1978 caramazza berndt pb

8/12/2019 1978 Caramazza Berndt PB

1/21

Psychological Bulletin1978, Vol. 85, No. 4, 898-918

Semantic and Syntactic Processes in Aphasia:A Review of the LiteratureAlfonso Caramazza and Rita Sloan BerndtThe Johns HopkinsU niversity

Recent investigations of lexical and syntactic aspects of language comprehensionin aphasia are reviewed. It is argued that these studies support theoretical as -sumptions concerning the functional independenceof various componentsofnormal language processing. Studies of the structure of the lexicon in aphasiaprovide support fo r componential theories of lexical semantics in that differenttypes of features of meaning can be selectively disrupted under conditionsofbrain damage. Studies of sentence comprehension support the existence of asyntactic mechanism that is independent of lexically based heuristic strategiesfor assigning meaning. There is evidence that these independent elementsoflanguage are subserved by different portions of the dominant hemisphere of thebrain. Focal brain damage can thus cause selective disruption of components,allowing the separation of elements that are highly integrated in the normaladult. Studies of aphasic language, therefore, provide a valuable source ofcon-straints on theoriesof normal language processing.

In a recent influen tial book, Fod or, Be ver,and Garret t (1974) s tated : It is .. . thesad t ru th that remarkably li t t le has beenlearned about the psychology of languageprocesses in normals from over one hundredyears of aphasia study (p. x i v ) . This viewmay not be far from the truth. Historically,the study of language dissolution has beenalmost exclusively th e province of neurolo-gists, whose concerns were not so much withthe normal organization of language pro-cesses as with the correlation between brainstructures and specific language behaviors.Although traditional work in aphasia haslargely failed to provide d irect information

During the writ ing of this review, Rita SloanBerndt was supported by Public Health ServiceTraining G ran t 5 TC X MH 06742-18 to the JohnsHopkins University.T he authors would like to thank Oscar Marin forhi s comments on an earlier version of the article.They owe a special debt to Edgar Zurif, who al-lowed them to borrow freely many of the ideasthat they have developed in collaboration withhim over th e past fe w years.Requests fo r reprints should be sent to AlfonsoCaramazza, D epa r t m en t of Psychology, T he JohnsHopkins U niversi ty, Bal t imore, Maryland 21218.

concerning the organization of normal lan-guage processes, it has established a solidfoundat ion from which more recent aphasio-logical research can address these questions.Wh a t is most important is that this earlywork has established that al though no twoaphasics are exactly alike either in thenature of the disorder or in the consequentlanguage per formance, it is the case thatthere are discernible pat terns of dissolution.Focal brain damage to the dominant hemi-sphere (left hemisphere in most r ight-handedindividuals) does not simply result in anoverall undifferentiated reduction of languagecapacity or a random constellation of symp-toms (Geschwind, 1970, 1 9 7 2 ; Luria, 1970 ) .On the contrary, it appears that differentparts of the brain subserve different linguisticfunctions.For psychologists, however, of more im-por tance than the brain/funct ion correlat ionis the ident i f icat ion of different types ofaphasias characterized by predictable constel-lations of symptoms. T he occurrence of suchwell-organized patterns of dissolution mayreflect natural divisions in the organizationo flanguage processes in normal adults. Armed

C opyright 1978 by the American Psychological Association, Inc. 0033-2909/78/8504-0898$00.75898


2/21

APHASIA 899with this well-established fact that languagedissolution follows predictable patterns, re -cent investigators of aphasic language havebegun to ask questions that are of concernnot only to aphasiologists but also to psy-chologists concerned with normal languagefunctioning.There are twoarguments that can be madeconcerning the importance of aphasia re-search to the study ofnormal processes. First,models of normal language functioningshould not be inconsistent with the dataobtained with aphasic populations. Since cur-rent theories of language are so underdeter-mined by data, it would be foolhardy toignore a source of constraint on permissibletheories of language. The second and moreimportant argument is based on the beliefthat there is ultimately more than a trivialconnection betw een psychological processesand the brain. This stronger claim is the onethat motivates this review. It is specificallyargued that in the ideal case, brain damageselectively impairs discrete components oflanguage, so processes that are highly inter-dependent in the normal adult can be moreclearly identified. Even in cases involvingcomplex patterns of dissolution, it is possibleby a judicious process of comparing variouspatient types to identify components of pro-cessing that are so closely intertwined in thenormal adult that they are irretrievable.Ultimately, research motivated on the basisof this claim should generate new hypothesesabout the structure of language processing.This review focuses on aphasia researchthat addresses tw o components that are in-volved in sentence comprehension: lexicaland syntactic processing. The basic issuesaddressed at the level of lexical processingconcern the question of the representation oflexical info rm ati on. At the syntac tic level,the concern is with determining whethersyntactic processes are independent of se-mantic an d heuristic processes in sentencecomprehension.

Comprehension in AphasiaResearch in aphasia has typically focusedon deficits in the productive capacities ofbrain-damaged patients, that is, on overt

speech behavior. Discussion of impairmentofthe ability to comprehend spoken languagehas been limited, for the most part, to im-pressions that a patient's comprehension isrelatively impaired or clinically intact.Such global descriptions of the phenomenonof language comprehension obscure the factthat a highly complex interactive system ofcomponents an d processes underlies the ex-traction of meaning from a particular config-uration of words. It is likely that damage todifferent parts of the brain will differentiallydisrupt these components of comprehension,just as the various elements of productivespeech are differentially affected.Several recent investigations have lookedmore carefully at the process of comprehen-sion in an attempt to disentangle the com-ponents involved and to determine whetherthey can be selectively impaired in compre-hension failure (Goodglass & Baker, 1976) .One result of these efforts ha s been the iden-tification of elementsof comprehension deficitthat are not readily apparent to clinical ob-servation. For example, the class of patientsusually said to have relatively intact com-prehension (those with anterior damage)1is now believed to be deficient in the abilityto integrate correctly understood lexicalitems into certain types of syntactic frames(Caramazza &Zurif, 1976). The impressiono f intact comprehension in the face of im-paired syntax may be produced by the pa-tient's adoption of communicative strategiesbased on redundant cues to meaning thatmight occur within the sentence or the sur-rounding context.It also appears to be possible to examinecomprehension breakdown in the class ofpatients who exhibit poor comprehension(those with posterior damage). On the basisof their speech production, these patients aresaid to lack an understanding of individuallexical items but to have retained an implicitThere are several classifications of aphasic

symptoms currently in use ; for present purposes thegeneral division between patients with focal damageto the anterior and to the posterior portion of thedominant hemisphere is used. Patients with anteriordamage include Broca's an d expressive types; thosewith posterior damage include Wernicke's and fluentaphasics.


3/21

900 ALFONSO CARAMAZZA AN D RITA SLOAN BERNDTunders tanding of the syntactic rules fo r com-bining words into grammaticalsentences.Al-though lexical disruption is undisputed as acause of the comprehension failure sufferedby these patients, it is not known to whatextent syntactic difficulties can be implicatedas a contributing factor, sincean unders tand-ing of syntax in the absenceof lexical com-prehension is difficult to measure. Neverthe-less, th e comprehension failures exhibited bythese patients remain amenable to furtheranalysis, since the lexical disruption experi-enced is rarely total (Goodglass &Geschwind,1976). Most often, some elementof aword'smeaning is retained, so a patient knows th eclass to which a word belongs (e.g., colorterms, body parts) without grasping its fullmeaning.Thus, th e meaning components thatare spared can be exam ined and comparedwith those that are lost in an at tempt toidentify th e elements involved in comprehen-sion failure.In general, therefore, there is reason toexpect that the process of comprehension canbe analyzed into component elements to pin-point the site of breakdown; that is, languagecomprehension is no more a unitary, all-or-none phenomenon than is the process oflanguage production, and there is no reasonto believe that comprehension disruptionshould be global. Rather, just as in produc-tive disturbances, brain damage should causedamage selectively.

Lexical Knowledge in AphasiaA great deal of research in aphasia hasbeen directed at describing the parameters ofindividual word loss associated with variousloci of brain damage (Geschwind, 19,67).Some red uction of available vocab ulary ap-

pears to be a very general symptom of alltypes of aphasia (Good glass & Geschwind,1976) . Loss of an individual lexical item ismost often viewedas a word-finding difficultyand may be signaled either by a patient'sinability to name a single item that is pre-sented or by his inability to find the rightword to express a message he has spontane-ously fo rm ulate d . These two types of word -finding difficulties, which necessitate di f-ferent levels of involvement of both sensory

stimulation and a knowledge of syntax,would be expected to present different typesof problems to the brain-dam aged speaker,and indeed they do (Geschwind, 1967; Green,1970). The present discussion is limited toproblems involving confrontation naming,that is, the patient's ability to provide averbal label for an object that is presentedto him.Since the main focus of this review is theproblem of comprehension impairment, it isnecessary to consider the possible parallelsan d divergences between th e ability tounder -stand a word and the ability to produce it.Such a comparison necessitates a selectivetreatment of language comprehension, how-ever, inthatit deals only with the referentialaspect of meaningwith the relationship be-tween th e word and the object or event towhich it refers. Comprehending a word'smeaning involves more than merely beingable to identify a referent (Alston, 1964);the sense an d reference aspects of meaningare discussed below.For present purposes, however, th e pro-cesses of both naming and lexical comprehen-sion ar e limited to the referential si tuation;in that way the similarities and differencesbetween the two can be readily compared.For expositional convenience, the processesinvolved in both naming an d lexical compre-hension are viewed as consisting of threeseparate stages. Each of these stages in-volves many other substages an d processes,any one of which may be of central impor-tance to a particular constellation of symp-toms. The purpose of this outline is not toprovide a comprehensive description of thepossible levels at which disruption can takeplace. It is instead an attempt to describesome general correspondences between whatmust happen for apatient to be successful ina naming task and in a comprehension task,and thereby to delimit the domain of interestof this review.In the typical naming task, a patient ispresented with a concrete object or a pictureand is asked to produce a name for thatobject .The ability to apply a name correctlywithin this limited context can be said toinvolve the following threeelements:1. An encoding stage, in which the stimu-


