The neglect of kinship in current anthropology and in the cognitive sciences is not far short of a scandal. Humans are the categorizing species, and kinship systems categorize our own most significant others, so reflecting fundamental forms of social organization. In small-scale societies, kinship forms the backbone of the political and economic organization, and even in complex societies, it plays a significant role (Kuper Reference Kuper2009). Moreover, kinship is one of two primordial foundations for relational cognition – the other being, as Jones notes, spatial relations (kinship identifies a person by a relation to Ego; spatial relations identify a place relative to a landmark, where a place is rather more abstract than a person). Advanced relational reasoning is the big divide between us and the other primates: Even our cousins the chimps have a hard time doing relational thinking (Penn et al. Reference Penn, Holyoak and Povinelli2008). It is likely that language and relational reasoning are mutually implicated, although which provides the foundation for the other is controversial (Loewenstein & Gentner Reference Loewenstein and Gentner2005). Many small-scale societies have such complex kinship systems that they have exercised professional mathematicians, while still being mastered by kids.
So, Jones has done us a big favor by putting kinship back on the cognitive science agenda with an interesting new twist. It has long been obvious that despite exuberent variation in the kinship systems of the world, there seem to be a limited set of underlying types (see, e.g., the important work by Godelier et al. [1998], unfortunately not referenced by Jones). Jones gives us a very neat way to think about this pattern in the diversity, in terms of differentially ranked principles for grouping and distinguishing kin. Previous approaches, such as componential analysis and reduction rule analysis (see Tyler 1969 for a range of approaches), have each captured part of the phenomenon but somehow have failed to give us an exhaustive way to think about the typology of kinship systems. A componential analysis, for example, might characterize cousin as +collateral, +same generation, unmarked for sex, while brother might be +lineal, +same generation, +male (Romney & D'Andrade Reference Romney, D'Andrade and Tyler1969), while a reduction rule analysis might tell us that a mother's brother's son's son is like a mother's brother's son, so counts as a kind of (less prototypical) cousin.
Jones' system uses just eight “faithfulness constraints,” or instructions, to group or individuate kin of different kinds, which are further constrained by three “markedness constraints” giving us implicational scales of the kind “if you distinguish different kinds of cousins, you should distinguish different kinds of siblings.” What this actually amounts to is a way of formulating metaconstraints over componential analyses of different kinship systems. If you reorder his constraints, some distinctions will be made and others not – so you can't have all your cake and eat it too. The markedness constraints will ensure that there is, for example, a greater lumping of distant kinsmen and greater splitting of close kinsmen. The whole system of Jonesian constraints then acts to constrain the possible componential analyses available to kinship systems. This looks as though it might be a major advance in the typology of kin term systems, although it will need testing on a representative sample of the exuberant variation (if there are 7,000 languages in the world, there are 7,000 kinterm systems, even if many of these will prove isomorphic – note though that Tamil, for example, has a dozen or more distinct kin term systems adapted to the different castes: See Levinson Reference Levinson1977).
Now, this system doesn't capture the notion of prototype kin and their extensions, so nicely captured in the reduction rule analysis of Lounsbury (Reference Lounsbury and Tyler1969), and argued for on psychological grounds by Malinowski. Here Jones plans an addition to the OT framework with a recursive application of the constraints. It is hard to see exactly what the implications for typology might be, but it must multiply the possibilities enormously, so undermining some of the attractions of the OT approach. It would seem better to keep separate the categories from the algorithms that assign referents to them, for the simple reason that we know that individuals use more than one manner of assigning kin to categories (Levinson Reference Levinson2006). This implies that the prototype structure of kin categories should form part of the primary category structure, not part of the assignment algorithm.
A major issue though is the psychological reality of all this. This is the Achilles' heel of OT analyses, which are typically derived by detailed comparison across languages: They are thus constructed with all the metalinguistic hindsight of the (hypothetically) omniscient analyst. It is quite unwarranted (despite the normal claims of OT) to think that any one native speaker has this kind of meta-knowledge in his or her head. To learn a kin term system a child starts off by learning “names” for those around her, then grasps the relational character (other kids have “moms” too), then learns the extensions – and kin terms are one domain of language where explicit instruction and correction are common (see Hirschfeld Reference Hirschfeld1989 for review). Thus an individual can learn to use the system without having an overview of how it works. Kinship systems are a wonderful example of culture as the second great invisible watchmaker – intricate systems without a designer. They are honed by cultural evolution, can be exceedingly complex like the Australian systems, and because they have to serve important social functions of regulating reproduction and social conduct, they are directly under selection by the social systems they must integrate with. Jones' constraints, if they correctly capture the typology of extant kinship systems, may do so not because the categories are antecedently in our mind, but because these are the ones that societies need to work with.
Jones' paper bristles with other interesting ideas. Particularly interesting, perhaps, is the idea that kinship systems have a digital character – unlike the analog (or gradient) nature of utility functions – because they are solutions to communicative coordination games. Here as elsewhere, Jones perhaps fails to draw a sharp enough line between kin term systems (linguistic in nature) and kinship systems (systems of reproduction, marriage, inheritance and authority). Note how the kin-selection metric (Hamilton's rule) does not map neatly onto any known kin term system – a Seneca cousin, for example, is treated as equal to Ego's sibling, even though the cousin shares 12.5% of his or her genes with Ego and the sibling 50%. In some ways the kin-term system will regulate the kinship system (e.g., in the kind of behavior appropriate to categories), but in other ways (e.g., inheritance) the two may part company. The kin-term system is part of the language and communication system, the kinship system is part of a social organization. Insofar as they are coincident, and kinship is digital, this is part of the magical power of language to construct the categories of our world (Levinson Reference Levinson, Gentner and Goldin-Meadow2003).
