Edward Sapir’s The Takelma Language of Southwestern Oregon is a grammatical description of a Native American language once spoken along the middle Rogue River and its tributaries in southwestern Oregon. It appeared as an extract from the Handbook of American Indian Languages (Bulletin 40, Part 2) issued by the Bureau of American Ethnology, and it is built on texts, vocabularies, and grammatical notes Sapir gathered largely from a single speaker. The book is a work of descriptive linguistics, not a narrative: its subject is the structure of a language, set out in the technical vocabulary of early-twentieth-century Americanist grammar.
“There are no known copyright restrictions in the United States on the use of the text.”
Sapir’s material came chiefly from Mrs. Frances Johnson, an elderly full-blood Takelma woman from a village on Jump-off-Joe Creek, whom he worked with at the Siletz reservation over a month and a half in the summer of 1906. He credits Franz Boas with encouraging him to abandon the ordinary rut of grammatical description and to arrange and interpret the facts in accordance with the spirit of Takelma itself. The book’s real arrangement follows the conventional plan of such grammars: an introduction, a long treatment of phonology, a still longer morphology moving through the verb, noun, pronoun, adjective, numerals, adverbs, and interjections, a conclusion, and two appendices—one a comparative table of pronominal forms and the scheme of seven voices in six tense-modes, the other specimen texts with analysis.
What the Grammar Sets Out to Show
The Language and Its Speakers
The book opens as a scholarly artifact before it opens as a study: library stamps from the Huntington Free Library and Cornell, a copyright notice, an archive.org address. Then the title page proper, identifying the work as an extract from the Handbook of American Indian Languages, Bulletin 40, Part 2, of the Bureau of American Ethnology, edited by Boas and printed by the Government Printing Office in Washington. A detailed table of contents lays out the whole architecture to come: phonology, morphology, the verb, noun, pronoun, adjective, numerals, adverbs, interjections, a conclusion, and two appendices—one a comparative table of pronominal forms and a scheme of seven voices across six tense-modes, the other specimen texts with analysis.
The introduction fixes the language on the map. Takelma was spoken in southwestern Oregon along the middle stretch of the Rogue River and its tributaries; together with an upland dialect it forms the Takilman stock of Powell. The name itself is practically identical with the native Da’^gelma’^n, “those dwelling along the river,” and Sapir sees no good reason to replace it with Powell’s variant. His sources are laid out with care: the myth texts he had already published through the University of Pennsylvania in 1909, grammatical material, eleven medicine formulas or charms printed in the Journal of American Folk-Lore, and a vocabulary of stems appended to the Texts. Manuscript notes gathered in summer 1904 by H. H. St. Clair, 2d, for the Bureau were placed at his disposal—mainly lexical, but useful on a point or two. Citations of the form 125.3 refer to page and line of the Texts.
The fieldwork itself was done at the Siletz reservation during a month and a half in the summer of 1906, under the Bureau’s direction. The informant was Mrs. Frances Johnson, an elderly full-blood Takelma woman from the village of DaYts’.asin or Daldani’^k’ on Jump-off-Joe Creek, a northern affluent of the Rogue; her mother came from a village on the upper Cow Creek. Her English was imperfect but she was exceptionally intelligent and good-humored. Sapir thanks Franz Boas for methodological advice and active interest, crediting him largely for the encouragement to abandon the ordinary rut of grammatical description and arrange the facts in accordance with the spirit of Takelma itself.
The Phonetic Character of Takelma
The phonology begins with an impressionistic placing of the language among its neighbors. In relative harshness or smoothness Takelma sits about midway between the rough languages of the Columbia valley and the northern California-Oregon coast—Chinookan, Salish, Alsea, Coos, Athapascan, Yurok—and the relatively euphonious languages of the Sacramento valley such as Maidu, Yana, and Wintun, inclining toward the latter. From the first group it differs chiefly by lacking voiceless l-sounds and velar stops; from the second by somewhat more complex consonant clusters, though these are strictly limited and hardly difficult in themselves. Like the Sacramento languages it has clear-cut vowels and abounds in long vowels and diphthongs, which together with a syllabic pitch-accent give a decidedly musical character—marred by a profusion of disturbing catches. The morphological and phonetic line between Takelma and the neighboring Athapascan dialects of Upper Umpqua, Applegate Creek, Galice Creek, and Chasta Costa is sharp.