4/21

APHASIA 901lus is perceived (either visually or throughsome other sensory modality) and the fea-tures that permit it s identification are ab-stracted from the total perceptual event.2. A central stage, in which the informa-tion abstracted from the stimulus in theencodingstage is integrated and mapped ontoa semantic representationa conceptualcategory of which the stimulus object is amember. A second mapping process thenrelates that semantic representation to aparticular lexical item, that is, the word thatis its name.3. A production stage, in which the lexicalrepresentation that hasbeen accessed is trans-lated intoa set of motor commands that makepossible the articulation of the correct phono-logical sequence.Comprehension tasks take several differentforms; the interest here is in tasks that assessunderstanding of individual lexical itemsrather than combinations of words. Apatientis most typically asked to point to a particu-lar item whose name is supplied aurally. Insome cases the referent object is part of arelatively open-ended set; that is, it can beanything in the immediate environmentbody parts,piecesof furniture, and so forth.Often, however, a set of items is presentedto the patient, and he must point to thereferent of the word presented. The stagesinvolved in this process can be described asfollows:1. An encoding stage, in which the acousticstimulus is encodedas a particular phonolog-icalshape.

2. A central stage, in which the phono-logical matrix representing a particular wordis mapped onto a semantic representationthat contains, among other things, descriptiveinformation about the physical form a re-ferent of the word wouldtake.3. A response stage, in which the descrip-tion of the referent object generated by thesemantic representation (in the form of amental image, a perceptual frame, a set ofpropositions, etc.) is compared with thepossible objects available in the environ-ment and a response is made based on thatcomparison.These parallel three-part schemes segmentthe processes of both naming and compre-

hending single words into stages that can bethought of as either peripheral or central tothe task at hand. The peripheral processesinvolve those aspects of functioning that linkthe central cognitive operations and linguisticstructures to the outside world. In theencod-ing phase of the naming paradigm and in theresponse phase of the comprehension task,the mechanism s involved are perceptual, mostoften visually perceptual. Both th e produc-tion phase of the naming task and the encod-ing phase of the comprehension task involveaspects of phonological processing.There is great potential for selective dis-ruption of one of these peripheral processessuch that the capacities of verbal productionan d comprehensionare differentially affected.The operationsthathave been labeled hereasperipheral involve finely tuned powers ofdiscrimination and integration within severaldifferent sensory modalities. Although disrup-tion at the level of these peripheral mecha-nisms has received a large share of the atten-tion of workers in the field (e.g., Geschwind,1965; Luria, 1973), the concern of this re-view is with the potential for disruptionofthe component that is central to both nam-ing and comprehension in the normal adult.This central processing phase, involving bothprocedural and structural elements, has onlyrecently been considered as the possible levelat which the aphasic patient's semantic sys-tem isdisrupted.At the core of all of the mechanisms andoperations involved in determining a word'smeaning there mustbesome form of semanticrepresentation of the informational elementsthat make up a par ticula r conceptual entity.Although no assumptions will be made at thistime about whether that information is rep-resented as sets of features, or propositions,or as a mental image, the assumption is madethat the processes of both understanding aword's meaning and finding the right wordforra particular situation involve accessingthis stored representation. Also included aspart of this central component are the pro-cesses involved in translating perceptual in-put (au d itor y, visual, etc.) into a form thatpermits access to the correct representation.In addition, another set of mapping pro-cedures is needed to relate the information


5/21

902 ALFONSO CARAMAZZA AN D R I T A SLOAN BERNDTcontained in the semantic representation tothe proper output mechanisms.

It is logically possiblethat either the map-ping procedures or the semantic representa-tion itself may be disrupted by damage tothe brain. Thus, it is possible that the map-ping procedures that relate input and outputto the semantic representation may be dis-rupted while the representation itself remainsintact. However, a pathological conditionaffecting the structure of the semantic repre-sentation would also have a disruptive effecton the procedures that mediate between therepresentation and the peripheral mecha-nisms, since a part of the information neces-sary for the mapping to take place would belost. Thus, an impairment at the level of thesemantic representation itself would appearto have the most generalized impact on apatien t's ability to process lexical inf orm a-tion.The foregoing discussion is intended tospecify the set ofphenomenaor , alternatively,the level of linguistic functioning with whichthis review is concerned.The domain of spe-cific interest is limited to the central mecha-nisms involved in extracting word meaning,and it is believed that the same processesand structures are par t of both naming andlexical comprehension. The next section ex -amines theextent towhich research inaphasiahas implicated these central processes asfactors in both naming an d comprehensiondeficits.Research on Naming Impairment

Virtually al l aphasics exhibit some impair-ment of naming ability, but such impairmentis rarely total (Goodglass & Geschwind,1976). This situation has led to many at-tempts to determine th e variables that in -fluence whether a given word will be lost.Factors such as word frequency (Wepman,Bock, Jones, & Van Pelt, 1956), form class(Marshall &Newcombe, 1966 ), picturability(Goodglass, Hyde, & Blumstein, 1969), se-mantic category (Goodglass, Klein, Carey, &Jones, 1966), and operativity (Gardner,1973) have all been implicated, to varyingdegrees, in naming disruption.One important an d stable result of this

effort has been the delimitation of two quali-tatively different types of naming disorder,which are highly correlated both with otherpatterns of symptomatology and with locusofdamage. One group of patients (those withfocal damage to the anterior portion of thedominant hemisphere) has many concreteand picturable nouns available and often per-forms quite well, albeit very slowly, in nam-ing tasks. For these patients, naming disrup-tion appears to bepar tof a more generalizeddisturbance, since other symptoms includelaborious articulation and a severe disruptionof productive syntax.A second group that can be defined on the'basis ofnaming performance displays a nota-ble deficiency o f concrete, picturable nounsan d most often fails to provide the correctname for an object, although their responsesare many times not far off target (Rin nert&Whitaker, 1973; Schuell & Jenkins, 1961).This impairment carries over into spontane-ous speech, which is characterized by fluentarticulation and intact syntax, but a lack ofcontent words. This type of naming impair-ment, termed anomia, is generally associatedwith lesions in the posterior portion of thedominant hemisphere.Although this standard classification ofaphasic symptoms captures the gross dif-ferences in naming disruption, there is someindication that more finely tuned techniquesof measurement are needed to isolate unitaryconstellations of symptoms, even within oneof the major classes as described above. Animportant recent review of the literature onaphasia has provided the following caveat toa summary of the two major classes of nam-ing disruption:Clinical impression suggests that word-finding diffi-culty is an inclusive term fo r several qualitativelydifferent defects that are not distinguished by ordi-nary naming tests. Thus, the patient who appears tohave dissociated th e sound from th e concept gives avery different impression from th e patient who isslow but sure or the patient who acts convincinglyas though he has recovered the inner sound of theword but has trouble in recovering th e articulatorymovements for it. (Goodglass & Geschwind, 1976,p. 403)

Such clinical impressions are valuable be-cause they underscore the complexity of theproblem; how ever, they are cited here for a


6/21

APHASIA 903different reason. Both the clinical and theexperimental literature on naming disruptiontend to account for the impairment in termsof peripheral mechanisms or retrieval failures.Thepossibilityof a disruptionof the semanticrepresentation itself is less often entertained.The passage quoted above mentions threequalitatively different types of naming dis-order in which the disrupted process is in-ferred to be a retrieval deficit (the first andsecond example) or a problem of articulation(the third example).T he apparent belief that naming disruptionis the result of retrieval failure, that is, thatthe semantic representation will be found tobe intact if it can be activated, is not lim-ited to clinical observations but pervadesthe experimental literature as well. Onemanifestation of this view has involved theattempt to differentiate the set ofitemsthatare likely to be named from those that arenot. The results of this differentiation areoften discussed in terms of the concept-arousing potential characterizing the itemsmost likely to be named.Gardner (1973) tested the ability of pa-tients with anterior and posterior lesions toprovide names fo rpictures of items thatwereeither operative, that is, discrete an dreadily manipulated, or figurative, that is,not easily manipulated. Both sets of wordswere equally picturable and occurred withequal frequency in the language. The namingadvantage obtained for the operative set wasexplained by the postulation that namingdepends on the capacity of the stimulus to

arouse some subset of the actions or sensoryexperiences usually involved in activity withthe object. Gardner argued that operativeitems were named most easily because theyaroused associations in several sensory mo-dalities, whereas the figurative items werelimited to the visual modality.More direct evidence concerning the role ofsensory modality in naming has been pro-vided by Goodglass, Barton, and Kaplan(1968) . These investigators found no d i f -ference in patients' ability to name objectsbased on whether they were presented forvisual, tactile, olfa cto ry, or aud itory naming .This result indicates that naming disruptionis not a function of peripheral problems at

