INTRODUCTION
The role of speech in human behavior has always been honored in anthropological principle, if sometimes slighted in practice. The importance of its study has been declaimed (as by Malinowski [1935]), surveyed with insightful detail (as in Sapir [1933)), and accepted as a principle of field work (see citations in Hymes 1959).

That the study of speech might be crucial to a science of man has been a recurrent anthropological theme. Boas (1911) came to see language as one in kind with ethnological phenomena generally (he interpreted ethnology as the science of mental phenomena), but revealing more of basic processes because more out of awareness, less subject to overlay by rationalization. Some anthropologists have seen language, and hence linguistics, as basic to a science of man because it provides a link between the biological and sociocultural levels. Some have seen in modern linguistic methodology a model or harbinger of a general methodology for studying the structure of human behavior.

American anthropology has played an important part in the progress of linguistics in this country, through the careers of Boas, Sapir, Bloomfield, and their students, and through the opportunities offered by American Indian languages. It has contributed to the development of particular techniques and concepts, and has used linguistics as a tool for other lines of research. In both respects, anthropology's involvement with linguistics has come to be shared now by psychology. Having assimilated modern advances in linguistics, many psychologists have contributed studies of considerable relevance and value in recent years. One need cite only the work of Charles Osgood, George Miller, and Roger Brown. Hybridization between linguistic concepts, and the technologies of the computers and experimental psychology, is producing perhaps the most rapidly growing sector in the study of speech, one with which anthropology must keep informed liaison.

Indeed, diffusion of the tools of modern linguistics may be a hallmark of the second half of this century. In the course of such diffusion, presumably three things will hold true: 1. the discipline of linguistics will continue to contribute studies of the history, structure, and use of languages; 2. in other disciplines, linguistic concepts and practices will be qualified, reinterpreted, subsumed, and perhaps sometimes re-diffused in changed form into linguistics; 3. linguistics will remain the discipline responsible for coordinating knowledge about verbal behavior from the viewpoint of language itself.

In any event, the joint share of linguistics and psychology in the burgeoning study of verbal behavior seems vigorous and assured. Has anthropology a share apart from some of its practitioners becoming linguists and psychologists, and apart from its traditional role as an intellectual holding company under the aegis of culture? Is the role of prime collaborator of linguistics among the sciences now to pass to psychology? Sheer weight of numbers may determine. It would be of no importance were it not for the value to linguistics and anthropology of a strengthening, not a relaxing, of mutual concern.

In one regard, there is no danger of lapse. Modern linguistics is diffusing widely in anthropology itself among younger scholars, producing work of competence that ranges from historical and descriptive studies to problems of semantic and social variation. Most such work is on well-defined linguistic problems; its theoretical basis is established, its methodology well grounded, and its results important, especially for areas in which languages rapidly dwindle in number. There is no need to detail the contribution which such work makes to anthropological studies, nor to argue its permanent value to linguistics proper. If anything, the traditional bonds between linguistics and anthropology in the United States are more firmly rooted now than a decade ago.

What may lapse is an opportunity to develop new bonds, through contributions to the study of verbal behavior that collaboration between anthropology and linguistics can perhaps alone provide. This is more than a matter of putting linguistics to work in the study of other scientific problems, such as cognitive behavior or expressive behavior. The role of speech in both is important, and has engaged anthropological attention: the cognitive problem in association with the name of Whorf, the expressive problem more recently under the heading of "paralinguistics." But to pursue these problems, and to try to give them firm anthropological footing, is to broach the study of a new problem area, one of which little account is taken.

There are indeed several underdeveloped intellectual areas involving speech to which anthropology can contribute. All are alike in that they need fresh theoretical thought, methodological invention, and empirical work, and have roots in anthropology's vocation as a comparative discipline. Among these areas are the revitalization of dialectology (perhaps under the heading of "socio- linguistics"); the place of language in an evolutionary theory of culture; the semantic typology of languages; and the truly comparative study of verbal art.¹ Fortunately, all those mentioned have begun to attract attention. For the anthropological study of behavior there is another area of importance, one that seems general, central, and neglected. It can be called the ethnography of speaking.

^1. Towards the first of these, see Gumperz (1961); towards the other three, see respectively, Hymes (1961c, 1961a, and 1960a [for the typology at the close of the latter]). Such developments will require rapprochement with established philological disciplines, which control much of the essential data.

