Edward Sapir spent his working life on languages that most of his contemporaries had never heard spoken, and the essays gathered here show why that labor mattered far beyond the field. This is a collection of his selected writings on language, culture, and personality, edited by David G. Mandelbaum and drawn from papers, encyclopedia entries, and reviews written across the 1910s, 1920s, and 1930s. Sapir was born in Germany in 1884, came to the United States as a child, trained under Franz Boas at Columbia, and died in 1939 while Sterling Professor at Yale. The pieces assembled here range from technical phonology to the theory of culture to the psychology of religion.
“There are certain terms that have a peculiar property.”
The book’s arrangement follows Sapir’s own three interests. A first part gathers writings on language—its nature, the sound patterns that organize it, the American Indian languages he documented in the field, and comparative work on Indo-European and Semitic. A second turns to culture: general theory, studies of American Indian life, and essays on literature and music. A third, on the interplay of culture and personality, carries his late thinking about how individuals carry and reshape the culture they inherit.
From Sound Patterns to the Shape of Culture
The Editor’s Frame and Sapir’s Career
The volume opens with editorial apparatus that establishes both the scope of the collection and the stature of its author. David G. Mandelbaum’s introduction presents Edward Sapir as a scholar of extraordinary range: a linguist whose command of languages ran from Germanic philology and Takelma to American Indian families, Jabo (learned from a speaker he found working in a Chicago bowling alley), and later Indo-European, Sinitic, and Semitic studies. Franklin Edgerton’s praise is invoked—Sapir could meet specialists on their own ground while surveying the whole terrain. The introduction insists that Sapir treated linguistics as a social science, language as an aspect of culture, and speech as something embedded in its social setting.
The biographical sketch traces a life from Lauenburg, Germany, where Sapir was born in January 1884, through his family’s migration to America when he was five, his father’s work as a cantor, scholarships at Horace Mann and a Pulitzer fellowship to Columbia, and the decisive meeting with Franz Boas. Fieldwork followed: Wishram in Washington, Takelma in Oregon for the doctorate, Yana at Berkeley, Ute trips and the Paiute informant Tony Tillohash, then Ottawa as chief of anthropology at the Canadian National Museum, with Nootka, Sarcee, Tlingit, Kutchin, and Ingalik work, a study of Chinese humor, and French-Canadian folk songs gathered with Marius Barbeau. Chicago came in 1925, where he was promoted to Professor of Anthropology and General Linguistics; Hupa and Navaho fieldwork followed, a second marriage to Jean McClenaghan, and in 1931 the Yale Sterling Professorship. A Rockefeller seminar on culture and personality occupied 1932–33; heart attacks struck in 1937–38; he died in February 1939. The honors are listed—an honorary doctorate from Columbia, the National Academy of Sciences, presidencies of the American Anthropological Association and the Linguistic Society of America—alongside anecdotes: Jenness’s story of Sapir holding children spellbound with three scraps of paper on human races, Hooton on his shyness and charm. His belief in human rights, his interest in Jewish affairs, and his late study of the Talmud round out the portrait, together with unpublished projects: Tocharian notes, a possible Sinitic–Na-Dene relationship, ethnological materials on Nootka, Yana, and Hupa, and the unfinished The Psychology of Culture. Language (1921), dictated in two months from jotted notes, is noted as a revelation to Hjelmslev. The volume excludes passages from Language, reproduces phonetic orthography as originally published, and thanks Jean Sapir, Philip Sapir, and Mary Anne Whipple; the bibliography was compiled by Leslie Spier and Philip Sapir. A 1941 memorial volume, Language, Culture, and Personality, had been edited by Spier, A. Irving Hallowell, and Stanley S. Newman. One psychiatrist called Sapir “an intoxicating man”; Hooton said his personality shed luster on the title “anthropologist.”
The Plan of the Collection
The table of contents lays out three parts: Language, with subsections on the nature of language, American Indian languages, and Indo-European and Semitic languages; Culture, covering a general view, American Indians, and literature and music; and The Interplay of Culture and Personality, followed by bibliographies of scientific papers and poems. The preface to Part One describes “The Nature of Language” as written for the Encyclopedia of the Social Sciences in 1933—concise yet covering the whole field, including linguistic drift and structural classification, phoneme terminology, psychological aspects, and international language. “Sound Patterns in Language” (1925) and “The Psychological Reality of Phonemes” (1933) stress configurational approaches and attack mechanistic, atomistic analysis; the former urges getting behind sense data to grasp intuitively felt forms, the latter opens with the theme that no entity can be defined as the mechanical sum of physical properties, drawing on Southern Paiute, Sarcee, Nootka, and English. “A Study in Phonetic Symbolism” (1929) uses experiment to show that symbolism constellates by unconscious or intuitive logic not necessarily based on normal functional experience. The 1931 phonetic-law paper uses Indo-European, Algonkian, and Athapascan materials to show regularity of change and conservatism within a pattern.
