Edward Sapir’s Language: An Introduction to the Study of Speech is a compact treatise on what language is, how it works, and why it matters — written not as a catalogue of facts but as an argument for a particular way of seeing. Published in the early 1920s, it grew out of Sapir’s work as a linguist and anthropologist, and it wears its learning lightly: the preface promises almost no technical symbols, not a single diacritical mark, and English examples wherever possible. Across eleven chapters the book moves from the definition of speech through phonetics, grammatical form, linguistic classification, and historical change to the relations between language and race, culture, and literature.
“Knowledge of the wider relations of their science is essential to professional students of language if they are to be saved from a sterile and purely technical attitude.”
Sapir’s central claim is stated plainly in his preface: the aim is to give a perspective on language rather than to assemble facts. He wants to show what he conceives language to be, how it varies across place and time, and how it connects to thought, history, race, culture, and art. The book’s authority rests less on exhaustive description than on the diagnostic value of linguistic form — valuable precisely because linguistic structure is unconscious and unrationalized, and therefore an unusually honest record of how the mind works.
From the Illusion of Instinct to the Pattern of Thought
The Preface and the Book’s Design
Sapir opens with a short preface, signed from Ottawa in April 1921, that sets the temper of everything to follow. He is not out to compile facts about languages; he wants to give readers a perspective on what language is. The book will say little about the ultimate psychological basis of speech and will supply only enough descriptive and historical material to illustrate principles. Its real business is to show how language varies across place and time and how it stands in relation to thought, history, race, culture, and art. He hopes the view will serve both professional students and that outside public which tends to dismiss linguistics as pedantry, since a sense of wider relations saves the specialist from a sterile, purely technical attitude. Among influential writers on liberal thought, he singles out Croce as one of the few who grasped language’s fundamental significance, above all its closeness to art, and he acknowledges a deep debt for that insight. Linguistic forms and processes, he adds, have diagnostic value for the psychology of thought and for the cumulative drift we call history or progress, precisely because linguistic structure is unconscious and unrationalized. He has avoided technical terms and symbols, uses no diacritical marks, and draws on English wherever possible, quoting exotic examples without apology. Friends who read the manuscript are thanked: Kroeber and Lowie at California, Wallis at Reed, and Zeitlin at Illinois. A contents list then lays out eleven chapters, from “Language Defined” through the sounds of speech, grammatical form and concepts, types of structure, drift, phonetic law, the influence of languages on one another, and the closing chapters on race, culture, and literature.
Speech Is Not an Instinct
The first chapter takes aim at the comfortable assumption that talking is as natural as walking. It is not. Walking is a biological function; the child is equipped by heredity to make the necessary muscular and nervous adjustments, and culture has almost nothing to do with it. Language is different in kind. A child is destined to talk only because he is born into a society that will lead him into its traditions. Remove society and he will still learn to walk, if he survives, but he will never learn to talk—never learn to communicate ideas according to a particular society’s traditional system. Transplant a newborn into an alien environment and walking develops much as before, while speech will be wholly at variance with the speech of his native place. Walking varies only within narrow limits, involuntarily and without purpose; speech varies without assignable limit between social groups because it is a purely historical heritage, the product of long-continued social usage, shifting as religions, beliefs, customs, and arts shift. Walking is organic and instinctive; speech is non-instinctive, acquired, cultural.
