I’d guess that the sibilants, consonant clusters, and/or vowel reduction would play a big role.