Sentence and Word Author(s): Leonard Bloomfield Reviewed work(s): Source: Transactions and Proceedings of the American Philological

Association, Vol. 45 (1914), pp. 65-75 Published by: The Johns Hopkins University Press Stable URL: http://www.jstor.org/stable/282688 . Accessed: 15/03/2013 11:32
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and extend access to Transactions and Proceedings of the American Philological Association.

http://www.jstor.org

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

Vol. xlv]

Sentenceand Word

65

VI. -Sentence
BY
PROFESSOR

and Word
OF ILLINOIS

LEONARD BLOOMFIELD

UNIVERSITY

of a language into distinctivesounds, their variations,and the like. When he has completedthis,he turnsto the analysis of the semanticstructure, -to what we call the morphology and syntax of the language, its grammatical system. The method generally pursued in this semantic analysis is admittedlya makeshift: we adhere to the process of syntheticdescription whichhas been developed out of the practice of the Alexandrine and Roman grammarians. Taking the single word as our unit, we name the big classes of words (parts of speech) and then describe the inflectionof each; there follows a hasty survey of such matters as derivation and composition; finally we discuss the uses and interrelations of the various inflected words in the sentence (syntax). This procedureis a makeshift, for it has long been recognized that the firstand original datum of language is the sentence,-that the individualword is the product of a theoretical reflection which ought not to be taken for granted, and, further, that the groupingof derived and inflected words into paradigms,and the abstractionof roots,stems,affixes, or otherformative processes,is again the resultof an even more refined analysis. It needs but little scientificreflection to make us realize that the grammarianought by no means to extract such products with magic suddenness, live and wriggling,out of the nafve speaker's hat. This has long been recognized. Wilhelm von Humboldt begins his discussion of polysyntheticlanguages (Uber die Verschiedenheit des menschlichen Sprachbaues, i, paragraph 17) by saying: "Wenn man, wie es ursprunglich richtigerist, da jede, noch so unvollstandigeAussage in der Absicht des Sprechenden wirklicheinen geschlossenen Gedanken ausmacht,vom Satze

THE first task of the linguisticinvestigatoris the analysis

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

66

Leonard Bloomfield

[I914

psychologic understanding has only confirmedthis great scholar's intuition. Since we have learned to distinguish between an investigator'slogical analysis after the fact and the actual psychic occurrence,and to observe the latterwithout confusingit withsuch logical analysis,we findit obvious and easily proved that in most of our speaking we are conscious of the whole sentence only, not of the words into which it may be divided. The experimentis easily made: one asks a speaker to tell how manywords he has used in the casual sentencejust spoken. The answer,if it comes at all, long in preparing,- and this with our will be surprisingly ceaseless training,throughout our reading and writing,in this form of linguisticanalysis. I need hardly refer to the II2, I, I ff.) fact,so well illustratedby Brugmann (Grzundriss, that in some cases we do not even upon reflectionsucceed in making a division into words: shall a German write Es komintZg Stande in two, three, or four words? Shall we writein stead of as two words or as three? In as mucl as in one, two, three,or fourwords? We have many instances of the writingof uneducated people (who lack the practice of wrong. is entirely copious reading) in which the word-division Hence we repeat to-day in more decided terms the quoted 3): "In allem Ubrigen (ausser der Semasiologie) hat eine d. h. auf die Natur des Objektes streng wissenschaftliche selbst gegrundeteDarstellung nichtvom Wort,sondern vom Satz auszugehen." Brugmann adds, however,that, for practical reasons, he mannerof exposition. In doing so he retainsthe traditional is followinga practice which, I believe, is universal. It is generallytaken for grantedby studentsof language that the wrong, need not traditionalprocedure,even if theoretically draw us into any errors: we shall go safely if we never, in a weak moment,make deductions which rest not upon the of our facts of the language, but merelyupon the peculiarity avoiding in not succeeded we have that I believe method.

bedienen, die Einheit des Satzes gar nicht, . . . "

ausgeht, so zerschlagen Sprachen, welche sich dieses Mittels

Increased

says (op. as whenBrugmann of Humboldt, dictum

Cit. 112,

1,

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

Vol. xlv]

