Ethnography: A neglected method of inductive linguistics

Finding a productive scientific perspective on the relation between grammar and culture can be difficult, as is shown by the decades old debate on the so called “Sapir-Whorf hypothesis”. It is proposed that linguistic research should make more systematic use of ethnography as a source of meanings entrenched in grammar. The cases presented illustrate how grammatical constructions are defined by cultural schemas, scenarios, and models; how non-linguistic (colinguistic) data resolve old problems in grammatical analysis. A potentially controversial claim is put forward that data on an archaic way of life provide clues to the nature of proto grammar. The cases examined include (1) a study of the global relations between language and culture implied by code-switching, (2) a fine-grained analysis of a semantic network that defines a compound affix, (3) the use of native language definitions to define a cultural model of sequences of thinking, feeling, and acting, (4) the use of previously published ethnographies to discover correspondences of cultural models to classes in classifier systems, and (5) the culture of a species of archaic human known to us only through archaeology and paleontology and the possible influences of this culture on the structure of proto grammar.


Cultural linguistics and ethnolinguistics
The purpose of this essay is to call the attention of those interested in cognitive and cultural linguistics to the need for ethnographic research in explanations of grammar.But before making an argument for more use of ethnography and more systematic use, it will be necessary to briefly introduce the theory of cultural linguistics, the roots of which we trace most directly to the thinking of Franz Boas and Benjamin Whorf, who both studied native American languages and developed approaches to linguistic relativity in the early to mid 20 th century (Palmer 1996).There are too many other important influences to discuss here.While similarities and differences with respect to the Lublin school are mentioned, a systematic comparison and critique is not on the agenda.On setting out to write this chapter, we had in mind an introduction that would sort out the subject matter of various linguistic disciplines, specifically linguistic anthropology, anthropological, cognitive, and cultural linguistics, and the two "ethnos" -ethnosemantics and ethnolinguistics.We would have used the notion of profile as it is similarly defined in cognitive and ethnolinguistics as that which is foregrounded or selected from a complex image or base concept, but preliminary reading led to the discovery that Jerzy Bartmiński (2009) had already written that essay, or at least one to which we could contribute little additional insight.Regarding "cognitive ethnolinguistics," he concluded that the term "highlights above all the connection of language with the community of speakers who use it, and secondarily (by implication) with the culture of that community" (p.8).Furthermore, he wrote of the enterprise as involving the " 'subjective reconstruction' of folk understanding." Further reading around the topic suggested that ethnolinguistics has much in common with cultural linguistics, except that cultural linguistics takes research into national value systems and linguistic worldviews to be possibilities afforded by a more general theory of language and culture rather than as central goals.The Lublin school of ethnolinguistics appears to have a humanistic preoccupation with discovering the values and presuppositions implied by usages of value-laden words and phrases in common use by communities of speakers.Cultural linguistics seems from my perspective to take a more scientific and objective interest in discovering how patterns of grammatical constructions are governed by culturally defined and value-laden imagery.But the distinction between subjective and objective is not always easy to make in the science of language, and the cultural linguistic emphasis on scenarios as important culturally defined images is much like Anna Wierzbicka's focus on scripts, except that cultural linguistics does not find it essential that scenarios be described by a semantic metalanguage consisting of a small inventory of universal terms.However,Wierzbicka (2014) has put forth cogent reasons for the practice, such as the advantage of making definitions understandable to native speakers as well as researchers.
The analytical tool kit of cultural linguistics gives prominence to the scenario, which is a type of linguistic meaning, a kind of concept 1 , so a definition is in order.The term evokes not only events, but also actors, goals, plots, and settings.Fillmore (1975: 124) wrote of "standard scenarios defined by the culture, institutional structures, enactive experiences, body image."In Fillmore's theory, scenarios are a kind of scene.Also important in Fillmore's theory is the frame, whichrefers to a set of words, phrases or other constructions related to one another by virtue of being connected to a particular part or perspective within a scene.Thus, a frame is essentially a symbolic system organized by conceptual metonymy.Consider Fillmore's famous example, "Get right back down from out of up in that tree"2 , which combines a discourse scenario of command (here including a tinge of admonishment) with a movement scenario based on spatial schemas of verticality (down, up), interior-exterior (out of, in) and nearnessremoteness (right back, from).It draws on lexemes and constructions from four frames, if you regard the spatial schemas as independent, or two frames if you regard them as belonging to a single model.Quinn (1985) spoke of the "scenario word" commitment, which is used by Americans in discussions of marriage.Evidently following Fillmore, Quinn regarded commitment as a social schema that serves as the base concept for a (metonymical) frame vocabulary (Palmer 1996: 125-126).