Sound Pattern as the Key to Sound Change
Sapir declines to address the general nature of sound change, arguing only that one must understand how a sound patterns to understand its history. Mechanical changes can readjust phonetic pattern and create new configurations—modern Central Tibetan b-, d-, g- against classical mh-, nd-, -ng’b-—but pattern feeling also hinders or stimulates change, so universally valid sound changes under like articulatory conditions cannot be assumed. Typical mechanical tendencies exist (nb > mb, -az > -as, tya > tsa), yet a complete theory must account for the orientation of sounds. In English, θ is felt close to f, not s; in Spanish, θ is near s, not f. Hence in Germanic θ never becomes or proceeds from s, while in certain Spanish dialects θ passes into s, and in Athabaskan θ often proceeds from s. English θ tends to vulgarize to f, never to s; Old Norse θ became t in Swedish and Danish. Such facts resist simple mechanical explanation.
Phonetic patterning also explains why people cannot pronounce foreign sounds they already possess: a Nootka Indian substitutes n for English ŋ and l, though he pronounces both in chants and songs, his feeling for stylistic character and the n-l equivalence preventing correct hearing. A speech sound is material for symbolic expression in linguistic context, not merely articulation or acoustic image. English j, ŋ, and ts occur only non-initially; tsa- and ŋa- are felt incredible, but ja- is easily acquired, its pattern mastered by implication. Sapir criticizes inserting phonetic graphs into descriptive grammars as confusing linguistic structure with a method of study, comparing it to inserting anecdotes into morphology or a painter transferring his unrevised palette. Phonetic phenomena are not physical phenomena per se; one must get behind sense data to grasp intuitively felt forms.
The Phoneme Against the Physical Sound
The second paper defines the phoneme as a functionally significant unit in a language’s sound pattern, distinct from the objectively definable sound. No entity in experience is a mechanical sum of physical properties; irrelevant properties are overlooked while one property may carry disproportionate sign value. The phoneme-sound distinction parallels defining a club: wood and shape do not define it; relational fitting into the total system does. Some linguists wrongly confine the phoneme to abstract discussion; Sapir calls this the reverse of realistic—the phonetician’s reduction of speech to physical processes is the true abstraction, while naive speakers actualize phonemes. Phonemic errors of naive speakers may be valuable evidence of dynamic phonemic structure. Teaching natives to write their languages reveals this: teachers project their own phonemic valuations and befuddle natives.
Five examples follow.
Chinookan Morphology and the Accentual Puzzle
Sapir’s analysis of Chinookan turns on a fine-grained puzzle of accent and form. The pronominal element combined with the transitivizing -gE- and with the indirective -grEZ-/-grEm- ought, on the surface, to behave in parallel fashion. Yet the indirect object is normally accented before the postposition -l- (“to, for”), producing forms like inia’lut, “I gave it to her,” and lA’loc, “it was to her.” If -l- and -gEl- were truly parallel, a’-l- (“to her”) and a’-gEl- (“for her”) should stand side by side; instead the language gives a’-l- but a-gE’l-. The conclusion is that the two are not morphologically parallel at all: -grE- displaces the pronoun and pulls the accent onto itself, a fact confirmed by its regularly voiced g rather than a voiceless k or affricated kx.
From this Sapir reconstructs an old demonstrative stem, -gd’-, predicatively related to the preceding pronoun, with the postposition attached enclitically. A cluster like i-gEl-, “for him,” originally carried the sense “him—that one—to, for him”; transitive clusters like t-gE’-n-, “they—me,” meant “they—that (it is)—me.” The old intransitive/transitive distinction was thus a matter of sentence idiom: “He goes” against “he that is who kills her.” This deictic -gd’- proves to be a reduced form of post-accentual -ka, the element of “only, just” visible in numerals, pronouns, and adverbs—Wishram i’xt-ka (“just one”), na’i-ka (“I”), a’x-ka (“she”); Lower Chinook na’m-ka (“I alone”), e’-ka (“thus”), kawa’t-ka (“soon”). All of these—ka, ka-, -ka, the tense prefix ga-, and -gE-—are reflexes of a widespread demonstrative ka (“that”) found across Penutian: Coos -ka, Takelma ga, Yokuts ka, Miwok i-ka, and Tsimshian -grE.