In one sense this area fills the gap between what is usually described in grammars, and what is usually described in ethnographies. Both use speech as evidence of other patterns; neither brings it into focus in terms of its own patterns. In another sense, this is a question of what a child internalizes about speaking, beyond rules of grammar and dictionary, while becoming a fullfledged member of its speech community. Or, it is a question of what a foreigner must learn about a group's verbal behavior in order to participate appropriately and effectively in its activities. The ethnography of speaking is concerned with the situations and uses, the patterns and functions, of speaking as an activity in its own right.

What the content of this area may be in detail, what a description of it as a system might be like-these things are hard to state, although I shall attempt it in this paper. Field studies devoted to the topic hardly exist, nor has there been much attention to what the theory and method of such studies would be.  Occasional information can be gleaned, enough to show that the patterns and functions of speaking can be very different from one group to another-how speech enters into socialization and education, for example, may differ strik- ingly. But the evidence is not enough to itemize all variables, or to show a system. Hence the orientation of what follows must be toward the field work that is necessary.

Why undertake such field work? The reasons are several: because the phenomena are there, ready to be brought into order; so that systematic descriptions can give rise to a comparative study of the cross-cultural variations in a major mode of human behavior (a "comparative speaking" beside comparative religion, comparative law, and the like), and give it its place in theory; for the contribution to other kinds of concern, such as studies of the formation of personality in early years.

I shall attempt to bring out the nature and problems of this area by indicating first that study of speech as a factor in cognitive and expressive behavior leads to concern with the ethnographic patterning of the uses of speech in a community. Then I shall sketch a descriptive framework for getting at such a patterning. A "notes-and-queries" survey of the role of speech in socialization will bring much of the content and method in the frame of one problem. Finally, I shall sketch the changes in theoretical perspective that underlie the whole.

SPEECH IN COGNITIVE AND EXPRESSIVE BEHAVIOR

The role of speech in cognitive behavior is an old concern of anthropology. In recent years discussion has most often had reference to Whorf's views. There is not space here to evaluate the ideas and studies that are pertinent, and I can only refer to two other papers (Hymes 1961a, b). It can be briefly said that there is no question but that speech habits are among the determinants of nonlinguistic behavior, and conversely. The question is that of the modes amounts of reciprocal influence.

If our concern is the role of phonological habits in the perception and interpretation of sounds, there exists an abundance of theory, technique, and experimental work. If our concern is the role of semantic habits in perception and interpretation of experience, there is no such abundance. Some experimental testing has been done (see comment in Hymes 1961b), but we cannot adequately investigate the role of semantic habits in ordinary behavior without knowledge of the semantic habits that are available to play a role, and such knowledge can be gained only by description in relation to native contexts of use. In other words, we need a semantic analysis that is a part of ethnography.

The need for such an ethnographic semantics has been pointed out before, and it is the theme of Malinowski's Coral Gardens and Their Magic Part 11. How to implement an ethnographic semantics, however, how to devise its methodology, largely remains. Malinowski saw clearly the need to analyze meaning in contexts of use, but his method amounted in practice to massive narrative. An ethnographic semantics may be bulky, but it need not be on principle interminable, nor endlessly ad hoc. It should be more than a narrative reflection of reality. It should be a structural analysis, achieving the economics of the rules of a grammar in relation to a series of analyses of texts.

In the past generation Jakobson and his associates have done most to develop such a structural semantics. In recent years a fresh wave of American interest has appeared in significant papers by linguists such as Haugen (1957) and Joos (1958), and by ethnographers such as Conklin (1955, 1962), Goodenough (1956a, 1957), and Lounsbury (1956). Here as in other studies there are two general approaches, as Jakobson has so brilliantly set forth; on the one hand to trace an item through all the various contexts in which it can occur, characterizing it in terms of its ability to co-occur with other items, and on the other to place an item within a set which can occur in particular contexts, characterizing it in terms of its substitutability for other items of that set. The two approaches have various names, such as the syntagmatic and paradigmatic axes (see Jakobson and Halle 1956). The first approach is essentially that of a concordance; the second approach can be termed that of a contrast within a frame, or better, contrast within a relevant (or valid) frame. Here I want to side with those who consider the latter the more fundamental of the two, since it validates the structural relevance of the items whose distribution is studied by the first approach, and adds information of its own; and assert that use of this fundamental "contrast within a frame" approach must lead linguistics into ethnography, and ethnography into analysis of patterns of speaking.