the level of processing sensory input. Rather,Goodglass et al. argued that the uniformityof namingabilityacrossmodalities supportsthe notion of a process which intervenes be-tween the perception of any stimulus and thearousal of its name (p . 494). Goodglass etal. cited the research of Bisiach (1966) ,which demonstrated that obscuring the per-ceptibility of an object adversely affectedpatients' ability to name it but did not af-fect ab ility to recognize it. In ad ditio n, if in-formation about the object was availablethrough several different sensory modalities,even though it was partially obscured, theprobability that a name would be providedwas greatly increased (North, 1971). Thus,North argued that the various sensory mo-dalities contribute information additivelytoward the arousal of an object's name.Again, retrieval of the (presumably intact)concept is believed to be disrupted.The research on sensory mod ality reportedabove did not discriminate between patientswith different typesof naming difficulties, butthere is some evidence that the amount ofinformation present in a stimulus item mightno t have the same importance for patientswith anterior and with posterior damage. Ina large-scale investigation of the effects ofsemantic category on naming and comprehen-sion d eficits, Goodglass, Klein, C arey, andJones (1966) found that the items with thegreatest info rmatio nal value (objects andactions) were not named more easily thanthe items with little info rma tiona l value (let-ters, numbers, and colors). Anterior aphasicsexhibited uniform ability to name across allcategories, and the posterior aphasics didmost poorlywith theobject names.The difficulties of posterior aphasics innaming concrete objects do not appear tobeamenable to aconcept-arousal ex planation.The most frequent nouns in the vocabulariesof these patients are nonpicturable, abstract,an d often very general terms that conveylittle information (e.g., something, any-body ) . Although early work explained thepresenceof this type ofwordas a simple over-representation of high-frequency words in thevocab ularies of these patients (Wep man etal., 1956), more recent work has argued thatsuch indefinites are not only of high fre-


7/21

904 ALFONSO C A R A M A Z Z A AN D RITA SLOAN BERNDTquency but are highly predictable elementswithin the context of a sentence (Goodglass,Hyde, & Blumstein, 1969). Since patients ofthis type ca n produce grammatical sentencesand have difficulty finding words, they main-tain fluency by substituting words that arehighly predictable and high in frequency butlow in informational content.The available literature on naming dis-orders has not provided a great deal of in-formation concerning the level at which theprocess of naming is disrupted. The workreviewed here favors a view that a conceptmediates between the level of sensory stimu-lation and the production of the word. N ospecific information is provided about howthat concept might be structured, and disso-lution of structure as a correlate of namingdisorder is not considered. Disruption is pre-sumed to be at the level of retrieval; that is,the pathology is believed to impose a limi-tation on the processes necessary to arouseth e concept, such that increased informationmay be necessary before the concept can beactivated.Nam ing and Semantic Structure

It is somewhat surprising that an explana-tion of naming impairment has historicallyavoided a consideration ofdisruption at thelevel of the semantic representation, since ithas long been recognized that the incorrectnames produced byaphasic patients are gen-erally not random (Head, 1926) . Rather, th eword that is generated is often related inmeaning to the target word. This phenome-non would seem to require an explanationbased on the representation of semantic in-formation, since that is what appears to beconfused. Nevertheless, early attempts toaccount for this out-of-focus naming in -voked retrieval difficulties. For example,The patient may not be able to perceive theword so well that he is able to understand it or torepeat it, but sufficiently enough so that the sphereof meaning to which it belongs is elicited, and thepatient may summon up another word belonging tothis sphere. (Goldstein, 1948, p. 91)

More recent research has focused on therelationship between the words that are in-volved in naming substitution. Schuell and

Jenkins (19S9, 1961) observed that the er-roneous naming responses of aphasic patientsresemble the word associations of normal in-dividuals. More recently, Rinnert and Whita-ker (1973) analyzed a large corpus of pub-lished data in an attempt to determine th esemantic relationship between what a patientsaid (or wrote) and the correct target item.When the two words (the word produced andthe target) were compared on the basis ofshared semantic features, it was found thatthe two terms tended to share features rep-resenting major semantic categories. Thefeatures on which they differed were specificfunctional distinctions. For example, substi-tutions of pen for pencil and pipe for cigarretain information about the superordinatecategory of writing implements or smokingmaterials, but confuse some specific but im-portan t distinguishing features. When thesedata were compared with word associationnorms fo r normal speakers (Postman &Kep-pel, 1970) , a striking similarity was found.Rinnert and Whitaker postulated that th eword associations of normal adults are basedon the same lexical organization that is dis-turbed inaphasia.If the naming errors produced by aphasicpatients are comparable to normal word as-sociations and if both can be taken as anindication of how the internal lexicon isstructured (cf . Whitaker & Whitaker, inpress), then it is reasonable to think thatlanguage-impaired patients could themselvesproduce word associations that might provideinformation ab out semantic structure. Adirect test of the extent to which normalword associations are impaired in aphasia wasreported by Howes (1 9 67 ) . A standard wordassociation paradigm wasused to elicit verbalassociates, an d results indicated selective dis-ruption of the associational network, basedon site of damage. Patients with anteriorlesions produced essentially the same set ofassociations as did a normal control group,although they responded quite slowly. Pa-tients with posterior damage, on the otherhand, produced erratic associations bearinglittle resemblance to the normal data. Thus,the patient class with the most severe word-finding d ifficulties displayed a serious d isrup-tion of associational structure.


8/21

APHASIA 905A study has recently been reported thatattempted to specify more precisely thedisso-lution of associational stru ctu re in aphasiaand to link that breakdown directly to nam-ing impairment (Goodglass & Baker, 1976) .Using a procedure designed to determine apatient's knowledge of the verbal associatesof a target word without necessitating theproduction of the associate, these investiga-tors attempted to determine how semanticstructure was affected in various types ofcomprehension impairment. In addition, theywere interested in finding out whether thestructure of the semantic network was a fac-tor in the patient's naming ability. Target

words were eight high-frequency and eightlow-frequency picturable nouns depicted oncards and presented visually for naming togroups of anterior and posterior aphasics,nonaphasic brain-damaged patients, and non-neurologically impaired controls. Thepictureswere shown a second time while a series of14wordswas read. Of these, 7 were unrelateddistractors and 7 bore specific types of asso-ciational relationship to the target. One ofthese was the name of the item, and othertypes of associations were the superordinatecategory to which the item belonged, thename of an attribute typically characterizingthe item, the name of another (coordinate)member of the same category, the name ofan action associated with the item (func-tional associate), and the name of a situa-tion or context in which the item would beexpected to be found (functional context) .Subjects were instructed to squeeze a re-sponse bulb whenever one of the aurallypresented word s remind ed them of thepictured target. Responses were analyzed interms of both reaction time and error rate,which revealed similar patterns.All subjects recognized the name of thepictured objects equally well. With regard tosemantic structure, th e nonaphasic controlsubjects and those with anterior lesions ex -hibited a similar pattern of associations. Forall three of these groups the clearest associ-ates were the superordinate category name,th e descriptive attribute, and terms relatedby situation or context (functiona l context).Although the other associates were also in-cluded in the semantic network by these two

groups, responses to these items were mark-edly slower than responses to the threestrongest associates.The subjects with posterior damage (andgenerally more severe comprehension deficits)differed from the other patients in their pat-tern of responding in that they had greatdifficulty recognizing both functional con-texts and functional associates as related tothe target. Goodglass an d Baker suggestedthat the failure to recognize this type ofassociation between words signaled a quali-tative change in the semantic organization ofthese patients. Further, they suggested thatthe inability to retrieve words in a namingtask may be inpart a function of breakdownof the semantic structure. Thus, eventhough al l patients could recognize the nameof the depicted ob ject, many patients hadnot previously been able to produce thatname. All patients responded most quicklyto associates of words they had been able toproduce, and the patients with posteriordamage were much less likely to respond toassociates ofword s they had failed to produce.Thus, the ability to produce a name for anitem seems to involve a rather completeunderstanding of the many elements that areassociated with that name.Goodglass an d Baker postulated that theretrieval of a name depends on the con-vergence of con curre ntly activated associa-tions that trigger the appropriate naming re -sponse. To the extent that a picture of anobject arouses an incomplete set of associa-tions, naming ability will be impaired. Thisposition seems to be somewhat different fromthe additive model of information accumula-tion mentioned above, in that now the se-mantic network that converges to triggernaming is presumably structured in accord-ance with principles that may predict thepattern of dissolution.That is to say, break-down in structure would be expected to fol-low an orderly pattern, with some elements(in this case functional correlates) beingdisrupted before others.

