Over a period of three years, I propose to investigate the Tongan conceptions of social relationships and how these same ideas are actively used in generating political actions. The way in which these notional representations are organized may point toward the presence of a Tongan cultural model. This cultural model may encompasses at least three "mental" modules, including the spatial relationships module, the conceptual structure module, and the action module (action as represented in the mind, see Jackendoff, 1997).

The Kingdom of Tonga, the only remaining Polynesian kingdom, is a constitutional monarchy. The current monarch, King Taufa'ahau Tupou IV, is the heir of a dynasty that goes back at least a millennium. The Constitution granted in 1875 by King George Tupou I sanctioned the format of the present political system (Latukefu, 1974). However, a recently born democratic movement is shaking the roots of this system (Lawson, 1996). The political debate has reached its highest points in the last few years, especially in the capital town of Nuku’alofa, but its reverberations can be heard throughout the country’s 150 islands and most remote villages. In spite of the high profile political discussion generated by the new democratic movement, the majority of Tongans do not think of democracy as the panacea for solving the problems connected with their status as a non-industrialized country or with any type of corruption they may attribute to their rulers. Tongans feel a sense of "fitness" between their cultural and historical milieu and their monarchy personified by the king (James, 1994, 2002; Lawson, 1996).

The budding democratic movement that characterizes the current political situation in the Kingdom of Tonga, the linguistic choices that Tongan, a Polynesian language, allows and Tongans make in expressing spatial relationships, and the "radiality" (the identification of objects from and toward a specified center) that structures Tongan mental representations of spatial relationships, make the Tongan linguistic and cultural milieu the ideal environment in which to carry out this project. In particular, this project seeks answers to the following general questions that contribute to several major areas of theory and long-standing concern in cognitive anthropology: 1) How are social relationships represented in the mind? 2) What is the relationship among representations of spatial relationships and the representations of social relationships? 3) Can this relationship be characterized as a "cultural model?"

While the mental representations of spatial relationships have been widely investigated (for a few examples see Miller and Johnson-Laird, 1976; Liben, et al., 1981; Levelt, 1982; Stile-Davis, et al.; 1988; Jackendoff and Landau, 1991; Eilan, McCarthy, and Brewer, 1993; Tversky, 1993; Pederson, 1995; Levinson, 1996, 2003; Olivier and Gapp, 1998; Golledge, 1999; Bennardo, 2002b), the mental representations of social relationships (beyond genealogical kinship) have not attracted the same attention (for a few exceptions see Ballonoff, 1974; Freeman, Romney, and Freeman, 1987; Krackhardt, 1987; Jackendoff, 1992b, 1999; Fiske, 1991; Hirschfeld, 1994; Haslam, 1994). The proposed research intends to contribute to the investigation of linguistic and mental representations of social relationships. In addition, this project relies on a newer, less conventional approach to cognition and a unique digitized database of Tongan geographic, linguistic, and ethnographic data.

This project proceeds from the vantage point of the P.I.’s long familiarity with Tongan and personal involvement with research on the representations of spatial relationships. A nexus between the two sets of representations (social and spatial) will be investigated as a "cultural model." A cultural model is here conceived as a partial homology among mental modules, a similarity in the structural organization of domains of knowledge (see the similar concept of "foundational schema" in Shore, 1996:53). The proposed research, then, also intends to contribute to the revived debate about cultural models in the last decade (D’Andrade and Strauss, 1992; Hutchins, 1995; Keller and Keller, 1996; Kronenfeld, 1996, 2000; Shore, 1996; Strauss and Quinn, 1997; Lehman, 2000). A cultural model is proposed as a major contributor to social action and eventually to political thinking, discourses, and choices.

Finally, innovative methodological tools have been prepared and will be used to gather, to store, and to analyze data to test a number of hypothesis that will illuminate mental intra and inter-modular relationships, and relationships between mental representations and behavior, all central issues in cognitive anthropology.

In my 22 months of residence in Tonga since my first visit in 1991, I became convinced that it was highly unlikely that Tongan politics would take the form of democracy, at least in the popular way it is conceptualized by Westerners. I reached this conclusion while I was looking elsewhere, investigating linguistic and mental representations of space. Why was my data leading me to conclusions about democracy and monarchy in Tonga? Because, I have come to believe, Tongan political ideas, actions, and engagements are inherently intertwined with the way that Tongans represent spatial relationships linguistically and mentally.

This project is part of a larger enterprise that is rooted in my doctoral and postdoctoral research, as well as in my current work. My detailed analysis of Tongan language has concentrated on spatial relationships, and led me into the investigation of an important aspect of Tongan thinking and meaning construction (Bennardo, 1996, 1999, 2000a,b). Spatial prepositions, directionals, and spatial nouns are used to produce linguistic descriptions of spatial relationships that indicate the choice of a specific point of view on the environment, a frame of reference (see Levinson, 1996).

Research conducted in a variety of cross-linguistic and cross-cultural contexts all over the world (see Hill, 1982; Hill, 1997; Levinson, 1996b; Ozanne-Rivierre, 1997; Pederson, 1993, 1995; Pederson and Roelofs, 1995, Senft, 1997; Bennardo, 2002b) shows that some languages and cultures express spatial relationships in habitual modalities. In other words, some speaking communities, culturally defined, exhibit mental and linguistic preferences for certain frames of reference in describing spatial relationships.

A frame of reference (FOR) is a set of coordinates (three intersecting axes: vertical, sagittal, and transversal) used to construct an oriented space within which spatial relationships among objects are identified (see Brewer and Pears, 1993 for a discussion of FOR). There are three major types of FOR: 1) relative; 2) intrinsic; 3) and absolute (see Levinson, 1996, for a typology of FOR, and Bennardo, 1996, for a revision of that typology). A relative FOR is centered on a speaker and it remains centered on the speaker when the speaker moves, for example, when one says, "The ball is in front of me." An intrinsic FOR is centered on an object and it remains centered on the object when the speaker or the object moves, e.g., "The ball is in front of the car." An absolute FOR uses fixed points of reference, e.g. north, south, east, west, as in "The town is south of the river."