Here I can only outline the argument. The paradigmatic approach requires discovering a relevant frame or context, identifying the items which contrast within it, and determining the dimensions of contrast for the items within set so defined. The approach has been successful for phonology and grammar, but only partly so for lexicon. Indeed, it is much disputed that a structural approach can be applied to the whole of a language, when the whole of vocabulary is considered. Yet it would be remarkable, and should be a source of embarrassment, if the paradigmatic principle fundamental to the core of language should fail us here. Recognizing this, linguists associated with the glossematic school have proposed modes of analysis of "content-structure" and defended the possibility of extending them to all of lexicon on principle. These modes may prove fruitful, despite theoretical criticisms, although some seem to smack too much of the ad hoc and arbitrary at present. In any case these approaches tend to stay within received bodies of linguistic data rather than to move outward into the exploration of speech behavior and use. Such exploration is essential, whether one is concerned with semantics delimited as dealing with designation and intension, or whether one is concerned also with what one might then term "pragmatic meaning," as the ethnography of speaking must be. (Cf. Firth's inclusion in his conception of "semantics" of this pragmatic dimension of meaning, which he places beyond lexicography in the province of "sociological linguistics" [1935:27].)

The need for such exploration is easy to see. One source of the present impasse in structural analysis of content is precisely the limitations of the contexts available in the usual linguistic materials. The usual corpus provides sufficient contexts for phonological and grammatical analysis, but for semantic analysis of only a few limited sets of frequently recurring elements, such as case-endings and prepositions. That is one reason Wells writes, regarding the possibility of structural analysis of items such as the Latin stem tabul-, "the only reliable method now available depends upon treating it as a member of some C[ontent]- paradigm. This we do not see how to do" (Wells 1957).

Scholars sometimes have been willing also to posit dimensions of contrast for a few other domains, apparently universal or "given," such as kinship terms, numerals, pronouns. But in fact even the seemingly most obvious domains can- not be taken for granted. It may sometimes be assumed that, although languages segment experience differently, what they segment is the same, as if it were a matter of different jigsaw puzzles fashioned from the same painting. But recent work shows that structural analysis of meaning must first demonstrate that a domain is a domain for speakers of the language in question. What the domain includes, what it excludes, what features define it and its elements, cannot be prescribed in advance, even for kinship (cf. Conant 1961) or color terms (Con- klin 1955). (The principle is generally true for cultural phenomena; cf. on residence rules, Goodenough (1956b], and on the structure of the family, Adams [1960].)

The exploration of native contexts of use to validate domains is the basis of the success of Conklin and Frake, and it points the way for the structural analysis of all of speech. All utterances occur contrastively in contexts, but for much of lexicon and most larger units of speech, the contextual frames must be sought not in the usual linguistic corpus, but in behavioral situations. One must reciprocally establish the modes and settings of behavior relevant to speech and sets of verbal items that occur within them, dimensions of contrast and rules of use, whether purely semantic  (designative) or concerned with other imports and functions, can then be found. (The sets would often not be perceived from a formal linguistic point of view, being formally diverse, e.g., a
set of greetings may range from "Hi" to "it's a damned good thing you got here when you did, Jack.")

The approach of course requires the structural analysis of the community in relation to speech that would constitute an ethnography of speaking. This approach is an answer to the problem posed by Hjelmslev (1957:283): "Une description structurale ne pourra s'effectuer qu’`a condition de pouvoir reduire les classes ouvertes a` des classes ferm`ees."

For understanding and predicting behavior, contexts have a cognitive significance that can be summarized in this way. The use of a linguistic form identifies a range of meanings. A context can support a range of meanings. When a form is used in a context, it eliminates the meanings possible to that context other than those that form can signal; the context eliminates from consideration the meanings possible to the form other than those that context can support. The effective meaning depends upon the interaction of the two. (Recently stated by Joos [1958], this principle has also been formulated by Buhler [1934:183] and Firth [1935:32].)

Important also is the point that the cognitive role of speech is no all-or-nothing, but a matter of what, where, and when. Speech is cognitively more important in some activities than others, some times more than others, for some persons more than others, for some societies more than others. The amount and kind of influence may change as between the child and the adult, and there are the obvious problems of the relative importance of their languages for multilinguals.