Cries, Interjections, and the Argument from Art
The obvious objection is the involuntary cry—the sound wrung out by pain or joy. Sapir grants that such utterances are instinctive, but insists they are non-symbolic: the cry does not stand apart and announce that an emotion is felt; it is an automatic overflow of emotional energy, part and parcel of the emotion itself. Such cries hardly communicate at all. They are not addressed to anyone, merely overheard, like a dog’s bark or footsteps or the wind; if they convey ideas, they do so only in the vague sense in which any phenomenon may. To treat “Oh!” as a speech symbol equivalent to “I am in great pain” is as allowable as reading the clouds as a symbol for “It is likely to rain”—a definition stretched until it means nothing. Conventional interjections like oh!, ah!, sh! must not be confused with instinctive cries. They are conventional fixations of natural sounds, differing widely from language to language according to each language’s phonetic genius, and they belong to speech in the cultural sense—no more identical with the cries than “cuckoo” and “killdeer” are with the birds’ calls, or than Rossini’s storm in the William Tell overture is a storm. Interjections and sound-imitative words stand to natural prototypes as art stands to nature. Family resemblances among interjections across languages prove no common instinctive base, any more than national styles of painting do: a Japanese picture of a hill and a modern European painting of the same hill both imitate the same natural feature, neither grows directly out of it, and they differ because of differing historical traditions and techniques. So Japanese and English interjections are suggested by a common prototype, yet diverge because they are built from historically diverse materials—linguistic traditions, phonetic systems, speech habits—while the instinctive cries themselves are practically identical for all humanity, like the human skeleton. Interjections, moreover, are among the least important elements of speech; the discussion matters mainly to show that even they, avowedly nearest to instinctive utterance, are only superficially instinctive.
Transfers of Speech Symbolism
From here Sapir turns to the many reduced and transferred forms of speech symbolism. There is “talking to one’s self,” or thinking aloud, where speaker and hearer are the same person; and there is silent speech, where sounds are not articulated at all. In silent thinking the auditory centers alone may be excited, or the impulse may reach the motor nerves and be inhibited in the muscles, or the auditory centers may be barely touched while the process plays out in the motor sphere—fatigue in the speech organs after intensive reading or thinking shows how common such motor excitation is. All these modifications are patterned on normal speech. The symbolism can also be transferred into other terms. Watching a speaker’s articulations makes a visual-image symbolism possible, though of limited value since not all articulations are visible; lip-reading by the deaf is a subsidiary method. The most important visual transfer is the written and printed word, with its accompanying writing and typewriting movements. Each written element answers to a specific spoken element; written language is a point-to-point equivalence to its spoken counterpart, “symbols of symbols,” yet so close that eye-readers may substitute it for spoken forms, with auditory-motor associations remaining latent. Visual symbols, he suggests, are to auditory symbols what money is to economic goods. Morse code transfers from written speech, making each letter a symbol of a symbol of a symbol, and operators may come to think directly in tick-auditory or tactile-motor terms. Gesture languages—of deaf-mutes, Trappist monks, parties in sight but out of earshot—are either one-to-one equivalences or imperfect transfers, like military gestures or the Plains Indians’ sign language understood across tribes of mutually unintelligible speech. Even vaguer symbolisms, at sea or in the woods, are intelligible only through silent translation into fuller speech. All voluntary communication aside from normal speech, then, is a transfer from, or involves, truly linguistic symbolism; auditory and correlated motor imagery are the historic fountain-head of all speech and thinking. The ease of these transfers shows that sounds are not the essential fact of language. The essential lies in classification, formal patterning, and the relating of concepts—language as the mold of thought.
Universality and Diversity
No people lacks a fully developed language.
Pitch Accent as a Grammatical Instrument
Sapir turns to pitch, arguing that differences of tone can carry grammatical weight just as stress does. The mere fact that pitch is phonetically essential in a language proves nothing by itself: in Chinese feng with a level tone means “wind” while the falling tone means “to serve,” and classical Greek distinguishes lab-on from gunaik-on, yet in such cases the tone belongs to the radical element or affix and is not doing grammatical work. Genuine grammatical pitch shows up where tone alone shifts a word from one category to another—Chinese chung level, “middle,” against chung falling, “to hit the middle”; mai rising, “to buy,” against mai falling, “to sell.” Such pairs are not common, and Chinese has no firm feeling for a tonal noun-verb distinction. In the Soudan, however, tone is fundamental: Ewe builds its reduplicated infinitive suhosubo and adjective suhosuho from suho, “to serve.” Shilluk uses tone to separate singular from plural (yit high, “ear,” versus yit low, “ears”), to mark three pronoun forms, and to distinguish forms of gwed-, “to write.” Tlingit on the southern Alaska coast puts the past in a low tone and the future in a high one; Takelma opposes falling hel, “song,” to rising hel, “sing!” Sapir’s conclusion is modest but firm: pitch is grammatical far more often than the habits of English speech would lead us to suppose.
