Fromkin Non Anomalous

8/11/2019 Fromkin Non Anomalous

1/28

The Non-Anomalous Nature of Anomalous Utterances

Victoria A. Fromkin

Language , Vol. 47, No. 1. (Mar., 1971), pp. 27-52.

Stable URL:http://links.jstor.org/sici?sici=0097-8507%28197103%2947%3A1%3C27%3ATNNOAU%3E2.0.CO%3B2-M

Language is currently published by Linguistic Society of America.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html . JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/lsa.html .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community takeadvantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.orgMon Feb 4 14:22:49 2008
http://links.jstor.org/sici?sici=0097-8507%28197103%2947%3A1%3C27%3ATNNOAU%3E2.0.CO%3B2-Mhttp://www.jstor.org/about/terms.htmlhttp://www.jstor.org/journals/lsa.htmlhttp://www.jstor.org/journals/lsa.htmlhttp://www.jstor.org/about/terms.htmlhttp://links.jstor.org/sici?sici=0097-8507%28197103%2947%3A1%3C27%3ATNNOAU%3E2.0.CO%3B2-M


2/28

THE NON-ANOMALOUS NATURE OF ANOlULOUS UTTERANCES

VICTORIAA. FROMKIN

University of California, Los Angeles

An analysis of speech errors provides evidence for the psychological rea lity oftheoretical linguistic concepts such s distinctive features, morpheme structureconstraints, abstract underlying forms, phonological rules, and syntactic andsemantic features. Furthermore, such errors reveal th at linguistic performance ishighly rule-governed, and that in many cases it is grammatical rules which con-st ra in or monitor actual speech production. While a model of linguistic compe-tence is independent of temporal constra ints, a model of linguistic performance

must provide information as to the sequencing of even ts in real time. To explainthe occurrence of particular kinds of errors , a specific ordering of rules is posited,which ordering may or may not coincide with the organiza tion of a grammar.

1 I n current linguistic and psychological literature a sizable number of arti-cles have appeared dealing with 'slips of the tongue' and errors in speech (seeReferences). This interest is not, however, of recent origin. Historically, speecherrors have been a source of humor as well as of serious study. In the sixteenthcentury, Rabelais utilized such errors to display his pungent wit; and in theCompleat gentbman (1622), Henry Peacham refers to a 'melancholy Gentleman'who says 'Sir, I must goe dye a beggar' instead of the intended I must goe buya dagger'? 'Spoonerisms' were uttered before and after the long happy life of theReverend William A . Spooner, who is credited as the originator of a particularkind of 'lapse'. I n fact, if one assumes that the origin of man and the origin oflanguage and speech were simultaneous, then a further assumption follows-that'spoonerisms' began with Adam.

Speech-error data have been studied as a source of historical linguistic change(Sturtevant 1917, 1947; Jesperson 1922; MacKay 1970d); as a means for under-standing the actual mechanisms of the speech production process (Lashley 1951;Boomer Laver 1968; MacKay 1969, 1970a; Hockett 1967; Fromkin 1968;

Nooteboom 1969); and to gain insight into psychological repressions (Freud1924). Speech errors have also been investigated in attempts to show the 'reality'of phonological units and rules, and the relationship between linguistic 'compe-tence' and 'performance' (Fromkin 1968, Green 1969). Freud, in his Psycho-pathology o everyclay life, questioned 'whether the mechanisms of this (speech)disturbance cannot also suggest the probable laws of the formation of speech'([I9241 1938 : 71). It is to that general question th at this paper is directed.

Robbins 1966 suggests th at the earliest lite rary example is found in Rabelais, in the fol-lowing: Il n'y a point d'enchantement. Chascun de vous l a veu. J e y suis mazstre pass6. Abrum, a brum, je suis prestre Mack.' Th e contrived error of transposing t h e m and p in mais-tre passt (past master) creates prestre Mack (priest Ma&), 'a monk whose name was synony-mous wi th simple or foolish'. In the same article , Robbins 457-8) cites a 'near-spoonerism'found i n the Lives of celebrities by Jo hn Aubrey (1626-1697), who, discussing a flir tation be-tween Sir Walter Raleigh and a young girl, has the wench's protest 'Sweet Sir Walter'


3/28

28 L A N G U A G E , V O L U M E 47 N U M B E R 1 (1971)

2. Every book and article which refers to speech errors is replete xvith examples.

The most extensive collection, an estimated 8,800 errors, appears in AleringerMayer 1895 and in Sleringer 1908. A rigorous statistical analysis of these errorsis contained in a number of articles by MacKay 1969, 1970a, b, d. This corpus ofGerman errors is augmented by errors in spoken Dutch noted by Cohen 1966, bymore than a hundred errors in English tape-recorded by Boomer Laver, andby other errors cited in various articles listed in the References.

I n this paper, while taking into consideration the extensive published data, Iwill primarily make use of a collection of speech errors collected by myself overthe past three years. More than six hundred errors were collected by myself, orby colleagues and friends who reported in detail errors which they either made orheard others pr ~ d u c e . ~ or each error which I myself noted, I recorded the nameof the speaker and the date, and where possible (particularly in the case of blends)the speaker was questioned as to what he had been thinking of saying. This isscanty information indeed when compared with the data recorded by Aleringerfor each error in speech which he heard. In true Teutonic style, he also included thebirthdate of the speaker, the educational background, the time of day, the s ta teof health and tiredness, the rate of speech etc. Sturtevant reports th at Meringerthus became the most unpopular man a t the University of Vienna; and since nocorrelations between any of th e above factors and the nature of the error were

found (MacKay 1970d), my own data-collecting omitted such information, inorder to protect my personal reputation while maintaining the scientific accuracyof the data. I t is important to note, however, that my method of data-collectinghas a built-in fault, since many errors occur when it is just not feasible to notethem, and unquestionably many errors made are not heard at all. The data-collection method used by Boomer Laver, in which they analysed tapes ofconference discussions, psychiatric interviews etc. for the errors which they con-tained, is free of this fault. Fortunately, however, there were no sharp discrepan-cies between the kinds of errors recorded by them and by myself. There are cer-tain kinds of errors included in my corpus which did not seem to occur among the

hundred or so errors recorded by them; but only included such errors whenheard and attested by other listeners, or when the speaker himself caught theerror and corrected it. I felt this precaution necessary to mitigate my own de-sire to hear certain kinds of errors.

The aim of this paper, then, is not to treat the errors in the corpus as a randomsample of all errors made, but to attempt an explanation for the errors which wererecorded.

DISCRETENESS F PERFORMANCE UNITS

3 Sturtevant defines a lapse or a speech error as an unintentional linguisticinnovation (1947 :38). Boomer Laver s definition echoes Sturtevant s : A

Most of th e examples cited in th e text will be from my own da ta . I n th e citation of ex-amples, the arrow is to be interpret ed a s spoken as . Th e pronunciation of t he utte ran ce willbe given in phonetic symbols withi nsquare brackets only when th e orthography may create


4/28

T H E NON-ANOMALOUS NAT URE O F ANOMALOUS UTTE RAN CES 29

slip of the tongue is a n involuntary deviation in performance from the speak-er s curre nt phonological, gram ma tical or lexical inten tion(4). Because suchunintentional or involuntary errors m ay resultin utterances which provoke

laughter, speakers and writers have also used the m intentionally. Such consciouscreations will no t be considered here, althou gh one finds th a t these intentionalerrors usually follow th e same rules as do non-intentional erro m 3

Me ringer was mainly intere sted in classifying th e kinds of errors which oc-curred in spontaneous speech; an d since his time, one finds in t he literature dif-ferent classification schemes an d varying terminology. I n Boomer Laver sclassification scheme, speech errors show a'MISORDERINO of units in the string,OMISSION of a unit, or REPLACEMENT of a unit (5). According to the m, the un itsso misordered, omitted, or replaced may be segments, morphemes, or words.

Nooteboom 1969 classifies segme ntal errors as phonemic speech errors an dnon-phonemic errors , including in th e la tte r classification meaningless com bins -tion of phonemes , morph emes (including affixes an d root morph emes), an d wholewords. Nooteboom dismisses th e possibility t h a t distinctive features behavemore or less (like) independent elements just as phonemes do , bu t H ock ett im-plies th e independen ce of such feature s (915).