The Limits of Phonetic Law
Sapir pauses over the difficulty of stating a phonetic law without comparative or older historical material. The form masga” (aorist third person of the stem masg-) serves as his test case. The final cluster -sk’ does occur in Takelma, as in dink!a”sk’ “long object lay stretched out,” so an irregularity must lie in one of the two forms; either masg’ or dink!asga” would be expected. Closer inspection shows the -k’ of dink!a”sk’ is a grammatical element added to the future stem dinklas-, whereas in masga” the -g- belongs to the stem and is not felt as a suffix. The conclusion: quasi-mechanical juxtaposition of grammatical elements does not follow the same phonetic lines as organic sound-complexes.
He then sorts the inorganic or secondary a-vowels into three kinds: the regular a inserted between two consonants that would theoretically form a cluster; the post-consonantal constant a of certain stems, such as wism-, that would otherwise end in impracticable clusters (this -a appears as -i under circumstances discussed later); and a connecting a joining consonantal suffixes to preceding consonants.
Double Diphthongs and Their Simplification
A “double diphthong” is defined as a syllable made of an ordinary diphthong, long or short, followed by a semivowel or by l, m, n—aiw, any, ain, alw, and the like. Those with initial short vowel have a value of three morae; those with initial long vowel have four and are called over-long. Such syllables are nearly always avoided, evidently from a rhythmic feeling, a dislike of heavy syllables containing three qualitatively distinct sonantic elements. They are regularly simplified by losing either the second or third element, shedding one mora. The exception is double diphthongs in -uy, which become dissyllabic by vocalizing y to i: ts!isiwi’k’ “he ran fast” beside k!’a-uya’s “fast runner,” yawi’t’e “I shall talk” beside yawsi’ja “you talked.” Double diphthongs ending in -w simply lose the -w, as in gal “eat it!” and galk’ “he ate it,” beside go’wa’n “I shall eat it.” All others are simplified by loss of the second vowel or consonant, with a glottal catch after it always preserved—gelhewe’ha’n “I think” beside gelhewe’hsin “he thinks,” imi’na’n “I sent him” beside imi’liam “he sent him,” and a run of parallel forms showing the connecting a.
Is Takelma Inflectional?
The grammatical discussion turns to classification. Sapir rejects the usual definition of inflection as involving words of indivisible psychic value in which stem and grammatical elements “chemically” coalesce. Multiplicity of phonetic symbols for a single concept is often cited as proof of inflection—Latin’s many plural endings—and stem variability, as in German hund- and hand-, is also adduced; both are far from a definition but usable as approximate tests. Takelma passes both. Possession by another is expressed by -xa, -a, -da (-fa), -t’, or nothing; “I” as subject by -t’e’ in the aorist intransitive, -t’e’ in the future intransitive, -n in the aorist transitive, -n in the future transitive, and -k’a’ in all inferential forms.
The Etymological Fog of Type 3
Sapir’s catalogue of verb types runs into a descriptive snag at Type 3. When the first consonant after the stem-vowel is a semivowel, liquid, or nasal, the line between Type 2 and Type 3 dissolves: a form like p’elega’^n, “I went to war against them,” can be read either as a Type 3 insertion of -e- into p’eleg- or as a Type 2 addition to a base p’el-g-. The distinction, he concedes, is historical rather than descriptive. What actually characterizes the Takelma aorist is a simple habit: the stem-vowel repeats itself immediately after the first consonant that follows it. Type 3 splits by quantity into short-vowel stems (t.’amy-, tslawy-, malg-, p’elg-, hants!-) and a smaller long-vowel group (uits’!-, gilib-, k’.alas-), and both aorist vowels come out short even when the stem-vowel is long—only rarely does the second stay long, as in leye^s- or ilyilHs’!-. Accent obeys Type 3 rules, rising or raised in dowalt’e^ and htn-giwip’, falling in hana’^s and uyu’^s-de^ because of the glottal catch.