Thirteen Concepts in One Sentence
To show how much grammar hides inside an ordinary remark, Sapir takes apart “the farmer kills the duckling.” Three concrete concepts stand out—farmer, kill, duckling—but farmer is really farm plus the agentive -er, and duckling is duck plus the diminutive -ling; these are derivational mediators, not separate ideas. Swap in man or chick, or takes for kills, and the structural mold holds: the same relational concepts appear three times over, in the relational word the, in the sequence of subject, verb, and object, and in the suffixed -s. Drop the and the sentence collapses; the and a fix definite against indefinite reference. “Farmer, kill the duckling” turns the statement into a vocative command. The farmers kill shows that -s carries singularity, personal relation (neither speaker nor hearer), and present time all at once. Reorder the words and the sentence becomes a question; interchange farmer and duckling and subject and object trade places. Counting it all up, Sapir finds thirteen concepts: three radical, two derivational, eight relational—reference, modality, personal relations, number, time. Function and form refuse to line up: the same suffixing serves both derivational and relational ends, one relational concept may be expressed twice, and a single -s weaves four relations together.
The Same Thought in German, Yana, and Chinese
Set beside English, German Der Bauer totet das Entelein binds definiteness to number, case, and gender, all borne by the articles, and splits “kill” into tot plus a causative vowel change. Yana arranges the sentence as “kill-s he farmer he to duck-ling,” its suffix vouching for the statement rather than marking number, and it has no gender at all. Chinese says “Man kill duck,” relying on radical words and position alone, and given the context that is enough. English, German, Yana, and Chinese all leave place and demonstrative relations unexpressed—an elaboration that would come naturally to a Kwakiutl speaker.
What Speech Cannot Do Without
Every proposition needs at least two radical concepts, Sapir argues, and relational concepts that make syntactic relations unmistakable; who is doing the killing cannot be dodged. Secondary relational concepts such as number and gender resemble derivational ones, yet in English, French, German, Latin, and Greek number “infects” other words through agreement. Latin illa alba femina quae venit packs four concepts into a word, though logic demands only case. French speakers feel no sex in un arbre or une pomme; English speakers feel no vivid present in “He comes,” which serves for future and general activity alike. Form outlives conceptual content, as drove and sank beside killed show, and the French or German choice between preterit and perfect is mere style. Language imposes rigid classifications—English’s three times, French gender, the Athabaskan and Yana verbs that differ for carrying pebbles and carrying logs—a “dogma of the unconscious.”
Form Without Function
A third source of non-significant form is mechanical phonetic process, which produces formal distinctions with no functional counterpart and accounts for much irregularity in declension and conjugation. In hats the -s truly symbolizes plurality; in selves a g-sound combines with the change of radical f to v—not a falling together of originally distinct concepts, as presumably with drove and worked, but a mechanical manifold of one formal element with no new conceptual growth. Sapir then revises his scheme into four types. Basic (concrete) concepts—objects, actions, qualities—stand as independent words or radical elements and involve no relation as such. Derivational concepts, less concrete, are affixed or worked by inner modification; they are irrelevant to the proposition as a whole but add an increment of significance to the radical, as -er marks the habitual subject and -ling a relation of attribution. Concrete relational concepts, more abstract yet not wholly devoid of concreteness, are affixed at greater remove and imply relations transcending the single word. Pure relational concepts are wholly abstract, relating concrete elements into definite syntactic form. Material content is types one and two; relation is types three and four. These schemes, he warns, must not be worshipped as fetishes; in exotic languages one may analyze words and still lack the inner feel that separates material content from relation. Types one and four are essential to all speech; two and three are common but avoidable, and three especially is a confusion of two and four. Speech’s illogical, metaphorical genius spans the gulf between the concrete and the purely relational with a continuous gamut, and a concrete concept may slide into the relational sphere while keeping its independence—as in Chinese and Cambodgian, where the verb “give” becomes the symbol of the indirect objective: “We make story this give all that person who have child.”