Fu rth er classification is not t he concern of this paper. Th e interest is rather inhow par ticular errors shed light on th e underlying u nits of linguistic performan ce,an d on th e production of speech. W ha t is apparent, in t he analyses and conclu-sions of all linguists an d psychologists dealing w ith erro rs in speech, is tha t, d e-spite the semi-continuous na ture of th e speech signal, there are discrete units a tsome level of PERFORM NCE which can be substituted, omitted, transposed, oradded. It should be stated here th at , were we to find no evidence in a ctua l speechproduction or percep tion for such discrete units, th is would be insufficient causeto eliminate discrete u nits in phonology or syntax . Th e fact t h at it is impossibleto describe the grammars of languages without such units is itself grounds forpostulating them in a theory of grammar. B ut when one finds i t similarly impos-sible to explain speech production (which must include errors made) withoutdiscrete performan ce units , this is fur the r substa ntiation of t h e psychological

Lewis Carrol l, in his preface t o he hunting of the snark 1876), disousses his 'portman-teau' words: '... let me take th is opportunit y of answering a question tha t has oft en beenasked me, how to pronounce slithy toves . Th e i in slithy is long, as in writhe ; and

toves is pronouncedso as to rhyme wit h groves . Again, th e first 0 in borogoves is pro-nounced like the 0 in borrow . I have heard people tr y to give i t the sound of the 0 inworry . Such is Huma n Perversity Hump ty- Dump ty's theory, of two meanings packed

into one word like a portmanteau, seems to me the right explanation for all. For instance,take the two wordsi'fuming and furious .Make up your mind t ha t you will s ay bot h wordsbut leave it unsettled which you will say first. Now open your mouth and speak. If yourthoughts incline ever so little toward s fuming you will sa y fuming-furious ; if they turn,by even a hair's breadth, towards furious , you will sa y furious-fuming ; bu t if you have

th at r ares t of gif ts, a perfectly balanced mind, you will sa y frumious .'I have quoted extensively from Lewis Carroll, not only because it is always delight toread or reread an y of his comments, but because i n this passage he st at es th at his 'port-manteaus' or 'blends' are possible i n natu ral speech and proposes a hypothesis as t o how theyoccur. As we shal l see below, however, these 'complete' blends are seldom found in just thi s


5/28

30 L A N G U A G E V O L U M E 47, N U M B E R 1971)

reality of such discrete units. I n other words, behavioral dat a of the kind de-scribed here may not be necessary to validate hypotheses about linguistic com-petence, but they certainly axe sufficient for such verification.

3 1 THE RE LITY OF THE SEGMENT OR PHONE.By far the largest percentage ofspeech errors of all kinds show substitution, transposition (metathesis), omission,or addition of segments of the size of a phone. These occur both within words andacross word boundaries, the latter case being most frequent in our corpus. Mostof these segmental errors are errors of anticipation, which is in keeping with theconclusions reached in t he literature. Simple anticipations result in a substitutionof one sound in anticipation of a sound which occurs later in the utterance, withno other substitutions occurring. The following examples illustrate such errors:

(1) a. John dropped his cup of coffee --t cuff coffeeb. also share --t sllsho share [DUO er]c. such observation --t sub suchd. delayed auditory feedback - audif auditorye. week long race - reek long racef. M-U values [em juw vseljuwz] --t [ ~ mjuw] valuesg. the third and surviving brother --t the sird and the bird the

thirdExamples la-e illustrate the substitution of one segment for another. I n If,

however, anticipating the [v], a segment is added where there is no segment in

the intended word. And in l g the error i s compounded: first the s is anticipated,and then, in an attem pt to correct the error, a later is anticipated.Perseverance errors are also not uncommon, as exemplified in the following:

2) a. I'm not allowing any proliferation of nodes --t proliperationb. John gave the boy --t gave the goyc. Spanish speaking people --t speaping peopled. irreplaceable --t irrepraceablee. Chomsky and Halle --t Chomsky and Challe

I t should be noted tha t one cannot unambiguously classify the error in 2c, since

it could be considered an error of either perseverance or anticipation. L4s shown byMacKay 1970d, the probability that errors occur when there are repeated pho-nemes is much greater than chance, and in this case the alliterative structure ofthe utterance seems to add to the substitution which occurs. As will be seen, thisis true of many of the errors to be discussed.

Classic Spoonerisms reveal a more complex error, in that there is a transposi-tion or metathesis of two segments. One possible interpretation is that such errorsinvolve an anticipation plus a perseverance, but it seems more likely that whatoccurs is a simple (or not so simple) switch in the linear ordering of the soundsintended. Such errors, attributed to Spooner, made him famous, as in his pur-ported admonition to a n undergraduate s tudent: 'You have hissed all my mysterylectures. I saw you fight a liar in the back quad; in fact, you have tasted a wholeworm' (Robbins).

Whether or not the notorious Reverend or his students sat up nights inventing


6/28

THE NOS ANOMALOUS NAT URE OF ANOMALOUS UTT ERA NCE S 31

such errors, attested errors reveal the same kind of metathesis, as is shown in theseexamples :

3) a. keep a tape-

teep a capeb. the zipper is narrow - the nipper is zarrowc. should serve --t [sud Barv]d. for far more --t for mar foree. with this ring I do wed - wing redf. I'm going to die young but I'll die less young --t yes lungg. in the past few weeks - fast pew [pjuw] weeks.

I n a number of cases, where the speaker catches his error, we cannot be surewhether a mere anticipation and substitution is involved, or whether a transposi-sition is caught before completed, as in the following examples:

4) a. Kathy can type - tathy Kathy can typeb. correct class of - collect correctc. s horn in the present slide shown in the pleasant I mean pres-

ent slided. greater pressure - [greyB~] greater pressuree. delayed auditory feedback - delayed audif auditory feedback

All the above examples reflect errors involving consonants. Vowels are also antici-pated, metathesized, etc., as shown below:

5 ) a. ad hoc [sed hak] --t odd hack [ad hsek]b Wang's bibliography Wing's babliographyc. tur n the corner - torn the kerner [tarn] [karnr]d. feet moving - [fuwt mijving]e. fish and tackle --t fash and tickle [fie81 [tik ]f the new Sony --t the no suny [now suwnij]g. place the stress - [PIES] he [strejs]h. dissertation topic [disytejgn taptk] - [disytagn tejpik]i. available for exploitation --t avoilable forj. prevailing temperature [prejvijlig]

k. the waterfall [wotyfal] isn't working-

isn't [wokiq]3 2 CLUSTERS S SEQUENCES OF DISCRETE PHONES OR SEGMENTS The above

examples show errors of transposition, substitution, omission, and deletion ofindividual segments, which may be either vowels or consonants. The error maybe either of anticipation (i.e., the interfering segment follows the error), of per-severation (i.e., the interfering segment precedes the error), or of transposition(i.e., the order of sound segments is changed). Further justification for assumingthat individual segments are units in speech performance is suggested by the factthat, in many errors where the intended utterance included consonant clusters,

only one segment of the cluster is involved:6) a. fish grotto -t frish gotto

b. fresh clear water flesh queer waterc. split pea soup --t plit spea soup


7/28

32 LANGUAGE, VOLUME 47, NUMBER (1971)

e. no strings attached -t no strings attrachedf at the Broadway stores the prices are --t spores the prices areg. in a split second --t slit secondh. that s a sticky point --t spicky pointi. a car with a stick shift [bttk stftI4

-4s seen in 6a, the intended jish grotto has been pronounced frish otto [frtb gaDo](the [Dl represents a voiced flap), the addition of an [r] in th e first word producingan initial cluster instead of the intended single segment. The substitution of thesingle [g] for the intended cluster [gr] may be explained by postulating that thecluster gr can be broken down into individual segments, [g] followed by [r].This being so, the individual segments can themselves be transposed. Similarly,th e error cited in 6b can be explained as an anticipation of the [I] in clear, causing

the replacement of the intended [fr] in fresh by [fl]. The substitution of [kw] in[knir], for the intended [ldir], may again be explained by an anticipation of the[w]in water. I t is of course true t ha t 6b may be simply an error in word substitu-tion, since JEesh is a word, as is queer. Such an explanation will not, however, ex-plain a number of the other examples given; i.e., [frtb] is not a word, nor is [gaDo],[pl~t] , spij], [blejk], [f ruw~d] , atraetgt], [spikij] etc. If we are seeking an explana-tion for such errors, it seems highly likely that we have here again single segmentalerrors, the difference being that the segments involved occur in consonant clus-ters.