Ethnography is the forgotten component
When one first encounters cognitive linguistics, one is exposed to the seductive concept of embodiment as the basis for conceptual metaphors of general use in discourse (Lakoff and Johnson 1980).Certainly English up and down as descriptions of mental states arise from such physical experiences as being sick and down in bed with the flu or healthy and up and about.Being a little slow, it took me some time to realize that not all 1 I regard schemas, scenarios, and models as types of concepts.All of them are more or less schematic, so when speaking generally, scenarios and models may be referred to as schemas, unless they are very concrete and fully specified, but schemas are not necessarily scenarios or models.They may be physical schemas, such as land/sea, hot/cold, or tall/short.Scenarios are action schemas, generally involving animate actors.Models are complex categories composed of more basic schemas, scenarios, and lower-level models.The categories themselves may blend into one another leading to usages that may appear to be inconsistent.Fuzzy analytic categories seem inevitable in a connectionist view of language, which itself models a very complex world.
schemas and models of possible interest to linguists are based on embodied physical experiences, and indeed that all experience is inescapably both (a) embodied and (b) defined and channeled by one's culture.To give a mundane example, whether your traditional dwelling is constructed of skins stretched over poles, of mud poked onto a stick frame with a roof of straw or corrugated iron, of lumber and plywood with a roof of asphalt and fiberglass shingles, or dug into the ground will affect your physical experience of heat, wind, rain, and light.Then it occurred to me that perhaps for many purposes, culture defined as shared knowledge and traditions was the more important source of the schemas and cognitive models that we find in language, not forgetting that language itself is a large part of our shared knowledge and traditions.The pervasive influence of both physical experience and culture on grammar has been recognized by Langacker (2014) and other cognitive linguists.Physical sensations on the one hand and culture on the other are mediated by psychology, specifically by the processes of construal that the cognitive linguists have described so effectively.Among dimensions of construal are bounded versus unbounded things and processes, figure and ground, schematicity versus specificity, profile and base, perspective taking, intersubjectivity, focus chains, discourse grounding, conceptual metaphors and metonymies, and blending.To these, Hill (2005: 157) might add, as would I, "universal human narrative competence." Yet something is missing from the program, and I think it is ethnography, that is, descriptions of culture obtained by a variety of methods, both linguistic and non-linguistic.Descriptions of rituals, subsistence practices, politics and warfare, child raising, and much more may reveal schemas, scenarios, and cultural models that surface as important linguistic categories.We are talking about what Bartminski (2009: 35) referred to as ad-linguistic data, which is "the socially entrenched, belief-based knowledge of the world, common to the speaker (sender) and the hearer (receiver)."3The need to apply ethnography to a theory of language was realized by the famous ethnologist Bronislaw Malinowski (Łozowski 2013: 363-364), but working in the early 20 th century, he lacked the tools of cognitive linguistics that help us to see the intersections of grammar and culture.
Since language is a massive part of human culture, we encounter a circularity.We can study the culture in language by studying language itself, or we can study it by observing other kinds of behavior.The circularity enters where we choose to limit our source of information to language.It is hard to make an argument that we can use language to find linguistic meanings that we call "culture" and then apply these cultural notions to explain the language in which we found them.We have to accept that these meanings are both cultural and linguistic.But we can at least say that there are strong correlations between types of constructions and the meanings that we find when we present definitions in the form of concepts, schemas or models.
Let us say that our goal is to understand constructions, by which is meant combinations of symbols, which themselves are linked elements of phonology and meaning.Constructions vary in levels of complexity from compound affixes on the low end to narratives and conversations on the high end, and they take many curious forms, such as infixes, circumfixes, and combinations of tonal patterns with reduplication, not to mention the myriad genres of narrative and conversation.
If we assume that some sort of cultural model provides the base concept of a construction, and that this model happens to be complex, observation of a single usage is insufficient to describe the model.One must either hear more talk pertaining to the model or observe other revealing cultural activity.Even a sizeable corpus of talk pertaining to a topic together with related non-linguistically sourced ethnography does not necessarily guarantee that one can define the meaning of a construction.For that, one requires multiple usages of the construction and its components.Then for each component and for the construction, one can abstract one or more central meanings and very likely multiple associations or conceptual metonymies.The end product, the meaning of a construction or a symbol (classifier, deictic, preposition, verbal affix, etc.), is often a network of semantic inclusion, similarity, and metonymy.Similar networks are found in the phonological domain, with the upshot that a symbolic construction articulates semantic and phonological networks.Let us now consider some examples that illustrate these analytical possibilities.

Zooming out to find culture in language
There are various ways to get at the cultural meanings in language.One can record texts, translate them, and then look for recurring themes, such as revenge, separation anxiety, or martyrdom.In viewing the Filipino film Dreaming Filipinos, we were perhaps unjustifiably startled to hear segments making liberal use of English words and phrases that seemed to frame scenarios belonging to models of class in one scene and colonialism in another.Did code-switching depend upon the topics under discussion, each topic representing a cultural model?Example (1) is a text from that study (Palmer 2002).In the scene, a high school graduate, Paul, is sent by his uncle Manny to visit a bored, but wealthy housewife of the kind that in the U.S. might be called by the popular term cougar : (1) Ladies man kaba, Paul?When Paul arrives at Toni's house, she is reading a magazine called Sexy Men.The conversation with Toni reflects a perception of leisure.Toni's language is marked by intersentential code-switching.Her first words to Paul are "You must be the pizza man."Paul answers in English "No, guess again."Toni switches to Tagalog Aha, ikawsi Paul 'Aha, you are Paul'.Paul answers "That's right" and Toni continues in Taglish: Sayang, nagugutom pa namanako.I'm Toni.Ladies man kaba, Paul? 'Too bad, I'm hungry.I'm Toni.Are you a ladies' man, Paul?' Thereafter, Paul retreats to Tagalog, while Toni continues to switch.She says, "Why don't you work for me as an escort?"Paul replies in Tagalog:Anonggagawinko? 'What would I do?' Tony says, You will dance with my friends.They are bored housewives.They pay good money to dance with people like you.Kaya magensayoka ng mabuti.'So you should practice very well.'I can make you a rich boy in no time.