Full and the Drift Toward Affixhood
Transitions run between types one and two, one and three, and less radically two and three.
The Case Against “Whom”
Sapir opens with a puzzle in English grammar: the slow death of whom. The objective forms that survive distinctly are me, him, her, us, them—and whom. The first five cluster into a solid objective group standing parallel to I, he, she, we, they. Whom stands alone, “psychologically isolated,” because the other interrogative and relative pronouns—which, what, that—draw no distinction between subject and object. A form group ought to mirror a function group, and whom violates that symmetry; it looks unesthetic, a pattern with gaps where its fellows should be. Since English has lost the power to coin new objective forms and cannot remodel which, what, that to match, the sensible remedy is to drop whom altogether and let who rejoin its flock.
Three further forces press in the same direction. The interrogative adverbs where, when, how are invariable and emphatic, and speakers feel that interrogative pronouns should be invariable too; the -m of whom drags on rhetorical effectiveness. Position reinforces the objection: the subjective/objective contrast in personal pronouns is bound up with word order—I see the man versus the man sees me—and no one says Him did you see? Yet the interrogative pronoun naturally comes first, so Whom did you see? conceals a conflict between two orders. Did you see whom? is too contrary to idiom; Who did you see? is where the language is heading. Finally, phonetics: the heavy, half-long whom refuses to trip lightly into a following syllable, and whom did sounds clumsy in a way what did and when did do not. Sapir treats this rhythmic instinct as a genuine linguistic determinant, the poet’s prosody being only a finer application of tendencies present in everyday speech.
The uneducated speaker who says Who did you see? without a twinge, Sapir suggests, has a sharper flair for the genuine drift than the grammarians. The four factors do not act separately; their energies canalize into one slight, largely unconscious hesitation. Sapir tests the analysis by elimination, building a scale of “hesitation values”: The man whom I referred to (factors one and three), The man whom they referred to (adding the phonetic factor), Whom are you looking at? (where the following vowel relieves the weight), and Whom did you see? at the top. The prediction is that Who are you looking at? might even sound faintly offensive to ears that welcome Who did you see?
Phonetic Law and the Nature of Drift
From grammar Sapir turns to sound change, asking what a “phonetic law” actually states. The vowel of foot is drifting from a long oo toward a short, open sound, with roof, soot, and hoop vacillating. No law can be stated with certainty in advance; there is a strong drift, and whether every old long-oo word will be swept along depends on conditions—perhaps words ending in voiceless consonants go first. Phonetic laws, he concludes, are not spontaneous automatisms but formulas for a consummated drift that begins at a psychologically exposed point and works through a gamut of analogous forms.
Tables tracing foot and feet, mouse and mice across fifteen centuries of English, with German parallels, show the same story: unaccented syllables reduced, radical vowels modified by following vowels, long mid vowels raised, old high vowels diphthongized. English and German today resemble each other more than either resembles its West Germanic ancestor, and the parallels root in a common pre-dialectic drift. Changes are regular—either mechanical replacement or “splitting,” where one sound divides in two. A single change threatens the phonetic pattern, and the remedy is analogous shifting of the other series: if p becomes b, then t and k move to d and g to restore the symmetry. This readjustment, Sapir says, is probably the most important tendency in the history of speech sounds, the central unconscious regulator. He dismisses “ease of articulation” and “faulty perception” as catchwords—sounds easy for us are hard for others—and admits the primary cause of drift remains unknown.
The foot:feet alternation shows how phonetics and grammar interlock. The o-to-e change was not originally a plural marker; it appeared in the dative singular too, and served unrelated functions in doom/deem, blood/bleed, food/feed. Only centuries later was it reinterpreted to signal number. Sapir detects three strands in phonetic change: a general dynamic drift, a readjusting tendency that preserves the pattern, and a preservative tendency that sets in when morphology is threatened. The Anglo-Saxon mind, he argues, was glad of the chance to shift functional burden onto vowel variation, letting the drift proceed without unsettling morphological contours; the slower German drift explains why German’s preservative change arrived three centuries later.