Such concern with speech in contexts of behavior leads toward analysis (if individual patterns in particular native situations. If, from a grammar, we not read off the role that speech habits play in present-day behavior, neither can we do so from an experimental situation novel to the culture. Nor can the assessment be made from compartmentalized accounts of speech habits and of other habits, compared point-for-point in some millennial future. The analysis must be made on the ground. We must know what patterns are  in what contexts, and how, where, and when they come into play. The maximum that "meaning is use" has new force when we seriously study the role of semantic habits in behavior.

In sum, description of semantic habits depends upon contexts of use to define relevant frames, sets of items, and dimensions of contrast. Moreover, persons and groups may differ in the behavior that is mediated by speech. Thus analysis of the role of speech in cognitive behavior leads into analysis of the ethnographic context of speech.

The same holds true for the role of speech in expressive behavior. Of course there is a cognitive aspect to expressive behavior, insofar as it presupposes the sharing of a code, so that semantic habits do not exhaust the cognitive role of speech. Likewise, there is an expressive aspect to the cognitive style of an individual or group, and in general, all speech phenomena can be interpreted by a hearer as expressive of a speaker. But expressive studies tend to emphasize speech as an aspect of personality, and to throw into prominence features of speech, such as tone of voice and hesitation pauses, that lie outside lexicon and grammar -phenomena which have recently been systematized in a preliminary way under the heading of "paralinguistics." (For a general survey of both cognitive and expressive aspects of personality, linguistically viewed, see Hymes [1961b].) The principal study to result so far from the work in paralinguistics, that of Pittenger, Hockett and Ranehy (1960) is based on the heuristic, if somewhat intuitive, use of the principle of contrast within a frame, applied to the unfolding of a psychiatric interview. Indeed, the main task confronting paralinguistics is to determine the import of the phenomena it has isolated by further study of their contrastive use in situations. In general, advances in analysis of the expressive role of speech also lead into analysis of the ethnographic context.²

Among other anthropological concerns which lead into such analysis, there is the aspect of culture change involving programs of fundamental education, concerned with literacy and multilingualism. In introducing new uses for indigenous forms of speech, and in extending foreign forms of speech into local contexts, the patterns and functions of speaking on both sides need to be analyzed, so as to anticipate points of congruence and conflict (cf. Weinreich 1953 and Hymes 1961c).

Now it is time to consider how the analysis of the ethnographic context of speech may be carried out. 'there are a number of lines of research whose goals overlap those of an ethnography of speaking, and whose results and method,, must contribute. Since these lines of research have so far not fused or had tl)c particular focus and scope that is of concern here, it is worthwhile, permit necessary, to take this opportunity to broach the descriptive problem, and to outline a method of approach. My way of getting at it is of course without prejudice to ways that prove rewarding to others. Approaches to ethnographic analysis devised under linguistic influence, although they may diverge, are likely to show strong resemblance at many points.³

² Mahl (1939) has discussed an "instrumental aspect of language" as constituting t gap in psychology. He argues that "The instrumental model is the more general and valid one for purposes of inferring emotional states from language behavior" (p.40) and that the instrumental model is more closely linked to behavior than the representational (cognitive, or lexicon-and-grammar focused) model. But a cognitive approach may be concerned with the effect of a speech-derived symbolic map on problem-solving, planning, and the like, and hence can also be called "instrumental," since it also deals with speech as tool-using behavior. In exploring the signalling of emotional states, Mahl deals with what will here be termed expressive function, and in pointing to the effect of this signalling of the behavior of others, he deals with what will here be termed directive function. His use of "instrumental" subsumes the two. I particularly value Mahl's analysis because he insists on "including the situational and/or the nonlexical contexts of messages" (p. 105) and in effect demands the equivalent of an ethnography of speaking in relation to the analysis of speech events for certain psychological purposes.

³ E. T. Hall, The Silent Language, is especially worthwhile. Details apart, my only reservation is that the 10 primary message systems, the 3 levels of culture, the 3 components of messages, the 3 principal types of patterns, and the 100-category map of culture should be taken more frankly as heuristic devices. In particular the 10 primary message systems seem but one convenient breakdown, rather than rooted in biology, and the components (set, isolate, pattern) and pattern types (order, selection, congruence) seem a valid but partial extrapolation of a linguistic model. Several such extrapolations, particularly those of Hall and Trager, of Jakobson, of Pike (1954, 1955, 1960), and of Uldall, have each their contribution to perspective, but none has yet carried the day. The Hall and Trager framework of components (set, isolate, pattern) converges in a noteworthy way with the trimodal framework (manifestation, feature, distribution modes) of Pike.