The omission of elements or segments in clusters also justifies the conclusionth at clusters are not unitary units of performance, as in these examples:

7) a. two hundred drugs two hundred [d~g z] .b. property th at excludes [~kskluwdz] property th at [ekskudz]

Errors involving final clusters show that they are also sequences of individualsegments, as in the following examples:

8) a. ta b stops --t tap [stabz]b. weeks and months -t [w~gks]andc. great risk --t great rist [r~st]d. french fried potatoes frend fried potatoese. there s a pest in every class - petf ar t of the fugue arg of the [fjuwt]

That some errors reveal the transposition of whole clusters is NOT evidence forthe fact that such clusters are indissoluble units. Such errors do, of course, occurvery often, as in these examples:

9) a. at the bottom of the pay scale a t the bottom of the [skej peyl]b. little island in Brit tany brittle island in litanyc. sweater drying --t dreater swying [dreDr swaj~g]

d. throat cutting coat thruttingSuch movement of whole clusters is but further evidence that the syllable is

For the speaker who made this erro r, [ ] followed by a consonant is not an unusual se-


8/28

T H E NON-ANOMALOUS NAT URE O F ANOMALOUS UTTER ANC ES

no t a single indissoluble u nit in speech production, b ut is itself composed of asequence of segm ents. This is attested b y th e fact th at a C V or a VC sequencewhich is part of a syllable can be involved in speech errors:

(10) a. pussy cat, -- cassy putb. foolish argument -- farlish ...c. a hea p of junk a hu nk of jeepd. s tress and pi tch -- piss an d stretche. lost an d fou nd [fawst] an d [land]

Example 10a shows th e monosyllable [ h t ] as a sequence of three segments[k+ae+t], with th e first two segments transposed with th e first two segme ntsof [p+ v+s +ij]. I n 10d, th e transposition which occurs can easily be explainedas

Another explanation is th a t th e wordpiss is substituted for stress (the reasonsfor such a substitutionI leave to Freud), an dstretch for pitch or instead, th at thespe ker star ted to say pitch and stress a n d the error is one of final consonant sub -stitutions. There are however, numerous examples which show errors involvingCV o r VC sequences which can not b e so explained.

3 3 AFFRICATES.he assu mption t h at clusters on a performance level shouldbe interp rete d a s sequences of consonants raises th e question of affricates .It isinteres ting to note th a t whiIe [str], [pl], [kr], [bl], [fr] etc., a s well as final clusters,revea l th e splitting of clusters into segm ents, no t a single example in m y owndat a, or th e English exam ples cited by others, shows a spli tting of [tEi] or [dgjin to sequences of s to p plus fricative:

(11) a. pinch hit -- pinch hitch, b ut no t *[pint his]b. pretty chilly - chitty pilly [t g~ tij rl ij]c. entire cha pter enchire [antsajr]...d. further surgery -- furge r [far r] surge ry

e. R a y Jackendoff Ja y Rackendofff. last cigarette Tim had in Jun e ... J im had in tuneg. in St . Louis Joh n said -- in S t . Jouis John said

We do no t find cases like 'St. [duas]', or 'St. [Zuis]'. One m ay assume th a t th e oldphonemic controversy, as to whether such affricates should be considered onesegment or two, is solved for linguistic performance, an d t h a t affricates shouldbe considered single segm ents in th e production of speech, for speak ers of Eng lish.

3.4. COMPLEXVOWELS. One finds a similar situation with diphthongs. If [ey]or [uw] or [ew ] are interp reted as a succession of V y or V w, one couldexpect th e non-glide section of th e diphtho ng to be subject to substitution w ith-ou t a change of th e particular glide. I n other words, one would anticipate th a tfeet mowing might be articulated s [fvyt miwving]. Th e examples in 12 show th a twhere vowel glide or [r] is involved, th e error always includes th e entiredi h h h l i h i ' li '


9/28

4 L A S G U A G E , V O L U M E 47 N U M B E R 1 1971)

(12) a. first and goal to go first and girl to gob. took part in the first - took pirt [part] in the farstc. dissertation topic --, [drsytaSn tejprk]d. we're going to have to fight very hard - we're going to have to

fart very [fayd]e. feet moving [futvt mijving]f. available for exploitation avoilable for ...g. soup is served - serp is [sutwd]

These examples are, of course, taken only from English, and the conclusionsregarding affricates and complex vowel nuclei have meaning only for En&sh.

I t is a fact th at one never finds an error which results in a 'non-permissible'sequence of, for example, front vowel plus back glide (e.g. [tw], [EW]), r of back

vowel plus front glide (e.g. [vj]); but this may have an alternative explanation,which is discussed below. One example above, (5i) 'available [avejlab]] avoilable[av~jlab ]', could be interpreted as a switch only in the non-glide portion of thevowel nucleus, as could all examples of errors which involve only tense frontvowels or tense back vowels. The errors involving both front and back diph-thongs, along with those involving a van-el followed by r cannot be explainedin this way, and seem to suggest that the complex vowels are single units, orthat errors which 'violate' phonological constraints are 'corrected' after thesubstitution occurs. (See below for discussion on this point.)

3 5THE

STATUS OF[q]

IN EXGLISH.Sapir 1925 and Chomsky Halle 1968present arguments for deriving [q] from an underlying sequence of /ng/. Their

phonological analysis is justified in itself. It is of interest, however, that be-havioral data, found in speech errors, indicate that, a t one level of performance,[q] may derive from the sequence of [n g]-or, because of the constraints whichchange [n] to [q] before a velar, the sequence of [q g]:

(13) a. sing for the man [sig ... mzn] - [stg ... m q ]b. Chuck Young [tE;ak jlq] - [ t S ~ g k ~ g ]c. shilling [Stliq] - shingle [Bigdld. cut the string - [lrant] the [strig]

A possible explanation for the [gl's in the actual utterances is to postulate that,prior to the execution of the articulatory commands, the following transpositionof segments has occurred:

a. [srqg mzn] [srgg ... mzqb. [ t Lk j q g ] [ tF ;qk ~Og]c. [Stltqg] - [&iggal]d. [knt] [strigg] - [kant] ... [stri0g]

If this highly speculative hypothesis can be demonstrated by other experimentaldata, the postulated phonological rule for English, g - 0 / n- , may be vali-

dated, in that when the nasal is deleted, the [g] emerges.The data can, however, be given an alternative explanation. Example 13a

may show persistence of the velar articulation from [stq], producing [maeq],and a simple loss of the nasality of the final velar in sing. In 13b, since in English


10/28

T H E NON ANOMALOUS NATU RE OF ANOMALOUS UTTERA NCES 35

transposition of oral vowel with nasal vowel, and a concomitant non-nasaliza-tion of the final nasal:

Example 13c may be similarly disposed of, but 13d cannot be so easily explained.The only explanation, other than that which postulates an underlying abstract/s tr~ng/, s to suggest that only the nasality of the vowel is anticipated-which,as we shall see below, is certainly possible. The examples are given, however,since they permit speculation as to the reality of the g in utterances containing[QI.

3 6 THE REALITY OF PHONETIC FEATURES Research on the perception ofspeech has shown that units smaller than the segment, i.e. properties or featuresof speech sounds, are 'perceived' and confused (Miller Kicely 1955; Wickelgren1965a,b, 1966). Thirty-two cases in the present corpus can be explained bypostulating that certain properties or features also constitute independent ele-ments in the production of speech. The fact that one finds no errors in whichconsonants and vowels are involved (i.e., vowels do not switch with consonants,etc.) may be explained by suggesting that true vowels ([+vocalic, -consonantal])constitute one class of segments in a performance model, as opposed to anotherclass composed of true consonants, glides, and liquids ([+consonantal] or[-vocalic]), but th at the segments which axe members of these two non-inter-

secting sets cannot be further analysed into independent features.As we shall see below, there are other explanations for why consonants andvowels do not 'interfere' with each other (e.g., are not transposed, anticipated,etc.) The data, however, suggest that while a HIERARCHY probably exists, otherfeatures are independently involved in speech errors:

(14) a. spell mother - smell [ b ~ e ]b. pity the new teacher - mity the due teacher mean nity

the poor teacher no - pity the new teacherc. bang the nail - mang the maild. Did you hear what Malcolm said - did you hear what balcolm -

Malcolm said?e. Cedars of Lebanon - Cedars of Lemadon

These examples show a change in the value of the feature [nasality], acting inmany cases independent of other features. I n 14a-b, the [-nasal] of [p] becomes[+nasal] (i.e. [p] - [m]). If the [m] of mother remained [+nasal], this examplecould be dismissed as merely an anticipation of the segment [m]. However,since [m] --+ [b], or since the value of the nasality feature in the [m] of moth rswitches from [+nasal] to [-nasal], all other features remaining the same, abetter explanation for the error is that what occurred was a single feature switch.