Type 4 and Its Intransitive -i-
Type 4, built as vowel plus consonant plus vowel in the stem and vowel plus consonant plus vowel plus -i- in the aorist, is intransitive, and the -i- seems bound up with that intransitive character. It drops in the causative in -n- (t’sLga.’^na’^n, “I make him cry”) yet survives in other transitives (Va.g&yagwa’^n, “I cry having it”)—a contradiction Sapir resolves by pointing to logical rather than formal transitivity. Two subclasses, 4a and 4b, attach to Types 2 and 1 respectively, with the -i- appearing only in the aorist: je’m’e’ “I shall return” against yeweK’e^ “I returned,” and loho’i’ “dead” against lohoife^ “I died.” The aorists carry a rising accent on the i-diphthong, and na- misbehaves elsewhere too.
Type 5, the Mirror of Type 2
Type 5 is morphologically awkward—effectively Type 2 turned inside out. The proportion yana- “go” : t.’an- “hold” ought to read yd”n- : t.’ana-, but phonetics would prefer the first and third terms reversed. The examples come in future-and-aorist pairs: wa^giwi’^n “I brought it to him” beside wageiwi’n “I shall bring it to him”; janfe^ “I went” beside yana’i’e^; lap’ “he carried it” beside laba’ “carry it!”; se^ba’^7! “I roasted it” beside sehe’n. Both stem-vowels stay short, the second regularly raised in imperatives (yana”, lebe”, odo^, woo^), and the long stressed aorist vowel takes rising accent—except odo- “hunt for,” which keeps a falling o” in the aorist (o’H’, never **o”f*) while its other forms follow the type. Sapir wonders whether two originally distinct paradigms, oM-/odo^- of Type 5 and o’^’d-/d’H!- of Type 6, were levelled together. A footnote adds that forms like lebe’n, with falling accent on the second vowel, only seem to contradict the rule, since falling accent regularly accompanies the personal ending.
The -w- and the Constant -a-
The -w- of transitive verbs looks at first like a marker of the indirect object, as wd°gim’^n “I brought it to him” suggests. Sapir doubts it: the -w- also stands beside the transitive connective -x- (al-sgalawi’^n against al-sgala’xhi^n, “I turn my head to look at you”), so it cannot be the morphological equivalent of -s-, and no distinct meaning can be pinned on either the intransitive or the transitive -w-. A sturdier feature is the “constant -a-”: verbs whose stem ends in two consonants keep an -a- in every form, even where the cluster could stand finally, apparently by analogy with the aorist (s-u^mfch “boil” keeps it because of aorist ts’lumu’^mt’a-, while s-omd- drops it because of s’omod-). This -a- is not the ordinary connective vowel, for it shifts to -i- when the object is not third person, in reciprocals and reflexives, and before non-agentive -x-: swedefgixi “he followed me,” da’^^minlk’^dAxbi “he taught you,” l-lesgi’Tcwif^ “touching himself.”
Repetition as a Grammatical Engine
Frequentatives, continuatives, and usitatives are made mostly by repeating all or part of the stem rather than by suffixing, with repetitive forms favoring a long vowel under rising pitch-accent. Since the shades of meaning cannot be sharply separated, Sapir treats the material formally. Type 13 reduplication builds on the verbal base, not the aorist stem—leb- and loh-, not lebe- and loho-—and functional suffixes follow the reduplicated complex while functionless ones drop away, which makes frequentatives a test for recovering the simple base. The aorist stem enters only where an initial fortis contrasts with a media, chiefly Type 8 (t.’oxox- against frequentative t!ox-o-t!ax-), and such verbs are of mixed type. A small subclass, 13c, omits the -Or- of the repeated base (wa-j&’D.&.gwa’n beside wo-yana-ina^wa’^n). One peculiar form, loha’lhi^’ “they used to die,”
Conditional Sentences and Their Modes
Sapir turns to conditional sentences, where the apodosis may open with the demonstrative ga, “that,” functioning as “then, in that case”—as in a string parting and the speaker declaring himself dead. General conditions of the “whenever” type belong here: arrows backed with sinew, a shaman’s slayer slain in turn, ordinary people killed when vengeance is exacted for a shaman. Contrary-to-fact conditions appear too: “if I knew all, I should tell it to you,” and “if that one had been there, I should not have killed him.” In that last example Tvaga, a demonstrative adverb summarizing the protasis, stands for “in that event, under those circumstances,” and is used with persons other than the third. A footnote adds that general conditions of past time or without time-limit take the subordinate for the protasis and the aorist for the apodosis, both verbs preferably frequentative or continuative.