In this view, then, brain damage can havea direct effect on the elements normally asso-ciated with a word's meaning, suchthat someelements are not aroused by a stimulus ob-ject. There is no clear indication of whether


9/21

90 6 ALFONSO CARAMAZZA AN D R I T A SLOAN BERNDTthese associational elem ents have b een lost inthe process of disruption or whether they maybe more difficult to retrieve for some reason.Within this context, it becomes interestingto consider the effect on naming of providingfunctional information with th e presentationo f the stimulus. Tha t is, will posterior pa-tients (wh o appear to have lost the functionalassociations of some words) improve in theirability to name if the functional informationis provided in the stimulus context?A recent study has attempted to assessdirectly the contribution of various types ofinformation residing in the to-be-named stim-ulus topatterns of naming disruption (White-house, Caramazza, & Zurif, in press). Thecentral question waswhetheraphasic patientscould attend to variation in several types ofperceptual elements, as well as functionalcontext information, in applying a label toan object. Using a paradigm designed for usewith normal adults (Labov, 1973) and sub-sequently modified for use with children(Andersen, 197 5), Whitehouse et al.designed24 stimulus pictures that were variations ofa modal cup. Both discrete (presence or ab-sence of handle) and continuous (height/width ratio) features were varied in thestimulus items to yield a set of 24 line draw-ings of containerlike objects.Functional context was provided by addi-tional pictures presented to the subjects alongwith one of the container drawingssothat thecontext picture showed some form ofpouringinto the container. A cup context depicted acoffee po t pouring coffee into the container;a bowl context showed a cereal box pouringout into th e container, and a glass contextpresented a picture of ice water being poured.In applying a name to these items, normalspeakers integrate stimulus information in amanner that underscores the interdependenceo f perceptual an d functional information(Labov, 1973) ; that is, when the perceptualinformation presents a clear member of oneo f the available categories (e.g., a cup, bowl,or glass) the naming decision is predicatedon that information alone, and shifts offunctional context do not affect the namethat is applied. However, if the perceptualinformation adds up to an object on theboundary between two categories (e.g.,

neither a clear-cut bowl nor cup), then func-tional information will be utilized in deter-mining the label to be applied.Therefore, tw o questions are addressed inpresenting this task to aphasic patients.First, it is important to determine whethersuch patients can use the perceptual infor-mation in a stimulus item to determinewhether that item is a clear or borderlinemember of a category. That is to say, per-formance should indicate the extent to whichaphasic patients are sensitive to the fuzzybo und aries between conceptual categories.Second, this task allows a straightforwardtest of the ability of aphasic patients to usefunctional information in determining a nameto be applied.The first question is addressed by analyz-ing th e consistency with which a particularname is applied to each stimulus item. Theclear (prototypical) members of each of thethree categories should be consistently named,whereas the borderline members should benamed inconsistently across subjects an dtrials. In general, the anterior aphasics in thisstudy exhibited this pattern; they namedprototypes consistently and appeared sensi-tive to the fuzzy boundaries between con-cepts. The posterior aphasics showed nosuch pattern an d either named al l items in -consistently or appeared to base their selec-tion of a name on only one perceptual fea-ture, for example, presence or absence of ahandle.The e f f e t of functional context on theapplication of a name is determined bynoting the number of times an item named inth e neutral condition is given a differentname when a particular context is provided.Again, the anterior aphasics appeared sensi-tive to the functional context provided, ap-propriately changing th e object's name withshifts in context. The posterior aphasics, onthe other hand, did not appear to be sensitiveto th e functional information that was pro-vided. Several posterior patients shifted thename inappropriately as often as appropri-ately; others failed to show an y namingchange as context was shifted but persever-ated in a unidimensional perceptually basedstrategy.These resu lts for the posterior aphasics


10/21

APHASIA 907provide compelling evidence that the seman-ticrepresentation of the elements comprisingthe meaningof the wordscup, g lass,and bowlis not well structured. Because the domain ofinterest w as constrained to the categ ory offood containers, because perceptual variationwas controlled, an d because functional infor-mation was explicitly provided in part ofthe task, it is difficult to believe that thenaming difficulty was the result of failure toactivate an intact representation. The possi-bility of retrieval was maximized in severalof the experimental conditions (e.g., when aprototypical cup was presented in the cupcontext) and yet these occurrences werenamed as inconsistently as the borderlinecases in an incongruent context. It seems aninescapable conclusion that these posteriorpatients did not have sufficient information. in thesemanticrepresentationoftheseitemsto differentiate between them.This review of some pertinent work onnaming disruption has been structured toshowthat there has beena discernible changein thinking about the level of processingthat is disrupted when a patient fails toprovide acorrect namefor an object. Namingimpairment is no longer viewedas a unitaryphenomenon; the level at which a namingresponse is disrupted may not be the sameacross all types of this difficulty. The majordivision based on symptomatologybetweenpatients with anterior and posterior damageappears to involve different degrees of in-volvement of the central processes outlinedearlier.Theimportant pointisthatthe possi-bility of disruption at the level at whichsemantic information is represented is be-ginning to be considered, especially for pa-tients with damage to the posterior portionof the dominant hemisphere.Comprehension: Sense and Reference

As mentioned above, little research hasbeen undertaken to disentangle the elementsthat may be involved in impaired lexicalcomprehension. However, unlike the situa-tion summarized here fo r naming disorders,there appears to be general agreement amongworkers in the area that comprehensiondeficits involve some kind of semantic dis-

ruption (Goodglass & Geschwind, 1976, p.406). This view has gained support from arecently reported study of phonological fac-tors in auditory comprehension that hasruled out the involvement of a phonemic-discrimination deficit as a major cause ofcomprehension disruption (Blumstein, Baker,& Goodglass, 1977) .Of course, several studies of naming re-viewed in the previous section implicitly in -volved comprehension. The semantic confu-sions exhibited by patients in a naming taskare also evidenced in comprehension, withthe result that a patient may point to his legwhen asked to point to his ankle (Goodglass& Geschwind, 1976). In addition, the wordassociation arguments used to explain nam-ing behavior are based onaspects of meaningcomprehension not usually implicated inreferential meaning, that is, the sense rela-tion of the word. The referential part ofmeaning specifies the information that isneeded to identify members of a conceptualcategory; the sense elements trace the rela-tionships among several different but relatedconcepts. Thus , hierarchical relationships,similarity relationships, functional relation-ships, to name but a few, are part of thesense of aparticular word meaning and com-prise the associational network, or semanticfield, discussed in the previous section.The point has already been made thatdisruption of these sense relations may be afactor innamingdisability; thestructure andorganization of sense elements within a se-mantic domain is also of interest in its ownright. That is, intact comprehension pre-sumably implies more than the ability torecognize an item as the referent of a givenword ; it implies a complete network of con-ceptual elements organized in such a waythat the relationships between conceptswithin a given semantic field are accessible.Two studies have been reported recentlythat attempted to ascertain the extent towhich these sense relations are disrupted inthe semantic organization of aphasic patients.Lhermitte, Derouesne, and Lecours (1971)presented patients exhibiting various sympto-matology with tw otasks that involved sortingcards containing printed words. In the firsttask, each of 12words was to be sorted into


11/21

908 ALFONSO C A R A M A Z Z A AN D RITA SLOAN BERNDTone of three columns based on its degreeo frelationship to a target word . For example,the target word fish would be closely relatedto words such as oceanand fisherman, some-what related to the words odor and cooking,and not at all related to the words chair andfantasy. In a second task,the ability of thesepatients to shift among various senses of apolysemous word was tested. Patients wererequired to sort seven words into one of twocolumns based on whether the word evokedsome sense of a target word. For example,the target word division has different sensesthat may be evoked by the words army andcalculationbut has nosense that is related tothe wordssky and couch.

Patients' incorrect responses were categor-ized into three groups: disruption of thehierarchical relationship among th e words,and narrowing and broadening of thesemantic field. Patients with all types ofsymptoms produced responses in these threegroups, but patients with anterior damagewere highly represented only in the groupmakinghierarchical errors. Errors of this typeindicate that category boundaries are some-what il l defined, but they do not representgross errors of association. Patients withposterior damage, on the other hand, dis-played a more severe disruption of semanticstructure, with a tendency to broaden theboundaries of the category.Another study directed at charting the dis-ruption of semantic organization in aphasiabegan with somewhat stronger assumptionsabout the structure of lexical knowledge(Zurif , Caramazza, Meyerson, & Galvin,1974). On the basis of recent attempts todescribe the organization of the lexicon inthe normal speaker (e.g., Fillenbaum & Rap-oport, 1971; Miller, 1967, 1969, 1972) , se -mantic information is assumed to be repre-sented as an internalizeddata structure basedon semantic features. These features, typi-cally mapped as single words (e.g., + male)or phrases (+ w ho is married) are taken torepresent the lexical information available tothe language user. They are meant to cap-ture the conceptual elements in a word , aswell as to define the range of semantic rela-tions into whicha wordcan enter with otherwords.

Zurif etal. used as stimulus items 12 high-frequency concrete nouns chosen to captureseveral different types of relationships amongtheir semantic features. One primary cri-terion was that the features of the wordschosen should be cast into a hierarchical ar-rangement. The central or dominating featureof the stimulus itemswas specifiedas hu-man, which provided a basic distinction be-tween two semantic domains, animal terms(shark, trout,dog, tiger,turtle,crocodile)andwords used to describe human roles (mother,wife, cook, partner, knight, husband).Components suchas + human and + ani-mal are very basic conceptual features ofmany entries in the mental dictionary andare thus relatively systematic in the lan-guage (Ka tz & Fodor, 1963). However,there are other features that are not centralto the word's definition but represent oftenidiosyncratic referential and affective infor-mation g rowing out of a speaker's ex periencein th e world rather than his knowledge of thelanguage (Miller , 1 9 7 2 ) . This distinctionparallels th e difference between de finingand characteristic semantic features thathas recently been included as an integral ele-ment of a model of normal semantic memory(Smith, Shoben, & Rips, 1974). T he wordsselected for this study were related throughoverlap of both defining an d characteristicfeatures.The aim of the study was to determinehow aphasic patients differ from normals intheir appreciation of the formal hierarchicalrelations between the words presented and toascertain whether th e defining and character-istic features of words were differentially im -portant to the organization of the lexicon inaphasia. To this end, aphasic patients withanterior and posterior damage, as well asneurologically normal control patients, werepresented with three of the stimulus words ata time and required to indicate which twowere the most similar in meaning.A ll possiblecombinations of the words were presented,and the patients' similarity judgments (bygroup) were analyzed using botha hierarchi-cal clustering technique and a nonmetr icmultidimensional scaling program to deter-mine the structural relationship that thepatients imposed.