Otherwise, no explanation is provided for the [m]-

[b] substitution.Example 14b illustrates the same phenomenon. [p], which is [-nasal], becomes[m], which is [+nasal], other features remaining in tact; and [n] is changed from[+nasal] to [dl, which is [-nasal].


11/28

6 LANGU AGE, VOLUM E 47, NU M BE R 1971)

-coronal], switches to [+nasal]; and [n] switches from [+coronal] to [-coronal].Even if one T\-isl~ed o explain the [m] of mail as a perseveration of the [m] whichhas occurred in [maer~], he substitution [b] - [m] u-ould be left unexplained. Theanticipation of the lowered velum which accompanies the following [n] is apossible explanation.

Example 14d represents a simple substitution, in the first segment of i l lalcolmfrom [+nasal] to [--nasal].

Th e following examples represent a change of value for th e feature [voiced]

15) a. What does the course consist of - what does the gorse consist ofb. referendum - reverendurnc. clear blue sky - glear plue skyd. reveal - refeele. define - devinef. big and fat --t pig and va t

I n these examples, only the value of the feature [voice] is changed, all otherfeatures remaining intact.

Other errors which appear to involve properties or features of whole sounds,rather than whole segments, are as follows:

16) a. pedestrian - tebestrian [p] - [t] and [dl - [b])b. scatterbrain [skseDrbrejn] - [spzeDrgrejn]c. spaghetti - skabettid. Death Valley [d ~f? ~ l i j ] - [fee tizelij]

I n 16a, only the value of th e feature [coronal]-i.e., only the P L C E of articula-tion-is changed. It is of course possible to argue that this is rather to be inter-preted as segmental transposition, with [p] - [t] in anticipation of the subsequent[t]. But what of [dl - [b]? f we explain [p] - [t] as a switch of labial [-coronal])to alveolar [+coronal]), then [dl - [b] is seen as the result of a change fromalveolar [+coronal]) to labial [ -coronal]).

Similar cases are seen in 16b-c. Again, one can suggest a segment transposition,particularly since the voicing feature of the [g] is neutralized after an [s]. Butthen how does one explain the [g] -- [b] switch? If a mere segment transpositionwas involved, we would expect [sgapetij] in 16c.

more complex error is seen in 16d: the switch from [dl -- [f] seems to be ananticipation of the subsequent [v]. The coronality of the [dl seems to influencethe switch from [v] to [dl, with the [+voice] and [+continuant] features re-maining.

I t is certainly true th at errors which involve a substitution of features are rare,compared to errors involving larger units. They nevertheless require some ex-planation, and one can only conclude that some features appear to be independ-ently extractable as performance units. Many segmental errors may also beexamples of such feature errors; but since they can also be accounted for as errorsof larger units, we are unable to conclude that individual features are independ-ently involved in all cases. However, the following examples show that feature


12/28

T H E NON-ANOMALOUS NATUR E OF ANOMALOUS UTTERANC ES 7

(17) a. extracted --t extrapted ([k] --t [p], or [+back, -ant] --t [-back,+ant]>

b. goofing off --t gooping off ([f] - [p], or [+cant] - [-cont])c. call the girl --t gall the curl ([k] --t [g], or [-voice] - [+voice])d. documentation --t documendation ([t] - [dl, or [-voice] -

[+voice])I n fact, most of the segmental errors can be so interpreted. I n the transpositionof brake luid to br ke fmid, one might suggest that what is involved is a trans-position of the feature [lateral] or [anterior] rather than transposition of the twosegments. If segmental errors are analysed as feature errors, we will find thatmany distinctive features other than those cited above do indeed represent areality, in speech performance.

This suggestion is supported by the findings of Nooteboom: 'In significantlymore cases than is to be expected in a random distribution the two elementsinvolved in a substitution error are phonetically similar to one another' (1969:121). MacKay found th at 'most pairs of reversed consonants differed in only onedistinctive feature (56 percent) and very few 2 percent) differed in all fourdistinctive features' (1969).5 This is in contradiction to the conclusion of Boomer

Laver 'that articulatory similarity is not an important determinant' in speecherrors, although they do note two exceptions: 'sequences of voiceless fricativesseemed to encourage mistakes of place of articulation, and (b) alveolar con-sonants showed a slight tendency to interact' (p. 8, and fn.) But they wereexamining errors to see if any particular features were involved more often thanany others. Their data were not analysed for the degree of similarity of the seg-ments involved. I t is interesting to note th at an analysis of jargon aphasia errorsalso shows that most errors involve no more than a confusion of one distinctivefeature (Green). Whether or not further analysis of substitution errors confirmsor contradicts the MacKay-Nooteboom conclusions regarding the 'similarity'of substituted segments, the only conclusion one can draw from the examples offeature switching given above is that at least some of the proposed distinctivefeatures are independent behavioral units.

But an examination of the errors, whether analysed as errors of whole seg-ments or of independent features, definitely shows a hierarchy and interdepend-ence of certain features. Thus, while there are errors showing just addition orsubtraction of nasality, one does not find a 'nasality' switch which results in avoiceless nasal. A t least for English, nasality and voicing seem to be interde-pendent features. This again prevents the occurrence of an 'inadmissible' soundin English.

The claim that certain features are independent units, which must be postu-lated as such in a model of performance, seems to contradict the earlier hypothesisthat segments (or feature complexes) are 'real' performance units. Actually,there is no contradiction. Features cannot exist except as properties of largersegments (just as segments, as we shall see, exist as parts of larger units). I nother words, in the generation of speech, there is a hierarchy of different-sized


13/28

8 LANGUAGE, VOLUME 47 , NUMBER 1 1971)

units. linear ordering of the segments (discussed below) occurs, and thislinear ordering may be disrupted. Since the discrete segments are specified byactual physiological properties (or neural commands), some of these propertiesor features may also get disordered, i.e. attached to other segments. Bu t theclaim that all distinctive features (as proposed by Chornskg Halle) are identicalwith phonetic properties that can in principle be independently controlled inspeech is not borne out by the da ta of speech errors. Unless controllable inspeech is defined in some new strange and abstract way, it n- odd appear th atwhatever the needs for certain separate phonological features may be, in actualspeech performance only certain of these phonological features have theircounterpart as phonetic feature^.^ Thus, while the two features [consonantal]and [vocalic] very nicely divide segments into four separate classes, needed for

phonology, the idea that such features have any independent phonetic realityseems highly improbable. To suggest that a substitution of a [p] for a [k] involvesthe PHONETIC substitution of [-anterior] for [+anterior], [-high] for [+high],[-back] for [+back], etc., is saying no more, on an articulatory level, than statingthat there is a change from a velar articulation to a bilabial articulation. Themotor commands to the muscles, specifying a bilabial or velar articulation,specify the part of the tongue to be raised, where it is to be raised, etc. I n otherwords, on a phonetic level a complex of the features [-anterior, +high, +back]is indissolubly a velar place of articulation, and one does not expect to find(and indeed, one doesn t find) a simple switch of the feature [+coronal], forexample, without other phonetic effects. I n the example pedestrian - tebestrian ,the error can be specified as a switch in the feature of coronality, but i t is obviousth at this feature is not independently controlled . What I am suggesting is th atsegments as feature complexes do exist; that some of these features or propertiescan be independently controlled, such as nasality, voicing, place of articulation

i considered as a single multi-valued feature) etc.; but that some propertiesare highly dependent on the existence of other properties of the segment. I t isthus tha t [delayed release] does not seem to be independent of affricates, and onecan only suppose that, on the neuro-physiological level, there is some command

for a stop closure combined with delayed release, which command cannot besplit into two segments. Tha t is, the command for the initial and final consonantsof church at one level of the generation of speech is a command for just such anaffricate. On the other hand, when one says did you as [d~dBuw], n rapid speechthe affrication occurs by a different process, i.e. by automatic and mechanicalmovements of the vocal organs. However, the results a t the level of muscle move-ments are identical.