Potential, Inferential, and the Particle wede
The potential and inferential modes differ from the aorist chiefly in the negative particle they take. Aorist statements are negatived by a’ni^; imperatives and futures cannot be negatived directly and must use the potential or inferential with wede prefixed, so that no morphological line separates a prohibitive “do not go!” from a negative contrary-to-fact apodosis “you would not have gone.” Wede is phonetically identical with the verb-stem wede- “take away,” hinting at a periphrastic origin—”Remove [all thought] that I go.” The inferential marks action not known through personal experience; a Coyote myth example implies the inference “I wasn’t dead, after all—I was really sleeping.” It recurs in rhetorical questions of anger, surprise, wonder, and discovery. One idiom pairs the interrogative gimfne “when, how long” with wede and the inferential to express repeated or continuous action (“he kept on sleeping”), and a similar non-negative future appears in “they never found him again.”
Verbal Nouns, Participles, and Nouns of Agency
Infinitives are formed by suffixes—intransitive -(a^)x, intransitive -k’wa (-gwa), transitive -ia (-ya)—with Class IV verbs taking -¥wa (-gwa), reflexives using -(d)efx- with possessive suffixes, and -cca- verbs using the bare stem with -xa. These verbal nouns may combine object and possessive pronoun, and purpose is normally expressed by infinitive plus the postposition ga^a^l “to, at, for”; some harden into specialized nouns such as p.’ala’Vwa “myth” and sana’Vwa “fight.” The active participle in -t’ is adjectival, marking a characteristic attribute and sometimes nearing a noun of agency—contrast “I am one who has killed his mother” with “I killed my mother.” The passive participle in -(a)k’w, -i^k’w is nominal, takes possessive suffixes, and refers to habitual activity; instrumental forms name implements (saw, pencil, scissors, gun, telescope), and forms in -k’w may derive from third-person possessives of relationship nouns. Passive participles in -xap’ (-sap’) name clothing—handkerchief, belt, pillow, shirt, vest, hat. Nouns of agency use four suffixes: -^s (verbal, incorporating objects, usable as a final clause), -sa’^, -si^, and -xi; the -sa’^ forms often carry disparagement, as in “one who always runs because of fear.” A handful of -i’ya forms from transitive stems—”burnt-down field,” “killed people”—look like attributive passives answering the active -t’.
The Anatomy of the Noun
A sharp formal line separates denominating from predicative elements, proven by distinct pronominal suffix sets; only a few stems serve as both noun and verb (paint/writing, song, shirt, urine), and pitch accent may distinguish the two. A noun may comprise stem, derivative element, noun-characteristic, and further increments. Among derivational suffixes, -p’a forms personal nouns (old woman, orphans, old men, poor people); -i’ appears in place-names (Klamath river, Applegate creek from “beaver,” and others from “elk,” “pine,” “rock”); -a’^{n) forms names of persons from a place, with loss of the derivative -k’—thus a person from Cow Creek, from the Klamath river, from the Illinois river, and the Rogue river Takelma themselves. A footnote argues that Somolu’k’ is better written with its accentual peak on the liquid, and that s-omol- is probably a frequentative of “mountain,” the name meaning “very mountainous region.” The suffix -gw- occurs in body-part terms (pancreas, face, shoulder, medicine-spirit, rectum, shoulder-blade), and -(a)n- is suffixal despite no definite function, as in board, hail, deer, speech, news, white duck. The difference between derivational -n and noun-characteristic -n- is that the former belongs to the absolute form while the latter appears only with grammatical increments—”board” against “in the road,” “my board.” The suffix -a shows in dissyllabic nouns like xu’ma “dry food” from jcu^m “dry,” and -s is probably derivational, especially after a consonant (mortar-stone, blanket, squirrel, moccasin, worm, moss).