12/21

APHASIA 909The major patterns in the data can besummarized as follows: The control patientsclearly clustered the human itemsseparatelyfrom the animal items. The anterior aphasics

also recovered the basic semantic featureo fhuman versus nonhuman. The posterioraphasicpatients, on the other hand, did notconvincingly separate the human items fromthe animal terms, despite having been ableto recognize the definitions of the nouns ona pretest.However, the anterior aphasics produced acluster within the domain of animal termsthat clearly differed from that produced bythe controls. The nonneurologically impairedpatients combined the items in terms ofshared species membership, discriminatingamong fish, reptiles, and mammals. The an-terior aphasics, in contrast, generated twomajor clusters: one consisting of sharkcrocodile, and tiger all ferocious, wild, andremote; and theother consisting of trout an dturtle, both partially edible and harmless.Within the framework of the distinctionbetween defining and characteristic featuresset out above, it seems reasonable to suggestthat by focusing on the species membershipof the items, the control patients appliedtechnical semantic features when judging th esimilarities an d differences among the mean-ings of the animal terms. In contrast, theanterior aphasic patients, by usinga featurebased on ferocity appeared to carry out thetask on the basis of incidental or character-istic features; that is, compared with speciesmembership, ferocity seems more tied to em-pirical knowledge (first- or secondhand) an dless dependent on an understanding of sys-tematic interlexical relations. Ferocity is notpart of the dictionarydefinition of these itemsbut is in large measure a function of extra-linguistic setting. Thus, even though it maybe argued that all semantic features ulti-mately derive from a knowledge of theworld, ferocity appears to be a much moreempirical orperceptualconcept than mammalfor example.This result suggests that compared withthe lexical structure underlying normal lan-guage use,the semantic representation in an-terior aphasia is more restricted in its rangeof conceptual integration. In effect, verbal

concepts in anterior aphasia appear to bemore tightly tied to affective an d situationaldata. Thus,itappearsas if the normal adulthas a numberof levels at which he can or-ganize his lexiconsome referentially practi-cal, others linguistically practicalwhereasthe aphasic primarily retains those featuresof wordsthatrelatetoperceivedor imaginedenvironmental situations.This study provides another piece of evi-dence that the semantic structures of an-terior andposterior aphasicsare qualitativelydifferent. What is more important, it definesan element of structural disruption for an-terior aphasics that has not emerged in testsof naming ability or referential comprehen-sion; thatis, thepartofword meaning tappedby the traditional tasks, namely, referentialmeaning, may be spared independently of thesense relations that provide the links be-tween lexical concepts.The investigations of lexical disruption inaphasia that have been reviewed here haveno t resolved the question of which processesor structuresare impaired whenapatient hasdifficulty with individual word meaning. Re-trieval problems and conceptual disorganiza-tion have both been implicated. For the mostpart, retrieval problemsin naming have beenviewed as a failure of the sensory input toprovide sufficient information to access asemantic representation. The fact that somewords are more successfully accessed thanothers is regarded as a function of the re-dundancy of information provided by theelicitor-stimulus. Thus, items that are con-crete and picturable, easily manipulated, andprocessed through several sensory modalitiesare most likely to be successful in activatingthe representation.

However, recentinvestigationsof thestruc-tureof the lexicon have raised the possibilitythat the same stimulus factors e.g., pictura-bility, operativity) are more than factorsaffecting retrieval. Rather, they are elementsthat lend themselves to referential encoding,which is the aspect of meaning that appearsleast subject to disruption. Thus, the patternthat dissolution of the lexicon takes mayinvolve selective disruption of elements, withthose features that have been acquired andexercised in many different contexts through


13/21

910 ALFONSO CARAMAZZA AN D RITA SLOAN BERNDTseveral sense modalities showing the greatestresistance to disruption.

There is evidence that this referential as-pect of the meaning of concrete nouns can beprocessed solely in the right hemisphere ofthe brain, which is not normally viewed asdominant in subserving linguistic processes(Geschwind, 1970). Recent studies of split-brain patientsindividuals whose interhemi-spheric commissures have been severed(allowing a relatively pure investigation ofhemispheric differences)have shown thatderivationally simple concrete nouns can becorrectly processed in the right hemisphere(Gazzaniga, Bogen, & Sperry, 1965). Simi-larly, studies with normal subjects in whichstimuli have been presented separately toeach visual half-field, have shown that high-frequency concrete nouns can be recognizedwhen processing is limited to the right hemi-sphere (Ellis & Shepherd, 1974; Hines,1976) .Again, the suggestion is made that high-frequency concrete nouns enjoy a specialstatus in the lexicon that allows resistance toloss in conditions of damage to the left hemi-sphere and appears to allow some level ofrepresentation in the right hemisphere. It ispossible that these easily picturable nounscan be recognized by the right hemispherebecause of the capacity fo r visual imagerythatis believed to behighly lateralizedto theright (C aramazza, Gordon, Zurif , & DeLuca,1976; Whitehouse, 1977) . However, as sug-gested above, the fact that the referents ofhighly concrete nouns can be experiencedthrough several sense modalities and can bemanipulated as well as visually perceived,may indicate thatthe meanings of those nounsare more diffusely represented in the brain;that is, the representation of items associatedwith a mult i tude of experiences in the worldmay not be as specifically localized as otherwords (such as abstract nouns) which areknown only through a linguistic channel.

Lexical and Sentence MeaningIt has been argued that the meaningof

an individual lexical item is a complex con-ceptual representation with aspects of bothsense and reference. In addition, the effect of

brain damage on this representation wasshown to be selective; that is , part of theword 's meaning may be lost while other as-pects are spared. In this section, discussionof the disruption of the semantic componentof comprehension is extended to includemeanings conveyed by combinations ofwords.There are many important differences be-tween lexical and sentence meanings, andthese differences may provide new potentialfor selective disruption. First, lexical andsentence meanings are not coextensive in thesemantic ideas theycanexpress.The semanticidea of predication, for example, can be ex-pressed only by a phrase or a sentence, notby a single lexical item. The suggestion hasbeen made (Luria, 1970) that specificallysyntactic meanings such as predication maybe selectively impaired by brain damage.A second difference between lexical andsentence meaning centers on the functionwords. Although these items are not believedto have a specific semantic representation(and have little importance as individualitems), they play a crucial role in conveyingsentence meaning by communicating suchimportant distinctions as definite an d indefi-nite reference (Goodenough, Zurif , &Wein-traub, 1977) .Most important for our purposes, lexicaland sentence meanings differ in the verynature of their semantic representations.Lexical meanings can be considered to havefixed representations, but sentence meaningsare novel, complex representations constructedby combining the meanings of single lexicali tems. This view is not universally held, butit has received considerable support in thel i terature, which is summarized here. Thereis some debate as to whether a lexical itemhas a fixed semantic representation for eachof its major senses (o r meanings) or whetherit is even possible to talk oflexical meaningoutside of a sentential context (Kempson,1975) . The meaning of a lexical item un-doub ted ly takes on different senses in d i f -ferent sentential contexts,yet it ispossible toconsider the meaningof a lexical item as theconceptual representation or rangeof concep-tual values that an item can take in variouscontexts. The position advocated here isthatword meaning involves an invariant core rep-


14/21

PH SI

reservation made up of the range of concep-tual values accepted by a linguistic commu-nity (C aramazza & Grober, 1976) .That coremeaning can be combined with other lexicalitems in the sentential matrix and in thecommunicationof an idea.There is agreement among psycholinguistson a number of issues that concern thestructure of the lexicon. First, the lexiconcan be structured along separate dimensionscorresponding to the four componentsof eachlexical itemphonological, graphemic, syn-tactic, and semanticeach of which is speci-fied as a matrix of primitive elements. The

representation of a single lexical item is pre-once acquired, remain unchanged over time.In fact, th e phonological, graphemic, andsyntactic representations of a lexical item,once acquired, remain unchanged over time.The semantic component may undergo somechanges, but these are of a relatively minornature, such as the addit ion of fine nuanceswithin a semantic network.Another important characteristic of thelexicon is that it is an open-ended system;that is, it consists of a very large numbero fentries with an indefinite upper bound. Inaddition, information specified for any lexi-cal item can be in any one of a combinationof sensory modalitiesvisual, auditory, tac-tile, and olf actory . This la tter characteristicsuggests that the information that specifiesthe semantic structure of a lexical item maybe more diffusely represented in the brainthan is information that is specified in asingle modality.