3.7. THE REALITY OF THE SYLLABIC UNIT. While it seems plausible to assume,as was done above, that units smaller than syllables (segments and features)

Chomsky r Halle are of course concerned wi th the grammatically determ ined aspec tsof th e signal. The occurrence or lack of occurrence of speech er rors involving phonetic fea-tur es are being discussed in t his articl e as they rela te t o a model of linguistic performancerat her tha n competence. However, when Chomsky Halle ta lk abo ut the se t of phonetic

h l b ll d h 295) d ff l f d h l


14/28

T H E NON-ANOMALOUS NAT UR E OF ANOMALOUS UTTERA NCES 39

have independent s tatus as behavioral units, this does not negate the possibilitythat syllable-size units are also units of speech performance. I n fact, all the

evidence from tongue slips supports such a view. Nooteboom (1969:119) sug-gests th at since 'the distance between origin and target (or the substituted seg-ments) does not generally exceed seven syllables, (and) since we know that theshort memory span of man may contain about seven units we might interpretour findings as an argument for the syllable to be a unit in the phonemic pro-gramming system7. Nooteboom 1969, MacI


15/28

40 LANGUAGE, VOLUME 47, NUMBER (1971)

h. pan-cakes - can-pakesi. neural mo-de-ling - neural mo-le-ding

Because of the co-articulation effects of segments within a syllable, it is impos-sible to omit the syllable as a unit of articulation, even if one were to ignore theevidence of the fixed order in the reversal, anticipation, or perseveration of seg-ments (Fromkin 1968).

There are of course many errors which involve the substitution, omission,replacement, addition etc. of one or more whole syllables, which further sub-stantiates the claim that syllabic units are real performance units:

(20) a. Morton and Broadbent point - Morton and Broadpointb. revealed the generalization - reeled the generalizationc. tremendously - tremenly

d. which I analyse as the following --, which I analyse as the followe. butterfiy and caterpillar - butterpillar and catterflyf opacity and specificity - opacity and specifityg. we want to reveal all the facts - we want to feel all

In many of the above, several factors are a t work. Some of these examples arehat are commonly called blends , as are the following:

(21) a. Did you bring your clarinola (a blend of clarinet plus viola)b. switch and changed - swinged [swindBd]c. importance of [adLajsnt] rules (a blend of adjacent plus adjoin-

ing )d. my data consists [monlij] - [mejstlij] (a blend of mainly plus

mostly )

4. The speech of jargon aphasics, as well as errors made by non-pathologicalspeakers, reveal that normal slips of the tongue and aphasic jargon utterancesare constrained by the linguistic system. One does not find phonemes (or morecorrectly, phones ) which are not found in regular utterances. For example, an

English speaker does not substitute a rounded front vowel in anticipation of arounded back vowel, nor a lateral click for a lateral liquid. Furthermore, onlypermitted segmental sequences occur. Wells 1951 stated this as his First lawof tongue slips: A slip of th e tongue is practically always a phonetically possiblenoise. It is obvious that Wells meant a phonetically possible noise in a particularlanguage. As I have stated in an earlier article (Fromkin 1968): The segmentsconstituting each syllable must have sequential ordering, so that only initialconsonants, vowels, and final consonants may interchange, IF AND ONLY IF THETE4iXSPOSITIOSS A R E I N KEEPING WITH THE PHONOLOGICAL RULES O F THE LAN

GCAGE (64). This First Rule appears to explain a Spoonerism attributed to

Spooner :sphinx in moonlight - minx in spoonlight.

What is of interest here is the transformation of the [sf] in sphinx to [sp] whenh l i d i h h [ ] Whil [ f]d i d lik hi


16/28

THE NON ANOMALOUS NATU RE OF ANOMALOUS UTT ERA NC ES 4

Webster's Third) are 'exceptions' to the regular morpheme-structure rule inEnglish which permits only voiceless stop obstruents after an initial s. [sfuwn-light] would thus not be a permitted sequence, and consequently [f] --t [p].

All the examples already cited include only permitted English sequences.Further examples will support the 'reality' of such constraints:

(22) a. play the victor --t flay the pictorb. tab stops - tap [stabs]c. plant the seeds [sijdz] --t plan the seats [sijts]d. bloody students [b hd ij stuwdants] t [bladant stuwdijz]

There are two ways of interpreting the error shown in 22a. One might suggesttha t i t is simply the manner of articulation (stop vs. fricative) which is switched.f such an interpretation is given, one must also add that, when the [v] is changedto a stop, the place of articulation changes from labio-dental to bilabial. Anotherpossible explanation is t ha t the two segments switch (p v , and that since [vl]is not a permitted sequence in English, the [v] is devoiced. This suggests thatthese phonological constraints, when learned, become behavioral constraintswhich occur AFTER the segmental transpositions occur.

A similar example is shown by 22b, in which the final consonant (or just thevoicing feature) of the first word is transposed with the penultimate consonantof the second word (or the final stem consonant, prior to the plural morphemeaddition). Again, the intended [ps] is changed not t o [bs] but to [bz], in keepingwith the phonological (and morphological) constraints of English.

Examples 22b-d represent another phenomenon. In these errors, the originalsyntactic s tructure of the phrases remains intact, in that the intended pluralnouns occurring as the last words of the phrases remain as words with pluralendings, despite the errors which occur; but the phonetic realization of the pluralmorpheme changes, as well as the preceding segments. Thus [stabs] and [sijts]can be explained simply as due to phonological or phonetic constraints, since[bs] and [tz] never occur as final clusters; but the error in 22d is more complex.[js] can occur in English as in Reese [rijs], mi e [majs], feast [fijst], face [fejs] etc.Bu t [ij+s] cannot occur when the final sibilant represents the plural morpheme.

One can then suggest that the phonetic representation of the plural morpheme isspecified prior to the automatic phonetic specifications which serve as the unitsfor articulatory commands. f this were not the case, one could not understandthe change of the [s] to [a] in [stuwdijz].

Further examples of the reality of morphophonemic rules are evidenced inerrors which include the alternation of the non-specific determiner a a n :

(23) a. a current argument [a knrnt argjumant] an arrent curgument[an arnt k~rgj uma nt]

b. an eating marathon --t a meeting arathonc. a history of an ideology t an istory of a hideologyd. an ice cream cone --t a kice ream cone

The changes a --t a n and a n t a indicate tha t, in the generation of speech, thesegmental errors or transpositions must take place PR OR to the actual neural


17/28

42 LANGUAGE, V O L U M 47, NUMBER 1 1971)

Such errors show the separation of morphophonemic rules and phonologicalrules. I n other svords, it is not a phonological rule which changes the a t o an,since there is no general restriction on vowel sequences like those of America is,Rosa and I . Thus the ordering of events must be as follows: (1) segmental errors,(2) morphophonemic adjustments, 3) P-rules.

The reality of the P-rules is attested by many of the errors cited above, e.g.(8a) [ t ~ btaps] --t [tsep stabz]. The transposition of the /b/ and /p/ must haveoccurred prior to the rule which constrains final clusters to be voiced or voiceless.I n 8b, [wijks an mnnOs] [wtrjlrs ..I can only be explained by the followingsequence: /wiks/ --t /winks/ - [wigks]. The tense I is not diphthongizedbecause i t occurs before a nasal, and the /n/ is made homorganic with thefollowing velar stop by a general rule.

