THE PSYCHOLOGICAL REALITY OF PHONEMES*

THE CONCEPT of the "phoneme" (a functionally significant unit in therigi~IY.definedpattern or configuration of sounds peculiar to a language),as .dIs~mct from that of the "sound" or "phonetic element" as such (anobjectIvely definable entity in the articu ated and perceived totality ofspeech), is be~oming more and more familiar to linguists. The difficultythat many stIll seem to feel in distinguishing between the two mustevent~ally disappear as the realization grows that no entity in humanex~enence.can be adequately defined as the mechanical sum or productof Its physIcal properties. These physical properties are needed of courseto give us .the signal, as it were, for the identification of the given entityas a functIOnally significant point in a complex system of relatednesses;but for. any given context it is notorious how many of these physicalpropertIes are, or may be, overlooked as irrelevant, how one particularproperty, ,possessing for the moment or by social understanding anunusual sIgn value, may have a determinedness in the definition of theentity that is out of all proportion to its "physical weight."

As soon, h.owever, as we admit that all significant entities in experienceare thus rev~sed from the physically given by passing through the filterof the functIOnally or relatedly meaningful, as soon as we see that wecan never set up a scale of added or changed meanings that is simplyc?ng~ent to the scale of physical increments, we implicitly make a dis-tmctIOn, whether we know it or not, between the phoneme and thesound in that particular framework of experience which is known aslanguage (actualized as speech). To say that a given phoneme is notsufficiently defined in articulatory or acoustic terms but needs to befitted into the total system .of sound relations peculiar to the language is,at bottom, no more mystenous than to say that a club is not defined forus when it is said to be made of wood and to have such and such a shapea?d. such ,and such dimensions. We must understand why a roughlys~ilar ?bJect, not so different to the eye, is no club at all, and why athIrd object, of very different color and much longer and heavier than thefirst, is for all that very much of a club.. Some linguists seem to feel that the phoneme is a useful enough conceptIn an abstract linguistic discussion-in the theoretical presentation ofthe f?rID of a language or in the comparison of related languages-butthat It has small relevance for the actualities of speech. This point of

* PUblis,~ed originally in Frenc~ under the title "La Realite psychologique desphonemes, Journal de Psychologle N01"male et Pathologique, 30 (1933): 247-265.

[46 ]

view seems the reverse of realistic to the present writer. Just as it takes aphysicist or philosopher to define an object in terms of such abstractconcepts as mass, volume, chemical structure, and location, so it takesvery much of a linguistic abstractionist, a phonetician pure and sim-ple, to reduce articulate speech to simple physical processes. To thephysicist, the three wooden objects are equally distinct from each other,"clubs" are romantic intrusions into the austere continuities of nature.But the naive human being is much surer of his clubs and poles than ofunnamed objects to be hereinafter defined in physical terms. So, inspeech, precise phonetic stations can be abstracted only by patient ob-servation and frequently at the expense of a direct flouting of one'sphonetic (one should say "phonemic") intuitions. In the physical worldthe naive speaker and hearer actualize and are sensitive to sounds, butwhat they feel themselves to be pronouncing and hearing are "pho-nemes." They order the fundamental elements of linguistic experienceinto functionally and aesthetically determinate shapes, each of which iscarved out by its exclusive laws of relationship within the complex totalof all possible sound relationships. To the naive speaker and hearer,sounds (i.e., phonemes) do not differ as five-inch or six-inch entitiesdiffer, but as clubs and poles differ. If the phonetician discovers in theflow of actual speech something that is neither "club" nor "pole," he, asphonetician, has the right to set up a "halfway between club and pole"entity. Functionally, however, such an entity is a fiction, and the naivespeaker or hearer is not only driven by its relational behavior to classifyit as a "club" or a "pole," but actually hears and feels it to be such.

If the phonemic attitude is more basic, psychologically speaking, thanthe more strictly phonetic one, it should be possible to detect it in theunguarded speech judgments of naive speakers who have a completecontrol of their language in a practical sense but have no rationalized orconsciously systematic knowledge of it. "Errors" of analysis, or what thesophisticated onlooker is liable to consider such, may be expected tooccur which have the characteristic of being phonetically unsound orinconsistent but which at the same time register a feeling for what isphonemically accurate. Such "errors," generally overlooked by the prac-tical field linguist, may constitute valuable evidence for the dynamicreality of the phonemic structure of the language.