The Tongan language differs substantially from Indo-European languages –including English–in the way in which spatial relationships are expressed. Instead of the eighty spatial prepositions used by English (Jackendoff, 1992b:108), for example, Tongan uses only three spatial prepositions. Many spatial descriptions are obtained by combining a spatial preposition and a spatial noun. For example, in English we say John is in the house, whereas in Tongan the location of John is expressed as ‘Oku ‘i loto fale ‘a Sione "Present at inside house John," 'i "at" being one of the three spatial prepositions and loto "inside" being a spatial noun. The prepositional phrase 'i loto fale functions as a verb. In addition, Tongan has five post-verbal directionals that express vertical movement (up and down) and radial movement from or toward a specified center/point (Bennardo, 1999). These latter three (mai, atu, ange) are monolexemic examples of the closed class type (grammatical) whose meanings are expressed by a whole prepositional phrase in English (e.g., mai, "toward a center/point").

When expressing spatial relationships linguistically, Tongans use prevalently the relative frame of reference (front-back and left-right axes centered on the speaker) in small-scale space (small objects very close to the speaker), but prefer the absolute frame of reference (fixed points of reference in the field of the speaker, i.e., east, west, north, south) to refer to large-scale space (any size objects at some distance from the speaker) (Bennardo, 2000b). Tongan speakers are the only other documented case–the other is the Hausa (Hill, 1982)–of frequent users of the translation subtype of the relative frame of reference (an object positioned beyond a tree that is in front of the speaker is considered "in front of " the tree) in both types of space (Bennardo, 2000b).

When representing spatial relationships in small-scale space in long-term memory, Tongans prefer the absolute frame of reference. The specific subtype of the absolute frame of reference that they use I have called "radial" (Bennardo, 1996, 2002a). A fixed point of reference in the field of the speaker is selected and objects are represented as from or toward that point. This preference for radiality does not exclude other forms of representations for spatial relationships.

After analyzing patterns of culturally salient exchanges such as the fakaafe "invitation," among others, it became clear that social events are organized in a radial manner as well (see Bennardo, 1996:chapter 7). Furthermore, exchange patterns, besides being realized radially in a way congruent with the preferred mental organization of spatial relationships, also provide the environment in which that spatial preference can be acquired (Bennardo, 2003; see also Morton, 1996; Nisbett, 2003:37 and 201).

My research also suggests that radiality (with the specific meaning I have indicated above) exists in the organization of three other Tongan knowledge domains including navigation, religion, and kinship (Bennardo, 1998, 2001, 2000c). The first two, traditional navigation and religion, are rooted in Tongan and Polynesian history and, although not practiced or believed any more, provide important evidence for structural connections among knowledge systems.

First, Polynesian navigation has been described at length (Buck, 1938; Golson, 1963; Lewis, 1964, 1972, 1974; Sharp, 1964a and b; Hilder, 1965; Gladwin, 1970; Finney, 1976; Kyselka, 1987; Feinberg, 1988; Turnbull, 1994). Hutchins (1995), after a review of various interpretations, states:

In other words, fixed points of reference (star points) chosen in the field of the navigator are later used to compute movement to and from them. I used Hutchins’ conclusions to argue for a further presence of radiality in a knowledge domain that is one of the most salient in the recent past–the last Tongan navigator died around thirty years ago–and the distant past of Polynesians (Bennardo, 1998).

Second, at the core of traditional Polynesian and Tongan religion is the concept of mana "power" (Handy, 1927; Gifford, 1929; Williamson, 1933). Described either as substance or process (Keesing, 1984; Valeri, 1985), as cause or effect (Hogbin, 1936; Firth, 1940), mana always implies coming into contact with supernatural forces by means of another human being– usually a chief–who acts as mediator. The supernatural ‘power’ radiates out of this person and brings good to individuals, to the land, and to crops if a number of procedures are followed, otherwise misfortune results. Thus, the practice of "binding" and its relationship with the concept of tapu ‘taboo’ can be understood (see Gifford, 1929; but also Shore, 1989). Again, we find radiality as an essential and pervasive concept in the religious domain.

Finally, kinship terminology has been extensively used to access mental organization of knowledge (see Goodenough, 1956; Lounsbury, 1956, 1969; Romney and D’Andrade, 1964; Lehman and Witz, 1974, 1979; Lehman, 1993; Kronenfeld, 1996; but see also Decktor Korn, 1974, 1978). I analyzed the Tongan terminology (a classificatory system) by entering it into an "expert system" called KAES, a computer application developed by Read and Behrens (Read and Beherens, 1990; Fischer, 1994; Read, 1997). This expert system analyzes relationships between kinship terms and extracts the basic logic behind them. It then displays the findings graphically. Accessing visually the spatial structure of the kinship knowledge domain provides immediate and uncontroversial verification of the hypothesis that radiality structures Tongan kinship terminology. In the summer of 2000, supported by a Faculty Research Grant and by a Dean Research Grant at University/College of Charleston, South Carolina, I worked with Dwight Read at UCLA on entering the Tongan data in the computer and initiating the analysis.

The complete analysis of the Tongan kinship terminology is being prepared for publication with Dwight Read. Relevantly, though, preliminary results indicate that the algebraic focus of the system is not ego. The whole system is rooted on the term tokoua ‘same sex sibling,’ and it is only from there that the totality of the system is accessible (Bennardo, 2000c, 2001). Basically, the Tongan kinship terminology is premised on a principle of radiality common to social exchanges, navigation, and religion. A center is chosen in the field (genealogical space) of the individual, this time the tokoua "same sex sibling," and other positions (kins) are described as from or toward that center. Also in kinship, then, I found important indications for a fundamental radial organization of a knowledge domain.