6 MacKay, Boomer Laver, and Nooteboom 1969 all investigate theinfluence of stress on errors in speech. Boomer Laver conclude that Theorigin syllable and the target syllable of a slip are metrically similar, in that bothare salient (stressed) or both are weak (unstressed), with salient-salient pairingspredominating 7). Nooteboom agrees with this conclusion, stating that Insignificantly more cases than is to be expected in a random distribution theelements involved in a speech error belong to stressed syllables (1969). Hedisagrees, however, with Boomer Laver s finding that Slips involve the tonic(primary stressed) word, either as origin or as target, with tonic origins pre-dominating. But from Kooteboom s o m data, the disagreement seems to be theresult of a misinterpretation of the difference between primary stress (tonicword) and salient stress. Differences between English and Dutch may also berelevant. MacKay finds that transpositions occurring within words appear insyllables with different stress, while in between-word reversals his findings cor-roborate those of Boomer Laver.

What seems to be of greater interest is that , when vowels or syllables or par tsof syllables or whole words are substituted or transposed, there is no change in

the stress pat tern or contour of the sentence. Boomer . Laver cite an examplein which a speaker, instead of saying how bdd things were , said how thingsbad were . It is evident that there was no transposition of the stress, despite thetransposition of the words. The following examples show the same phenomenon(an acute accent [ I represents primary stress, as does 1 above the vowel; agrave accent [ I represents non-primary stress-secondary or tert iary; a 2 abovethe vowel represents secondary stress, and a 3 tertiary stress).

(24) a. hammer and sickle -- sickle and hammer

b. peoples park --t [pa ruz pijp]

c verge of a nervous breakdown --t nerve of a vergeous breakdown2

d. he s been around a long time - he s been long around time1


18/28

T H E NON ANOMALOUS N ATUR E OF ANOMALOUS UTTE RAN CES 4

1

computere

f. examine the eyes of the horse - examine the horse of the eyes8 8

g. broke the crystal on my watch - broke the whistle on my crotch

h. in the theory of phonology - in the phonology of theoryExamples 24e and 24h show that , while the word position of primary stress inthe phrase is not transposed, the stressed syllable of the word in isolation is thesyllable which receives the sentence stress. That is, if the primary stress is to beplaced on 'laboratory', i t is placed on the first syllable; and if it is to be placedon 'computer', i t is placed on the second syllable.

Thus i t seems that two aspects of stress must be accounted for: first, the word

stress moves with the word itself (i.e. the syllable of the word which receivesmain stress in isolation also receives the primary stress when the word is moved) ;second, the stress contour of the phrase is fixed by the syntactic structure of thephrase itself, and must be generated independently of the word order in theutterance.

One may then suggest tha t the word stress is stored as part of the articulatoryspecifications of the stored unit 'word', but that th e sentence or phrase stress andover-all intonation contour is generated separately, as part of what BoomerLaver c d he 'tone-group'. would therefore agree with them th at 'The pivotalrole of the tonic word in slips suggests that it s phonological, syntactic and se-mantic prominence is matched by an analogous neuro-physiological prominence,coded in the brain as a part of the articulatory programme' 8), and furtherthat 'the tone group is handled in the central nervous system as a unitary be-havioral act, and the neural correlates of the separate elements are assembledand partially activated, or 'primed' before the performance of the utterancebegins' 9). However, in the construction of a model of linguistic performance, i tis necessary to specify the nature, i.e. the syntactic structure, of this tone group,for the 'priming' of the 'tonic' syllable depends on the syntactic structure of theutterance.

The suggestion that the stress placement on words is fk ed in the lexicon doesnot mean that one cannot, or should not, attempt to generalize stress assignmentrules n the phonology of English. I n fact there may be some evidence fromspeech errors that not only in a grammar of competence, but also in the actualstored lexicon, words (or perhaps formatives) are stored in a more abstract formthan by their actual articulatory specifications. There are speech errors whichdisplay a movement of stress, and in certain cases a change in the vowel qualitiesdepending on where the stress is placed:

(25) a. This can viewed altdrnately alternatively - no Alternatelyb. similarly simil ralij]c. homogeneous - [hom dZanfjas]d. in favor of [ham a - homo fjnias]e. syllabff - syUbification [srlabif - sd4bafakej6nlf opAcity and specificity opscity and sp6cifity


19/28


20/28


21/28

46 L A S G U A G E VOLUbfE 47, N U M B E R (1971)

(29) a. l l y data consists [mo~vnlij] [mejstlij] ... (mainly/mostly)b. I swindged [s~vindid] .. (switch/changed)c. It s a lot of [ba] [bra31 (bother/trouble)d. She s a real iswip] chick (sninging/hip)e. it s a [spajratrv] (spirant/fricative)f. a tennis [~Olar] piaperlathlete)

Such errors seem to support Carroll s assumptions. speaker has in mind somemeaning which he wishes to convey. In selecting words, it appears that he ismatching semantic features. Where there are a number of alternative possi-bilities, rather than making an immediate selection, he brings them both into abuffer storage compartment, with their phonological specifications. Either aselection occurs at this point, or the IT-ords are blended, resulting in the above

kind of errors.The literature and my o nn data attest the fact that, besides the phonological

s i d a r i t y in substituted words, errors often involve semantic features in com-mon, or substitution of antonyms, i.e. words having the same features withopposite values:

30) a. I really like to --hate to get up in the morningb. It s a t the bottom I mean top of the stack of booksc. This room is too damn hot coldd. the oral writ ten par t of the exam

Xooteboom presents a number of examples which seem to involve a semanticsnitch from the space to the time dimension (1967: 14) as in the following:

(31) a. the two contemporary, er sorry, adjacent buildingsb. during the apparatus, er behind the apparatusc. the singular, sorry, the present time

Evidence from aphasia studies also show that substituted words often fallinto the same semantic class, as in cases where patients will read tree for flower,night for dark, spoon for fork, liberty for democracy etc. (Marshall Newcombe1966, Luria Vinogradova 1939, Jakobson 1966). Such errors provide importantevidence as to the storage of vocabulary and the generation of speech.

8 1 THE LEXICON When one learns a language, he learns among other thingsa vocabulary. Judging both from errors of speech and from speakers ability toform new words by adding derivational affixes to stems (e.g. He s a real com-puterish type ) and by inflecting newly coined words in keeping with the rulesof the language (e.g. 22d, [stuwdijz]), it seems plausible to assume that thestored lexicon consists of stems and a x e s , as well as idioms, compounds, wholewords etc. Given the higher than chance probability that prefixes and suffixes

are involved in syllable errors (JlcKay, unpublished), one can further assumethat, even if w-ords are stored with their a x e s , the stem and affix have a sepa-rate status. Thus it is not unlikely th at grouping is stored as group ing,which permits a substitution of m nt for the affix ing The fact tha t one does not


22/28

THE NON ANOMALOUS NAT URE OF ANOXALOUS UT TER AN CES 7

find stems substituting for or transposing with affixes further justifies theirseparate status.

Since phonological or phonetic specifications, semantic features, and syntacticword-class features all play a role in the speech errors that occur, it is obvioustha t vocabulary items must be stored with such features indicated. But we cannotsimply assume that there is one dictionary-like storage starting with all wordsbeginning with A and ending with all words beginning with with other featuresgiven. Semantic errors show that words are selected to convey certain meaningsas specified by their semantic features. And for literate speakers the listing mustalso specify the orthography, t o account for the ability of people to play 'geog-raphy', a game in which one must name a country, river, city etc. beginning withthe same LETTER with which the previous word ended: thus, Passaic ends with

the letter c, pronounced [k], and the next player can offer Charleston, whichbegins with c pronounced [ti] . The relationship between orthography and soundmust be accounted for. Crossword puzzles, double-crostics, and the 'tip of thetongue' phenomenon (Brown McNeil 1966 also at test this fact. For example,it is often the case tha t in trying to remember someone's name, forgotten at themoment, a speaker gill say, I know it begins with a C.' The name may beCohen, which begins with a C pronounced [k]. And of course a game like 'geog-raphy' is further evidence for the storage of words in semantic classes.

One may then suggest t ha t t he vocabulary is stored in a thesaurus-like latticestructure. I t is possible to conceive of this network as a listing of all the stemsand affixes in some fixed phonological order, each one with all of its feature speci-fications, and each one with a particular address. The separate semantic sectionof this lexicon may then be divided into semantic classes, with semantic featuresunder which are listed, not the particular vocabulary item, but the addresses ofthose items which satisfy the features indicated. One might suggest also that thelistings under the semantic headings are grouped under syntactic headings suchas [+noun], [+verb] etc., to account for the proper grammatical selection in thegeneration of utterances.