Exclusives, Plurals, and the Deictic Apparatus
The exclusive suffix -fa marks selection of one item among mutually exclusive alternatives; with adjectives it seems only to form comparatives and superlatives, since “this is smaller” really means “this is small, not large like that.” It is idiomatic, best rendered by stressing the English word—”boy” as male, not female; the older brother doing nothing while the younger gets into trouble; “that one is biggest.” An inorganic a is added to keep -t intact. The plural suffixes -i’an, -han, -k.’an are never syntactically necessary: Takelma generally leaves number to context, and the plural is indefinite with a third-person possessive, since plurality may refer to possessed, possessor, or both—hence it likely expresses reciprocity, agreeing phonetically with the verbal reciprocal -an-.
A Language of Prefixes, Stems, and Shifting Vowels
The heart of the description is the verb, and the verb is built by a small set of grammatical processes: prefixation, suffixation, infixation, reduplication, and ablaut of both vowels and consonants. Sapir catalogues an elaborate system of body-part prefixes—da- for mouth, s’in- for nose, gwen- for neck and back, t- for hand, used, he says, with almost pedantic frequency—alongside local prefixes and an instrumental wa-. Verb stems fall into sixteen formal types distinguished by how the aorist stem is formed, and the aorist, the most-used tense-mode, generally shows the amplified form. Derivation runs through frequentatives and usitatives, causatives, comitatives, indirectives, reflexives, reciprocals, and non-agentive forms, each with its own suffix and its own phonetic behavior.
The phonology is equally detailed. Takelma sits, in Sapir’s estimate, about midway between the harsh languages of the Columbia valley and the more euphonious languages of the Sacramento valley, inclining toward the latter: it lacks voiceless l-sounds and velar stops, but it has clear-cut vowels, abundant long vowels and diphthongs, and a syllabic pitch-accent that gives it a musical character—marred, he adds, by a profusion of disturbing catches. Much of the book’s finest-grained argument concerns what he calls inorganic or secondary a, the simplification of double diphthongs, and the way phonetic law bends around grammatical structure. One telling case: the final cluster -sk’ does occur, so either masga” or dink!a”sk’ must be irregular; Sapir resolves it by observing that the -k’ in the second is a grammatical element added to a future stem, and concludes that quasi-mechanical juxtaposition of grammatical elements does not follow the same phonetic lines as organic sound-complexes.
Why Sapir Calls Takelma Inflectional
Sapir devotes real argument to a question of classification. The usual definition of inflection—words of indivisible psychic value in which stem and grammatical element “chemically” coalesce—he finds far from adequate, but usable as an approximate test. Takelma passes both of its rough checks. It shows multiplicity of phonetic symbols for a single concept: possession by another is expressed by any of several suffixes, “I” as subject by five different endings depending on mode and transitivity. And its stems vary more than German or Greek: do’m-, dUm’m-, tlomom-, tlumu’- all belong to “kill.” The typical Takelma verb form is thus a firm phonetic and psychic unit with the earmarks of inflection, and the language should be called inflectional. Labels like “polysynthetic” and “incorporative” do not conflict with this, Sapir insists, because they describe the treatment of concepts rather than the method of word-building; classification depends on your point of view. He warns, too, against taking sex gender or case as criteria—inflection concerns method, not subject-matter.
What the Book Leaves Behind
The closing sections turn to the small machinery of speech: particles that never begin a clause but attach postpositively, among them a quotative found only in myth narration and an interrogative that often implies an emphatic negative; interjections classed into terms of address, set calls, animal cries, and song burdens. The conclusion characterizes Takelma once more as inflective and incorporating, compares it with Athapascan, Klamath, Maidu, and Yokuts, and lists its typical American and special grammatical traits. Appendix B then lets the language speak for itself, giving the myth “The Origin of Death” with interlinear notes and translation, followed by the opening of “How a Takelma House Was Built.” For a language with no known close relatives beyond its upland dialect—together forming the Takilman stock of Powell—the book remains the fullest record of its grammar, and a model of Boasian description written from the inside of the language rather than imposed upon it.