It was at this juncture that I decided to posit radiality as a Tongan cultural model, and to look for such an organizational principle in the representations of social relationships to investigate the consequences of this possible cultural model for political actions. This decision was also influenced by two other bodies of literature: one about a variety of proposals suggesting radiality in many aspects of Eastern (e.g., Nisbett, 2003), South-East Asian (e.g., Kuipers, 1998), Micronesian (e.g., Ross, 1973), and other Polynesian societies (e.g., Shore, 1996; Herdrich and Lehman, 2002); and one containing current ideas about the content of a "cultural" component-module of the mind (e.g., Jackendoff, 1992a, 1994; Pinker, 1997; Talmy, 2000b) that is orchestrated around the mental representations of social relationships (i.e., kinship, group membership, dominance).

This proposal follows from research that suggests a modular organization of the mind (Chomsky, 1972, 1986, 1995; Fodor, 1983; Hirschfeld and Gelman, 1994; Sperber, 1996; Jackendoff, 1997; Talmy, 2000a and b; Smith, 2003). Different human faculties call for different parts of the human mind to be specialized in their internal organization and generative capacity. Formal aspects of language are considered a module of the mind and this module is also internally modularly organized itself (Chomsky, 1972, 1986, 1995). Jackendoff (1997:44) proposed a modularity of the mind based on a linguistic and structural homology, that is, in all the modules, the representation of knowledge is propositional like in the linguistic module. In his proposal a conceptual structures module feeds into and is fed directly by a set of substantive propositions. This conceptual module is also at the center of the interaction (input/output) with other modules such as auditory information, emotion, smell, action, and spatial representation.

The spatial representation module is also centered in relation to other modules such as the action module, the visual representation module, and the haptic representation module. An important aspect of Jackendoff’s representational modularity proposal is this positing of a separate module for spatial representation. He says, "... certain types of visual/spatial information (such as details of shape) cannot be represented propositionally/linguistically. Consequently visual/spatial representation must be encoded in one or more modules distinct from conceptual structures..." (Jackendoff, 1997:43).

I build on Jackendoff’s proposal and argue for the existence of a relationship–a partial homology in Tongans’ preferential way of organizing knowledge–between the spatial representation module, the conceptual structures module, and the action module (see Talmy, 2000a, b, for a similar proposal). It is this overarching relationship among these cognitive modules that I call the Tongan ‘radiality’ cultural model. Contrary to canonical modularity proposals (see Fodor, 1983; and also Hirschfeld and Gelman,1994), then, I suggest that inter-modular communication (spatial, linguistic, and social cognition) is conducive to some minimal structural similarity in knowledge organization and maybe processing (see Gopnik, 2001). I also suggest that a specific preference in the organization of knowledge in the spatial relationship module (e.g., radiality) points to a similar preference in the organizations of knowledge domains in the conceptual structures module (e.g., social relationships), and the action module (e.g., political choices).

Tongans position themselves socially in a distinctive way. In everyday conversations when trying to define their position in the social hierarchy, Tongans often make initial reference to a high status person as a fixed point of reference. They then trace their personal position from that person/point. Similarly, in a fono "official meeting," an individual's status is indicated and determined by the "distance"–calculated in units represented by intervening individuals–from the highest status person present, for example, the local village chief, a noble, or the king (Bott, 1972; Marcus, 1980). This is true at the village, island, and national levels.

The flow of power seems to be conceptualized from a "higher" focus point down to the "lower" position of the individual. In contrast, in democracy–at least in the popular notions of the term I am familiar with–the flow of power is conceptualized as going from the individual to the "higher" elected representative. Moreover, the Tongan vertical hierarchy is transposed on the horizontal plane and rendered as physical "distance" from a chosen focus person as in the seating arrangements in the fono (see Gifford, 1929; Bott, 1972; Marcus, 1980). This conceptualization of social hierarchy and social relationships I term "radial." Thinking radially to locate objects in space implies looking for a fixed point of reference and describing the object to be identified as positioned from/toward that point. It must be noticed that the specific way in which Tongans position themselves socially and the official arrangement of people in the fono represent a sub-case of radiality as instantiated in a single vector, away from one point or toward it.

Radiality as a cultural model is used to think and reason with (D'Andrade, 1989; Johnson-Laird and Byrne, 1991; Lehman, 2000) and consequently I propose that this way of thinking contributes to the generation of social action, linguistic behavior, and political choices. This cultural model is not suggested as the only generator of (linguistic and socio-political) action, but it is simply hypothesized to be used in thinking and eventually affect behavior. Individual variability in the instantiation of the model is expected, both at the processing level (thinking) and at the realization level (behavior). In other words, thinking with a cultural model or the presence of a cultural model does not eliminate individual variability but it illustrates one source that defines the variability itself (see Keesing, 1987:377; Kronenfeld, 1996:17-20, and 2002; Shore, 1996; Strauss and Quinn, 1997:83; Lehman, 2000, for discussion).

Approach to Cognition. The proposed research requires not only the adoption of a different way of looking at cognitive architecture but also a different way of looking at cognitive processing. Both symbolic and connectionist approaches fall short of explaining the complexity of inter-modular interactions, the former because of the unnecessary reduplications of computations necessary to explain them, and the latter because of the limited applications investigated so far that account for larger mental functioning (for discussion see Strauss and Quinn, 1997).