The topic of sentential semantics has re-ceived a great deal of attention in thepast 10years, with considerable debate about howth e processes that are involved should becharacterized (Katz, 1972 ; Lakoff, 1971). Forpresent purposes, it is most important to notethat there is a combinatorial operation impli-catedinsententialsemantics thatis not pres-ent in any formulation of lexical semantics.It is assumed here that the syntactic systemis structured along the lines of a generative-transformational system. That is, the syn-tactic system (like the phonological system)consists of a closed set of rules that can berecursively applied to produce an unlimitednumber of sentence structures. In this sys-

tem, no specific sentence structure need berepresented in the human mind; instead, itis the syntactic rules that produce th e sen-tence structures that are stored.The major point is that syntactic rulesconstitu te a finite and relatively small set ofoperations that are acquired early in life andundergo no major changesinadult life. Fur-thermore, unlike the lexical system, the rep-resentation of syntactic information is notmodality specific, even though it is primarilyacquired through th e auditory system. Thesetwo considerations lead to the suggestionthatsyntactic operations are less diffusely repre-sented in the brain than islexical knowledge(cf. Zurif &C aramazza, 1976).In light of these differences between thelexical and syntactic systems of language, itshould not be surprisingthat they appear tobe functionally independent upon ananalysisof language dissolution an d that the systemsare subserved by distinct brain areas.Thatis,focal brain damage to the dominant hemi-sphere does not result in an across-the-boardreduction of language capacity (cf . Schuell &Jenkins, 19S9) but instead gives rise to re-markably stable patterns of abnormal per-formance. In fact, it could be argued thataphasia probably offers th e strongest evi-dence for the functional independenceof lexi-cal and syntactic operations (Marin, Saf-fran, &Schwartz, 197 6), as well as evidencefor th e independence of syntactic and heur-istic processes (Caramazza & Zurif, 1976).The next section provides a review of someof the literature that addresses the questionof the disintegration of the ability to usesyntactic operations in language performance.Agrammatic and Paragrammatic Speech

Over a century of research has led to someagreement on the identification of two majorforms of aphasic speech that are primarilycharacterized byabnormal syntactic organiza-tionagrammatic and paragramm atic speech.Agrammatic speech has a strikingly tele-grammatic form in which syntax seems to berestricted to a single declarative form, func-tion words are infrequently present, andverbs, when used, are most often uninflected.Frequently associated symptomsar e effortful-


15/21

912 A L FO N SO C A R A M A Z Z A A N D RITA SLOAN BERNDTnessof speech and distorted articulation.Thisconstellation of symptoms is usually presentwith lesions of the anterior portion of thedominant hemisphere. Early explanations ofthis form of behavior most often attributedthe telegrammatic nature of speech to aneuromuscular problem. The belief was thatthe strain of speaking is so great that thepatient speaks asyntactically in order toeconomize effort (Lenneberg, 1973). An al-ternative view (Jakobson, 1964; Luria, 1970)attributes the problem to a more central dis-ruption: Agrammatism results from a dis-turbance of the mechanismsthat subserve theability to structure a string of words syn-tactically (see also Zurif &C aramazza, 1976) .Paragrammatic speech seems to involve no tso much a restriction of syntactic organiza-tion as the inappropriate jux taposition oflexical items. The speech of these patients ismarked by facility of articulation and bymany long sequencesof wordsin a varietyofgrammatical constructions. However, the out-put isinformationallyemptyInd efinitenounphrases are often substituted for an appro-priate noun, and when a noun of specificreference is chosen it is often the wrong one.Because these patients can produce a varietyof syntactic forms, it is often assumed thatsyntactic impairment is minimal. However,it should be noted that the situation may besomewhat more complicated inthat the inap-propriate juxtaposition of lexical items oftenresults in serious grammatical distortions,such as category violations (e.g., the use ofa noun in a verb or adjective position) andselectional restriction violations (e.g., the useof an animate noun in a sentence that re -quires an inanimate noun). Nevertheless, thespeech of these patients generally gives theappearance of being syntactically well struc-tu red . The lesion site most often associatedwith disturbances of this type is the posteriorpart of the dominant hemisphere.As in the case of lexical disruption, a dis-proportionate amount of attention has beenpaid to verbal production, with neglect ofdetailed analysis of comprehension perfor-mance. Furthermore, most of the early workrelied primarily on clinical data, with littleemphasis on experimental considerations.However, since the mid-1950s, Goodglass and

his collaborators have done much to improvethe situation, and in the last several years amore balanced view of the problems of syn-tactic breakdown has emerged. The follow-ing section presents a review of some of thework that has addressed the question of theeffect of focal brain damage on syntacticmechanisms.Research on Syntactic Breakdown

Production. Detailed studies of both free-conversationdata and controlled experimentaltests have revealed several factors that areimplicated in agrammatic and paragrammaticspeech. These factors include a disruption oflexical availability (primarily associated withparagrammatic speech) and the more gram-matically based difficulties, such as problemswith prosody, morphology, and syntax (typi-cally found in agrammatic speech).In several articles, Goodglass and his as-sociates have shown that agrammatic apha-sics d i f f e r from paragrammatic aphasics onseveral dimensions of syntactic ability. Onestudy (Goodglass & Hunt , 1958) investi-gated aphasic patients' ability to producethe inflectional markers for the possessivecase and the plural form ofnouns. In Englishthese markers are phonologically indistin-guishable because the same allomorphic vari-ations /s, z,az/ can indicate both the pluralan d possessive forms. Consequently, any d i f -ferential effects in the ability to producethese morphemes for the plural and posses-sive cannot be attributed to a peripheral dis-order, but must be considered as the result ofa more central, syntactic disorder.Goodglass and Hunt found that the pos-sessive marker was omitted nearly twice asoften as the plural, suggesting that syntacticoperations can be differentially affected byfocal brain damage. Unfortunately, there areseveral alternative hypotheses as to the basisfor the differential effects. Grammatical com-plexity, redundancy, and frequency of usagecan all account for the observed effect. Inaddition, unselected aphasic patients weretested in this study, so it is difficult to makeany inferences about site of lesion an d func-tional breakdown .Goodglass (1968) replicated these findingsan d extended the inquiry in two specific


16/21

APHASIA 913respects.First,patientswere subdivided intoagrammatic and paragrammatic groups onthe basis of a length-of-utterance criterion(Goodglass, Quadfasel, &Timberlake, 1965).Second, themorphological changesthat markverb tenses were included in the task. Therewere no differences between agrammatic an dparagrammaticaphasics on the order ofdif-ficultyamong morphological markers: Pluralswere easier than possessives and the thirdperson singular /s/ for the verb. However,there was a marked difference betweenaphasic groups on their ability to producethe allomorphic variations /z vs.az/ and /dvs.ad/ forplural and tense markers, respec-tively. Specifically, paragrammatic aphasicsha d more difficulty with the more complexsyllabic allomorphs /az andad/, just aschil-dren do (Berko, 1958), and agrammaticaphasics displayed theopposite pattern. Thisresult suggests that the effect of anteriorbrain damageonlanguage ability at the syn-tactic level is more than just an overall re-duction in syntactic capacity; instead, syn-tactic operations are impaired selectively(see also Goodglass & Berko, 1960).

There isalso evidence for differential per-formance of agrammatic an d paragrammaticaphasics at the level of prosody. Goodglass,Fodor,and Schulhoff (1967) used a sentencerepetition task to assess the effect of stresson the production of a word. The resultsclearly indicated that anterior aphasics aremuch morelikelytoomit anunstressed wordthan ardposterior aphasics.

Two other studies by Goodglass's group,Meyerson and Goodglass (1972) and Good-glass, Gleason, Bernholtz,and Hyde ( 1 9 7 2 ) ,have further charted th e syntactic abilitiesofagrammatic aphasics. Both studies were con-cerned with a single agrammatic patient'sability to produce syntactically well-formedstructures. However, Meyerson and Good-glass (1972) analyzed speech obtained infree conversation, whereas Goodglass et al.(1972 ) attempted to elicit specific syntacticforms. The major finding was that syntacticability, as specified by the presence or ab-sence of transformational rules, is quiteclearly compromised in anterioraphasia.This brief review has tried to show thatfocal braindamage may selectively affect the

syntacticsystem independently of the lexicalsystem. However, the force of this argumentis attenuated by the claim that anterioraphasics, though agrammatic in their output,display good comprehension, suggesting thattheir agrammatic speech may simply be theresult of some peripheral disorder that doesnotimpairsyntacticknowledge. However, theclinical claim that anterior aphasics usuallypresent nearly normal comprehension doesnotnecessarily mean that thepatientmay not be agrammatic in comprehension as well. Toillustrate this point, assume that when ananterior aphasic is presented with the sen-tence The boys are coming tomorrow, heprocessesthesentenceonly enough torecoverthe sequence boy come tomorrow. On thebasis of this recovered sequence, he may beable to understand the intended message,especially since there is usually sufficientcontextual information to help disambiguatepotential ambiguities. Thus, the dissociationof production and comprehension ability inanterior aphasia may be more apparent thanreal. In fact, in the last several years, evi-dence has accumulated that the comprehen-sion abilities of anterior aphasics closelyparallel their agrammatic productions.Comprehension, Caramazza and Zurif(1976) assessed comprehension capacities inaphasics with anterior and posterior lesionsof the dominant hemisphere. The task em-ployed in this study was a sentence-picturematching task.Patientswerepresented orallywith a sentence and were asked to choosefrom tw opictures the one that depicted th eproposition described by the sentence. Therewere two major manipulations that are ofinterest here: One was the type of sentenceused; the other was the type ofinformationcontrast depicted by the picture pair. Theimportant variable in the sentence materialwas whether the center-embedded sentencewas reversible or nonreversible, that is,whether the lexical items contained in thesentence permitted one or more readings ifon e ignored syntactic constraints. Thus, inthe reversiblesentence Thelion that thetigeris chasing is f a t , th e lexical items alone per-mit the readings The tiger is chasing a fatlion, The lion is chasing a fat tiger, The fattiger is chasing th e lion, an d The fat lion is


