Kyh/Materials/Karuk: analysis - embedded hierarchies
HBC using Multiple embedded hierarchies
- (1) Simple hierarchies
- a. 1/2>3 >>
- b. pl>sg >>
- c. Obj:pl>Subj:pl >>
- d. Subj:sg>Obj:sg
The hierarchy of arguments is comprised of four hierarchies (see (1)), embedded in one another (or ranked in sequence). The first is a simple prototypical person hierarchy, with non-third person arguments always outranking third-person arguments in the competition for verb agreement. The second is a simple prototypical number hierarchy, where a plural argument will always outrank a singular argument. Finally there are two opposite grammatical-role hierarchies, ranking objects higher in interactions where both arguments are plural, and subjects where both arguments are singular. The simplest way to understand this HBC structure is the following algorithm:
- (2) Parsing Algorithm
- a. Select higher ranking argument on highest hierarchy [initially 1-a]
- b. If both arguments are equal on this hierarchy (or if it's irrelevant), use next hierarchy on the list
- c. Repeat
In other words, the first hierarchy is used for HBC, but if it is inconclusive (1/2 > 2/1 or 3 > 3) the second is used. When that is inconclusive (pl > pl or sg > sg), the third or fourth is used as applicable.
These hierarchies can be embedded within one another, producing the following composite hierarchy:
- (3)Composite Hierarchy
The exception to this rule is that a 1sg object will always be agreed with. However, this is only a partial exception, as 1sg object agreement is accompanied by partial subject agreement for [+pl] subjects. Nonetheless, this appears to be a defective structure, possibly indicating an earlier complete subject or object agreement pattern for 1sg objects. The lack of overt subject agreement for sg > 1sg may simply be a strategy to avoid redundancy (no ambiguity is created regarding the number of the subject.) Besides the case of 1sg objects, the hierarchies in (1) and (3) explain the entire agreement pattern, assuming the appropriate specification or underspecification of the agreement affixes (for instance, ʔu- must be (under)specified as [+3], marking subject agreement for 3sg > 3sg and object agreement for 3sg>3pl). However, this seems far less of a stretch than analyses which assume each and every one of the affixes marks both subject and object.
Issues (a.k.a "problems")
nu- and ni-: the 1st-person subject hitch
In the positive paradigm, nu- is a prefix that appears in agreement with 1pl subjects as well as one 1sg subject (with 2sg object); in the further two cases of 1sg subject agreement, ni- is used instead. ni- can either be analyzed as a portmanteau for 1sg > 3, or it can be analyzed as a specific 1sg subject agreement marker, with nu- acting as a default 1st-person agreement marker. For the latter analysis one must assume that in the cases of 1 > 2sg, agreement with the subject is defective and limited to person only.
- A possible solution would be to lean on Iconic Number Theory and specify nu- using the system of multiple privative number features.
The same applies for the optative and negative paradigms, where different forms are used in the same pattern. The negative series, which uses na- for 1sg subject agreement (the same form as marks 1sg objects) and kin- for the other 1st-person agreement markings (the same form as marks 1pl objects.) This synchronicity seems to lend a bit of support to the analysis that the 1 > 2sg forms are simply underspecified for number. In such an analysis, kin- could simply be specified as a default 1st-person agreement marker and na- as a default 1sg agreement marker. In the alternative analysis, at least one separate na- morpheme would have to be posited.
In the cases of 1 > 2 and 3 > 3, subject agreement occurs where the object is singular and object agreement where the object is plural. Additionally, in the 1 > 2sg cases we may want to assume the subject agreement is underspecified for number. For the 3 > 3sg cases, this would certainly do no harm, and would avoid redundancy (3sg > 3sg must use a subject agreement prefix that has to be unspecified for number and role at any rate). A pattern emerges: subject agreement, when determined by the lower number-based hierarchies in (1), is unspecified for number. problem/exception: 3sg > 3sg or 3pl > 3sg actually has to have number specified
Intransitive subject agreement
Intransitive forms are identical with 3sg-object forms, in which subject agreement invariably takes place. This is consistent with (1) and (3), in which a 3sg object is the lowest possible type of argument on the hierarchy(ies). This leads to the conclusion that 1/2/3 > 3sg is the most unmarked transitive interaction, naturally lending itself to use for intransitive agreement.
The suffix -ap
Possibly to create further contrast between 2nd person object and subject marking (which, almost invariably, use the same prefixes, differentiated only for number – the only exception is ku:[A+2+pl–MOD]), the suffix ap:[P+2] is appended. This could be interpreted as a defective inverse marker parasitic to 2nd-person objects (this contrasts with Macaulay's (1992) analysis of -ap as a defective inverse marker only in that Macaulay did not point out that it appears exclusively and consistently in inverse relations with a 2nd-person object.)
Beside the ap:[P+2] morpheme, a second, phonologically identical suffix is appended in some negative verb forms: ap:[+NEG]//[+pl].
This form appears only where agreement with a [+pl] argument is found: 2pl/3pl > 1sg; 2/3 > 1pl; 2pl > 3; 3 > 3pl; 3pl > 3sg (underlined is the argument with which the verb agrees; in the case of 2pl/3pl > 1sg, this agreement is only partial, and secondary to object agreement).
In 1pl > 2/3, where one would expect negative -ap, the suffix is blocked by the special status of [+1]. Possibly as an alternative strategy for negative marking, different prefix forms are used for 1st-person subject agreement than are used in the positive paradigms. These appear to be complementary strategies for marking negative polarity in the cases of subject-agreement which agree with a first person argument. It should be noted that only 2sg/3sg > 1sg cases remain identical and lacking an overt suffix throughout all paradigms.
- Relativized Parse
- Parse X y/z
- Parse features of type X from a head with features y when adjacent to a head with features z
- Parse Φ [+1]/[+2]
- Parse Φ [+2]/[+1]
- Parse Φ [+1]/[+3]
- Parse Φ [+2]/[+3]
- Parse Φ [+pl]/[+sg]
- Parse Φ [P+pl]/[A+pl]
- Parse Φ [A+sg]/[P+sg]
- Parse [+pl] [A+2]/[P+1]
- Parse [+pl] [A+3]/[P+1]
|PrsΦ1/3, PrsΦ2/3, Prs[+pl]A2/P1, Prs[+pl]A3/P1 ≫|
(This has to be completed for suffixes, negative marking in general, and the 1st-person subject agreement synchronicity.)
Cf. Morpheme list
- ka:[A +pl] / na:[+1 –pl]
- kan:[A +1 –pl +mod]
- ni:[A +1 –pl]
- na:[+1 –pl]
- nu:[A +1 +pos]
- kin:[+1] ([–3]?)
- ku:[A +2 +pl –mod]*
- kiik:[+2 +pl]
- ∅:[A +2 –pl +mod]
- ∅:[+2 –pos]
- kin:[P +3 +pl]
- kun:[A +3 +pl]*
- kam:[+3 +mod]
- *These morphemes are replaced by null forms in the negative paradigm.
- ap:[P +2]
- ap:[–POS] // [+pl]