In contrast to these more conventional approaches to cognition, I adopt an approach that has been called "radically intentional" by Keller and Lehman (1991:272, note 1; see also Lehman, 1985; Keller and Lehman, 1993). This approach considers knowledge domains as theories, and concepts–units of knowledge–as generated within these theories (for similar positions see Murphy and Medin, 1985; Medin, 1989; Gelman, Coley, and Gottfried, 1994; Gopnik and Wellman, 1994; Keller and Keller, 1996; Gopnik, 2001). The internal computations of these theories are not a number "of binary features in a matrix whose dimensions are nothing but such features" (Keller and Lehman, 1991:288), but a number of relations, including cause-and-effect, that are possible given the presuppositions of the theory. In other words, theories are computational devices; that is, given a set of assumptions, a number of theorems can be obtained (generated concepts can be considered theorems). Theories are also recursive computational devices. Once theorems have been obtained, they may function as principles for other theories. Considering knowledge domains as theories and concepts as theorems (and due to recursiveness also mini-theories) explains how they can come together to become cultural models. This is possible because they share this basic intra- and inter-structure or nature.

Within this theoretical framework, culture is a favored cognitive activity, computational in nature, based on a universal set of axioms to arrive at theorems that in their turn become axioms for specific domains. These theorems/axioms are at the same time individual and collective due to their recurrent and preferred state in a community. It is in this way that cultural models are part of a culture and at the same time they are individually pursued (for a similar reasoning, see Shore, 1996; Strauss and Quinn, 1997; Lehman, 2000). The privileged position eventually attributed to the spatial relationship cognitive module is justified by the fact that solving computations regarding space is not only universal, but it also takes place very early in the cognitive development of an individual thereby offering a basic paradigm for subsequent elaboration in other knowledge domains (Piaget and Inhelder, 1956; Clark, 1970; Moore and Timothy; 1973; Lakoff and Johnson, 1980; Liben, Patterson, and Newcombe, 1981; Stiles-Davis, Kritchevsky, and Bellugi, 1983; Cohen, 1985; Lakoff, 1987; Pick, 1993; Fauconnier, 1994; Bowerman and Levinson, 2001; but see also Strauss and Quinn, 1997; Carey, 2001).

Research Hypotheses. Others have called the presence of an overarching similarity in the organization of mental representations of knowledge "scripts," "schemas," "mental models," or–in anthropology–"cultural models" (Schank and Abelson, 1977; Casson, 1983; Johnson-Laird, 1983; Brewer and Nakamura, 1984; D ’Andrade, 1985; Holland and Quinn, 1987; D’Andrade and Strauss, 1992; Shore, 1996; Strauss and Quinn, 1997; Kronenfeld, 2000; Lehman, 2000). This project will test the hypothesis that radiality is an important feature of the mental representation of three knowledge domains for Tongans and, as such, can be termed a Tongan cultural model.

The assumption of radiality as a Tongan cultural model leads to the following related sub-hypotheses about the Tongan cultural realm that this project will test.

These three sub-hypotheses are conducive to a variety of other hypotheses that justify each of the various gathering activities included in the methodology proposed to collect data (see methodology section below). Finally, if all the above sub-hypotheses are supported, one may also conclude that preferences in mental representations of spatial relationships can be indicative of a similar organization in other knowledge domains. In other words, since space is one of the universal ontological categories (Piaget and Inhelder, 1956; Miller and Johnson-Laird, 1976; Jackendoff, 1983), the way in which it is structured or preferentially represented in the mind may be a reliable indicator of a cultural model.

The data that will be collected to test sub-hypothesis #1 are about social networks. The social networks data consist of information collected by means of questionnaires (for an example see Burkett, 1998), interviews (for an example see Wellman and Wortley, 1990), and/or structured observations (for an example see Bernard, Killworth, and Sailer, 1980, 1982; Freeman and Romney, 1987) about villagers' perceived and actual frequency of interactions with relatives and/or co-villagers. Such data illustrate the nature, structure, and composition of these imagined and actual social interactions in public arenas (Freeman, White, and Romney, 1989; Scott, 1992; Wasserman and Faust, 1994).

The analyses of social networks data (e.g. estimating cliques, centrality, and density measures) highlight exactly these internal organization/s or structure/s (McCarty, 2002). Radial organizations (star graphs) or vectorial subtypes (line graphs)–always centered on an individual different from the one providing the information–can be detected and are expected to be found (for different, but related works on networks see Hage and Harary, 1991, 1996). The finding of circle graphs, graphs with low and uniform measures of centrality for all members, would undermine my hypothesis. However, any network structure found will contribute to the overall project of testing the social environment represented by the networks against cognitive, linguistic, kinship, and geographic location data about these same networks (Krackhardt, 1987).

In the same way as maps of an environment drawn by subjects are compared to the geographic reality of that environment (see Gould and White, 1974; Downs and Stea, 1977; Tversky, 1993, 1996; Golledge, 1999; Bennardo, 2002a), the obtained social networks will represent the (social) environment against which comparisons can be made. The content (linguistic form and linguistic meaning) of the interviews about social relationships and political actions, as well as the language produced during the transcriptions of those interviews (see Duranti, 1997:107, for the value of such data), will be compared to the results of the analyses of the complete social network survey and observational data. Similarly, the results of the experimental tasks (cognitive network data) described in the Methodology section below will be compared to the surveyed network data. Partial homologies (e.g., radial and vectorial organization with other than speaker as center or apex) between these types of data will be considered useful indications of specific mental representations of social networks. Only a few centers/apexes may exist embedded in individuals around which and out of which social and political relationships are organized and represented mentally.

All the social networks data gathered will be entered in the "Digitized Tonga" database (see Methodology section below). The availability of the networks data in the database makes the content of the interviews much more reliable for speculations about cognitive representations (see Romney and Weller, 1984; Lehman, 2000). In fact, interviewees immediately confront the consequences of their statements about social networks as these compare to those in the database, and deeper understanding or heightened consciousness of one's mental constructions can take place among both interviewer and interviewee. If such deeper awareness does not occur, repair strategies in dealing with discrepancies between informants' statements and the network "reality" presented by the computer will still provide valuable insights.