DESCRIPTIVE ANALYSIS OF SPEAKING
The descriptive focus is the speech economy of a community. The scope is all behavior relevant to a structural ("emic," in Pike's terminology) analysis of this. The approach is not to consider behavioral reality a pie and the speech economy a unique slice. It is a question of an organizing perspective on a social reality that is the same for differing analytical frameworks. I believe that structural analysis in this particular framework will be of value in its own right and will feed back into analyses from other perspectives.

By structural analysis is meant more than the placing of data in an articulated set of categories. Such placing is a necessary starting point, and also a desired outcome, when systems that have been individually analyzed are studied comparatively. But for the individual system, structural analysis means a scientific and moral commitment to the inductive discovery of units, criteria, and patternings that are valid in terms of the system itself. An illustration is the interrelation between phonetics as a starting point, the phonemic analysis of a given language, and the use of the results of that analysis in general linguistics, e.g., in phonemic typology; or, ethnological categories as a starting point, the ethnographic analysis of, say, the residence rules of a community, and the use of the results of that analysis in a comparative study. The categories presented here for an ethnography of speaking must be taken as ways of getting at individual systems, as analogous to a phonetics and perhaps part of a practical phonemics. The intent is heuristic, not a priori.

The point seems obvious, but experience shows it to be easily mistaken. Let me put it another way. What would be an appropriate improvement, or correction, of what follows? Not an argument that there really are 3, or 8, or 76, factors or functions of speech-in general. That would be equivalent to arguing how many phonemes there really are-in general. The problem, of course, is how many phonemes, or factors and functions, there are in some one determinate system. What the range in number of factors and functions may be, what invariants of universal scope there may be - answers to these questions may perhaps be glimpsed now, but must wait for demonstration on the structural analyses of many systems. An appropriate improvement or correction, then, is one that contributes to that job, that makes of this paper a better practical phonetics and phonemics.

It can be asked: to what extent is analysis from the perspective of speaking itself valid structurally to a given case? Activity defined as speaking by one group may be defined as something else by another. But differences of this sort are themselves of interest. Some behavior will be organized and defined in terms of speaking in every group, and the import of this behavior may be missed if not investigated as such. Only a focus on speaking answers the structural question, and provides data for comparative study of the differential involvement of speaking in the structure of behavior in different groups. In one sense, a comparative ethnography of speaking is but one kind of comparative study of the utilization of cultural resources.

Note that the delimitation of the speech economy of a group is in relation to a population or community, however defined, and not in relation to the homogeneity or boundaries of a linguistic code. If several dialects or languages are in use, all are considered together as part of the speech activity of the group. This approach breaks at the outset with a one language-one culture image. Indeed, for much of the world the primary object of attention will not coincide with the units defined as individual languages. The patterning of a linguistic code will count as one among several analytical abstractions from verbal behavior. In cultural terms, it will count as one among several sets of speech habits. The specialization of particular languages or varieties to particular situations or functions, and the implications of each for personality, status, and thinking, will be a normal part of a description. Standard analysis of each code will of course be necessary, but the broader framework seems more "natural," indeed, more properly anthropological. The structure of this argument also applies if the focus of attention is not a population but an individual personality.⁴

⁴ Aberie (1960) argues that language has been an inadequate model for culture-and-personality studies, having only two terms, the individual and the sacred cultural pattern, whereas a third term, the cultural system in which persons participate but do not share, is necessary. In Aberle's terms, I am saying here that the two-term model is inadequate for linguistics studies as well. "Ethnography of speaking" involves a speech equivalent of "cultural system".

A necessary step is to place speaking within a hierarchy of inclusiveness: not all behavior is communicative, from the viewpoint of the participants; not all communication is linguistic; and linguistic means include more than speech. One can ask of an activity or situation: is there a communicative act (to one- self or another) or not? If there is, is the means linguistic or non-linguistic (gesture, body-movement) or both? In a given case, one of the alternatives may be necessary, or optional, or proscribed. The allocation of communication among behavior settings differs from group to group: what, for example, is the distribution of required silence in a society-as opposed to occasions in which silence, being optional, can serve as a message? (To say that everything is communication is to make the term a metaphor of no use. If necessary, the wording could be changed to: not all behavior is message-sending ... not all message- sending is linguistic ... etc.) The allocation of communicative means may also differ. For any group, some situations must be speech situations, some may be, some cannot be. Which situations require writing, derivative codes of singing, whistling, drumming, non-linguistic uses of the voice or instruments, or gesture? Are certain messages specialized to each means?

