words which begin in a vowel or a glide from the dataset as they will fuse with the preceding vowel in the carrier.