For Toni, the bored housewife, the language of money and opportunity is English.The language of effort, practice, and instruction is Tagalog.Toni tempts Paul to become like Manny, living the easy life with no real purpose.
Other conversations take place with Paul's older friend Mang Sammy, who once worked for the deposed dictator Ferdinand Marcos.Their conversations revolve around Filipino feelings of inferiority vis-à-vis America and Filipino-Americans.The question What's wrong with the Filipino? is asked repeatedly, but never in Tagalog.One mixed-language scene contained the terms American, Philippine History, school, exchange professor, teacher, assignment, mag-research (2 instances), expert, mentality, vindictive, idol (2 instances), practical (2 instances), freedom of expression, papa-surveillance, Sir, marshal law baby,and September.It also contained the acronym B-day, for birthday.In general, their conversations could be said to highlight the impact of American imperialism, which began at the beginning of the 20th century and lasted about half a century.It took the form of cultural and economic imperialism imposed by American air and naval bases, Americandesigned school and medical systems, American support for the dictator Marcos, and American corporations.
Dreaming Filipinos provides an example of a global relation between cultural models and mixed-language frames in which the English consists largely of salient lexemes reflecting the Filipino absorption with American consumerism and subjection to American imperialism.Further ethnographic and ethnohistorical research would clarify the background models and the relationship between frames and models.

A granular search for culture in a collection of lexemes
We take another case and zoom in tightly to view grammatical fine structure in the form of word construction.The data comes from a combination of Tagalog-English dictionaries and computer searches of a corpus consisting of translations of items from various media including novels, newspaper articles, and videos (Palmer, Rader, and Clarito 2004).Example (2) from the "Showbiz" page of Kabayan Online is a fairly typical line with interlinear morphology and translation: 4 (2) JC-Melissa-Camille love triangle, ka-cheap-an lang . . .part-cheap-loc just 'The JC-Melissa-Camille love triangle is a cheap affair . . .' Example (2) contains a circumfix construction: ka--an, which we gloss with the prototype spatial sense of place.By labeling it as a circumfix, we are making an assumption that the form has been grammaticized.The construction appears in many place names.The suffix -an is a locative.The category ka--an and the examples are taken from a more comprehensive study of the prefix ka-, which we define as a partitive, that is, a grammatical morpheme that has the sense that its substantive stem contributes a part of its base model (sometimes the whole) to the meaning of the full construction with ka-.The stem may be nominal, verbal, or adjectival.One could call kaa metonymic affix.It is like the suffix -er in the English word screwdriver (Panther and Thornburg 2001).Constructions with ka--an are very common.A mixed genre corpus of 607 lines of text produced 68 examples, even without counting terms with additional prefixes before ka--an or those in which ka-undergoes metathesis with a verbal prefix, e.g.kina--an, <ni-ka--an.The instances can be sorted into subsets based on common meanings rendered in small caps according to the convention in cognitive linguistics (Table 1).The definitions of the categories in Table 1 could doubtless use refinement.With a larger sample, we would expect to find that the categories display additional subcategories, but they constitute a beginning in our project of using language to discover the culture behind the constructions.We also seek an explanation for the category structure.The notion of place offers itself as the prototype category because it is elemental, possibly even innate, and it occurs with high frequency in the definitions of ka--an terms.In Figure 1, it is given the bold outline of a prototype.Next we may ask why the second notion, abstract quality, belongs to the construction.Terms in this category have glosses such as 'arrogance,' 'goodness,' 'truth,' and 'inner being'.We suggest that they permit conceptual localization, in time or space or both, of an otherwise unbounded segment of the qualities brought to the construction by the roots or stems.A more literal gloss of the circumfix would then be 'localization of quality.' We now have a rudimentary semantic network for ka-an.It includes the schematic category locative, a central, concrete subcategory place, and abstract quality, which is similar to place in that it is also, perhaps metaphorically, localized (Figure 1).This characterization remains descriptive and hypothetical, but clever tests of associations might be devised.For example, is there a general locative grammar that also applies to ka--an constructions covering abstract qualities?