Stress, Fusion, and the Sentence-Word
Two forces, Sapir argues, shaped Chinookan’s irregular morphology: a strong movable stress accent, and the vowel loss and fusion that followed it, welding originally independent elements into complex assemblages—an old analytical sentence petrifying into a synthetic sentence-word. The shift of 9 to tc antedated the incorporation of certain elements into the verb, and the tense prefixes came late, since Lower Chinook tense prefixes differ considerably from the upper dialects; old temporal particles coalesced with pronominal prefixes separately in each dialect. These elements never fused phonetically with the early transitive forms for “he” (itcd-) and “she” (ag9-). In Wishram i-g-i’-ux, “she made him,” the prefix i- does not palatalize -g- to -tc-, because it entered the verb complex after the palatalizing force was spent; had it been prefixed while g- was still agd-, ig-agd’- would have arisen. Likewise Lower Chinook atcE’tax, “he made them,” arose before the tense prefix a- joined the complex, since a- could not have palatalized original -gd-.
Phonetic Drift Across the Northwest Coast
Sapir then widens the lens. Similar sounds and processes can characterize independent languages within a continuous area, pointing to phonetic influence through bilingual speakers. The change of g/k to dj/tc by palatalizing i/e is too general to carry much weight, but the shift of k-sounds to tc-sounds occupies a nearly continuous stretch from northern Vancouver Island south to the Columbia mouth. All Nootka dialects turned original Wakashan palatal k-sounds—preserved in Kwakiutl—into palatal sibilant affricatives; Kwakiutl g- and k- appear in Nootka as tc. Many Salish dialects made the same alteration, though Sapir, citing Boas, cannot map its distribution. Lower Chinook’s k- corresponding to ordinary Wishram k- is a comparatively recent dialect development before i-vowels, unrelated to the older Chinookan change of g to tc.
Male and Female Forms of Speech in Yana
The notes break off just as Sapir turns to Yana, where some languages distinguish male from female forms of speech—a thread taken up in the pages that follow.
Judeo-German Against Middle High German
The comparative study of Judeo-German opens with a challenge to the claim that final voiced stops survived as sonants. Sapir explains them by analogy, not phonetic retention. Middle High German tac/tages/tage was levelled to tag/tages/tage; when final -e dropped, g could not become surd, giving Judeo-German tog against Modern German tak. So too veg from wec, by analogy with weges/wege. He cites op from ahe, where no paradigmatic levelling occurred and final h became p by regular law, and avek (“away”) as adverb against veg as noun—the adverb, unassociated with the noun, underwent no levelling. Such frequency of final voiced stops is foreign to German dialects but paralleled in English and Swedish.
The Vowel Scheme
Sapir then sketches the main lines of vowel change, treating each Middle High German vowel in turn and explaining apparent exceptions by paradigmatic analogy, umlaut, lengthening in open syllables, broadening before r plus consonant, unrounding, monophthongization, and diphthongization. a remained in closed syllables but lengthened to â (then open o) in open ones, with analogical o in closed syllables explained through the tac/tage paradigm; â regularly became o, sometimes preserved where Modern German shortened to a (nox against nach)
Specialization and the Age of Culture Traits
Sapir opens his treatment of American Indian cultures by asking how a trait’s elaboration can betray its age. Great specialization of form or technique, he suggests, points not to the antiquity of the particular form but to the antiquity of the type of action behind it. The Chilcat blanket, whatever its present shape, presupposes a long development out of simpler origins; Tlingit weaving of mountain-goat wool must reach back to respectable antiquity. The argument works best when the same culture complex is compared across tribes: as a rule, the complex is oldest where it is most elaborated. The sung myth of the Yuman tribes of the Colorado, far richer than the restricted dialogue portions found among the Southern Paiute, suggests the Paiute borrowed the practice from the Yuman; the intensive agriculture of the Iroquois, set against their Algonkian neighbours, implies the latter learned the art later.