In the course of many years of experience in the recording and analysisof unwritten languages, American Indian and African, I have come tothe practical realization that what the naive speaker hears is not pho-netic elements but phonemes. The problem reaches the stage of a prac-tical test when one wishes to teach an intelligent native, say one who can

read and write English reasonably well and has some intellectual curi-osity besides, how to write his own language. The difficulty of such atask varies, of course, with the intelligence of the native and the intrinsicdifficulty of his language, but it varies also with the "phonemic intuitive-ness" of the teacher. Many well-meaning linguists have had disappoint-ing experiences in this regard with quite intelligent natives without eversuspecting that the trouble lay, not with the native, but with themselves.It is exceedingly difficult, if not impossible, to teach a native to takeaccount of purely mechanical phonetic variations which have no pho-nemic reality for him. The teacher who comes prepared with a gamut ofabsolute phonetic possibilities and who unconsciously, in spite of all histraining, tends to project the phonemic valuations of his own languageinto what he hears and records of the exotic one may easily befuddle anative. The native realizes when what he is taught "clicks" with whathis phonological intuitions have already taught him; but he is made un-comfortable when purely phonetic distinctions are pointed out to himwhich seem real enough when he focuses his attention on them but whichare always fading out of his consciousness because their objective realityis not confirmed by these intuitions.

I have selected for brief discussion five examples of phonemic versusphonetic hearing and writing out of many which have come to me in thecourse of my experience with natives and students. In each of these, itwill be observed, we have clear evidence of the unconscious reinterpreta-tion of objective facts because of a disturbing phonological preparednessnot precisely adjusted to these facts.

1. When working on the Southern Paiute language of southwesternUtah and northwestern Arizona I spent a little time in trying to teach mynative interpreter, a young man of average intelligence, how to write hislanguage phonetically. Southern Paiute is an unusually involved lan-guage from the phonological standpoint and, as my point of view at thattime stressed phonetic accuracy rather than phonemic adequacy, I doubtif I could have succeeded in teaching him well enough to satisfy mystandard even if I had devoted far more time to the effort than I did.Ai; an example of a comparatively simple word I selected prJ'{3a' "at thewater" (voiceless labial stop; stressed long a; voiced bilabial spirant;unstressed short a; final aspiration). I instructed Tony to divide the wordinto its syllables and to discover by careful hearing what sounds enteredinto the composition of each of the syllables, and in what order, then toattempt to write down the proper symbol for each of the discoveredphonetic elements. To my astonishment Tony then syllabified: pa-, pause,pa'. I say "astonishment" because I at once recognized the paradox that

T was not "hearing" in terms of the actual sounds (the voiced bilabialony . . . ) b' f{3 was objectively very different from the mltlal stop ~t. m~~ ~ a~etymological reconstruction: pa': "water" plus postposItiOn -pa .at.The slight pause which intervened after the stem was e~~ugh to dIvertTony from the phonetically proper form of the postposItion to a theo-retically real but actually nonexistent form.

To understand Tony's behavior, which was not in the least due to merecarelessness nor to a tendency of the speakers of this language "to con-fuse sounds," to quote the time-worn shibboleth, we must have recourseto the phonology of Southern Paiute. The .treatment of the stoppedconsonants may be summarized in the followmg table:

POSTVOCALIC

INITIAL3. Geminated

I. Spirantized 2. Nasa1ized a. After b. After un-voiced vowel voiced vowel

Labial p fJmp p' P

Dental t rnt t· t

Guttural k "Y,.,k k' k

Labialized guttural kw oyW ,.,kwk'w kw

The postvocalic forms of the stops of t~es 1, 2? and 3a are .furthermodified before an unvoiced vowel, the VOICed spIrants becommg un-voiced spirants (0, R, x, xW),1 and the nasalized and ge~~ted st~psbecoming aspirated (mp', p"; nt', t·'; 11k', k"; 1/kW, k' W). It IS ~possl?lehere to give a systematic idea of the phonologic processes ~hich b:m.g