17/21

914 ALFONSO C A R A M A Z Z A AN D RITA SLOAN BERNDTchasing th e tiger. However, in the nonreversi-ble sentence Th e bicycle that the boy is hold-ing is broken, selectional restrictions on thelexical items permit only the reading The boyis holding a broken bicycle. Four types ofpicture contrasts were used, three of whichcould be solved using lexical info rm atio nalone. The fourth contrast required the pa-tient to recover the correct syntax of thesentence in order to make the correct choice.There was a clear effect of contrast typefor the agrammatic (anterior) aphasics, whoresponded correctly about 90 % of the timeon the lexical contrasts but only about 10of the time on the syntactic contrasts. Theparagrammatic patients showed no reliablepattern but were significantly poorer in over-all performance than the agrammatic pa-tients. Furthermore, the performance ofagrammatic and paragrammatic patientsshowed a clear and consistent differencebased on whether the sentence was reversibleor nonreversible. The agrammatic patientsperformed at the 90% level of accuracy onnonreversible sentences, but their perfor-mance dropped to chance with reversible sen-tences. The paragrammatic patients' perfor-mance was unaffected by this manipulation.These results clearly show that the agram-matic aphasic's ability to comprehend sen-tences is seriously compromised. What ismore important, the locus of disruption ap-pears to be at the level of syntactic process-ing.This limitation of anterior aphasics' abilityto process syntactic information has emergedin several other recent studies of comprehen-sion (Marin, Saffran, & Schwartz, 1976;Scholes, 1978) and in studies of linguisticintuitions on the structure of sentences (An-dreewsky & Seron, 1975; Zurif, Caramazza,& Myerson, 1972; Zurif, Green, Caramazza,& Goodenough, 1976; Ulatowska & Baker,Note 1).This last set of studies indicates that theanterior aphasic's intuitions about his lan-guage are as agrammatic as his output. Zurifet al . ( 1972) req uired patients to jud g ewhich words of a sentence went best to-gether. The patients were presented withtriads of words from the sentence (whichwas always in their view) and were asked to

point to the two words that were most closelyrelated. These judgments of proximity wereanalyzed using a hierarchical clustering pro-cedure (Johnson, 1967; Levelt, 1970) to ob-tain for each sentence a treelike hierarchicalstructure. The implicit hierarchical organiza-tion that emerged for the normal controlgroup corresponded to the familiar parsingtree of the surface structure of a sentence.The structures that emerged for the anterioraphasics, on the other hand, were quite dis-tor tedThe aphasics grouped together onlythe content words and for the most partignored or inappropriately placed the func-tion words.Thus, it appears that th e anterioraphasic's agrammatism extends to his intui-tions about language.Several recent studies of the linguistic abili-ties of individuals who had undergone surgi-cal removal of the cortex of one cerebralhemisphere (hemidecortication) during in-fancy have further supported th e functionalindependence of syntactic processes (Dennis& Kohn, 1975; see also Dennis &Whitaker,1976) . Two groups of matched subjects,functioning with only a right or a left hemi-sphere, were presented with six spoken sen-tences varying in syntactic form. Subjectswere required to select a picture from tw oalternatives that depicted the meaning of thesentence. Subjects with an intact right hemi-sphere had increasing difficulty comprehend-ing the sentences as syntactic complexityincreased. Thus, fo r example, right- and left-hemisphere groups processed simple activesentences equally well (measured by percentcorrect andlatency torespond),but the groupwho had undergone removal of the cortex ofth e left hemisphere performed significantlyworse with more complex constructions, es-pecially the passive negative. These twogroups were comparable in verbal IQ , andboth were capable of processing the informa-tion contained in simple sentences. The func-tions underlying comprehension of syntacticforms, how ever, had apparen tly not devel-oped comparably in the two groups. Thesesyntax-related functions thus appear to havesome measure of independence from otherlanguage capacities, and an intact dominanthemisphere seems to be necessary fo r theirfull utilization.


18/21

APHASIA 915Nonlinguistic Cognitive Operations:Memory and Syntactic Processing

The foregoing treatment of comprehensionperformance has proceeded as if the onlymechanisms involved were language-specificones. It is obvious, however, that there arenonlinguistic cognitive operations that comeinto play in processing a sentence fo r com-prehension. Lexical meanings involve retrievalmechanisms thataccessprestructured concep-tual entities, but sentence meanings addition-ally involvethe operation of a syntax,a sys-tem of rules that works over time to deter-mine the logical relations that hold amongthe lexical items in a sentence. The opera-tions of a syntactic system necessarily impli-cate the functioning of a working memorythat temporarily stores parts of a sentencewhile the remainder of the sentence is beingencoded.To the extent that memory mechanisms areimpaired in patients with damage to thedominant hemisphere, comprehension perfor-mance willbe disrupted (C ermak &Moreines,1976; Lesser, 1976; Saffran &Marin, 1975).Consequently, it is critical that a specificmemory deficit be ruled out as a basis forthe abnormal comprehension performance intheanterior aphasic in order to mairitain theposition that structures of the anterior partof the dominant hemisphere subserve syntac-tic operations.It is unlikely that a single memory deficitcan account for the findings that have beenreported, particularly the disruption ofaphasics' metalinguistic judgments, whichwere performed outside the demands imposedby real-time processing. A direct test of thiscontention was provided by a recent investi-gation of sentence memory in aphasics withanterior and posterior brain damage (Cara-mazza,Note 2). The specific purpose of thisstudy was to determine the level to whichsyntacticprocessing in a memory task is car-ried out by agrammatic aphasics. This ap -proach wasmotivated by the contention thatgrammatical morphemes are processed atsome level of auditory representation but notbeyond (i.e., not to a level of meaning rep-resentation) and also by the assumption thatmemory for a sentential item is a function of

the degree of processing at different levelsthat the item receives at the time of input(the depth-of-processing hypothesisC raik& Lockhart, 19 72 ; C raik & Tulving, 197S) .It would seem that the relatively shallowprocessing of functors would result in lessstable memory traces fo r these items relativeto content words, which would most likelyreceive more elaborate processing.The experimental paradigm employed amodified memory-probe technique. The pa-tients were presented orally with a sentenceand immediately afterward were givena probeword from that sentence. The patients' taskwas to produce the word from the sentencethat followed this probe item. If a patientfailed tomakearesponse or produced a soundthat could not b e readily id entified as one ofthe wordsin the target sentence, he wasgivena multiple-choice alternative. T he object wasnot to penalize the patient for any diffi-culties he may have had in the implementa-tion of speech, but to obtain a sensitivemeasure of his memory. A response wasscored as incorrect if it was an item otherthan the oneimm ediately following the probeword.Results were analyzed in terms of transi-tional error probability (TEP), that is, theprobabilityofrecallingan itemotherthan theon e that actually followed the probe in thesentence. Two different patterns of resultswere expected depending on whether the pa-tients were affected by a generalized memorydeficit or a specific syntactic processing im-pairment. In the case of a generalized mem-ory deficit, TEP would be expected to bequalitatively the same as that for a normalcontrol group, but with a higher level oferrors. If, on the other hand, th e patientscould not adequately process syntactic mor-phemes, then the TEP pattern should bemarkedly different from that of the normalcontrol group. The difference should be pri-marily in the relative effectiveness of func-tion words to probe content items versus theeffectiveness of content words to probe func-tion words.The results clearly supported thehypothe-sis that agrammatic aphasics cannot ade-quately process grammatical morphemes. Thefunctors seem to be processed to a level that


19/21

916 ALFONSO C A R A M A Z Z A AND RITA SLOAN BERNDTdoes not permit the words themselves to beretrieved but does allow their traces to bereinstated. This study thus provides furtherevidence that focal brain damage can selec-tively impair syntactic processing while spar-ingsemantic processing.In this section, a number of recent studieswere reviewed that clearly point to the func-tional and neurological independence of syn-tactic and semantic processing in sentencecomprehension. Data obtained from studiesof sentence processing in aphasia are impor-tant for a number of reasons. First, they es-tablish a clear brain/function relation inwhich the anterior portion of the dominanthemisphere subserves syntactic processes.Second, they demonstrate that in sentencecomprehension, syntactic processes are inde-pendent of processes such as strategies basedon lexical comprehension. Third, they dem-onstrate the psychological reality of linguis-tic processes by showing that although theseprocesses interact with other cognitive opera-tions (e.g., memory), they can be selectivelyaffected by brain damage. Even though theseconclusions are of a rather general nature atthis time, they form th e basis on which moredetailed experimental studies can be con-duc ted .