Since th e 'tip of the tongue' phenomenon suggests that speakers recall the

number of syllables-the metrical beat of the word-a further division underthe full phonological listing is suggested. In other words, it is not impossible toassume that ll monosyllables beginning with the same phonological segmentconstitute one block, followed by disyllables, etc.

The error cited in 30a might then occur in the following way: the speakerwishes to say (at least on a conscious level-we leave the unconscious motiva-tions to be explained by others) I really h te to get up in th mornin At the pointin the generation of the utterance prior to the selection of the words, in the'slot' representing hate the features [+verb, -desire I occur, and an addressfor a word is sought from the semantic class which includes [ desire]. But

either because of unconscious uishes or due to a random error, the address for averb with the feature [+desire] rather than one specified s [-desire] is selected,and the item at that address called forth with its accompanying phonologicalfeatures turns out as [lajk] rather than [hejt].


23/28

8 LANGU AGE, VOLUM E 47, NUM BER 1971)

The complexity of the stored lexicon is enormous, and i t is obvious that thereare too many lacunae in our knom~ledge o suggest anything more than thekinds of sub-parts or components it must contain. I have suggested an indirect-addressing system above with nothing to justify this accept a vague appeal tostorage simplicity. It seems plausible to assume, however, that any model of alexicon must include the following sub-parts

(a) A complete list of formative5 with all features specified, i.e. phonological,orthographic, syntactic, and semantic.

(b) A subdivision of the phonological listings according to number of syllables.This is necessitated by the fact that speakers can remember the number ofsyllables of a word without remembering the phonological shape of the syllables.I t is also suggested by the fac t th at one can get a subject to produce a list of one-,

two-, or three-syllable words.(c) A reverse dictionary sub-component, to account for the ability of speakers

to produce a list of words all ending in a particular sound or letter.(d) A sub-component of phonologically grouped final syllables, to account

for the ability of speakers to form rhymes.(e) Formatives grouped according to syntactic categories, to account for the

errors noted above, and the ability of speakers to list nouns, or verbs, or adverbson command, as well as the more important ability to form grammatical sen-tences.

(f) Formatives grouped according to hierarchical sets of semantic classes.(g) Words listed alphabetically by orthographic spelling.Furthermore, it seems plausible to assume that all these components must be

intricately linked in a complicated network.This highly speculative, oversimplified model of the lexicon is suggested as a

first approximation to what must be a most complicated storage mechanism.Wh at seems certain, however, is that any model of the lexicon must account forthe observed types of errors, which require the specification of various kinds ofproperties which we have called phonological, syntactic, and semantic features;no lexicon consisting of a single listing of items can explain what occurs.

8.2. THE GENERATION OF UTTERANCES. I t seems quite evident from all theexamples of speech errors cited above that , in the production of speech, i t is nottrue that anything goes, or th at speech performance obeys no rules, or tha t theerrors are totally random and unexplainable (see discussion of this in Frornkin1968 . While we may not be able to explain as yet the exact mechanisms involvedin speech errors, the errors made are not only highly constrained, but provideinformation about speech performance which non-deviant speech obscures. Inother words, if we had no record of errors in which consonant clusters are splitinto segments, are would not be able to justify the assumption th at clusters inperformance are strings of individual discrete segments.

Any model of speech performance must therefore account for the kinds oferrors which do occur. Such a model must account for the following:

(a) th at features, segments, syllables constitute uni ts in the production of ah


24/28

T H E NON ANOMALOUS NATURE OF ANOMALOUS UTT ERA NCE S 9

(b) that segments are ordered within a syllable, and that only segments simi-larly ordered are involved in the same error;

(c) that 'root morphemes may be interchanged but root morphemes and anaffix cannot take each other's places' (Nooteboom 1967: 16), or th at words of thesame syntactic or morphological class usually interchange with each other;

(d) th at intonation contours (including the placement of primary stress)remain fixed, and are generated separately from the individual word stresses;

(e) that morphological constraints and phonetic or phonological constraintsmust occur a t different times in the production of an utterance;

(f) th at non-permissible phones or phonetic sequences do not occur;(g) th at errors may be semantic in nature, as in the case of blends or word-

substitutions involving similar semantic features; and

h) th at the similarity of the phonological form of words appears to play arole in word substitutions.

To account for such phenomena we may suggest the following (over-simplified)order in the actual generation of an utterance:

STAGE . A 'meaning' to be conveyed is generated.STAGE 2. The 'idea' or 'meaning' is structured syntactically, with semantic

features associated with parts of t he syntactic structure. For example, if aspeaker wishes to convey the fact tha t 'a ball' rather than 'a bat' was thrown bya boy, the utterance ball was thrown or alternately e threw a bdll is structureda t this stage. If he uses the second structure, pa rt of the features specified for thefinal nouns must include [+emphasis] together with the features selected for'ball,' i.e. [-animate, -human, +count, +round, +used in games etc.] Thissuggests that the STRUCTURE itself is put into buffer storage prior to actualarticulation of the utterance; this would account for the switching of noun fornoun, verb for verb etc., when such transpositions occur.

STAGE 3 The output of Stage is thus a syntactic structure with semanticand syntactic features specified for the word slots. I n order to explain the factthat 'the tone group is handled in the central nervous system as a unitary be-havioural act' (Boomer Laver, g), one can suggest that the intonation contour,

with the placement of primary stress, occurs at this stage. Since a transpositionof words in the utterance will cause a transfer of primary stress to the mainstressed syllable of the word in a given position, one can posit that only theposition of the primary stress is indicated a t this stage, and not the particularsyllable. That is, the generation of t he sentence intonation contour must occurprior to the selection of t he words themselves.

STAGE 4. We now have in t he buffer a syntactic phrase with semantic featuresindicated, and with sentence stress assigned. A lexicon look-up now occurs;the semantic class sub-section of the lexicon is st consulted, with featuresbeing matched, and the direction is obtained to go to a certain address in theover-all vocabulary. The item in the specified address is then selected, this wordbeing specified as to its phonological segments, which are identified and orderedinto syllabic units. At this stage in the process, errors resulting in a choice of a' g' d S h i l th t hi g f l f


25/28

LANGUAGE, VOLUME 47, NUMBER 1 1971)

meaning t o be conveyedR

syntact ic-semant ic

I

Lexicon

s t r u c t u r e s w i t hsernant i cclasses

t c t a l v o c a b u l a r y( a l l f e at ur es )

r imary s t r e s sand in tonat ions p e c i f i e d

go t o 1010 _ _ j iOlC: word speci f iedas t o f e a t u r e s

- s y l l a b i c o r - e r of segmnts

strings of segmentsd iv i de d i n s y l l a -b le s - - syn tac t i c /phonological F'sspec i f i e d

f u l l y s pe c if i edphonetic segmentsi n s y l l a b l e s

L

U T T E R N C E

norram 1 Utterance generator

semantic features, resulting in a wrong address being specified. Or the correctaddress may be specified, hut different address substituted which is in thevicinity of the intended address. Thus, if t he intended word is l ik and the


26/28

T H E NON-ANOMALOUS NATU RE O F ANOMALOUS UTTE RAN CES 5

and the produced word is present, the correct address is obtained, but the wrongaddress selected, given that pressure and present have addresses in the same sec-tion of the vocabulary. This would be due to the phonological similarity of theirfirst three segments. This process thus results in a string of phonological segments,each segment specified by certain features or properties and also specified as tosyllabic order, with the syntactic bracketing remaining intact. But it is at thisstage, when the string of phonological segments is put into the buffer, that arnis-ordering of segments may occur. In other words, as the segments are sentinto the short-term memory buffer, segment 1 of syllable may be substituted forsegment 1 of syllable 4. I am not concerned at this stage with an explanation ofwhy and how this occurs, but with the fact that it can occur at this stage withoutdisturbing the syllabic ordering. It is also here that whole syllables or parts of

syllables may get transposed or misplaced. These errors must occur before Stage5 which is where the morphophonemic rules or constraints take over.

STAGE . The morphophonemic constraints of the language at this stagechange, if necessary, or perhaps spell out , the phonological shapes of mor-phemes. The segmental errors must occur before this stage to account for thealternations of a an and s z of the plural.