about the sound interchanges within a given articulatory senes, but It ISimportant to know that the spirantized, nasalized, and geminated stopscan occur only in postvocalic position and that they are largely deter-mined by the nature of the element (stem or suffix) w~ch prec~e~ themand which may be said to have an inherently spirantIzmg, nasali~mg, .orgeminating force. The stem pa'- is a spirantizing stem, ~d th.e .splrantlz-ing of a theoretical *-pa' "at" to -{3a' is parallel to the sprrantlzmg of p-:r-"trail" to -{33'- in such a compound as pa'-{33'-, "water-trail." In otherwords, the language is so patterned that examples of type p3'-: -{33'- le~dtc the proportion *pa': -{3a'2 and, while *pa' "at" does not actually eXIstas an independent element but must always be actualized in one of thethree possible postvocalic forms, its theoretical existence suddenly comes

1 W represents voiceless w. ",. r2 This theoretical ·pa' occurring only as -fJa, -mpa, -p'a 10 po(stvocal)Cposition, is not to be confused with secondary _pal (type 3b) < -p'a' type a.

The phonetic orthography is more complex and, in a sense, more ade-quate, but it goes against the grain of the language in one importantrespect, for it identifies the second p in type 5 with the initial p, which isphonologically unsound. The phonologic orthography, on the other hand,is useless for one who has not mastered the phonology of the language, asit leads, or seems to lead, to incorrect pronounciations which would havethe cumulative effect of making the language, so read, entirely unintelli-gible to a native. To a slightly schooled native, however, there can beno serious ~mbiguity, for the phonetic forms result from the phonologiconly by the application of absolutely mechanical phonetic laws of spiran-

a The analogy to French liaison and, still more, to the three types of consonan-tal treatment in Old Irish (spirantized or "aspirated," nasalized or "eclipsed,"and geminated) is obvious,

• A represents voiceless a.

to the light of day when the problem of slowly syllabifying a word ispresented to a native speaker for the first time. It then appears that the-{la' of speech behavior, as a self-contained syllabic entity without im-mediately preceding syllable, is actually felt as a phonologic pa', fromwhich it differs in two important phonetic respects (voiced, not voiceless,consonant; spirant, not stop).

All this has an important bearing on the construction of a maximallycorrect orthography of Southern Paiute, if by "maximally correct" wemean, not most adequate phonetically, but most true to the sound pat-terning of the language. Af3 it happens, there is reason to believe fromboth internal and comparative evidence that the spirantized form of aconsonant is its normal or primary form after a vowel and that thenasalized and geminated forms are due to the emergence of old nasal andother consonants that had disappeared in the obsolete form of the preced-ing element.3 It follows that the postvocalic -{l- is more closely relatedfunctionally to a simple initial p- than is the postvocalic -p- (after un-voiced vowel), which must always be interpreted as a secondary form of-p'-. These relations are summarized in the following table of theoreticalnonfinal forms,

tizing, alternating stresses, and un~oicing..It is no~ necessary to d.ealwith these laws here5 but we can indIcate their operatIOn by the followmg

table of theoretical final forms:

Obviously in such a language as this, spirants, whether voiced or voic~less and voiceless vowels are not phonemes but are mere~ ~honet~crefl~xes of stopped consonants and voiced vowels under fix ynamlc

dit' Long consonants and long vowels are sub-phonemes. Thecon IOns. ed t ) andformer are the resultants of simple phonemes (stop~ cons~na~ s .the operation of certain phonologic (and morphologIC) latenCIes~ ~lv~nsyllables, present or formerly present. The l.atte~ are phon~l~~~~:~

I ble into short vowel plus short vowel, I.e., mto two sy:~~~:ngth (moras), of which the second .begin~ with a zero conso~ant.

Southern Paiute, then, is a language 1ll whICh an unusually SImple

. '1' E S . The Southern Paiute Language,i They are described m detal In . aplr, d S . 65 (1930)

Proceedings of the American Academy of Arts an menccS, .8 M is voiceless m. .

papapapa'pa'papa'pa'pap'apap'a'pa'p'apa'p'a'mapapamapapa'mapa'pamapa'pa'mapap'amapap'a'mapa-p'amapa'va 'map'apamap'apa'map'a'pamap'a'pa'map'ap'amap'ap'a'map'a'p'amap'a'p'a'