Survivals and Their Limits
The seventh method of chronological reconstruction is the study of survivals, a tool the evolutionary ethnologists used almost abusively. A genuine survival is not a remnant conjured from general theory but an obscure or isolated belief, custom, or myth-episode standing out of context, its full content seemingly lost and no longer in intelligible relation to the rest of the culture. Such elements are typically customs or beliefs blindly accepted without rationalization; taboos often belong here. The nucleus of the Nootka puberty rite for girls—rigidly prescribed ceremonial acts whose meaning the Indians no longer understand and make no attempt to explain—is a survival complex, undoubtedly older than the rest of the ceremonial. A survival may also hark back to a superseded practice of daily life, such as ceremonial entry into the house through the smoke-hole. Yet survivals can deceive: an element borrowed from a tribe where its setting is plain may, detached, look like an isolated relic of great age, or may seem a survival merely because the descriptive data needed to explain it were never recorded.
Distribution as a Measure of Time
The most powerful method, in Sapir’s view, infers chronology from geographical distribution. Distribution is generally continuous, halting at barriers like mountain ranges or deserts, sending spurs along navigable streams and coast lines, and tending toward a compact land mass with a more or less defined centre where the element is most elaborately or most typically developed. Agriculture in aboriginal America spreads from the heart of South America through Central America and Mexico into the Pueblo country and on through the gulf region and Mississippi valley, with its centre probably in the valley of Mexico. The quadrangular wooden house on corner posts reaches from the Tlingit of southern Alaska down to northwestern California, centred perhaps on the coast of southern British Columbia. The Sun Dance, an elaborate but clearly defined ritual complex, is found among all typical Plains tribes and their neighbours east and west, its centre seemingly among the Arapaho and Cheyenne.
Diffusion, Centre, and Periphery
The reasoning runs that a phenomenon originated once in its area and spread by gradual borrowing from tribe to tribe, the borrowed element undergoing associative influences until it appears least typical at the periphery and most typical, or historically oldest, at the cultural centre. Sapir concedes the complications: spread may be faster in one direction, so the culture centre lies far from the geographical one or even near a powerful barrier; the oldest form may be so modified at the centre that it looks more typical at a distance; movements of population may distort the picture. Still, the general validity of the theory of culture diffusion must be granted. The tribe at the cultural centre developed the element first; those removed were affected later, those at the periphery last. The Carib and Arawak, and the Pueblo Indians, probably became agriculturists considerably later than the advanced peoples of Mexico; the Mandan and Iroquois later than the Pueblos; outlying Southern Paiute and southern Ojibwa bands only recently. The quadrangular house of the Hupa and Yurok represents a later period of diffusion than the more elaborate structures of the Kwakiutl. Roughly, the larger the territory a trait covers, the older the trait: agriculture is probably earlier in America than the massive quadrangular house, and both are certainly older than the Sun Dance.
Language as Evidence for Migration
Linguistic evidence, Sapir argues, can fix the direction of population movements, and Athabaskan is his neat instance. Athabaskan languages occupy three isolated areas—a vast northern region from interior Alaska toward Hudson Bay, a Pacific strip in southwestern Oregon and northwestern California, and a southern area in Arizona, New Mexico, and western Texas. Geography alone would treat the three dialect groups as equivalent major divisions, but the southern and Pacific groups are each clearly homogeneous, while nothing shows the northern dialects form a single equivalent group; the divergence between Carrier and Loucheux is no less profound than between Chipewyan and Navaho. The historical centre of gravity therefore lies in the north, and the other two regions were peopled by southward-moving tribes. The case is clinched by the fact that Athabaskan is one of three major divisions of Na-dene, alongside Haida and Tlingit on the northwest coast—so the stock’s centre of gravity is emphatically northern, and it is impossible to think of Athabaskans spreading north from California or the southwest. Sapir adds a counter-example: the theory of Eskimo dispersion from west Hudson Bay would not have won such ready acceptance had linguistic differentiation been appreciated, since Eskimo divides sharply into Eskimo proper and Aleut, Aleut is confined to Alaska, and many distinct Eskimo dialects are spoken there—placing the earliest centre of dispersion in Alaska.
Contact Written into Grammar
Striking phonetic and morphologic similarities between neighbouring, genetically unrelated languages must, where not fortuitous, stem from assimilatory influence—speech habits carried over when speakers acquire a displacing foreign language, or peculiarities created in one language on the model of another through bilingual individuals. On the West Coast, velar consonants, voiceless laterals, and glottalized stops recur across Na-dene, Tsimshian, Kwakiutl-Nootka, Chemakum, Salish, Chinookan, Lower Umpqua, and Coos; morphological parallels include sex gender in Coast Salish and Chinookan, numeral classifiers and reduplication in Tsimshian and the Kwakiutl-Nootka group, and instrumental verb prefixes across Maidu, Shoshonean, Washo, and Shasta-Achomawi. Such phenomena presuppose very long contact, and so indicate the relative age of contact or of former contact now disrupted. Tsimshian, though culturally closer to the Pacific-coast Na-dene, shows morphologic resemblances to the southern languages that point to much earlier contact with Kwakiutl and Salish than with Haida and Tlingit—making the bulk of the culture peculiar to Haida, Tlingit, and Tsimshian much more recent than that earliest contact.