The distribution of acts and means of communication in the round of behavior is one level of description. Patterns of occurrence and frequency are one kind of comparison between groups. Much more complex is the analysis of the communicative event itself. (In discussing it, I shall refer to speech and speaking, but these terms are surrogates for all modes of communication, and a descriptive account should be generalized to comprise all.) Let me emphasize again that what I present is not a system to be imposed, but a series of questions to be asked. Hopefully, the questions will get at the ingredients, and from the ingredients to the structure of speaking in a group.

There seem to be three aspects of speech economy which it is useful to consider separately: speech events, as such; the constituent factors of speech events; and the functions of speech. With each aspect, it is a question of focus, and a full description of one is partly in terms of the rest.

Speech Events
For each aspect, three kinds of questions are useful. Taking first the speech events within a group, what are instances of speech events? What classes of speech events are recognized or can be inferred? What are the dimensions of contrast, the distinctive features, which differentiate them? (This will include reference to how factors are represented and functions served.) What is their pattern of occurrence, their distribution vis-a-vis each other, and externally (in terms of total behavior or some selected aspect) ?

One good ethnographic technique for getting at speech events, as at other categories, is through words which name them. Some classes of speech events in our culture are well known: Sunday morning sermon, inaugural address, pledge of allegiance. Other classes are suggested by colloquial expressions such as heart-to-heart-talk, salestalk, talk man-to-man, woman's talk, bull session, chat, polite conversation, chatter (of a team), chew him out, give him the lowdown, get it off his chest, griping, etc. I know no structural analysis. Clearly time material cannot be culled from a dictionary alone: instances and classes of speech events may be labelled by quite diverse means, not only by nouns, but also by verbs, phrases, and sentences. In response to the question, "Nice talk?", a situation may be titled by the response "Couldn't get a word in edgewise."

Insofar as participants in a society conceive their verbal interaction in terms of such categories, the critical attributes and the distribution of these are worth discovering.

Take "cussing out," a Wishram Chinook's English label for a class of aboriginal speech events. A set of verb stems differentiates varieties of "cussing out." What alternative events (linguistic or non-linguistic) are possible in the same situation, such as dismissal or beating? With regard to factors, who cusses out whom, when and where, in what style or code, about what? With regard to function, is there an aesthetic element, are speakers rated as to ability, what does "cussing out" do for speakers, what effect is expected or follows for hearers? What is the role of "cussing out" in maintenance of social system, cultural values, personality systems? (The analysis of Hausa roka (praise singing) by Smith [1957] is an interesting work along these lines, as is Conklin [1959].)

An interesting question about speech events concerns what can serve to close them, or to close a sequence within one.

Factors in Speech Events
Any speech event can be seen as comprising several components, and the analysis of these is a major aspect of an ethnography of speaking. Seven types of component or factor can be discerned. Every speech event involves 1. a Sender (Addresser) ; 2. a Receiver (Addressee) ; 3. a Message Form; 4. a Channel; 5. a Code; 6. a Topic; and 7. Setting (Scene, Situation).⁵

⁵ In what follows I am most immediately indebted to Roman Jakobson's presentation of factors and functions in his concluding remarks to the Conference on Style held at Indiana University, April 1958, sponsored by the Social Science Research Council. The published statement identifies six factors and corresponding functions (Jakobson 1960). Jakobson's rich discussion should be carefully read. I have also pervasive debts to Kenneth Burke, Kenneth L. Pike, Sinclair (1951) and Barker and Wright (1955).