The second category of abstract qualities, experiential states, has a similar derivation to the first as a localized instance of some otherwise diffuse state of being.The fourth category of custom and collective activity resemble qualities in that they submit to localization.The category seems a bit more concrete in that it involves actors.Customs are behavioral rather than experiential.They are dispersed in time or space or both, and so can be sensibly parceled out and localized in a ka--an construction.We theorize that the function of ka--an is to create a conceptually localized instance of phenomena that are perceived as insubstantial, distributed, or diffuse: places, qualities, persistent experiences, customs, or collective activities.Ka--an often nominalizes, but it can also create verbs.
The fifth sense of ka--an, collections: material, tools, assembled item, can be seen as analogous to the others, except that instead of getting a part or essence from an insubstantial phenomenon, it predicates creating an object by assembling (localizing) parts.Or, metonymically, it is the material or the tool used for making something.
We have so far said little about the ka-in ka--an, which covers a complex semantic network independently of -an.The schema and subschemas of ka-are attributed not to the prefix alone, but to constructions with various classes of linguistic roots and stems.Words with the ka-prefix fall into two semantic classes: partial and experientialWithin the partials, individuation is ambiguous between one of a collection and part of a mass.Within the experientials, we find abstract quality, comparison of equality, intense experience, and participant in shared experience.We are also tempted to include recent completive ka-, within the experiential senses, because it emerges on the realis side of the inchoative moment, and so represents an aspect of subjective experience that is both existential and temporal.We can see the potential for meanings such as abstract quality and intense experience to submit to localization in the ka--an construction, and in fact, they obviously do in senses 2 and 3 of Table 1.It is instructive to compare katákot-tákot 'frightful' (<takot 'fear') with katakutan 'state of being afraid, to fear (x).'We begin to see fascinating glimpses into Ka-tagalog-an culture in the notions of comparison of equality and participant in shared experience.
We have proposed that ka-abstracts and profiles some aspect of the base conceptualization of the root -a part cut from the whole, a member of a group, an endpoint in a process, a result of provocation, a look-alike, or a generic feature.Some might see this attention to co-occurrences of ka-with the content of groups of stems and roots as the rediscovery of selection restrictions and feature-based models.From our point of view, the features commonly deployed in the selection restriction rules of generative linguistics are actually unduly abstract representations of cultural schemas.In defleshing cultural models by reducing them to skeletal feature sets, one loses sight of the flexibility, adaptiveness, creativity, and playfulness that we encounter in so much of actual discourse, which plays with a much richer set of possible connections.
The same observation would apply to a rigorous application of the natural semantic metalanguage.The network organization of meaning and the multiple connectedness of even small function words and affixes is well established (Bybee 1985;Sandra and Rice 1995;Palmer 1996Palmer , 2003b;;Rice 2003;Palmer, Rader, and Clarito 2004).Knowing the networked organization of language makes me reluctant to fully embrace definitions based on semantic universals or any claim for the atomic discreteness and semantic independence of any word.It is not that very general categories do not exist.It is just that they are unlikely to be sufficient to define nuanced and subjective usages in a natural language.We suspect that a proposed universal may indeed share an important schema or a few qualities with a word in every other language, but we doubt that its core sense fully defines it in every language.Words in use seem to be always enmeshed in complex networks of meaning that lie on continua of entrenchment and conventionality -embedded in traditions at one pole and situated in novel discourses at the other.To deny connections of any proposed universal to culture-specific or context-situated meanings seems likely to lead to definitions that native speakers would find somehow incomplete because they are stripped of known connections and possibilities for creative discourse.Even embodied meanings are always enmeshed in culture (Langacker 2014).

Building coherent cultural models from linguistic data
The method of collecting instances of a single type of construction from dictionaries and texts has revealed schemas and glimpses of cultural models, but one would like to draw pictures that are more coherent and complete.Suppose, for example, that we set out to understand the Katagalogan model of emotion and causation, as we did for an edited collection Talking about Thinking Across Languages (Palmer, Goddard and Lee 2003).An inspiration for this study was the publication of an English folk model of the mind by D'Andrade (1987,1995).D'Andrade found "a major direction of causation" described as action > perception > thought ( >feeling) > wish > intention > action.The parentheses around feeling show that it is optional in the sequence.
We set out to discover a comparable sequential model in the emotional language of Tagalog video melodramas.Our method of exploring the problem was to watch melodramas with two Tagalog speakers, Marica and George, who were studying at the University of Nevada at Las Vegas.During scenes in the videos when emotions appeared to surface, we transcribed and translated the discourse.From the translated texts we collected terms and constructions that appeared to mention or reveal some kind of thinking or feeling.Then, we asked questions in Tagalog to obtain definitions in Tagalog.The kinds of generic questions that we could ask were suggested by the students.Outfitted with native language questions, we could discover categories inherent in Tagalog definitions and avoid introducing distortions arising from our own linguistic worldview.The value of this approach was demonstrated in the landmark study by Casagrande and Hale (1967) of Papago (O'odham) definitions.The method has been developed for use in ethnographic fieldwork by Spradley (1970).We soon found that we needed to expand our project to include a wider spectrum of cognitive experience, including thinking and feeling.Examples (3)-( 6) illustrate how closely related these categories are in Tagalog talk about emotions.Examples ( 5) and ( 6) show a sequence of two mental events.Example (6) shows how outside events may influence thought, feelings, and physical experience.We found that the definitions could be sorted into a few categories of experience (Table 2).Furthermore, with a listing of native language definitions, we could readily trace a Tagalog model of cognitive experience and action (Figure 2).