Sentenceand Word

67

this pitfall,-that some of the currentdoctrineof linguistic science is a transferenceof our own process of analysis into our beliefsabout the course of linguistichistory, and, as our process of analysis is, admittedly, not in harmony with the facts of speech, but,in a sense, diametrically opposed to them, the transference may (and, I think,sometimesdoes) lead to false conclusions. A bit of the older historyof our science well illustrates what I mean. The personal verb-forms of the Indo-European languages were easily analyzed, as soon as people began to reflect upon such things,into personal endings attached to a stem(e.g. &w-,ut&&-' &saz-at,older and West Greek &8&-Tt). Proceeding fromthis analysis and taking for granted that it representeda historicsynthesis, Bopp identified the personal endings with old formsof the personal pronouns. Similarly he saw in the i-suffix of the future and optative the root (itself,of course, the creature of a similiar analysis) of the verb ire. These theories were given up not only because Bopp's specificexplanations were in conflictwith the ascertained sound-developments of the languages concerned,but also because we realize that Bopp was inspiredchiefly by the feeling that our analysis of formsis necessarily in accord withtheirhistoricorigin, -and we know now that this feeling was wrong. When scholars to-day speculate upon the origin of the personal verb-inflection they turn rather to an and suppose that the endings of these adaptation-theory formshave come, more or less accidentallyto their personal meaning; so, forinstance,Hirt,L.F. XVII,36. That is to say, the grammaticalanalysis of a given stage of a language must not lead us into thinkingthat the formsare the result of a correspondingconglutination. Bopp's error is a thing of the past,' but if a genius like Bopp could fall into such an error,it is obvious that we, too, need the correctiveof an occasional analysis in the psychologicallyjustifieddirection, proceedingfromthe sentence,the
1 The Editor remindsme of Professor Fay's articles; needless to say, I do not agree with their tendency. It is fairto add, however, that I know Fay's " Returnto Bopp " onlyfrom in Idg.Jahrb.ni. the summary

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

68

Leonard Bloomfield

[I914

concretedatum,to the less and less explicit articulationsin the sequence of speech. A serious errorthat has outlived the agglutinationtheory of the sentence. The ancients, for whom is our definition grammar was an ancillary discipline of logic, necessarily looked upon the sentence as a combinationbuilt up out of words. Dionysios Thrax 2 defined the sentence as vre?s this: Oratio est ordinatio dictionum congrua, sententiam perfectamdemonstrans. It is Wilhelm Wundt who, in his is understand the psychology of the thing, this definition a sentence,says Wundt,is the linguisticexprestopsy-turvy: sion of the voluntaryanalysis of a total experience into its parts,which then stand in logical relationto one another. does not distinguish It has been objected thatthisdefinition a sentence froma word,such as Tpt'wOVW,which also involves an analysis of the total experience which it expresses. We are face to face, then, with the problem of distinguishing between the analysis made by a speaker who says Tpeds 7o'Saq et'ov and that made by one who says Tpti'ovs,-between sentence and word. Meanwhile we cannot retain even Wundt's definitionof the sentence, for it implies an articulation of the sentence into parts which we have no right to look upon as essential or universal. The assumption that every sentence must break up into two or more
independent Voikcrpsychologie,I,
2, 234 Xdewteo- V9VOater,&davotav a"ToTeX'

Priscian3 translated 871jXovioa,

ff., first showed that, when we

(in Gercke and Norden's characterizedby Kretschmer rightly f.) as a ves55 die , 1 in Altertumswlsscnsclzaft2 Einleitung it was to which tige of the old rationalizingview, according built up out of such components. Kretschmerdefines the -that is, sentence as the linguisticexpression of an affect, of emotionalvolumeand tension. Perhaps of an up-and-down we should do betterto say that the sentence is the linguistic expression of an affectinvolvinga single total experience, for an affectof higher order may be accompanied by the
2 T iXvm7, par. II.
3

and

logically articulated !-

components

is

De ArkeGramm.II,

15.