PHONOLOGIC OaTHOGRAPHY

1. pallA2. pa{3a'3. pa',pA4, pd',Ba' "water-at"5. pap'A6, pApa'7. pa'p'A8. pa'p'a'9. ma{3d

We see at once that dini "this one" behaves like a word ending in asmooth vowel (witness contraction of i + i to an over-long vowel and

9 The lack of a tone mark indicates that this syllable is pronounced on the

middle tone.10 a,a is an over-long a, consisting of a long a' followed by a weak rearticulated

a. Syllables of this type result in Sarc~e from, contraction of 011final vow.els v.:ithfollowing suffixed vowels. The change m qualIty from -i to -a.' IS due to hlstoflcalfactors, -ni "person" is an old

unaffected l of -la), while dini "it makes a sound" acts as though the fin 1~ow;l had ~ voiceless consonantal latency, which registers partly ~- (- -la passmg, as always, to -la), partly as -t'-.

~t is clear that, while John was phonetically amateurish, he was phono-10gI?ally subtle and accurate. His response amounted to an index of thef~e~g that dini "this one" = dini, that dini "it makes a sound" =dmi, and that this -ni' = -nit'. John's certainty of difference in th ff b' t' 'd . , e aceo o,lec Ive 1 entIty IS quite parallel to the feeling that the average

Enghshman would have that such words as sawed and soared are notph~neticallyidentical. It is true that both sawed and soared can be pho-~etICally :epresented ~ s;rd,12 ~ut the -ing forms of the two verbs (saw-mg, soar~~), phonetIcally 8:r-1,V and so'r-iv, and such sentence sandhiforms a~ Saw on, my boy 1" and "Soar into the sky 1" combine to producethe feelmg that the so'd of sawed = so'-d but that the so'd of soared =so'r-d. In the o~e c~s~ zero = zero, in the other case zero = r, Among~ucated but ImgmstICally untrained people who discuss such mattersdIfferences of orthography are always held responsible for these d'ff -

ff I' Th" 1 erences 0 ee mg, 18 IS undoubtedly a fal acy, at least for the great massof people, and puts the cart before the horse, Were English not a writtenlanguage, the configuratively determined phonologic difference betweensuch doublets as sawed and soared would still be "heard" as a colI t'ill

' , ec IveUSlOn, as a true phonetic difference,III: The most. successful American Indian pupil that I have had in

practICal phonetIcs is Alex Thomas, who writes his native languaN tk 13 ' h h ge,00 a, WIt t, e utmost ~uency and with admirable accuracy. Alex'sorthography, as IS natural, IS phonologic in spirit throughout and it islargely from a study of his texts that I have learned to estimate at itstrue value the psychological difference between a sound and a phoneme.~yonewho knows the phonetic mechanics of Nootka can easily actualizehIS orthography. ~hus, M,14 phon?logically parallel to si or ni, is actuallypronounced 1}E, WIth. a v0.w~l whICh is much nearer to the e of Englishmet than to tha~ of S'tt. ThIS IS due to the peculiar nature of the laryngealconsonants, whICh favor an a-timbre and cause the following vowels i andu to dr?p to E and a respectively. The orthographies M and 1}u are entirelyunambIguous because there can be no phonologically distinct syllables oftype 1}E and 1}a,

Another mechanical peculiarity of Nootka is the lengthening of con-sonants after a short vowel when followed by a vowel. This purely

~~ Th~s~ remarks apply to British, not to normal American, usage14 T~iS is spoken on the west 90ast of Vancouver Island, B. C, .

1) is a VOiCeless laryngeal spirant, almost identical with the Arabic hti, ,

mechanical length has no morphological or phonological significance andis ignored in Alex's orthography. His hisi'k and Msa' are, then, to benormally pronounced his'i'k' and 1}E8'a', It sometimes happens, however,that a long consonant, particularly s' and s', arises from the meeting oftwo morphologically distinct consonants (e,g., s + s > s' or s+ s > S,or, less frequently, s+ SOl' S + s> s,), In such cases the long consonantis not felt to be a mechanical lengthening of the simple consonant butas a cluster of two identical consonants, and so we find Alex writing, forexample, tsi'qsit'lassatlnii5 "we went there only to speak," to be analyzedinto tsi'qsitl-'as-sa-('a)tl-ni, The s of -'as "to go in order to" and the s of-sa "just, only" keep their phonologic independence and the normalintervocalic -S'- of -'as'atl is interpreted as -ss-, Similarly, kwis-sila "todo differently," to be analyzed into kwis-sila, It does not seem, however,that there is an actual phonetic difference between the -s- (phonologically-s-) of such words as tlasatl "the stick takes an upright position on thebeach" (= tla-satl), pronounced tlas'atl, and the -S'- of -'assatl above, Hereagain we have objectively identical phonetic phenomena which receivedifferent phonologic interpretations,