Reversing direction: Using ethnography to understand grammar
To this point, we have only attempted to find cultural schemas, scenarios, and models in linguistic data.We did this with mixed-language lexical frames, constructions, and native-language definitions of lexemes.But would it be possible to reverse the process?Suppose we delineate cultural models in non-linguistic ethnographic data, perhaps combined with unsystematic and linguistically naive translations of descriptive native-language texts.Can we then use the models to explain grammatical constructions?Explain here means to show that particular constructions correlate with particular components of cultural models.For example, cultural models might be induced from observations of such activities as marriages, healing practices, bush schools, child raising practices, games, economic trade, food gathering and preparation, or warfare.All of these would provide scenarios and schemas that are salient by virtue of the emotions that they evoke or the amount of time and energy that they require.Then one could attempt to relate schemas and scenarios to some set of constructions, say, to those involving noun classifiers or ergative verb forms.For example, Samoan allows the ergative case to be applied to actors in narratives.Duranti (1994) transcribed the discourse in a series of council meetings in Western Samoa.He found that the ergative was used only when participants were receiving credit or blame, or where "the power of certain individuals or groups to affect others through their actions or to cause or initiate events is at least acknowledged" (Duranti 1994:26).The senior orator made more use of the ergative than did more junior councilmen and the ergative is applied to actions of the Almighty, as in (7) (7) e fa'alava e le Akua mea 'uma ta caus+enough erg at Lord thing all 'The Lord makes all things sufficient.' Without the understanding of Samoan authority roles and rules of discourse, it would be hard to describe this bit of grammar.A similar analysis of voice in the discourse of a Tagalog melodrama appears in Palmer ( 2006).Another instance from Samoan is provided by Shore (1996: 269), who described the social distinctions overlain on the spatial frame of tai 'seaward' and uta 'inland' in the village of Matavai, Safune.To summarize, "Tai is the realm of women, light, clean, and formal, where there is civil life, social control, and good speaking.Uta is the realm of men, dark and dirty, but intimate, where it is uncivilized, village laws are inoperative, and there is bad speaking" (Palmer 2007: 1063).The point is that using orientation language in Safune would require consideration of the social context, much as the speakers in Samoan council meetings take context into account in their usage of the ergative case.There is very likely a set of constructions utilizing orientation frames in talk about social scenarios.
In the Navajo language of the southwestern U.S., there is a system of control marked by the prefixes yi-and bi-, which were for a time simply translated with the pronouns him, her, or it and characterized as subjectobject inversion, as though their function were purely a matter of syntax.But in fact they can not be understood or properly used apart from the culturally specific knowledge of a control hierarchy involving inanimate objects and animate beings such as animals, infants, and adult humans (Witherspoon 1977).An extensive discussion and explanation of why constructions with bi-should be regarded as designating a scenario with a controlling patient is provided in Palmer (1996).
Where grammar that predicates discourse or social interaction is the object of study, as in these instances of ergative case in Western Samoan, voice in Tagalog, and control in Navajo, it is relatively easy to inspect narratives or to record discourse and interaction simultaneously and to correlate types of situations with usages.But there are grammatical phenomena where this is not possible.Noun classifiers provide an example.
Many languages use grammatical devices such as classifier nominals or affixes to assign more concrete nouns to sets.The gender systems of French, German, and Polish provide convenient examples.There is an old debate as to whether the assignments are arbitrary or semantically motivated.Often a few nouns in the set appear to have something in common, perhaps a flat shape, flexibility, or animacy, but the preponderance of examples appear to be unexplainable.One of the first to propose a semantic explanation was Meillet, who in 1923 observed that in Indo-European languages, feminine gender is extended from female living things to other entities construed as animate, such as the earth and trees.Cases like this seem to beg for ethnographic research to discover why some cultures regard the earth and trees as animate while others, such as most contemporary Euro-Americans, do not.For example, one might look into the variety of animate characters in folk narratives and the economic or ritual uses of animate and inanimate objects.Yet many contemporary linguists appear to have given up on semantic explanations and proceed as though assignments of nouns to sets in classifier languages is arbitrary, even while evidence accumulates that such assignments are a matter of how speakers are thinking about things and construing their relationships.