We have now reached the stage where automatic phonetic and phonologicalrules take over, converting the sequences of segments into actual neuro-motorcommands to the muscles in the articulation of the utterance.

The above stages may be diagrammed as shown in Figure 1. It must beemphasized that the various black-boxes are highly schematic, and whatactually occurs in them is outside the concern of this paper. Rather, the attemptis to show a possible ordering of events in the production of an utterance whichcan account for non-deviant utterances, as well as for utterances containingerrors in speech?

REFERENCES

APPLEGATE, . B. 1968. Segmental analysis of articulatory errors under delayed auditoryfeedback. POLA reports, 8.) Berkeley: University of California.

BOWDEN, . H. 1899. A study of lapses. Psychological Review, Monograph Supplement3, no. 4.

BOOMER, 1968. Slips of the tongue. British Journal of Dis-. S., and J. D. M . LAVER.orders of Communication 3.1-12.

BROADBENT, . E. 1966. The well ordered mind. American Educational Research Journal3.281-95.

BROWN,., and D. MCNEIL. 1966. The tip of the tongue phenomenon. Journal of VerbalLearning and Verbal Behavior 5.325-37.

CARROLL, EWIS. 1876. The hunting of the snark. London: Macmillan.CHOMSKY, 1968. The sound pattern of English. New York: Harper., and M . HALLE.

Row.COHEN,A. 1966. Errors of speech and their implication for understanding the strategy of

language users. Instituut voor Perceptie Onderzoek, Annual progress report, 1.Eindhoven. To appear in Models of speech, Proceedings of the International Con-gress of Psychologists, Moscow.)

FREUD IGMUND.924. Zur Psychopathologic des Alltaglebens. 10th ed. Leipzig: Inter-


27/28

52 LANGUAGE, VOLUME 47 NUMBER 1 (1971)

nationaler Psychoanalytischer Verlag. (English version in Basic writings of Sig-mund Freud , ed. by A. A. Brill. New Yo rk: Modern Library, 1938.)

FROMKIN,V. A. 1968. Specula tions on performance m odels. Journ . of Ling. 4.47-68.GREEN,E. 1969. Phonological and grammatical aspects of jargon in an aphasic patient:

a case study. Lang. and Speech 12.80-103.HOCKETT,C. F . 1967. Wh ere the tongue slips, there slip I. T o honour R oma n Jakobson,

2.910-36. (Jan ua linguarum , series major, 32.) T he Hag ue: Mou ton .JAKOBSON . 1964. Tow ards a linguistic typo logy of apha sic imp airments. D isord ers of

language, CIBA Fou nda tion Symposium, 21-46. Lond on.. 1966. Linguistic ty pes of aph asia. Brain function, ed. b y E . C . C artere tte, 67-91.

Berkeley Los Angeles: Un iversity of California Press.JESPERSEN, 1922. Language: its nature, development, an d origin. London: Allen a nd

Unwin.KO ZH EVN IKO V, 1965. Speech: articulation and perception. A., and L. A. CHISTOVICH.

(revised). (Joint Publications R esearch Service, 30,543.) Washin gton: U .S. De par t-ment of Commerce.

LASHLEY,K. S . 1951. Th e problem of serial order in behavior. Cerebral mechanisms inbehav ior, ed. by L. A. Jeffress, 112-36. New Y ork : Wiley.

LA VER , . 1969. T he de tection an d correction of slips of th e tongu e. W ork in progress no.3 D ep t. of P honetics an d Linguistics, Edinb urg h University, 1-13.

LURIA,A. R., and 0 S. VINOGRADOVA.959. An objective investigation of t he d ynam icsof sem an tic sys tems . B rit . Jo urn . of Psych . 50.89-105.

MACKAY,D . G. 1969. Forw ard an d backward masking in motor systems. Kybern etik6.57-64.

. 1970a. Spoonerisms: the a na tom y of errors in the serial order of speech. T o app ear inNeuropsychologia.

.

1970b. Context dependent stuttering . To appear in Ky bernetik.. 1970c. Spoonerisms of children. To appear in Neuropsycho logia.. 1970d. Sound change an d errors in speech. Unpublished.

MARSHA LL,. C., and F. NEWCOMBE.966. Sy nta ctic and seman tic errors in paralexia.Neuropsycholog ia 4.169-76.

MERINGER,R . 1908. Aus dem Leben der Sprache. Berlin.-, and K. M AY ER . 895. Versprechen un d Verlesen, eine p sycl~olog isch inguistische

Studie. Vienna.MILLER,G. A. 1962. Decision units in t he perception of speech. I R E T ransaction s on

Informa tion T heory, IT-8 , 81-3.- an d P . E . NICELY. 955. An an alysis of perceptual confusions amon g some Eng lish

conson ants. Jo ur na l of th e Acoustical Soc iety of America 27.338-52.NOOTEBOOM,. G. 1967. Some regularities in phonemic speech errors. (Instituut voor

Perceptie Onderzoek, Annual progress report,2. Eindhoven.. 1969. The tongue slips into patterns. Nomen: Leyden studies in linguistics and

phonetics, ed. by A .G. Sciarone et al., 114-32. Th e Hagu e: Mouto n.ROBBINS,ROSSELLHO PE. 1966. Th e warden s word play: tow ard a redefinition of th e

spoonerism. Dalhousie Review 46.457-65.SAPIR,EDWARD.925. Soun d p atte rn s in langu age. Lg. 1.37-51.S T U R T E VA ~ ,. H. 1917 Lingu istic change. Chicago: Un ivers ity of Chicago Press.

. 1947. An introductio n t o linguistic science. New Ha ven : Yale University Press.WELLS,R. 1951. Predicting slips of th e tongue. Yale Scientific Mag azine, Decem ber, pp .

9-12.

WICK ELGR EN,. A. 1965a. Acoustic similarity an d intrusion errors in short-term m emory.Jo urn al of Ex perim enta l Pscyhology 70.102-8.. 1965b. Distinctive features and errors in short-term memory for English vowels.

Jo ur na l of th e Acoustical Society of America 38.583-8.. 1966. Distinctive features an d errors n short-term memory for English consonants.


28/28

You have printed the following article:

The Non-Anomalous Nature of Anomalous UtterancesVictoria A. Fromkin Language , Vol. 47, No. 1. (Mar., 1971), pp. 27-52.Stable URL:http://links.jstor.org/sici?sici=0097-8507%28197103%2947%3A1%3C27%3ATNNOAU%3E2.0.CO%3B2-M

This article references the following linked citations. If you are trying to access articles from anoff-campus location, you may be required to first logon via your library web site to access JSTOR. Pleasevisit your library's website or contact a librarian to learn about options for remote access to JSTOR.

References

The Well Ordered MindD. E. Broadbent American Educational Research Journal , Vol. 3, No. 4. (Nov., 1966), pp. 281-295.

Stable URL:http://links.jstor.org/sici?sici=0002-8312%28196611%293%3A4%3C281%3ATWOM%3E2.0.CO%3B2-C

Sound Patterns in LanguageEdward Sapir Language , Vol. 1, No. 2. (Jun., 1925), pp. 37-51.Stable URL:http://links.jstor.org/sici?sici=0097-8507%28192506%291%3A2%3C37%3ASPIL%3E2.0.CO%3B2-K

http://www.jstor.org

LINKED CITATIONS- Page 1 of 1 -
http://links.jstor.org/sici?sici=0097-8507%28197103%2947%3A1%3C27%3ATNNOAU%3E2.0.CO%3B2-M&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0002-8312%28196611%293%3A4%3C281%3ATWOM%3E2.0.CO%3B2-C&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0097-8507%28192506%291%3A2%3C37%3ASPIL%3E2.0.CO%3B2-K&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0097-8507%28192506%291%3A2%3C37%3ASPIL%3E2.0.CO%3B2-K&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0002-8312%28196611%293%3A4%3C281%3ATWOM%3E2.0.CO%3B2-C&origin=JSTOR-pdfhttp://links.jstor.org/sici?sici=0097-8507%28197103%2947%3A1%3C27%3ATNNOAU%3E2.0.CO%3B2-M&origin=JSTOR-pdf

Date post:	03-Jun-2018
Category:	Documents
Upload:	ichigo-knight
View:	238 times
Download:	0 times

Fromkin Non Anomalous

Documents