IV. In the earlier system of orthography, which Alex was taught, theglottalized stops and affricatives were treated differently from theglottalized nasals and semivowels, The former were symbolized as p!,t!,k!, k!w, q!, q!w, ts!, it:! (= is), and L! (= it); the latter as 'm, 'n, 'y, and'w, The reason for this was traditional. The glottalized stops and affrica-tives, as a distinctive type of consonants, had been early recognized byDr. F, Boas in many American Indian languages and described as"fortes," that is, as stops and affricatives "pronounced with increasedstress of articulation." The type 'm, 'n, 'l, 'y, and 'w was not recognizedby Dr, Boas until much later, first in Kwakiutl, and described as consist-ing of nasal, voiced lateral, or semivowel immediately preceded by aglottal closure. The orthography for these consonants (later discoveredin Tsimshian, N ootka, Haida, and a number of other languages, but notas widely distributed as the so-called "fortes") suggested their mannerof formation, but the orthography for the glottalized stops and affrica-tives was purely conventional and did not in any way analyze their for-mation except to suggest that more energy was needed for their pro-nunciation,16 As a pure matter of phonetics, while the Nootka glottalized

15 I have slightly modified Alex's orthography to correspond to my presentorthography, but these changes are merely mechanical substitutions, such astl for L, and in no way affect the argument, q is velar k (Arabic 1);), tl is a lateralaffricative, tl its glottalized form.

16 This, incidentally, is not necessarily true. In some languages the glottalizedstops and affricatives seem to be somewhat more energetic in articulation thanthe corresponding unglottalized consonants, in others there is no noticeable

:zted0ps and at~ricatives are roughly parallel in formation with the glottalsonan IC consonants the t d -

glottalized p, for instanc~, ou~ ;::s::t ~nan~~~not be ~ntirelY. so. In achronous closure of lips and glottal cor~s a cl ~er .p., ~here Is.a syn-produced between the two, there is a sudden re~:=se~; t~ea~be~ IS thusmoment of pause and then th I Ip C osure, arelease of the lip (or other oral)ec~~s::i~f ~he glottal closure. It is thethat gives consonants of this t th . a vance.of the glottal closureacter.J7 On the oth h d' ype. elr superficIal "click-like" char-

er an ,10 a glottalIzed m 0' h'l .and glottal closure are synchronous as befo; ~r m, w 1 e the IIp closurebe released at the point of initial sonancy 0;'th~he g~ttal~losure ~usttherefore, Vmay be analyzed into p +' h'l' m. oug y speaking,,+ S h ' W 1 e m may be analyzed intI ha;'i~;ri= ~~~~g:::~~~~:::~et:=~~iversus 'm, there!ore.' whic~purely phonetic grounds. on, was not unjustIfied on

We now come to the intuitive phonologic test whether 'and'consonants of the same type or not Al led' P maretype V and is very readily (our e~rli:: ~r:nd to, WrIte co~s~~ants of(earlier plapli), isa'ak "stream" (earlier ~"a'ak) t;.)' e.g., pap:' "ear"volunteered ml in such words as 'ma":' ,', 0 my surpnse Alexsister]," which he wrote m'am'iq I m:hqsu the older [brother orevidence here for the phon~lo :c :~lit 0 er word~, we had valuablenants which included both tYP~ '( 'thY o! a glottahzed class of conso-t ' . P WI pnor release of oral clos ) dype m (WIth prior release of glottal closure) A phonolo' 11 ure. anort~ographywould require Vand rh (or pI ~nd ,) 0 glCa y consls~ntnatIve's phonetic "i norance" .m.. nee more, a naIvethe scientist's "kn~wled "p;~vedhPhonol~glC~llYmore accurate than

ge. e p onologlC Justification for Alex's

difference so far as "stress of art' I . ".la~guages that I have heard (Sar~~~ au:o~ h·IS concerned. In the AthabaskanVOiceless stops and affricatives (of tvpe t'U k