A landmark case study is provided by the four classifier sets of Dyirbal, a language of Queensland, Australia.Following Dixon (1982), Lakoff (1987) proposed that the classifier balan marks a category whose central member is human females.Other members of the noun set are linked to it by the sun myth: In Dyirbal mythology, the sun was a woman.Other members of the class were birds (mythical females) and plants and animals who either appeared in the myth or were seen as somehow similar to fire (they were hot or they had stingers).Fire belongs to the class because it belongs to the same domain of experience as the sun.(Palmer 2006: 33) Lakoff (1987: 95) also noticed category chaining: "Central members are linked to other members, which are linked to other members, and so on."The links that he mentions appear to be metonymies: the sun is linked to sunburn, which is linked to the sting of the hairy mary grub.Others are prototype to variant relations; the sun is the prototype woman.Objections to the theory are discussed in Palmer (2006).
Inspired by Lakoff's insights, which evidently required ethnographic data, we applied the approach to the study of noun classifiers in ChiShona, a Bantu language of Zimbabwe.In Bantu languages, classes vary in number from about a dozen up to 21, depending on how they are arranged.Their main distinguishing feature is a class prefix, which may take male and female forms.The noun prefixes are concordant with sets of pronouns, demonstratives, and verb prefixes.Hence, a class can be said to be defined by whatever meaning may accrue to a small set of concordant operators.
Researchers had been struggling with Bantu classifiers for decades.The famous paleontologist Louis S. B. Leakey, who had been raised speaking Kikuyu, proposed that the classes were ranked by spiritual value, and that in Guthrie class 5/6, "every single word in this class is an object which is used, or has been used until recently, in connection with religion, magic or ritual or some other form of ceremonial" (Leakey 1955: 13).Notice that verification of Leakey's assertion would require substantial ethnographic research and his assertion might well prove to be true of all the classes.Other researchers proposed semantic explanations, but in general, the explanations have not adequately covered a full set of nouns dominated by a single classifier, and researchers have either sought the preponderance of their explanations within grammar or in highly abstract schemas resembling features (Palmer 2006;Palmer and Arin 1999).
Our approach involved reading all the available ethnographies pertaining to Shona culture.In the ethnographic descriptions, we found cultural scenarios that we thought would be salient in Shona thinking.We listed 11 scenarios that could possibly govern the inclusion of nouns in classifier sets (Palmer and Arin 1999).Scenario 10, for example, is very general and subsumes scenarios 1 and 2. Together, the three scenarios account for the preponderance of terms in Guthrie's class 9/10.Scenario 10.There is a scenario of protection in which the central participants are dominating protectors, protected ones, and the victims of domination.Scenario 1.The spirits of ancestral chiefs live in the bodies of lions (mhondoro).Scenario 2. The chiefly ancestral spirits (mhondoro) reign over both the things of the wild and human affairs.They are the protectors of the land and the wild animals.
Class assignments and the species of animals entered by the spirits of ancestral chiefs vary among the Bantu peoples.
Further research uncovered new scenarios and an ethno-ecological schema that enabled us to propose an explanation for Shona class 3/4 that covered most of the members and did so, in our opinion, convincingly.This group of scenarios and schemas (3, 12-15) can be regarded as a cultural model pertaining to the role of the ancestors and the function of the mortar and pestle in providing food and health.In its use of scenarios, the model bears some similarity to the cultural definition of STRZYGOŃ 'a ghost, spectre assuming various forms' presented in Bartmiński (2009: 69).
3. The spirits of ancestral chiefs bring rain, thunder, and lightning.12.People pray to the ancestors.13.Grain is pounded daily with a mortar and pestle.14.Doctors cure with herbal medicines that are ground in a mortar and pestle.15.Trees, shrubs, and herbs are associated with coolness, moisture, and medicine.
Our explanation of ChiShona class 3/4 (m: mu-, f: mi-) departs from Lakoff's explanation of Dyirbal noun classifiers in having not one central category per class, but several.Each of the listed scenarios can be regarded as a central category, so the class has multiple central categories.The class includes the related terms muchengamuchenga 'abundance of grain' and muchanganherera 'general rain.'There are prototypes such as muti 'tree, medicine' and schematizations such as mudzukwa 'tall, straight object (e.g.tree; skyscraper).'Many members are related by conceptual metonymies, as muhwi 'pestle' and muchaka 'meal from green mealies,' both from scenario 13.So also is muzhandwa, which may refer to the act of crushing or to plagues that strike down animals or people.Notice that even the central categories relate to one another, as 13 is related to 14 by a similar mechanical process and 3 is related to 12 by ancestors as shared actors.Spatial schemas undergo secondary schematizations, such as end-point transformations to ends of paths, beginnings, final times, worn-out objects, and ends of objects as in mufika 'tapered end of axe or hoe blade.'It appears that class 3/4 of ChiShona is almost fully explained by reference to a small group of central categories, which take the form of scenarios or models.Constructions in class 3/4 may designate (1) objects named in the central categories, in which case we feel justified in treating them as prototypes, (2) schemas (abstractions from the prototypes), (3) objects related to the central categories by similarity or conceptual metaphor, or (4) objects related to the central categories by metonymy.The model, which we refer to as a polycentric category, is described in detail in Palmer andWoodman (1999) andPalmer (2006).Here, we simply note that linguistic data alone would be insufficient to explain the grammar of classifiers.The semantic structure of the category is best revealed by ethnography.Thus, a noun class is very much like a frame in Fillmore's sense.The lexemes or constructions that constitute a frame are relative to the cultural model.