The way in which social network analysis is used in this project to assess mental representations is not new (see Romney and Weller, 1984). What is new is the embedding of such an analysis within a rich ethnographic investigation. It has been known for a long time in anthropology (see Gardner, 1976; Bernard, et al., 1984; Boster, 1985; Lehman, 2000) that information provided by informants needs control for individual manipulations, lack of knowledge, and personal gain (among others). While sampling and a variety of statistical analyses attempt to keep these possible distortions in the data under control, the availability of the social networks data in a digitized form during the interviews provides an important and necessary control for the appropriate acquisition and interpretation of the linguistic data.

Social network analysis is a well established field in the social sciences and requires expertise, training, and experience that go beyond my personal level of expertise with the subject. I have already contacted Dr. Charles Cappell, Department of Sociology, NIU, whose expertise and experience in social network analysis particularly fits the present investigation. We will work together on the preparation of the research tools needed–questionnaires, sampling procedures, and data storage. He will also suggest and conduct the most appropriate statistical analyses (e.g., cliques, centrality, density) that are necessary for the type of questions we expect these data to address. His collaboration is essential for the completion of the investigation.

The data that will be collected to test sub-hypothesis #2 include electoral counts and answers to questionnaires about political choices by co-villagers during local and national elections (see Romney and Weller, 1984; Freeman and Romney, 1987, on ways to enhance the reliability of informants). This last type of information will be processed statistically and will produce a sufficiently accurate description of villagers’ political choices while making unnecessary the asking of questions about interviewees' personal choices. Moreover, data obtained during interviews with each villager (see below) can be compared with the results obtained by the questionnaire. Thus insights about mental representations of political action can be achieved if discrepancies or similarities result between "actual" behavior (as perceived by co-villagers) and expressed behavior (as presented in the interview).

I expect the local figures that constitute the apexes or centers of local social networks to be the targets of the villagers' preferential choices. I also expect traditional kinship ties to be weaker than typically suggested for Tongans (Gifford, 1929; Gailey, 1987; Small, 1997). Thus, new social parameters such as religious affiliation and/or economic status (participating in the constitution of social networks) may be shown to have a more prominent role than they used to have in the social fabric of Tongan life.

The data that will be collected to test sub-hypothesis #3 consist of transcripts of structured and unstructured interviews with elected officials and village residents, public and private speeches, and everyday conversations. Samples of other texts already available, both my previous transcriptions and other texts, e.g., newspaper and magazine articles, will also be used.

Linguistic data have typically been assigned a privileged place when inquiring into the mind (Chomsky, 1972; Miller and Johnson-Laird, 1976; Dougherty, 1985; Lakoff, 1987; Pinker, 1997; Olivier and Gapp, 1998; Bowerman and Levinson, 2001). The way in which information or meaning is organized and expressed linguistically is regarded as a reflection of mental organization of knowledge (see for example Talmy, 2000a and b; Strauss and Quinn, 1997). I will conduct interviews in which I will inquire about and discuss social relationships, the current political situation (monarchy vs. democracy), and political choices. The interviews will be conducted away from election time (in June 2004 and 2005) and during election times (in February-March 2004 and 2005), both local (2004, election of village officer) and national (2005, election of representatives to the Parliament). The saliency of the topic during election times may facilitate the conducting of the interviews besides having other potential effects (e.g., inhibitory, enhancing) that need to be checked. All interviews will be videotaped in order to obtain a thorough collection of the linguistic, paralinguistic, and contextual features of these events (see Duranti and Goodwin, 1992; Duranti 1997). I will also videotape (with the help of a graduate assistant or Tongan assistants) language production in a variety of situations such as fono "official meetings," and formal and informal family and communal gatherings, especially during election time. All these linguistic materials will be transcribed in the field with the help of Tongan assistants.

Measures of the extent of linguistic radiality will be correlated with measures representing the vectorial or radial structure of the respondent’s cognitive and surveyed networks. Briefly, I expect respondents with cognitive networks that reflect high levels of radiality will have higher uses of directionals and radial discourse patterns (centered around other than speaker and with use of radial metaphors). We will be able to explore how a respondent’s location in the surveyed network possible affect discourse patterns: do peripheral members use directionals more or less than central members?

The types of data described (social networks, political choices, and linguistic) represent multiple entries into the structure of mental representations of social relationships sought for in the investigation. These data, will be the ground in which a partial structural homology, i.e., radiality, between the representations of spatial relationships, the representations of social relationships, and the representations of political action will be sought.

Time Frame and Collaborations: Three types of activities are involved in this project: the acquisition of data during fieldwork in Tonga; the digitization and the analyses of data at the home institution; the publication of the results as articles in professional journals and as a book with a press to be identified. The acquisition of the linguistic, geographical, and social networks data will be conducted over a period of three years, 2004-2006, including six months of fieldwork–6 weeks (February-March) and one summer month in 2004 and 2005, and one month in summer 2006. During the first period of fieldwork (2004), I will collect the first part of the social network data, and conduct some interviews. During the second period of fieldwork (2005), I will mainly collect linguistic data and finish the social network data collection. In the third year (2006), I will conduct the final interviews. Analyses of the data during these three years will be conducted at Northern Illinois University.

Collaboration with Charles Cappell, Department of Sociology, and Kurt Schultz, School of Art, both at Northern Illinois University, is deemed as fundamental for the success of the project. Cappell will collaborate in the preparation of the questionnaires about social networks, in the preparation of these data for statistical analyses, and in the selection of the appropriate statistical application/s and procedures that best suit the goals of the research project.

Cappell is the Director of the Social Science Quantitative Research Laboratory (SOCQRL, a lab funded by the National Science Foundation, DUE 9551910) at Northern Illinois University. This lab contains all the needed software for this project. Several social network programs are available (e.g. UCINET, STRUCTURE, and STOCNET). In addition, the lab supports the analysis of digitally recorded interview and observational data. Three qualitative data analysis programs, HYPERRESEARCH, ATLASti, and NVIVO, can be used to code the raw digital data, create categories, and produce semantic models capturing the structure behind the observations. Dr. Cappell will serve as the consultant on the use of these applications and will train the graduate students working on this project to use this software.