C' It:)HuP~, Navah~) th~ a~pirated

act.er than the correspondintT

giott'l' d' ,s are ar more fortIS" m char-no ." a Ize consonants (e (k' t)necessary correlatIOn between lar neal ..g.,.' ,S. There isles~, gJottahzed; or any of these with rs gj .type of artICubtlO~ (voiced, voice-lems). So far as Nootka is concerned -r d~JlOn) and force of artIculation (fortisstops and a~ricatives (Boas' "fortes',,) w~ no.t s~em to me .that the glottalizedfro~ the .0rdInary stops and afTricative I re s~Tficantly different in emphn..sislogICal difference of emphatic and no;~mnhUt~ anJuages as recognize 11 phono-glottaliz.ed consonants, there is no reason J'h: tl~ea~ itaf· thde same time, possessappear In both emphatic and nonem h f f goa Ize. consonants may notshown, :some of the North Caucasic lag ~al~s orm. As Pnnce Trubetzkoy hasemphatic and nonemphatic glottalized gt g ,ad a m.att~r of fact, possess both

17 These consonants are a a I's ops. an ~ffncatIves.Jones. There is another ap~~r:~tlt YI IdentICal With the "ejectives" of Danielaffricative in which the ~ral and glo[t ISs Icommon, type of glottalized stop or

18 The pronunciation of 'm 'n' a de ,eases ar.e synchronous.(') plus m, n, w, and y is rejected byWt'haenNo ytkas a slmp!e sequence of glottal stopo a ear as Incorrect.

"error" is not difficult. Consonants of type Vare entirely analogous toconsonants of type 'm for the following reasons.

1. Each occurs at the beginning of a syllable and, since no word canbegin with a cluster of consonants, both p and 'm are felt by Nootkaspeakers to be unanalyzable phonologic units. In other words, the glottalstop can no more easily be abstracted from 'm than from V·Similarly,the affricatives and glottalized affricatives are phonologically unanalyz-

able units.2. All consonants can occur at the end of a syllable except glottalizedstops and affricatives, glottalized sonantic consonants ('m, 'n, 'y, 'w),semivowels (y, w), nasals (m, n),19 the glottal stop ('), and h. This rulethrows consonants of type 'm more definitely together with consonants

of type V·3. Many suffixes which begin with a vowel have the effect of "harden-ing"20 the preceding consonant, in other words, of glottally affecting it.Under the influence of this "hardening" process p, t, k become V, i, ~,while m and n become 'm and 'n. For example, just as the suffixes '_a'a

21

('-a"a) "on the rocks" and '-a1J,s "in a receptacle" change the stemwi'rwp- "to stay, dwell" to wi'nav- (e.g., wi'nava'a "so stay on therocks") and wik- "to be not" t.o wik'- (e.g., wi~1),s "to be not in a re-ceptacle, a canoe is empty"), so t'lum- (alternating with t'1 ~p-) "to behot" becomes t'lu'm- (alternating with t'lup-)(e.g., tlu'ma"a "to be hoton the rocks" and ilu'rnal),s "to be hot in a receptacle, there is hot water";compare iluPi'tS1), "summer, hot season" = parallel ilup- + '-i"tel),"season") and kan- "to kneel" (e.g., kanil "to kneel in the house") be-comes ka'n- (e.g., ka'na1J,s "to kneel in a canoe"). As there seem to be nostems ending in h or " the group 'm, 'n, 'w, 'y22 is left over as functionallyrelated to the group m, n, w, y in the same sense as the group exemplifiedby V is related to the group exemplified by p. Morphology, in otherwords, convincingly supports the phonologic proportion p:V = m: 'm. Itis maintained that it was this underlying phonologic configuration thatmade Alex hear 'm as sufficiently similar to Pto justify its being writtenin an analogous fashion. In other languages, with different phonologicand morphologic understandings, such a parallelism of orthographymight not be justified at all and the phonetic differences that actually

U m and n may be followed by a murmured vowel of i-timbre which is a re-duced form of a, u, or i. Syllables or half-syllables of type m

lor n

lare preceded

by i, an assimilated product of a, u or ii' inl and iml result therefore, in part, fromsequences of type ama, umi, anu. Simp e -em or -an become -ap, -at.

so A term borrowed from Boas' equivalent Kwakiutl phenomenon.SI The symbol' indicates the "hardening" effect of a suffix.II The phonologic details involving 'wand 'y and their relation to wand y

and other consonants are too intricate for a summary statement in this place.

ob~ain .between 'm and pwould have a significantly different psychologicweIghtmg.