You have
Access
The neglect of kinship in current anthropology and in the cognitive sciences is not far short of a scandal. Humans are the categorizing species, and kinship systems categorize our own most significant others, so reflecting fundamental forms of social organization. In small-scale societies, kinship forms the backbone of the political and economic organization, and even in complex societies, it plays a significant role (Kuper Reference Kuper2009). Moreover, kinship is one of two primordial foundations for relational cognition – the other being, as Jones notes, spatial relations (kinship identifies a person by a relation to Ego; spatial relations identify a place relative to a landmark, where a place is rather more abstract than a person). Advanced relational reasoning is the big divide between us and the other primates: Even our cousins the chimps have a hard time doing relational thinking (Penn et al. Reference Penn, Holyoak and Povinelli2008). It is likely that language and relational reasoning are mutually implicated, although which provides the foundation for the other is controversial (Loewenstein & Gentner Reference Loewenstein and Gentner2005). Many small-scale societies have such complex kinship systems that they have exercised professional mathematicians, while still being mastered by kids.
So, Jones has done us a big favor by putting kinship back on the cognitive science agenda with an interesting new twist. It has long been obvious that despite exuberent variation in the kinship systems of the world, there seem to be a limited set of underlying types (see, e.g., the important work by Godelier et al. [1998], unfortunately not referenced by Jones). Jones gives us a very neat way to think about this pattern in the diversity, in terms of differentially ranked principles for grouping and distinguishing kin. Previous approaches, such as componential analysis and reduction rule analysis (see Tyler 1969 for a range of approaches), have each captured part of the phenomenon but somehow have failed to give us an exhaustive way to think about the typology of kinship systems. A componential analysis, for example, might characterize cousin as +collateral, +same generation, unmarked for sex, while brother might be +lineal, +same generation, +male (Romney & D'Andrade Reference Romney, D'Andrade and Tyler1969), while a reduction rule analysis might tell us that a mother's brother's son's son is like a mother's brother's son, so counts as a kind of (less prototypical) cousin.
Jones' system uses just eight “faithfulness constraints,” or instructions, to group or individuate kin of different kinds, which are further constrained by three “markedness constraints” giving us implicational scales of the kind “if you distinguish different kinds of cousins, you should distinguish different kinds of siblings.” What this actually amounts to is a way of formulating metaconstraints over componential analyses of different kinship systems. If you reorder his constraints, some distinctions will be made and others not – so you can't have all your cake and eat it too. The markedness constraints will ensure that there is, for example, a greater lumping of distant kinsmen and greater splitting of close kinsmen. The whole system of Jonesian constraints then acts to constrain the possible componential analyses available to kinship systems. This looks as though it might be a major advance in the typology of kin term systems, although it will need testing on a representative sample of the exuberant variation (if there are 7,000 languages in the world, there are 7,000 kinterm systems, even if many of these will prove isomorphic – note though that Tamil, for example, has a dozen or more distinct kin term systems adapted to the different castes: See Levinson Reference Levinson1977).
Now, this system doesn't capture the notion of prototype kin and their extensions, so nicely captured in the reduction rule analysis of Lounsbury (Reference Lounsbury and Tyler1969), and argued for on psychological grounds by Malinowski. Here Jones plans an addition to the OT framework with a recursive application of the constraints. It is hard to see exactly what the implications for typology might be, but it must multiply the possibilities enormously, so undermining some of the attractions of the OT approach. It would seem better to keep separate the categories from the algorithms that assign referents to them, for the simple reason that we know that individuals use more than one manner of assigning kin to categories (Levinson Reference Levinson2006). This implies that the prototype structure of kin categories should form part of the primary category structure, not part of the assignment algorithm.
A major issue though is the psychological reality of all this. This is the Achilles' heel of OT analyses, which are typically derived by detailed comparison across languages: They are thus constructed with all the metalinguistic hindsight of the (hypothetically) omniscient analyst. It is quite unwarranted (despite the normal claims of OT) to think that any one native speaker has this kind of meta-knowledge in his or her head. To learn a kin term system a child starts off by learning “names” for those around her, then grasps the relational character (other kids have “moms” too), then learns the extensions – and kin terms are one domain of language where explicit instruction and correction are common (see Hirschfeld Reference Hirschfeld1989 for review). Thus an individual can learn to use the system without having an overview of how it works. Kinship systems are a wonderful example of culture as the second great invisible watchmaker – intricate systems without a designer. They are honed by cultural evolution, can be exceedingly complex like the Australian systems, and because they have to serve important social functions of regulating reproduction and social conduct, they are directly under selection by the social systems they must integrate with. Jones' constraints, if they correctly capture the typology of extant kinship systems, may do so not because the categories are antecedently in our mind, but because these are the ones that societies need to work with.
Jones' paper bristles with other interesting ideas. Particularly interesting, perhaps, is the idea that kinship systems have a digital character – unlike the analog (or gradient) nature of utility functions – because they are solutions to communicative coordination games. Here as elsewhere, Jones perhaps fails to draw a sharp enough line between kin term systems (linguistic in nature) and kinship systems (systems of reproduction, marriage, inheritance and authority). Note how the kin-selection metric (Hamilton's rule) does not map neatly onto any known kin term system – a Seneca cousin, for example, is treated as equal to Ego's sibling, even though the cousin shares 12.5% of his or her genes with Ego and the sibling 50%. In some ways the kin-term system will regulate the kinship system (e.g., in the kind of behavior appropriate to categories), but in other ways (e.g., inheritance) the two may part company. The kin-term system is part of the language and communication system, the kinship system is part of a social organization. Insofar as they are coincident, and kinship is digital, this is part of the magical power of language to construct the categories of our world (Levinson Reference Levinson, Gentner and Goldin-Meadow2003).