The set of seven types of factor is an initial ("etic") framework. For any group, the indigenous categories will vary in number and kind, and their in- stances and classes must be empirically identified. For example, Sender and Addresser, or Receiver and Addressee, need not be the same. Among the eastern Chinookan groups, a formal occasion is partly defined by the fact that the words of a chief or sponsor of a ceremony are repeated by a special functionary to the assembled people. In general, the categories of these two factors must be investi- gated in terms of the role system of the group studied. Moreover, depending upon beliefs and practices, the categories of Senders and Receivers variously overlap the membership of the human group. The coming of a flock of ravens brought warning for the Kwakiutl, and, indeed, there was a corresponding category of Receiver: an individual whose afterbirth had been eaten by ravens could, as an adult, perceive raven cries as one or another of a limited set of Kwakiutl utterances. A stone is one type of potential Sender among the Fox. Infants may or may not be counted as a class of potential Addressees and talked to; they were so counted among-the Mohave and Tlingit, who thought infants capable of understanding speech. (The practice with infants and pets varies in our own society.) The form of a Message, or the typical form of a class of Messages, is a descriptive fact that becomes significant especially as an aesthetic and stylistic matter, whether in relation to the resources of a code (Newman [1940] has shown that Yokuts and English stand in sharp contrast), to a particular context (Riffaterre [1959] takes this relation as fundamental to analysis of style), or to a particular referential content (as when some linguists find that the modifier "Trager-Smith" fits their sentence rhythms better as "Smith-Trager").

Cross-cultural differences in Channels are well known, not only the presence of writing, but also the elaboration of instrumental channels among West African peoples such as the Jabo, the whistling of tones among some of the Mazatecs of Mexico, etc.

It has already been noted that the Code factor is a variable, given a focus on the speech habits of a population. The range is from communities with different levels of a single dialect to communities in which many individuals command several different languages. The presence of argots, jargons, forms of speech disguise, and the like enters here. Terms such as "dialect," "variety,” “vernacular,” “ level," are much in discussion now (see Ferguson and Gumperz 1960, Hill 1958, Kenyon 1948). It is clear the status of a form of speech as a dialect, or language, or level, cannot be determined from linguistic features alone, nor can the categories be so defined. There is a sociocultural dimension (see Wolff 1959, on the non-coincidence of objective linguistic difference and communicative boundary), and the indigenous categories must be discovered, together with their defining attributes and the import of using one or another in a situation. Depending on attitude, the presence of a very few features can stamp a form of speech as a different style or dialect.⁶

⁶ The phenomena which Voegelin treats as "casual" vs. "noncasual" belong here. Voegelin (1960) sees the need for an empirical, general approach to all forms of speech iii a community, discussing their variation in number and kind between communities, and the situational restrictions on their use. His discussion takes "casual" as a residual, ,unmarked category, whereas the need is to assume that all speech manifests some posi- tivtly marked level or style, and to discover the identifying traits. He generalizes that neither formal training nor specialized interest contributes to proficiency in casual and that judgments of proficiency are not made, but evaluations of proficiency among the Menominee (Bloomfield 1927) and the Crow (Lowie 1935) show that his implication of "casual" is misleading. Indeed, for some groups, most utterances might have to be classed in Voegelin's terms as "noncasual," for training in proper speaking is intensive and proficiency stressed (e.g., the Ngoni of Nyasaland and many groups in Ghana).

The Topic factor points to study of the lexical hierarchy of the languages spoken by a group, including idioms and the content of any conventionalized Utterances, for evidence and knowledge of what can be said. To a large extent this means simply that semantic study is necessary to any study of speaking. An Ethnography of speaking does also call special attention to indigenous categories for topics. One needs to know the categories in terms of which people will answer the question, "What are they talking about?", and the attributes and patterns of occurrence for these categories. The old rhetorical category of topoi might go here as well.

The Setting factor is fundamental and difficult. It underlies much of the rest and yet its constituency is not easily determined. We accept as meaningful such terms as "context of situation" and "definition of the situation" but seldom ask ethnographically what the criteria for being a "situation" might be, what kinds of situations there are, how many, and the like. Native terms are one guide, as is the work of Barker and Wright (1955) to determine behavior settings and to segment the continuum of behavior.⁷