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

Vol. xlv]

Sentence and Word

69

The phonetician,firstof all, tells us that physical difference there is none. His ear tells him,and the difficulties of orthographicseparation above referredto prove to him,that there are in an utterance no pauses to indicate its structure. We have the proof in ourselves whenever we hear people speaking a language which we do not understand,for it is then beyond our power to findthe word-divisions. What is it, then,that enables us to analyze utterancesinto words and morphologicelements? To begin with, it is not any reflectionof the speaker's. Even people who have studied language and may be to an abnormalextentconscious of the facts of speech, uttermany sentences every day withoutthe least reflectionupon their analysis. As a writeron logic puts it, we ought to writeall our sentenceswithhyphensbetweenthewords; a phonetician would say that we ought not to indicate the word-division at all. The division of the sentence is not a reflective one; it is a matterof implication, and is due to the associational connectionsof the parts of the sound-sequencewhich constitutes the sentence,- as it were, to their connotation. A Latin sentence such as exibant is, like every sentence, primarily and so far as any logical reflection on the part of speaker or hearer may be concerned,a unit. The various parts of this sound-sequence, however, have been heard and uttered by the speaker (or the hearer) in other sentences and have, in these other earlier occurrences,always corresponded to an element of meaning which is present also in this new experience accompanying the sentence exibant. All these past occurrences of parts of the present sound-sequenceexercise upon the latter the subtle force known to psychologistsas

or compounded suchas Tpw7roVS. flected, derived, word,

utterance of a succession of sentences, each of which correspondsto a subordinateup-and-downmovementof the emotional curve. What we most need is not, however, a definition of the sentence or of the word,-we have a very decided naYvefeeling for these units, -but rather an underbetween a succession of words, standing of the difference such as Tpe's vro"asge'xovand what we feel to be a single in-

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

70

Leonard Bloomfield

[I9I4

simultaneousassociation or fusion: theygive them a tone of whichwe, forour purposes,may speak of as their recognition meaning. Thus the firstpart of the sound-sequenceexibant owes its value to earlier utterances(heard and spoken from and the like; the next childhoodon), such as excessit, C:r7igit, sound, to such as abfrcm, re(llmtls, and the like; the next such as has occurredalso in numerousutterances, sound, -b-, in all of which it correspondedto regUbat,videbit,condbitur, of action,past or future; the -aa vague notionof continuity has occurred also in regebat,eram,fucrat, parallel with a is one of the semantic element of past time; the -nt,finally, has been and in the language, mostfamiliar sound-successions heard and spoken innumerabletimes in sentences that expressed an event in which more than one actor, including neitherspeaker nor hearer,performedan action or was the goal (object) of an action, e.g. dolent,conaztur,de1ecta;itur, and so on. Now, though all this dissectionis far too exeuint, that implication-values clumsy to do justice to the intanigible involved in the speaking and automatically are immediately or hearing of the sentence exibant,yet we can be sure that the meaning of this sentence to a Latin was due to these veryassociations,forwe knowthat in language the sentences which a speaker may utter are not confinedto those which new he has actually heard before,but may consist of entirely of the habitual speech-elements. A speaker of combinations Latin who happened never to have heard the form extbat could use it, and use it withoutthe slightestconsciousness anizusince he had many times heard exabant, of innovation, we In very may, other words, bat, arn4bant,and so on. clumsily,indicate the associational values in the sentence cxfbant by dividingit into ex-ib-a-nt. If, now, we take the corresponding English sentence, we finda similar associational habituationof Yejuwrbowizydwt, the different parts of the sound-sequence. Yej has occurred
in ;Yejddn jdit, eseJdsow,and the like, where also there was

a third person plural actor; wr in such expressions as wj-

wrwejtiy, juwwi-Vr, and so on; gow in letsg5w, downtgzw,

and the like; iy, expressiveof continuedaction,in hizjrajtiy,

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

Vol. xlv]

