Lama:  Frequently-Asked Questions

Why don't na: sewa 'cow ran' and wa:l ra sewa 'husband's friend ran' have downstepping in the last syllable?

Because in IPA transcription, the downstep mark is used only where it's necessary. The H tones on the final syllables are lower in pitch than the H tones on the initial syllables--the registers have been lowered by the intervening L's. But this is a predictable (allophonic) effect, and IPA doesn't transcribe it.  In general terms:  IPA transcribes downstep, but not predictable downdrift.

Is there any constraint such as "*Struc" for tone? I am trying to figure out a constraint that is violated by some of the faithful winners?

Sure, but I would just save time by assuming that some faithful winners are simply perfect.

Why does the Start file have Hl L defeat H LL for underlying /HL L/?

Note first:  Hl L and H LL depict exactly the same thing phonetically, as you can determine by working out the pitch registers and tonal targets.  So this is one of the delicate cases that arises in autosegmental phonology:   there are multiple "right" answers (i.e. which derive the phonetically correct result) but the answer favored by the analysis depends on the constraints and their ranking.  I.e. it's a hidden structure problem.

In the analysis I worked out, the constraint banning tones that have no phonetic target is outranked by the Kirchnerian conjoined constraint that prevents spreading from happening to the tones of underlying HL.

I got an answer that works, and I didn't use the Kirchner counterfeeding recipe.  Is that ok?

Yes, in principle.  Probably you are (intentionally or not) using the "hidden structure" recipe, advocated by Smolensky and others.  That is, you've got candidates that are pronounced the same, but differ in terms of hidden stuff (floating tones, uninterpreted tones) in the formal representation.