⁷ Jakobson treats the last two factors (his Context and Referent) together as one factor. To stress my descriptive concern with factors, I eschew the theoretically laden term "Context" for a factor here, retaining "Setting" (cf. Barker and Wright 1955) with "Scene" (Burke 1945) and "Situation" (Firth 1935, following Malinowski) as alter- natives. As factors, I distinguish Setting and Topic because the same statement may have quite different import, as between, say, a rehearsal and a performance. In one sense, it is simply a question of what one has to inventory in describing the speech economy of a group. Settings and Topics seem to me to involve two obviously different lists, and lists on the same level as Addressers, Addressees, Channels, etc. Put otherwise, "Who said it? Who'd he say it to? What words did he use? Did he phone or write? Was it in English? What'd he talk about? Where'd he say it?" seem to me all questions of the same order. With functions I cannot avoid using "Context." I agree with Jakobson that referential function involves context (as an earlier section makes plain), but find this no difficulty if a function may be defined in relation to more than one factor. I also agree with Jakobson that all aspects of a speech event are aspects of context from one point of view, but I have argued that all aspects may be viewed in terms of any one factor;  and the level at which all are aspects of context merges all, not just context and reference, while the level at which the others are distinct seems to me to distinguish context all(i reference as well, as I hope the illustrations, especially the literary ones, show. Certainly if reference is less than the total import of a sentence, then shifting the line "And seal the hushed casket of my soul" from early in the sonnet "To Sleep" to its close (as manuscripts show Keats did), enhanced the effect of the line and its contribution to the poem, without changing its reference.

Some of the import of these types of factors will be brought out with regard to the functions of speech. With regard to the factors themselves, let us note again that native lexical categories are an important lead, and that contrast within a frame is a basic technique for identifying both instances and classes, and for discovering their dimensions of contrast.

Given the relevant instances and classes for a group, the patterning of their distribution can be studied. One way is to focus on a single instance or class, hold it constant, and vary the other components. As a sort of concordance technique, this results in an inventory, a description of an element in terms of the combinability of other elements with it. As a general distributional technique, this can discover the relations which obtain among various elements: whether co-occurrence is obligatory, or optional, or structurally excluded. Some- times the relation will hold for only two elements (as when a certain category of Receiver may be addressed only by a certain category of Sender), sometimes for several. The relation may characterize a class of speech events.

In this way we can discover the rules of appropriateness for a person or group. (And indications that such rules have been violated are of special help in discovering them.) From a linguistic (Code) point of view, such rules may account for variance in the speech material on which a description is based, explaining why some grammatically possible utterances do not occur (e.g., to illustrate each type of factor: because the informant is not an appropriate Sender, the linguist not an appropriate Receiver, a different choice of words or order is preferred, the sequence is sung, and cannot be dictated apart from that mode of channel, the sequence indicates a speech variety or level which the informant avoids or must not use, the topic is tabued, the situation which would elicit the utterance has never occurred or been imagined, such a thing is said only in a context to which the linguist has no access). From an ethnographic point of view, the discovery of such rules of appropriateness is of practical importance for participant observation, and it is central to the conception of speaking as a system. One way that patterns of speaking constitute a system is in virtue of restrictions on the co-occurrence of elements.

Relevant data have been noted by ethnographers, especially as incident to lexical items of interest, such as kin terms. Linguists have taken account of such data when intrusive into the formal code, as when different morphemic shapes or different paradigms are used according to the sex of the speaker and hearer. (Haas 1944 is the best treatment.) The participants in speech may then be admitted as environments for use of the principle of complementary distribution, and the different forms treated as lexically or grammatically equivalent; but such data are likely to be regarded as a frayed edge of grammar rather than as an opening into the broader system of speaking. (Such facts have sometimes served as casements for vision of different men's and women's "languages," but serious characterization of speech differences between men and women in a society hardly exists.)

A descriptive analysis of patterns of speaking in terms of indigenous instances of the constructive factors of speech events is worthwhile in its own right, and it feeds back into prediction and inference about behavior. Given a speech event in the limited sense of a concrete message, frequently the main interest is in what can be told about one or more of its constituent elements. What can be told about the Sender, either as to identity (age, sex, social class, and the like) or as to motive, attitude, personality?; what can be told about the Receiver, including his or her likely response?; about the Context (including antecedent circumstances, verbal or non-verbal) ?; and so on. (For the fieldworker or learning child, the question may be what can be told about the Code; for the communications engineer, what can be told about the Channel.) We may consider relations between elements, or consider all as evidence about a certain one.

The saliency of this focus is of course that it is what we often have to work with, namely, text of one sort or another. Inquiry of this sort is common in and out of science. But in our own society the success of such inquiry presupposes a knowledge of the relations--diagnostic, probabilistic-that obtain among the constitutive elements of speech events. We share in the patterns of speaking behind the text or message, and can to some extent ask ourselves, what would be different if the Sender were different?; if the Sender's motives were different?; and so on. In another society this contrast-within-a-frame technique must appeal to an explicit analysis of patterns of speaking.