Jones' analysis of Seneca kinship is inadequate. He restricts his OT analysis to Ego's and the first ascending generation and within these generations to only parents, aunts, uncles, siblings, and first cousins. Seneca kinship terminology covers the infinity of possible kin relations in Ego's and the two adjacent generations. To account for more distant relatives, Jones suggests, but does not provide, a set of rewrite rules intended to reduce more distant kin relations to those just listed. The OT analysis correctly classifies the restricted set of kin relations in Ego's and the first ascending generation to which it is applied, but fails on the close kin of the first descending generation (children, nephews, nieces). The full model, relying on the kind of reduction rules Jones suggests, cannot correctly classify the more distant kintypes.
For his OT analysis of close kin, Jones does not consider the first descending generation, that of Ego's children and niblings (nieces and nephews). Here the OT analysis fails. Ego's sister's child is a member of Ego's matriline, regardless of Ego's sex. So according to Jones's OT constraint ranking, in which DISTINGUISH MATRILINE outranks all other relevant constraints, Ego's sister's children are classed with Ego's children. In Seneca they are not. A man's sister's child is not classed with his children because a man and his sister are of opposite sex. Matriline membership is irrelevant. The classification of a man's sister's child as a cross relative exemplifies a semantic distinction that governs the classification of all of the infinity of kintypes in Ego's and the two adjacent generations in Seneca. Parallel collaterals are merged terminologically with Ego's lineals or siblings (henceforth, for convenience, “lineals” tout court) of the corresponding generation; cross relatives are terminologically distinguished from Ego's lineals. This pervasive distinction, to which we now turn, is implicit in Lounsbury (Reference Lounsbury and Hunt1964) and explicit in Kay (Reference Kay1975).
The Iroquois cross-parallel distinction is given in (1).
-
(1) Lineal relatives (including siblings) are Iroquois parallel. Two collaterally related kintypes are parallel iff the furthest separated, same-generation nodes on the genealogical path connecting them are of the same sex. Collateral relatives that are not parallel are cross.
The definition is most easily unpacked in terms of the traditional terminology of English cousin numbering. The children of siblings are first cousins, the children of first cousins are second cousins; the children of nth cousins are (n+1)th cousins. These are the “integral” cousins, cousins of the same generation. For nonintegral cousins, two relatives are ith cousins j times removed iff one is an ith cousin of a jth generation ancestor of the other (j>0). What counts for the Iroquois cross-parallel distinction, is the relative sex of the “ith” cousins (or siblings). Iff that pair of integral cousins or siblings are of the same sex, Alter is a parallel relative of Ego.
Looking back at the subset of relatives analyzed by OT, mother's sister and father's brother are parallel, and mother's brother and father's sister are cross because of the relative sexes of those sibling pairs. No mention of matriline membership or of minimizing-kinterm-sets is necessary. Similarly, in Ego's generation, the first cousins whose respective parents are of the same sex are parallel and therefore merged with siblings; those whose respective parents are not of the same sex are cross and therefore not so merged. Again issues of matriline membership and/or minimizing cousin versus sibling categories are otiose. In the first descending generation, alter is classed with Ego's children (parallel) iff he or she is the child of Ego's same-sex sibling. Classifying in this generation on the basis of matriline gets the facts wrong.
Seneca kinship terminology extends outward from Ego indefinitely in Ego's and both adjacent generations. Jones is aware that even in Ego's and the first ascending generation classification of more distant relatives, for example, second cousins, on the basis of matriline membership doesn't work. He proposes an unspecified set of reduction rules, which, successively applied, are intended to eventually rewrite longer kintype expressions to the small set analyzed by OT. (Since the OT analysis on offer does not work for any descending generation, we must imagine a new OT analysis that works for a small subset of first descending generation kintypes, to which an infinite set of kintypes will be reduced by the reduction rules.) These rules are by definition local. “It is convenient (but maybe not absolutely necessary) to assume that, with each move, at most one pair of adjoining elementary kin types changes” (sect. 4.1, emphasis added). The single example given reduces a genealogical path by application of two rules, each replacing two adjacent nodes with a single node that shares either the sex or the generation of one of the nodes it replaces. But no set of such local rules can account for the Seneca cross/parallel facts because the Seneca cross/parallel distinction is based on the relative sex of two arbitrarily separated nodes, the two most distant same-generation nodes. For example, Ego's father's mother's father's mother's brother's daughter's son's son's child is terminologically classed as Ego's sibling because Ego's father and alter's father are of the same sex. Nothing else matters.
One could of course imagine rewrite rules that looked at nonadjacent nodes. Or one might perhaps add some condition(s) on strictly local rules that would have this effect. But any such moves would just come down to a needlessly complicated notational equivalent of the definition of Iroquois cross-parallel given in (1). That definition effects the accurate classification of all Seneca kin in the relevant generations, without recourse to (a) separate formalisms for close and different kin, (b) ordering of constraints (OT), or (c) reduction rules. No set of local reduction rules of the kind Jones hints at can do the job because the key dependency is nonlocal. Additionally, the OT analysis of the finite subset fails on the facts in the first descending generation. The rejected model, l'il ol' componential analysis, gets the whole job done accurately with less machinery (Kay Reference Kay1975; Lounsbury Reference Lounsbury and Hunt1964).