Schultz is responsible for the 3-D rendering and for the programming of the Director application that makes possible the analyses of the linguistic data (digital videotapes) in synchrony with the spatial data (3-D of Houma). His contribution will make possible the innovative analyses to be conducted with the data (digitized interviews) gathered in the field.

Social Network Analyses:

This project will collect three basic types of social network data: 1) observed interactions in public places using digital video which will be coded; 2) complete network survey data where each respondent is asked about their connections to all other villagers over several relational contents, e.g. influence, support; and 3) ego-centered cognitive networks where each respondent is asked to group and draw the relations between members of village.

All of the network data will be analyzed using the latest software incorporating the statistical and mathematical methods of network analysis (UCINET, STRUCTURE, STOCNET). First, the reported and recorded interactions between members of the village will be translated into matrix form: incidence matrices, including both symmetrical relations (friendship, conversation) and asymmetric relations (influence over, seek help from). The underlying structure of the resulting surveyed and observed networks will be recovered using both graph theoretic and statistical methods.

The full network of all members will be drawn as a digraph model. Network properties such as reachability and distance will be computed. Additional ego based measures derived from the network structure will be computed, such as centrality and connectedness (indegree and outdegree) of each actor. The distance matrix representing the network distance between actors will be analyzed using metric multidimensional scaling to produce a graphical image of the dimensions needed to capture the relational distances. I expect the networks to be shaped like star or line graphs and not like circle graphs. Additionally, cognitively perceived networks will be produced by asking respondents to cluster and draw networks involving members of their village. From these diagrams, digraphs will be produced and a corresponding set of network properties generated.

In network analysis, these dimensions captured from the network data have been shown to correlate, or be predicted by, other properties of the actors, such as dimensions captured from other networks. These other non-directional networks will be obtained from residence of subjects (distance of each individual's house from all the other houses), from their positions within the kinship structure (e.g., kainga "extended family"), and within the cultural groups in the village (e.g., unmarried, elder).

It may also be more feasible to reduce the full set of network members into fewer, structurally equivalent nodes, so that digraphs of these reduced clusters can be recovered. The resulting location of each member in each of several types of networks can then be correlated and compared. By treating each network of a different type of relation as an analytically distinct basis of network generation, one can then statistically test to what extent the different networks overlap.

The fundamental evaluation of the major premise of this project will involve the tests regarding the extent to which the "radial" spatially based structure is reproduced in these other relational network representations (with special attention to vectorial instantiations). These tests will be carried out in at least three ways: 1) using the set of distances between actors in the spatially based network to predict their distance in another substantive based network; 2) using the actor’s position in the spatially based cluster or clique to predict the actor’s membership in a cluster based on a different sociometric dimension, e.g. political access; and 3) comparing the respondent’s network characteristic, such as centrality or position in multidimensional scaling solutions, across different substantively based networks.

The data collected, stored, and analyzed with the help of this database will facilitate the acquisition of appropriate insights into the hypothesized cultural model. For example, the houses of the cliques' members can be color coded and their distribution in the village can be displayed. This distribution can be compared with the distribution of kinship groups, cultural groups, and religious affiliation groups (all these groups can also be color coded and displayed). Thus, important conclusions can be reached regarding the possible relationships between these various socio-cultural factors. Is the distribution of the houses in the village related to kinship groups? Is religious affiliation related to the distribution of cliques over the territory? Are the centers/apexes of the social networks identified located in (relative) central places in the village? In other words, the relationship between distribution in space, socio-cultural grouping, and social networks can be explored.

Two Field Sites: Two communities will be involved in this project: the village of Houma and the Tongan capital town of Nuku’alofa. These field sites are chosen as representative of two different Tongan environments: the village of Houma, a small subsistence community (172 residents), on the northern island of Vava’u; the capital town, the only middle-sized town (ca. 20,000 residents) with an intense political and economic life, on the island of Tongatapu, in the homonymous southern Tongatapu archipelago.

All the adults (18 years and older) in the village of Houma will be interviewed and observed. The observation of informal interactions during the working hours, at the end of the day, and on week-ends will occur at fixed intervals of two or three days and for periods of two or more hours. Observations of formal gatherings (e.g., fono ‘village meeting’) will also be conducted and digitally recorded. Both activities require the help of Tongan assistants that I have already secured. I will also closely observe the final stages of the local election of the town officer of Houma and the post-election period during spring 2004 and, similarly, those of the national election of representatives to the Parliament during spring 2005.

In Nuku’alofa, I will conduct interviews with one of the leaders of the democratic movement, Samiuela ‘Akilisi Pohiva (see Hoponoa, 1992; James, 1994), an elected representative to the Tongan Parliament. I have already met several times with him during my previous residence in Tonga, and he has informally agreed to devote some of his time to talk with me about democracy and political networks. Interviews with some of his relatives, friends, and political followers will also be conducted. The questionnaires will be administered with the help of Tongan assistants. The analyses of these data will not have the support of the ethnographic data available for the village of Houma. However, it will be extremely relevant to see if rising to power has any relevant effect on the conceptualization of social relationships, i.e., conceptualizing social relationships by means of a structure centered on ego instead of one centered on other.