Such a system raises an interesting possibility.The models governing the classifier systems of Bantu languages may be thought of as culturally significant background information.When classifiers are used in conversations or narratives, they have the function of raising the profile of the background information and delimiting the background model for the topics under discussion.Hence, they become a part of the discourse ground and the subjective experience of engaging in discourse.Since they are obligatory, they have the function of engaging participants in discourses structured by salient cultural values (Palmer 2014).A speaker of a Bantu language necessarily imposes his or her ethnic ideology on the discourse.In this respect, classifiers function much like code-switching in Tagalog.Evidently, there is more at stake than grammar alone.
7. The culture and proto grammar of archaic humans5 Language may have originated with a human ancestral population that lived in Africa, Europe, and Asia during a period beginning between 800K ya 6 and 600K ya and ending between 130K and 97K ya, a period that spans the gap between Homo erectus and Homo sapiens.The scientific name for this ancestor from the Middle Pleistocene is Homo heidelbergensis.One might believe that little of interest could be said about the grammar of a proto language developing during this period, accepting that one could even establish the possibility of any language appearing this early.It requires just a few thousands of years for a single language spoken by two separated communities to diverge to the point where the kinship of the evolving branches is no longer evident, so comparisons of living languages are unlikely to reveal much of any proto language.Many scholars assume that such things as body decoration, cave paintings, and rock art provide the first signs of symbolism and therefore of language some 100Kya to 200K ya.Yet there is evidence from physical anthropology that sometime during the early-middle to latemiddle Pleistocene heidelbergensis evolved the vocal and auditory apparatus necessary for speech.Also encouraging is the fact that a rough picture of the culture of heidelbergensis can be built up from archaeological evidence.Raw materials for stone tools became increasingly widely scattered over the time period until by 250K ya they were being transported as far as 340 km, even when closer sources were at hand.The distribution suggests that economic exchanges, very likely involving language, were taking place between bands.The tool kit of heidelbergensis provides evidence of a hunter-gatherer lifeway.It included bifaced hand axes, cleavers, scrapers, choppers, and picks made of stone.There were also spears and there is evidence that heidelbergensis made use of fire.In both Europe and Africa, they hunted the local varieties of large animals.Evidence from tooth wear argues for a diet that included roots, stems, and seeds.The evidence of punctured skulls suggests cannibalism.
The cultural evidence for the hunter-gatherer lifeway provided by archaeology and paleontology together with some very slim phonological evidence provides a few clues about the kind of proto grammar that heidelbergensis may have spoken.There is evidence that humans are born with the ability to produce four kinds of syllables consisting of three CV patterns and one CVC.The same four patterns occur in a proposed proto language based on syllable distributions found in recorded languages (MacNeilage and Davis 2000).With these syllables as our only constraints, we can generate more than 13,000 distinct one and two word utterances.Elaborations and refinements such as attention to word order, syllable reduplication, and distinctions between prefixes and suffixes would expand the possible utterances into what would appear to be a very adequate phonological basis.
We have theorized about what heidelbergensis might have been talking about.We surmise that Heidelbergensian proto semantics involved some of the same patterns and processes that interest cognitive linguists, cultural linguists, and ethnolinguists studying contemporary languages: conceptual metaphors; complex concepts; models of spatial orientation; action chains and scenarios; morphology, polysemy, and grammaticization; and iconicity in phonological networks.These semantic and phonological patterns and processes may have taken less complex forms than those of contemporary languages, but there is no reason to think they were absent.The following possibilities are elaborated in Palmer et al. (2014).
Many historically known and contemporary hunter gatherers have some version the conceptual metaphor mating is foraging, which is based on the common, but not universal, division of labor between men as hunters and women as the gatherers of plant foods.The metaphor can be seen in folktales where a spouse is a game animal and in marriage rituals in which the family of the male provides game and the family of the female provides plant foods.It seems likely that the conceptual metaphor would have occurred to heidelbergensis even prior to the development of proto language.
The middle Pleistocene foraging cultures were simple relative to modern industrial cultures, but they were not simple in an absolute sense.Individuals had to consider the dangers of hunting and gathering together with the vagaries of weather.Their toolmaking technologies and their methods of hunting required considerable skills.They traveled throughout a territory in search of game and plant foods.They must have worried about injuries, illnesses, and the crises of childbirth.They may have dreamed or day-dreamed about animal spirits, beloved parents, lost children, and all the scenarios that go with them.Thus, they had the mental material at hand for the emergence of polycentric categories each consisting of a small set of related scenarios in which the plots involved such abstract sequences as persecution and revenge, pursuit and conquest, scarcity and abundance, separation and reunitement, and such concrete sequences as killing game animals, running from human attackers, wooing, making tools, and invoking spirits to heal the sick.
