This is all too complex for me.

I don’t have the source with me, but I recall a paper about listening to various languages under different signal/ noise thresholds. If I recall correctly, languages like German that have multiple declensions were about to better able to parse noisy samples because of the redundant information. Sorry for not having the source off hand though.