Experimental Tasks: I will administer a set of experimental tasks to investigate ego-centered cognitive networks (Freeman, Freeman, and Michaelson, 1988, 1989). One of these tasks involves transferring mental information about social space–kinship, social relationships, and social networks–into graphic forms. I will first present subjects with a set of cards containing the names of all the adults co-villagers (around 90), including the subject. Then, I will ask the subject to sort the cards in significant groups. Finally, I will ask them to draw (create a diagram of) the group to which they have assigned themselves. I expect them to put themselves in a peripheral positions in the drawing and to put in the center one of the few apical or central individuals from the already obtained social networks. The usefulness of this type of task in eliciting mental representations of spatial knowledge is widely recognized (Ittelson, 1973; Gould and White, 1974; Tuan, 1974; Downs and Stea, 1977; Golledge and Stimson, 1997; Tversky, 1981, 1993, 1996; Franklin, Tversky, and Coon, 1992; Taylor and Tversky, 1992; Lloyd, 1997; Golledge, 1999; Bennardo, 2002a). Information about mental representations will be found not only in the products of the activity (drawings), but also in the timing and strategies used in producing the drawings (see Bennardo, 2002a).

Other examples of experimental tasks are a memory task and a free recall task. The first one involves asking individuals to remember the co-participants in an event. I expect their recall to be skewed toward few participants, especially those who occupy central or apical positions in the social networks the subjects belong to. A similar task I administered in 1994 yielded preferences toward powerful individuals over visibility (e.g., people sitting in front of subject), neighborhood (closeness in the event and in residence), and kinship ties (Bennardo, 1995, 1996). However, no control against membership in social networks could be done then. The nature of this task will limit the number of people to whom I will be able to administer.

The other one requires individuals to provide a list of people to whom one is related for a stated reason, e.g., neighborhood, subsistence/business, religion, politics (see Romney, 1989). Given the assumption (to be tested) that people first remembered or recalled are more prominent, these tasks provide salient information about representations of social relationships. I expect people first remembered to be those occupying central positions in the social networks the subjects belong to. Since this task is used as a further check of my hypotheses, I will administer it only to a limited sample of the whole adult population of the village.

Research Assistance and Equipment: Most of the described activities require some fluency in Tongan, fluency that I have acquired through formal studies and during my residence in the Kingdom. They also require the assistance of a graduate student for the electronic recording of the data, both audio, video, and digital and for conducting some of the social network observations. The assistance of native Tongan speakers for the transcription of the linguistic data is also necessary. My own linguistic and cognitive audio/video/digital laboratory at the Anthropology Department at Northern Illinois University already has available most of the necessary equipment: a professional cassette recorder, an audio transcriber, an analog video camera, a portable VCR, an analog dual-cassette VCR with TV monitor, a scanner, two printers (one portable), two desktop computers (Mac G4), and various computer applications to handle the digitized data. The other equipment necessary and for which I am seeking funding is a digital videocamera, a laptop computer, and a LCD projector.

The Department of Anthropology at Northern Illinois University has committed the availability of a graduate research assistant during the final phase of the proposed project, year 2006. However, the amount of data to be processed and analyzed requires the assistance of a graduate assistant throughout the entire period of the research, both at the home institution and in Tonga. The College of Liberal Arts and Sciences will make available one undergraduate research apprentice (on a voluntary basis) during all of the semesters involved.

I anticipate that my investigation will provide the following insights into the organization of Tongan social and political knowledge domains: The mental representations of social and political networks will be organized radially (not ego-centered), and political action will be heavily influenced by the radial organization of these representations. Thus, when considering any political act, the radiality of these mental representations will have a great impact on Tongans decision making process, both at the planning stage and at the implementation stage. For example, kinship ties may be overlooked when "powerful" centers acquire prominence by virtue of their capacity to attract and structure a large network of relationships (see Lansing, 1991). Positions with such power may be rooted in history and tradition, as a local chief or a noble is (see Marcus, 1980). Other such positions maybe the result of new opportunities created by the process of modernization of the country and the introduction of a market economy (for example, a rich entrepreneur of commoner rank; see Gailey, 1987).

The way in which Tongans linguistically describe social and political relationships will point to a radial mental representation of those relationships as indicated by the extensive linguistic use of lexemes expressing radiality, i.e., directionals (Bennardo, 1999). Important questions include, for example, how much spatial language is used in these descriptions? How much use of the Tongan directionals expressing radiality is made? How does this frequency of use of spatial language compare to other types of linguistic texts? Is there any specific linguistic structure that is frequently occurring in these descriptions? Are there any culture-specific discourse strategies? Are these linguistic forms expressed in common metaphors? The answers to these questions guide the search for the organization/s of the conceptual content/s expressed linguistically.

The distribution of the subsistence plots will be organized radially around the village and in direct relationships to the distribution of the houses in the village (see Shore, 1996:271, for a Samoan case). Traditionally, the spatial organization of Tongan plots followed a similar pattern (Maude, 1971; James, 1995). This radiality is based on more than one center out of which only a few vectors radiate. I expect the mental representations of social relationships to be organized radially in a similar way. Only a few centers exist embedded in individuals around which and out of which social and political relationships are organized and represented mentally. Kinship relationships may intertwine with these networks and occasionally overlap with them.

Social networks will be structured radially (including vectorial organizations, i.e., line graphs), with central members always someone other than the individual providing the information. Each individual has a radial representation of these social networks with prominent members assigned to central roles (Wasserman and Faust, 1994:172). The centrality of these members maybe be determined by three factors: high status in kin group, high social status, and high economic status. These three variables may contribute individually to the generation of a center or they may overlap.

The congruence or dijunctures among these types of data– linguistic, geographical, social networks, and political choices–will provide ample material to test my original hypothesis: the existence of radiality as a Tongan cultural model and, as a corollary, that preferences present in the spatial representations module are indicative of a similar organization of the content of knowledge domains in the conceptual structure module and in the action module.