Sentence and Word

71

almw4jtiy,and so on; awt, in value like the Latin ex-, in and many other utterances; and here, komanawt, hararndwt, as in the Latin sentence,these parts are in ordinaryspeech by no means drawn into the focus of the attentionor explicitlydistinguished, but are rather,by the associative effect - one mightalmost say, mutely, of theirearlier occurrences, - symbolic of the meaning. Their utterance in certain situations of experience, and the reproductionof a corresponding meaning whenever they are heard, is a matterof habit, not of explicitagreemnent or reflection. There are, however,occasions when we utter such a sentence with a full and explicit insistence upon some one part of it, and thus show,a consciousness of its divisioninto parts and try, indeed, to arouse the same consciousness in the hearer. Suppose that an elementof the situationis in doubt or in question,for instance,the identityof those who went out. Then we say ;ej wrgowiyawt. Here it is no longer the mere implicit associational value of the sound-groupYej that lends it meaning; our attention, like a vivid spotlight, focuses this part of the utterance,singling it out from the rest; and the hearer's attention, by the loud tone and other phonetic features,is drawn to it. We may similarly, if the time of the occurrencebe in question,accent the wr and say Yej wr gowiyawt,and so on. One element,however,of those found by analysis to make up this sentence, we cannot so emphasize, namely the sound-groupiy, expressive of continuityof action. Its associational value is clear, but apperceptive value it can never have: it never falls into the focus of the attention. Besides this habit of never clearly considering the elementiy, we have another limitingits use: it is spoken -after an element expressive of action,such as gow or and it never rajt, to which it lends the meaning of continuity, occurs in any other connections. This, moreover, is true of all the parts of the Latin sentence which we have examined, exzbant.4
4 The sound-group ex-, to be sure, does occur in other connections, such as ex urbe venif,but it has then a different value, expressing spatial relation with regard to an object, not direction of movement: it is then a preposition, not an adverb.

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

72

Leonard Bloomfield

[1914

between a formativeelement and a This is the difference word,of course: both recur as the expressionof a constant element of meaning, but the formativeelement is bound to certainpositionswith regard to the other elements,while the word may occur in all kinds of connections; and, above all: while both occur usually as associativelydeterminedparts of (clearly a sentence,the word may be focused by the attention element never rises to this apperceived), while the formative explicitrecognition. It is a commonplaceof psychologythat,of these two forms of the structure of experience,the associative or passive, and the apperceptive or active, the formeris the primaryand usual one, the latter the more developed and rarer. The greaterpredominanceof associational processescharacterizes forus the mental habits of savages (sympatheticmagic, and the like; cf., in connection with language, Levy-Bruhl,Les and M arrett, fonctiousmentalesdans les socie,'s infirieures, Jespersen'sProgressin Language is, of course, Anthropology; familiar),the course of dreams,and morbidmental processes der Psychologie). Opposed to all these, (Wundt, Gruzndriss the higher phases of mental life, such as sane thinkingor scientificreasoning, are characterized by the frequent and wheneverthe occasion demands,to apperresort, unhesitating ceptive focusingof parts of an experience. as the agglutinaIt would be strange if linguistic history, tion theoryassumed, showed us a retrogressive development, -a developmentfrom forms of speech which allowed not only of associative but also of occasional apperceptive distoward formswhich moved only and always in the tinction, dim realm of associative reminiscence. As a matterof fact, whereverwe know it, shows us progressin linguistichistory, the directionfromassociative toward apperceptive structure. we express by a sepaWhere in Old English one said gad6u't, rate word both the actor and the tense: ;Yej a:r gowiy awt; where in Latin one said Romam it or Romam vadit,one uses in French a separate word for the directionand another for the actor: il va a' Rome. in this respect, between Latin or Old The differences,

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

Vol. xlv]