Many such scenarios would involve action chains involving agents, instruments, and objects.Action chains become the basis of language-specific forms such as the absolutive and ergative case markers found in many languages.Hence, we would expect some reflection of action chains in proto grammar.Langacker (2014: 46) proposed that in important aspects of human experience there is a control cycle consisting of six "force-dynamic phases": relaxation >stimulation > preliminary activity > build up of tension > release through action > resulting control > relaxation.The cycle represents a "fundamental rhythm inherent in moment-to-moment living."We would expect to see some reflection of the control cycle in proto grammar.
Hunter gatherers typically move seasonally within known territories.They must know their territories and remember their routes and camp sites.The ability to give directions becomes useful, whether by pointing, drawing, or speaking.The routes and the sites become invested with knowledge of significant things and events such as waterholes, successful or unsuccessful hunting and gathering forays, encounters with other bands, births, deaths, and places for obtaining materials for toolmaking.Historically known foragers have cycles of songs that are sung at particular sites, creation stories that explain the origin of sites, and stories of events that occurred at them.Through socialization, singing and telling stories, places become invested with social and moral significance (Shore 1996: 269;Shieffelin 1976: 30 et passim;Basso 1984Basso , 1990)).Hence, we can surmise that place names, deictics, directional adpositions, and perhaps travel narratives and songs would have been important in proto language.
Protolanguage would have evolved much as modern languages do, by means of innovation and grammaticization.In grammaticization, phonological constructions may contract or words for concrete objects may acquire new abstract meanings, as when head is abstracted to an endpoint as anadjective (headwater ) or partial (head of ).Grammaticization often occurst hrough subjectification, the process by which expressions may, through fre-quent usage, shift the locus of predication from an objective entity to a speaker's experience in discourse (Palmer 2014).That is, a speaker's experience is shifted from offstage to onstage, while the former object of predication is backgrounded or even elided.It is very likely that utterances of protolanguag eunderwent subjectification.It is likely that the iconicity of syllable reduplication and vowel lengthening was important in protolanguage as it is in so many contemporary languages, and that protolanguage gradually built up symbolic networks based on similar phones and connected meanings (Bybee 1985, Palmer 1996, Occhi 1999, Tuggy 2003).Much more could be said about protogrammar.We have not mentioned anaphora, subject reference, evidentiality, possession, or many other things that one might think every language must provide for.A hypothetical protogrammar would have to include discursives, i.e. lexemes and constructions that predicate about ongoing discourse (discourse markers, indices, etc.), because as soon as protolanguage could sustain discourse, it would need to comment on its own performance vis-à-vis the imagery of preferred discourse.As we learn more about heidelbergensis culture, it may be possible to expand the margin of plausible protolanguage with more specific hypotheses.

Summary and conclusions
The intent has been to use examples taken from research projects to illustrate the articulation of theory and method for cultural linguistics and, by implication, for its sister discipline ethnolinguistics.In our experience, cultural linguistic theory applies readily along a continuum of constructions ramifying from compound affixes on up to narratives.Whereas many linguists and linguistic anthropologists use their grammatical and cultural competence and insight to elucidate meanings emergent in discourse (Occhi 2014) or entrenched as systems of metaphor (Yu 2009, Kövecses 2005), we have focused on using culture to explain grammar.We find cultural models, or their components, in texts and systematic arrays of constructions.Hence, source materials for linguistic ethnography may take the form of texts, dictionaries, and thesauri, each offering different perspectives on cultural models.For fine-grained analysis it is often useful to arrange grammatical constructions in tables and subdivide categories by morphology and by semantics according to schemas and scenarios rather than features.A feature set may work well enough to describe the vocal tract and phonology, but it is too limiting for the realm of linguistic meaning.We also find cultural models in non-linguistic ethnographies, which we believe to be underutilized.
Ad-linguistic (or: co-linguistic) data obtained by participant observation or from films, ethnographies, and recordings can reveal otherwise obscure cultural models underlying grammatical structures such as noun classifiers and hierarchies of animacy.We found that the noun classifiers of ChiShona reveal underlying models in the form of polycentric categories.The finding suggests a new perspective on Bantu classifiers as complexes of concordant constructions that raise the salience of important cultural themes in discourse.This assertion should apply as well to classifier languages that lack category concordance.This discovery of ideology hidden in plain sight as grammar might be of particular interest to ethnolinguists.
We have attempted to use the ethnography of prehistory to deduce some parameters of the language of the Middle Pleistocene human ancestor Homo heidelbergensis, whose culture we can know only through archaeology and paleontology.While the details of heidelbergensis' linguistic abilities will probably always remain obscure, we can take some comfort in the fact that cultural linguistics puts us in a position very much like that of the first makers of stone tools.They were developing language in concert with their foraging culture.We should be developing linguistic theory in concert with ethnographic methods.As with archaic humans, much more lies within our grasp than we could have imagined.

Figure 1 .
Figure 1.Semantic relations among categories of ka--an

Figure 2 .
Figure 2. A Tagalog model of experience: Causal sequences in experiential statements.(Based on Figure 1 from Palmer 2003a: 274)

Table 1 .
Senses of compound affix construction ka--an