V. In a course in practical phonetics which I have been giving for a~umber of years I have so often remarked the following illusion of hear-mg on th~ part of students that there seems no way of avoiding a generalphonologIc theory to explain it. I find that, after the students have beentaught to rec?gnize the glottal stop as a phonetic unit, many of them~nd to hear It after a word ending in an accented short vowel of cleartImbre (e.~., ~, E, e, i). This illusion does not seem to apply so often towor~s endmg m a long vowel or an obscure vowel of relatively undefinedqualIty (a) or an unaccented vowel. Thus, a dictated nonsense word like8mE or pild would occasionally be misheard and written as 8mE' andp,ild' but ther~ seems far less tendency to hear a final glottal stop in wordslIke pila or ptld:. What is the reason for this singular type of "overhear-ing?" Is it enough to say that students who have learned a new soundlike to play with it and that their preparedness for it tends to make themproject its usage into the stream of acoustic stimuli to which they areaske? to attend? No doubt such a general explanation is a correct dy-n~mlC form~a so far as it goes but it is not precise enough for a phonolo-gIst because It does not take sufficient account of the limitations of theillusion.

It must be remembered that the language of my students is English.We may therefore suspect that the illusion of a final glottal stop is dueto some feature in the phonologic structure of English. But English has~o glottal stop. How, then, can English phonology explain the overhear-mg of a consonant which is alien to its genius to begin with? NeverthelessI. believe that the students who projected a final glottal stop into th~dictated wo~ds were handling an exotic phonetic element, the glottalstop, according to a firmly established but quite unconscious phonologicpattern. It requires both the learning process, with its consequent alertprepar~ess t.o re~ognize what has been learned, and English phonologyto explam the IllUSIOn. If we study the kinds of syllables in English whichmay normally constitute an accented monosyllabic word or an accented(or secondarily accented) final syllable of a word we find that they maybe classified into three types: '

A. Words ending in a long vowel or diphthong e.g. sea flow shoe reviewapply. ' , , , , ,

B. Words ending in a long vowel or diphthong plus one or more consonantse.g., ball, cease, dream, alcove, amount. '

C. Words ending in a short vowel plus one or more consonants, e.g., back,fill, come, remit, object.

The theoretically possible fourth class:

D. Words ending in a short vowel, e.g, French fait, ami; Russian xllrllh'

does not exist in English. English-speaking people tend to pronouncewords of type D in a "drawling" fashion which tr~nsfen: them to t!~~ A(e.g., ami' for ami). Observe that the apparently mconsistent pOSSIbIlityof e. nonfinal accented syllable ending in a short vowel (e.g., fidd~, bu~ter,double pheasant) is justified by the English theory of syllabIficatIOn,which'feels the point of the syllabic division to lie in the following con-onant (d, t, b, z, in the examples cited), so that the accented syllables of

the words really belong phonologically to type C, no~ to t!pe D. Inter-vocalic consonants like the d of fiddle or z of pheasant, m spIte of the factthat they are not phonetically long, are phonologically "flanking" or two-faced in that they at one and the same time complete one syllable andbegin' another. Should the point of syllabic divis~on s~ift b~ck"of th~consonant, the preceding vowel at once lengthens.m spIte of It~ .shortquality (type A), and we thus get dialectic Amencan pronunCIatIOns .ofwords like fiddle and pheasant in which the accented vowel keeps Itsoriginal quality but has been lengthened to the unit length of "longvowels" of type feeble, reason, and ladle. .