Sentence and Word

73

English and the modern languages are of interestbecause of the accessibilityof the historic relation and all that it implies,but the structureof Latin or of Old English is not so widely different from that of our speech. If actor, action, and tense are there expressed in one word,we findin other languages not only these elements, but also objects, directand indirect,and other featuresof the experience,all expressed without the possibility of a single apperceptive articulation, that is, in one word. Thus in the Fox language (Jones and Michelson,Bulletin40, U. S. Bureau ofEthnology), pyit(`kwdwdwa, ' He brings home a wife' (pydte 'hither, home,' 'kzwvw'long hair, woman,' a 'her,' wa 'he'), nimdwinAtutAmawawa, 'I shall go and ask him for it' (ni ' I, in future action,' mdwi 'go,' nAtut 'ask,' Amaw 'it, as secondaryobject,' d and wa, both referring to animate third person). It is interesting to notice that the first and most important divisionwhich logical reflection has always demanded of the sentence,namely,that into subject and predicate,is one of the rarest,and, where we know the history, one of the latest, to receive a corresponding word-division in the sentence: in Latin, Greek, and Sanskrit subject and predicate are usually both expressed in the verb-form:in Slavic and most of the Romance languages both possibilitiesare open (Italian canta or ella canta, Polish spiewa or ona 4piewa). In our languages we have in some cases the choice between the two methodsof expression,one by a single word and one by a succession of several words. Thus, we may speak of a horse-tamer or a tamerof horses. The former kind of expression obviously analyzes itself into the elements horse and tamer,and linguistic scholars, taking for granted that our analysis corresponds to the historicoccurrence,are wont to assume that such compound words are the productof a coalescence of independentwords. This assumptionmeets with a very significant difficulty: the fartherback we go in tracing the historyof our languages, the less resemblancedo the parts of such compounds bear to the individualwords from which it is supposed that they were derived. Thus in An-

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

74

Leonard bloornfield

[1914

words, e.g. Greek lcv3po"; Icv&dvetpa,7raTrpo7raTpo'. The most [7wwc3a,uosa, widespread of these deviations,the type of 'w7ros,

fromany differ cient Greek the parts of the word i7rwo'a,juos, any actual from differs occurringforms; 'w7roindependently 'tamer,' is a type of formation form of '77ros(, and -8ajuo's the -ptrarelyfound outside of compoundwords. Similarly, one Any word. independent of rpt'7rov does not occur as an (Grundwho reads Brugmann's section on noun-compounds riss, II2, 1, 49 ff.) or the second volume of Wackernagel's monumentalSanskrit Grammarwill be impressedby the endof accidental or secexceeding all possibilities less deviations, fromindependent of composition-stems ondarydevelopment,

is so obtrusivethat it has given rise to the supplementary theorythat these compounds go back to a time (postulated Od hoc) when uninflectedstems were used as words, and used, the compounds compel one furtherto assume, in the value of any and everycase-relation. So Brugmann(Grundriss, II2, I, 78); upon this theoryJacobi has built his specuzi undNebensats. i lations in Compositm Needless to say that

the whole assumptionthat compound words are historically the resultof a coalescence has no othersupportthan the circumstance that we analyze them into elements more or less closely resemblingsingle words,-exactly as Bopp analyzed certainelementsmore or less out of the personal verb-forms closely resemblingpersonal pronouns. In neither case does the analysis justifya historical assumption. Quite on the the less do the back we go into history, the farther contrary, we have every words: elementsof compoundsresemblesingle reason to believe that the compoundwords of the Indo-Euroin which pean languages representan older type of formation meanings that are now usually expressed in several words were still merged into one word whose divisionshad only an - a word comparable to the formations associative identity, of breaking up of the American languages. The possibility we are accustomed units which into those smaller the sentence was of later dewords to to look upon as corresponding simple acfrom actor of exactlyas thepossibility separating velopment, times. tionin Latin orGermanicspeech has developedin historic

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions

Vol. xlv]

Sentence and Word

75

The compound word remained in use where its meaning had undergone transferenceor specialization and differed, accordingly,from that of the now more favored collocation of simple words. This accounts for the persistence of such types as the so-called exocentriccompounds, '/cV7rEposI ' having quick wings,' English long-nose'one who has a long nose,' and, in general, for our habit of using compounds where we mean somethingmore specificthan what would be expressed by a collocation,e.g. blue-bird, as opposed to blue bird. To recite the evidence for this view would be to tell the entire storyof compound words in the Indo-European languages.5 So much,however, is certain, that,hereas elsewhere, the course of linguistichistoryhas been from associational articulationof the utterance toward apperceptive structure; and that the grammarian's dissection of words, though of infinite practicalvalue, must not mislead us into thinkingthat language is really a pasting together, by means of hyphens or a similar agency, of the elements which this dissection may reveal.
6 An interesting and hope some dayto finish. task,whichI have undertaken

This content downloaded on Fri, 15 Mar 2013 11:32:35 AM All use subject to JSTOR Terms and Conditions