Some General ConclusionsThis review was begun with the statement

that traditional studies of aphasic languagehave not contributed appreciably to ourunderstanding of normal language processes.It wassugg ested that this situation may havearisen in part because of the aphasiologist'sconcern with brain/function relations as op-posed to language processes for their ownsake. Paradoxically, then, this review has ledto the conclusion that a clear formulation ofthe functional organization of the dominanthemisphere can lead to important insightsinto th e structure of normal language pro-cesses; that is, even though there is logicallyno reason why the study of language break-down cannot proceed independently of con-cerns about brain structures, it is an empiri-cal fact that there is a strong correlationbetween brain structures and linguistic pro-cesses. This is a substantive neurolinguistic

conclusion that is of heuristic value for psy-cholinguistics in that it provides strong con-straints on the generationofhypotheses ab outlanguage processes. Thus, knowingthat dam-age to a particular area of the brain affectsa specific language component can lead tothe formulat iono f testable hypotheses aboutth e structure of that component and theinteraction of that component with other cog-nitive processes. For example, on the basis ofprior observations that damage to the an-terior part of the dominant hemisphere af-fects syntactic processes, hypotheses could betested concerning the operation of syntacticmechanisms in sentence comprehension. C ara-mazza and Zurif ( 1976) wereable to demon-strate that anterior aphasics achieve a rela-tively high level of sentence comprehensionthrough the application of lexical and heur-istic processes. This result lends support toth e hypothesis that heuristic procedures canindependent ly assign a semantic interpreta-tion to a sentence. What isinteresting isthatthis latter hypothesis is one of several com-peting hypotheses that, according to Fodor,Bever, and Garret t ( 1974) , cannot b e differ-entiated on the basis of data obtained instudies with normal subjects.

It must be emphasized, however,that con-cern with issues of brain/function relationsis not a necessary component of the study ofaphasic language. Psycholinguistic investiga-tions can proceed quite inde pend ently of suchissues, concentrating instead on patterns oflanguage dissolution in an attempt to definethe componential structure of language pro-cesses. The material reviewed here showsthis to be not only a feasible but a fruitfulendeavor: Lexical, syntactic, and heuristiclanguage processes, as well as other cognitivecapacities, can be investigated in relativeisolation.

Reference Notes1. Ulatowska, H., & Baker, W. Linguistic study ofprocessing strategies in right- and left-brain dam-

aged patients. Unpublished manuscript , Universityof Texas, 1975.2. C a ram az z a ,A. Sentence processing in aphasia. InE. Zurif ( C h a i r ) , Language processing: A neuro-psychological perspective. Symposium presented atthe meeting of the Intern ational Neuropsychologi-


20/21

APHASIA 917cal Society, Santa Fe, New Mexico, February1977.

ReferencesAlston, W. P. Philosophy of language. EnglewoodCliffs, N.J.: Prentice-Hall, 1964.Andersen, E. S. Cups and glasses: Learning thatboundaries are vague. Journal of Child Language,197S, 2, 79-103.Andreewsky, E., & Seron, X . Implicit processing of

grammatical rules in a classical case of agramma-tism. Cortex, 1975, 11, 379-390.Berko, J. The child's learning of English morphol-ogy. Word, 1958, 14, 150-177.Bisiach, E. Perceptual factors in the pathogenesis ofaphasia. Cortex, 1966, 2, 90-95.

Blumstein, S. E., Baker, E., &Goodg lass, H. Phono-logical factors in auditory comprehension inaphasia. Neuropsychologia, 1977,IS, 19-30.Caramazza, A,, Gordon, J., Zurif ,E.,& DeLuca, D.Right-hemispheric damage and verbal problemsolving behavior. Brain and Language, 1976, 3,41-46.Caramazza, A., & Grober, E. Polysemy and thestructure of the subjective lexicon. In C . Rameh(Ed.), Semantics: Theory and application. Wash-ington, B.C.: G eorg etow n U niversity Press, 1976.Caramazza, A. , & Zurif, E. B. Dissociation of algo-rithmic an d heuristic processes in language com-prehension: Evidence from aphasia. Brain andLanguage, 1976,3 ,572-582.Cermak, L. S., &Moreines, J. Verbal retention defi-cits in aphasic and amnestic patients. Brain andLanguage , 1976,3, 16-27.Craik, F. I. M., & Lockhart, R. S. Levels of pro-cessing: A f ramework fo r memory research. Jour-nal of Verbal Learning an d Verbal Behavior,1972, 92 , 149-154.Craik, F. I. M., &Tulving, E. Depth of processingand the retention of words in episodic memory.Journal of Experimental Psychology: General,1975, 104,268-294.Dennis, M., &Kohn, B. C omprehension of syntax ininfantile hemiplegics after cerebral hemidecortica-tion: Left-hemisphere superiority. Brain and Lan-guage, 1975, 2, 472-482.Dennis, M., & Whitaker, H. Language acquisitionfollowing hemidecortication: Linguistic superiorityof the left over the right hemisphere. Brain an dLanguage, 1976,3,404-433.Ellis, H. D ., &Shepherd, J. W. Recognition of ab-stract and concrete words presented in left andright visual fields. Journal of Experim ental Psy-chology, 1974, 103,1035-1036.Fillenbaum, S., & Rapoport, A. Structures in thesubjective lexicon. New York: Academic Press,1971.Fodor, J. A. , Bever, T. G., &Garrett, M. F. Thepsychology of language: An introduction to psy-cholinguistics and generative grammar. New York:McGraw-Hill, 1974.Gardner, H. The contribution of operativity to nam-

ing capacity in aphasic patients. Neuropsycho-logia, 1973, 11 , 213-220.Gazzaniga,M.S., Bogen, J. E., &Sperry,R. W. Ob-servations on visual perception after disconnectionof the cerebral hemispheres in man. Brain, 1965,

88 ,221-236.Geschwind, N. Disconnexion syndrome in animalsand man. Brain, 1965, 88, 237-294; 585-644.Geschwind, N. The varieties of naming errors.Cortex, 1967, 3, 97-112.Geschwind, N. The organization of language an dthe brain . Science, 1970, 170, 940-944.Geschwind, N . Language and the brain. ScientificAmerican, 1972, 22 (4), 76-83.Goldstein, K. Language an d language disturbances.N ew York: Grune &Stratton, 1948.Goodenough, C ., Zurif, E., & Weintraub, S. Apha-sics' attention to grammatical morphemes. Lan-guage an d Speech, 1977,20 , 11-19.Goodglass, H. Studies on the grammar of aphasics.In S. Rosenberg & J. Koplin (Eds.), Develop-ments in applied psycholinguistic research. N ewYork: Macmillan, 1968.Goodglass, H., & Baker, E. Semantic field, naming ,an d auditory comprehension in aphasia. Brainan d Language, 1976, 3, 359-374.Goodglass, H., Barton, M. E., &Kaplan, E. Sensorymodality an d object naming in aphasia. Journalo f Speech an d Hearing Research, 1968, 11 , 488-496.Goodglass, H., Berko, J. Agrammatism and in-flectional morphology in English. Journal ofSpeech an d Hearing Research, 1960, 3, 257-267.Goodglass, H., Fodor, I., & Schulhoff, C . Prosodicfactors in grammar: Evidence from aphasia.Journal of Speech an d Hearing Research, 1967,10 , 5-20.Goodglass, H., & Geschwind, N . Language disorders(aphasia). In E. C . C arterette & M. F riedman(Eds.) , Handbook of Perception (Vol. 7). NewYork: Academic Press, 1976.Goodglass, H., Gleason, J. B. , Bernholtz, N . A., &Hyde, M. R. Some linguistic structures in thespeech of a Broca's aphasic. Cortex, 1972,8, 191-212.Goodglass, H., & Hunt , J. Grammatical complexityand aphasic speech. Word, 1958, 14, 197-207.Goodglass, H., Hyde, M. R., & Blumstein, S. Fre-quency, picturability, and the availability ofnouns in aphasia. Cortex, 1969, 5, 104-119.Goodglass, H., Klein, B., C arey, P., & Jones, K. J.Specific semantic word categories in aphasia. Cor-tex, 1966,2, 74-89.Goodglass, H., Quadfasel, F. A. , & Timberlake, W .H. Phrase length and the type and severity ofaphasia. Cortex, 1965, 1, 133-153.Green, E. On the contribution of studies in aphasiato psycholinguistics. Cortex, 1970, 6, 216-235.Head, H. Aphasia an d kindred disorders of speech.London: Cambridge University Press, 1926.Hines, D. Recognition of verbs, abstract nouns andconcrete nouns f rom th e left and right visualhalf-fields. Neuropsychologia, 1976, 14, 211-216.


21/21

918 A L F O N S O C A R A M A Z Z AA N D RITA SLOAN BERNDTHowes, D . Some experimental investigations oflanguage in aphasia. In K. Salzinger& S. Salzinger(Eds . ) , Research in verbal behavior and someneurophysiological implications. N ew York: Aca-demic Press, 1967.Jakobson , R. Towards a linguistic typology ofaphasic impairments. In A. V. S. DeReuck & M.O 'Connor (Eds . ) , Disorders of language. L o n d o n :Churchil l , 1964.John son, S. C. H ierarchical clustering schemes.Psychometrika, 1967, 32, 241-254.Katz , J. Semantic theory. N ew York: Harper &Row, 1972.Katz, J. J., & Fodor , J. A. The s t ructure of asemant ic theory. Language, 1963, 39, 170-210.Kempson, R. M. Presupposition and the delimitation

o f semantics. Cambridge, England: CambridgeU

1978 caramazza berndt pb

Documents