We are now prepared to understand the illusion we st~rted WIth. Suchwords as 8mE and pila are unconsciously tested as pOSSIble memb~rs .ofclass A or class C. Two illusions are possible, if the hearer is to be a VI?tlmof his phonologic system. Inasmuch as a final accente~ sh~rt vowelI~ anunfamiliar entity, it can be "legitimized" either by pro)ectI.ng ~ength mtoit (misheard sme and pild' fall into class A) or by pro)ectmg a finalconsonant after it (class C). We shall call this imaginary consonant "x"and write 8mEX and pilax. Now the fact that one has added the glottaltop to his kit of consonantal tools leads often to the. tem?tation to solve

the phonologic problem symbolized as 8mEX and ptzax. m terms of theglottal stop and to hear 8mE' and pila'. The glottal stop IS the.most ~realor zerolike of consonants to an English or American ear and IS ad~rablyfitted, once its existence has been discovered, to serve as the p~oJ.ectedactualization of a phonologically required final consonant of mmimUIDsonority. The illusion of the final glottal s~op is essentially the illusion ofa generalized final consonant ("x") needed to classify the dictated wordsinto a known category (type C). Or, to speak more analytically, Englishphonology creates the groundwork (-x) of the synthetic illusi~n, while thelearning process colors it to the shape of -'. The ~rror of hearmg a glottalstop where there is none, in words of type ~, IS fu~damentallY a moresophisticated form of the same error as hearmg a dIctated final glottal

stop as p or t or k, which occurs fre uentl . .acquiring of a phonetic techniq q y m an earlier stage of the

T ue.he danger of hearing a glottal sto wh h .

long vowel or diphthong lOS of p den t e dICtated word ends in acourse ren ered very lik I b

that such words conform to a comm E I' h un e y y the facth on ng IS pattern (typ A) Threason wythe error does not so easil ° • ° e . e

ending in an unaccented short vowel : occ~r m ,h~ar!ng dIctated wordstoo, conform to an English patt \:.,0:, 8uoh) IS that such words,allowed a vowel in thl's pos'tO e:n, oug the range of the qualities

I IOn IS not as gre t hcovered by a follo . a as w en the vowel iswmg consonant (e.g., idea, very, follow).

A STUDY IN PHONETIC SYMBOLISM*

'1'RE SYMBOLISM of language is, or may be, twofold. By far the greaterportion of its recognized content and structure is symbolic in a purelyreferential sense; in other words, the meaningful combinations of vowelsand consonants (words, significant parts of words, and word groupings)derive their functional significance from the arbitrary associations be-tween them and their meanings established by various societies in thecourse of an uncontrollably long period of historical development. Thatthese associations are essentially arbitrary or conventional may be seenat once by considering such a proportion as

phonetic entity 'boy': idea (or reference) 'boy'- phonetic entity 'man': idea (or reference) 'man.'

In passing from the notion of 'boy' to that of 'man' we experience a defi-nite feeling of relationship between the two notions, that of increase insize and age. But the purely phonetic relationship of 'boy' : 'man' takesno account of this. So far as the referential symbolism of language isconcerned, the words 'boy' and 'man' are discrete, incomparable ph9neticentities, the sound-group b-o-y having no more to do with the sound-group m-a-n, in a possible scale of evaluated phonetic variants, than anyrandomly selected pair of sound-groups, say 'run' and 'bad,' have to dowith each other.

This completely dissociated type of symbolism is of course familiar;it is of the very essence of linguistic form. But there are other types oflinguistic expression that suggest a more fundamental, a psychologicallyprimary, sort of symbolism. l As examples may be given the interrogativetone in such a spoken sentence as "You say he's dead?" in comparisonwith the simple declarative tone of the corresponding "You say he'sdead"; further, the emphatically diminutive ee of teeny as contrastedwith the normal i of tiny. In both of these examples the phonetic differ-ence is undoubtedly felt as somehow directly expressive of the differenceof meaning in a sense in which the contrast between say 'boy' and 'man'is not. We may call this type of symbolism 'expressive' as contrastedwith the merely 'referential' symbolism which was first spoken of. It

• Journal of Experimental Psychology, 12 (1929): 225-239. Publication of theBehavior Research Fund, the Institute for Juvenile Research, Chicago (HermanM. Adler, Director), Ser. B, No. 132. For valuable suggestions in the preparationof this paper I am indebted to Professor H. A. Carr, University of Chicago.

1 For the two symbolic layers in speech, as in all expression, see Edward Sapir,"Language as a Form of Human Behavior," English Journal, 16 (1927): 421-433.

[61 ]

100101102103104105106107