The Phoneme as a Psychological Fact
Sapir’s linguistics rests on a distinction that looks technical but carries a whole philosophy. The physical sound a speaker produces is not the unit the mind hears. What speakers actually operate with is the phoneme—a functionally significant unit defined by its place in a system, not by its acoustic properties. He makes the point through his interpreters and informants: Tony, a Southern Paiute, reconstructs a word etymologically when asked to divide it into syllables; John Whitney, a Sarcee, insists two words differ by a “t” that no instrument can find, and turns out to be right about a real phonological contrast; a Nootka pupil writes glottalized stops and glottalized sonants with the same mark because, to him, they belong to one class. Naive speakers, Sapir argues, are the true realists. It is the phonetician who reduces speech to physical processes who is guilty of abstraction.
The same principle governs his account of sound change. Whether a language shifts one consonant toward another depends on how that sound patterns in the system—English θ drifts toward f, Spanish θ toward s, because each language feels the sound differently. Sapir uses this to attack the assumption that identical articulatory conditions produce identical changes everywhere. He also insists that phonetic systems bear no correlation with physical environment or social life: neighboring peoples with closely related cultures can differ radically in sound, and unrelated languages sharing a region can converge. Language, he concludes, changes subconsciously and conservatively while culture changes consciously and fast, so a language may preserve the imprint of a culture long vanished.
Form, Causation, and the Completeness of Every Tongue
Against the practical American temper that saw language as a mere tool for getting thoughts across, Sapir argued for the study of language as form. Every language, whether Eskimo or Hottentot, is formally complete: it possesses a full phonetic system and a set of prepared forms into which any expressible experience can be fitted without inventing new structures. Moving between languages, he suggests, is like moving between coordinate systems—the world is the same, the orientation differs. Vocabulary is another matter. Eskimo lacks a word for “causation,” but that is a gap in the speakers’ culture, not in the language’s form; the language can express the causative relation perfectly well, and many “primitive” languages do so with a relentlessness modern European tongues have lost. Nootka, he notes, has no verb without its precise causative counterpart. The capacity to feel and express such relations is unconscious and develops early; the ability to conceive causality abstractly is conscious, laborious, and late. Primitive people may express what they cannot consciously formulate.
Culture, the Individual, and the Psychiatric Turn
The essays on culture and personality carry the argument into anthropology and psychiatry. Sapir rejects the notion of culture as a superorganic whole floating above persons. Its true locus, he writes, is in the interactions of specific individuals and in the meanings each abstracts from those interactions; “society” is itself a construct people use to interpret their own behavior. Every individual represents at least one sub-culture, often several. This leads him to a sharp critique of psychoanalysts who treated neurotic symbolism as a regression to primitive racial inheritance. The resemblances between neurotic ritual and primitive ritual are real, but they exist because all humans are psychologically primitive at bottom, not because civilization inherits archaic ideas. Anthropology’s gift to psychiatry is a healthy skepticism about “normal behavior”—the discovery that adjustment takes an enormous variety of forms.
His essays on religion and on the group extend the same temper. Religious sentiments—community with a necessary universe of values, the feeling for holiness, the sense of sin—persist in sophisticated people long after their rational justifications lapse, and Sapir treats them as psychologically necessary rather than as errors to be corrected. His analysis of groups sorts them by whether they are defined physically, by purpose, or by symbolic role, refusing to let a single abstraction do the work.
The Legacy of a Restless Mind
Sapir’s influence ran through several disciplines at once. Boas credited him with shaping phonemic principles; Kluckhohn said his psychiatric insights reshaped anthropological postulates; Hjelmslev called Language, dictated in two months from jotted notes, a revelation. The culture-and-personality movement that flourished in American anthropology after his death owed much to his insistence that culture lives in individuals. His unfinished The Psychology of Culture was completed posthumously by students, and a memorial volume appeared in 1941. Colleagues remembered a shy, charming man who could hold children spellbound with three scraps of paper, and who late in life turned to the Talmud. The essays here remain the clearest record of how one scholar made language, culture, and mind into a single field of inquiry.