You have
Access
Jones' analysis of Seneca kinship is inadequate. He restricts his OT analysis to Ego's and the first ascending generation and within these generations to only parents, aunts, uncles, siblings, and first cousins. Seneca kinship terminology covers the infinity of possible kin relations in Ego's and the two adjacent generations. To account for more distant relatives, Jones suggests, but does not provide, a set of rewrite rules intended to reduce more distant kin relations to those just listed. The OT analysis correctly classifies the restricted set of kin relations in Ego's and the first ascending generation to which it is applied, but fails on the close kin of the first descending generation (children, nephews, nieces). The full model, relying on the kind of reduction rules Jones suggests, cannot correctly classify the more distant kintypes.
For his OT analysis of close kin, Jones does not consider the first descending generation, that of Ego's children and niblings (nieces and nephews). Here the OT analysis fails. Ego's sister's child is a member of Ego's matriline, regardless of Ego's sex. So according to Jones's OT constraint ranking, in which DISTINGUISH MATRILINE outranks all other relevant constraints, Ego's sister's children are classed with Ego's children. In Seneca they are not. A man's sister's child is not classed with his children because a man and his sister are of opposite sex. Matriline membership is irrelevant. The classification of a man's sister's child as a cross relative exemplifies a semantic distinction that governs the classification of all of the infinity of kintypes in Ego's and the two adjacent generations in Seneca. Parallel collaterals are merged terminologically with Ego's lineals or siblings (henceforth, for convenience, “lineals” tout court) of the corresponding generation; cross relatives are terminologically distinguished from Ego's lineals. This pervasive distinction, to which we now turn, is implicit in Lounsbury (Reference Lounsbury and Hunt1964) and explicit in Kay (Reference Kay1975).
The Iroquois cross-parallel distinction is given in (1).
(1) Lineal relatives (including siblings) are Iroquois parallel. Two collaterally related kintypes are parallel iff the furthest separated, same-generation nodes on the genealogical path connecting them are of the same sex. Collateral relatives that are not parallel are cross.
The definition is most easily unpacked in terms of the traditional terminology of English cousin numbering. The children of siblings are first cousins, the children of first cousins are second cousins; the children of nth cousins are (n+1)th cousins. These are the “integral” cousins, cousins of the same generation. For nonintegral cousins, two relatives are ith cousins j times removed iff one is an ith cousin of a jth generation ancestor of the other (j>0). What counts for the Iroquois cross-parallel distinction, is the relative sex of the “ith” cousins (or siblings). Iff that pair of integral cousins or siblings are of the same sex, Alter is a parallel relative of Ego.
Looking back at the subset of relatives analyzed by OT, mother's sister and father's brother are parallel, and mother's brother and father's sister are cross because of the relative sexes of those sibling pairs. No mention of matriline membership or of minimizing-kinterm-sets is necessary. Similarly, in Ego's generation, the first cousins whose respective parents are of the same sex are parallel and therefore merged with siblings; those whose respective parents are not of the same sex are cross and therefore not so merged. Again issues of matriline membership and/or minimizing cousin versus sibling categories are otiose. In the first descending generation, alter is classed with Ego's children (parallel) iff he or she is the child of Ego's same-sex sibling. Classifying in this generation on the basis of matriline gets the facts wrong.
Seneca kinship terminology extends outward from Ego indefinitely in Ego's and both adjacent generations. Jones is aware that even in Ego's and the first ascending generation classification of more distant relatives, for example, second cousins, on the basis of matriline membership doesn't work. He proposes an unspecified set of reduction rules, which, successively applied, are intended to eventually rewrite longer kintype expressions to the small set analyzed by OT. (Since the OT analysis on offer does not work for any descending generation, we must imagine a new OT analysis that works for a small subset of first descending generation kintypes, to which an infinite set of kintypes will be reduced by the reduction rules.) These rules are by definition local. “It is convenient (but maybe not absolutely necessary) to assume that, with each move, at most one pair of adjoining elementary kin types changes” (sect. 4.1, emphasis added). The single example given reduces a genealogical path by application of two rules, each replacing two adjacent nodes with a single node that shares either the sex or the generation of one of the nodes it replaces. But no set of such local rules can account for the Seneca cross/parallel facts because the Seneca cross/parallel distinction is based on the relative sex of two arbitrarily separated nodes, the two most distant same-generation nodes. For example, Ego's father's mother's father's mother's brother's daughter's son's son's child is terminologically classed as Ego's sibling because Ego's father and alter's father are of the same sex. Nothing else matters.
One could of course imagine rewrite rules that looked at nonadjacent nodes. Or one might perhaps add some condition(s) on strictly local rules that would have this effect. But any such moves would just come down to a needlessly complicated notational equivalent of the definition of Iroquois cross-parallel given in (1). That definition effects the accurate classification of all Seneca kin in the relevant generations, without recourse to (a) separate formalisms for close and different kin, (b) ordering of constraints (OT), or (c) reduction rules. No set of local reduction rules of the kind Jones hints at can do the job because the key dependency is nonlocal. Additionally, the OT analysis of the finite subset fails on the facts in the first descending generation. The rejected model, l'il ol' componential analysis, gets the whole job done accurately with less machinery (Kay Reference Kay1975; Lounsbury Reference Lounsbury and Hunt1964).