Diffusion Between Languages
Can sounds and structures cross from one language to another? Sapir gathers cases: Suabian nasalized vowels near French territory; Dutch and Flemish unaspirated stops with a French-like metallic quality; Russian’s dull yeri echoing Ural-Altaic; and, most striking, a glottalized stop series shared by at least four unrelated stocks from southern Alaska to central California. Independent invention is hard to believe. Sounds and manners of articulation, he concludes, spread over continuous areas like culture radiating from a center, entering a language’s drift at the borderlands through bilingual suggestion—provided the innovation aligns with native tendency, since a language preserves its sound patterning, not its sounds.
Morphology proves far more resistant. English’s productive -ize and -able are derivational conveniences, not structural imports; a true morphological borrowing would be English adopting a French-style synthetic future, and no such case can be shown. The old picture of the Normans playing nine-pins with Anglo-Saxon is rejected: Danish developed analytic structure with no foreign pressure at all, and English, flooded with French loanwords, changed fast yet stayed true to its own type. Sapir offers three inferences—morphological influence may be too slow to see in recorded history, the favorable conditions may never have arisen, or languages simply do not easily remold one another. Where striking structural traits are distributed across supposedly unrelated languages—Semitic and Hamitic, Athabaskan, Tlingit, and Haida—he suspects vestiges of a former community of type rather than borrowing, and lacking convincing cases of deep morphological diffusion, he ascribes major concordances to autonomous drift. Language, he says, is probably the most self-contained and massively resistant of all social phenomena; easier to kill it off than to disintegrate its individual form.
The Concepts That Carry the Argument
The book’s intellectual spine is a set of distinctions Sapir builds with unusual care. Speech is not instinctive but cultural: a child is predestined to talk only because it is born into a society that will lead it into a tradition, whereas walking develops regardless. Even interjections, the sounds nearest to raw instinct, turn out to be conventional fixations shaped by each language’s phonetic habits — art, not nature. From there Sapir separates the sound from the significant element: the individual sound is clay, not brick; the real units are words and grammatical elements, radical and affix, and the word itself is the first thing that genuinely “exists” in speech.
His most influential conceptual move is the fourfold scheme of grammatical concepts — basic (concrete) concepts, derivational concepts, concrete relational concepts, and pure relational concepts — with the observation that classes I and IV are essential to all speech while II and III are common but avoidable. This scheme underlies his attack on the old isolating/agglutinative/inflective hierarchy, which he condemns as a classification built on sentimental values: the linguist who treats Latin and Greek as the highest type is like a zoologist who sees the organic world as a conspiracy to evolve the race-horse. Every language, he insists, is a form language; the supposed “formless” languages merely fuse syntactic relations with notions of another order.
Why the Argument Still Matters
Sapir’s book endures because it reframed linguistics as a study of patterning rather than of sounds and rules. The idea of an “inner” phonetic pattern — a language’s ideal sound-system persisting long after its actual sounds change — and the notion of a structural “genius” or basic plan underlying a language’s surface features gave later thinkers a vocabulary for talking about linguistic form as a whole. His account of drift, in which phonetic change, readjusting tendency, and preservative tendency work together to keep a language’s pattern intact while its sounds shift, remains a model of how to think about unconscious historical process.
The closing chapters widen the frame. Sapir argues that language is probably the most self-contained and massively resistant of all social phenomena — easier to kill off than to disintegrate — and that the grand concordances of phonetic pattern and morphology are better ascribed to autonomous drift than to borrowing. Culture and language, he concludes, are not causally related: culture is what a society does and thinks, language a particular how of thought. That distinction, and the book’s final turn toward language and literature, is where Sapir’s perspective earns its keep — a perspective that saves the professional from sterile technicality and gives the outside reader a reason to care about a subject too often dismissed as pedantry.