New interesting questions may emerge, if less pervasive congruence is found than that hypothesized. For example, does the spatial relationships module feed only a limited number of other modules? Maybe only one? Which one and why? Why are certain modules or domains of knowledge within a module, i.e., kinship as a domain of the conceptual structures module, partially homologous in their structural organization with the spatial representation module and not others? What are the other domains of knowledge that are (partially) homologous? Is the way in which social relationships are represented mentally related or not related to political action/s? Is the radial model more evident for some Tongans than for others? Why? What kind of knowledge then guides or generates political action/s? How reliable is linguistic information in addressing and assessing issues of mental representation and architecture? The nature of the data collected and the results of the extensive analyses planned will eventually provide the appropriate context for finding answers to these important questions about human cognition and cultural variability.

The expected findings are relevant for three different areas of investigation.

In the investigation of the human mind, they help in assessing questions about cognitive architecture. What kind of cognitive architecture is supported by the cross-modular homologies I call a radiality cultural model? Are these partial homologies the result of a similar cognitive computational engine in each module or dictated by the structure of the information that is processed? In other words, is radiality a feature common to the various types of information processed, e.g., social relationships and geographical distribution of plots and houses, or is it a result of a common way of processing and storing this information? And how is this commonality achieved within the cognitive architecture? Cross-modular partial homologies like the expected ones in this investigation can make the role of the spatial relationship module within the cognitive architecture become more prominent than suggested so far (Miller and Johnson-Laird, 1976; Jackendoff, 1997; Talmy, 2000a and b).

The expected results contributes substantially to current research about the modular organization of mental architecture. In fact, while it may support and confirm specific internal organizations for each cognitive module, it also contributes evidence for how the modules are necessarily related. Furthermore, these findings would go beyond Jackendoff’s (1997) suggestion of a spatial representation module as separate from the conceptual structures module. The internal organization of the spatial representation module may uniquely provide an insight into the organization of other modules such as the conceptual structures module, the linguistic module (see Talmy, 2000a and b for a similar position), and the action module.

An innovative methodology is being used that allows to analyze a number of digitized data (geographic, linguistic, and ethnographic) in a comparative fashion. The simultaneous availability of these diverse digital data (ethnographic information, linguistic digital video recording and transcripts, social networks graphs, 2-D GIS Imaging, and 3-D imaging), both in the field and during the analyses, is new and experimental and may allow to obtain insights that could not be achieved before.

In the investigation of political action, they provide a key in the interpretation of a political stage where an historically hierarchical and very stratified society is coping with the concept and possible implementation of democracy. Democracy is practiced and lived by those same people that brought to the islands the Christian religions that are so much a part of the contemporary Tongan landscape, the new market economy that is slowly uprooting the inherited social order, and so many new commodities now widely present but not yet widely accessible to many Tongans. This new political system is intertwined with a specific way of conceptualizing power and social relationships–individuals are equally empowered before willingly appointing/electing a representative to act on their behalf. How are Tongans coping with this unexpected import? Changing habitual thinking patterns is not as effortless as exchanging a long knife for a mechanical saw or a lawn mower. The expected findings of the proposed investigation can function as a search light in the intricate web of issues and connections Tongans are facing in their minds and their daily lives.

Regarding the Tongan cultural milieu, they should provide insights into the thinking patterns and political behavior of a population caught between isolation and the world as a global village, between tradition and innovation, and between monarchy and democracy. The Tongan scholar Futa Helu states that "[T]ongan culture never had any instincts for human rights and that the one value, namely equality, the most important of moral goods, did not occur in traditional Tongan culture" (Helu, 1999:35). His insights into Tongan culture are highly regarded throughout the Pacific. Intuitive insights based on personal experiences are of great value; however, systematic collection and analyses of linguistic, ethnographic, and experimental data do contribute another way of understanding. I believe that the findings of this investigation provide a solid background against which speculations about aspects of Tongan thinking and culture can be tested and explained in ways that are scarcely transparent to indigenous cultural members as well as "objective" outsiders.

Gaining insights into preferential patterns of Tongan thinking and political actions opens the way to a better understanding of Polynesians in general. Despite their varied and specific distinctiveness, Polynesian cultures (e.g., Tongan, Samoan, Tahitian, Marquesan, Maori, Hawaiian) share fundamental similarities. For example, radiality, or "point field" as the authors call it, has been documented as a thinking pattern in American Samoa that leads to controversial land disputes (Herdrich and Lehman, 2002; but see also Ross, 1973:118; Shore, 1996:270). Moreover, diasporic Polynesians are starting to crowd several neighborhoods in a number of U.S. towns and cities (especially in the U.S. West Coast). Several studies have begun to document their daily struggles in dealing with the American cultural milieu (e.g., for Tongans see Small, 1997). The proposed investigation contributes greatly to a better understanding and explanation of habitual thinking patterns that generate daily and political actions. These patterns may often lead to possible conflicts among Polynesians and/or between Polynesians and other ethnic groups. Sometimes these conflicts may be avoided by knowing the source that generates the behavior–specific mental representations of knowledge.

The way in which a traditionally hierarchical society like the Kingdom of Tonga is coping with newly introduced democratic ideas has implications for similarly situated cases around the world. We are all aware of the number of issues generated in societies struggling to handle changes from a "traditional" hierarchy –either local or imposed by colonial powers–to new forms of government including democracy. Despite the large and extensive literature available, rarely has research been conducted with the breadth of data and through the methodology that has been proposed for this investigation.

The world is witnessing a large number of peoples facing very similar issues–peoples negotiating historical constraints and present problems, tradition and modernization, self-preserving political choices and all-changing ones, authority and justice, individual gains and social losses, common good and self restraint. The findings of the proposed project will be significant for many other situations in which these issues are debated. They will provide another way to understand, another key to interpret, another approach to predict political actions as rooted in one’s mental representations of knowledge. Beyond the theoretical and empirical cognitive, linguistic, and ethnographic contributions, then, the proposed research will also create a new, original, and productive "space" to be filled by other researchers interested in the dialogue between tradition and modernity.

Back to A Tongan Cultural Model