Rice on Athapaskan Verb Chapts 1, 2 , 7 and 8

Note: the inexplicable 'G' in the first few examples from Ahtna in chapter 7 stands for an affix that Kari calls 'Gender', something that Rice treats elsewhere as a noun classifier. Basically, these verbs require a classifier prefix that agrees with whatever their internal argument is — if you're roasting a round thing, you'd use G prefix A, if you're roasting a long skinny thing, you'd use G prefix B, etc. Kari includes 'G' with the lexical entry for the verb because not all Ahtna verbs require the use of such a prefix, only some do.

1. What morphological process is described as both characteristic of templatic morphology and difficult for a templatic account to deal with?

2.    Hupa is a good example of a language where the two voice markers Rice describes — causative and middle — appear to behave compositionally with respect to each other. Because of the morphophonemics of the lg, the two never show up overtly together. Which Hupa examples show that combining the middle with a causative produces a different result than combining the causative with the middle? What problem for the templatic approach does Rice perceive in these facts?

3. Give one pair of verbs for which their co-occurence with the -d marker or the £-marker (sorry, that's the closest I can get to the voicless-l symbol in html) is not compositional — that is, that show these valence/voice markers aren't always doing the semantically compositional job that they're does elsewhere, but are sometimes just the product of a morphological subcategorization.