BRAIN AND 35, 24-65 (1988) LANGUAGE Dissociation of Inflectional and Derivational Morphology GABRIELE Istituto di Neurologia, MICELI’ Universita Cattolica, Roma, Italia AND ALFONSO CARAMAZZA Cognitive Neuropsychology Laboratory, The Johns Hopkins University A patient is described who makes morphological errors in spontaneous sentence production and in repetition of single words. The great majority of these errors were substitutions of inflectional affixes. The patient did make some derivational errors in repeating derived words but almost never made such errors for nonderived words. The inflectional errors for adjectives and nouns occurred mostly on the plural forms for nouns and adjectives and on the feminine form for adjectives. For verbs, inflectional errors were produced for all tense, aspect, and mood forms. There were no indications that these latter verb features constrained the form of inflectional errors produced. The results are interpreted as support for the thesis that morphological processes are located in the lexicon but that inflectional and derivational process& constitute autonomous subcomponents of the lexicon. 0 19% Academic Press, Inc. A number of reports in the literature have attempted to deal with various aspects of morphological processing in brain-damaged patients. Thus, for example, Patterson (1980; 1982) has provided detailed analyses of “derivational” errors in oral reading (we place derivational in quotation marks because Patterson did not distinguish between inflectional and ’ Also at the Istituto di Psicologia, Consiglio Nazionale delle Ricerche, Roma, Italia. The research reported here was supported by NIH Grant NS23836 and a grant from the Consiglio Nazionale delle Ricerche, Italy. We thank Bill Badecker and Cristina Burani for their comments on an earlier version of this paper and Kathy Yantis for her assistance in the preparation of this manuscript. We are especially grateful to F.S. for his unfailing patience and cooperation throughout all phases of this research. Requests for reprints may be addressed to either author: Alfonso Caramazza, Cognitive Neuropsychology Laboratory, Cognitive Science Center, The Johns Hopkins University, Baltimore, Maryland 21218, U.S.A., or Gabriele Miceli, Servizio di Neuropsicologia, Istituto di Neurologia, Universita Cattolica, Largo A. Gemelli 8, 00168 Roma, Italia. 24 0093-934X/88 $3.00 Copyright All rights 0 1988 by Academic Press, Inc. of reproduction in any form reserved. INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 25 derivational errors in her analyses); Caplan, Kellar, & Locke (1972) analyzed the pattern of inflections in a patient who produced neologistic stems; and, Goodglass and his associates (Goodglass, 1976; Goodglass & Berko, 1960; Goodglass, Gleason, Ackerman, & Hyde, 1972), Gleason (1978), and DeVilliers (1978) have all reported detailed analyses of inflectional errors in the speech of agrammatic patients. These results (as well as others; see Caramazza, Miceli, Silveri, & Laudanna, 1985; Coltheart, 1985) converge nicely with experimental results in psycholinguistic research with normal subjects which suggest the need to represent lexical information in morphologically decomposed form. (See Butterworth, 1983; Cutler, 1983; Henderson, 1985; Taft, 1985, for recent reviews of the literaturenote that not all these authors endorse the view that the lexicon represents lexical information in morphologically decomposed form.) However, major issues remain unresolved, both with respect to the question of how morphological processing may be affected in various conditions of brain damage and with respect to the question of how lexical information is represented and accessed. To properly address questions concerning morphological processing, we need a reasonable characterization of the morphological component of the language processing system. Fortunately, after a period of relative neglect, there has recently been within linguistics a strong resurgence of interest in morphology (e.g., Anderson, 1982; Aronoff, 1976; Jackendoff, 1975; Lapointe, 1979; Lieber, 1980; Matthews, 1974; Scalise, 1984; Selkirk, 1982; Siegel, 1974; Williams, 1981). Although many issues remain unresolved some general principles have emerged which can be used to guide research in morphological processing. We will assume that all morphological operations-both inflectional and derivational-are located in the lexicon (e.g., Lapointe, 1979; Lieber, 1980; Scalise, 1980; Selkirk, 1982; Williams, 1981). However, following Anderson (1982), Aronoff (1976), and Scalise (1984), we will further assume that inflectional and derivational operations constitute functionally distinct processes (see Fig. 1). Briefly a major distinction is drawn between the Derivational Processes Component (DPC) and the Inflectional Processes Component (IPC). Various arguments have been presented in support of this distinction. Some of these are briefly summarized here: (i) Inflections never change the syntactic category of the word to which they are applied whereas derivations may do so (e.g., boy -+ boys; division -+ divisive; inflectional and derivational, respectively); (ii) Rules of inflection are applied only after any rule of derivation may have been applied (e.g., beauty + beautify -+ beauti$ed); (iii) Rules of the IPC and DPC are sensitive to different properties of the base to which they are applied (e.g., syntactic category and semantic features); (iv) The rules of the IPC and DPC affect the root morpheme to which they are applied in different ways (e.g., derivations always change the meaning of the root morpheme to which they are applied- 26 MICELI AND CARAMAZZA LEXICON FIG. 1. Organization of lexicon. The lexicon has three components: A set of root morphemes, a set of derivational affixes and associated rules of affixation (DPC), and a set of inflectional affixes with associated rules of affixation (IPC). kind --, kindness, boy + boyhood-whereas inflections never do so); and, most generally; (v) “Inflectional morphology is what is relevant to syntax” (Anderson, 1982; pg. 587), but the same cannot be stated for derivational morphology. An important feature of the model of the lexicon adopted here is that lexical stem representations (base plus derivational atlixes) and inflectional affixes interact with different kinds of information in the syntactic component of language--only the latter are sensitive to the morphosyntactic representation specified at the S-structure level (e.g., Anderson, 1982). Stating this claim in terms of a process model of sentence production (e.g., Garrett, 1980), lexical stems are accessed through their semantic specification for insertion at the Functional Level Representation while inflectional affixes are selected through morphosyntactic information for insertion at the Positional Level Representation (see also Stemberger, 1985; Dell, 1986; for alternative models of sentence production which nonetheless assume morphologically decomposed lexical representations). Given these differences in the nature of the IPC and the DPC it seems natural to consider these differences to be reflected in processing distinctions in the language system. In such a case it ought to be possible, in principle, to selectively disrupt one or the other morphological component of the lexical system. As already noted there are some indications to suggest that the inflectional component of morphological processing may be disrupted selectively, implying the functional autonomy of this subcomponent of morphological structure. The speech of so-called “agrammatic” patients is supposed to be characterized by the selective omission (or misselection) of inflectional affixes and free-standing grammatical markers. However, it is not clear whether deficits of this general type reflect damage to a lexical or a syntactic component of the language production system-either hypothesis INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 27 is consistent with the performance of the loosely characterized category of agrammatism. One could maintain that damage to a component of the lexical system which “stores” free-standing grammatical markers and inflectional affixes could result in the pattern of speech classified as agrammatic; alternatively, one could place the locus of damage in the processing device that specifies the syntactic frame of the sentence to be produced (see Caramazza & Berndt, 1985, for discussion). Thus, although in either case the deficit is defined in terms of morphological features of language structure, the underlying cause of the impairment may be assumed to be located either in the lexicon or the syntactic processing component of language production. Of course, for any one “agrammatic” patient the underlying deficit may be to either or both components of processing. One way to distinguish between specifically lexical and syntactic deficits in morphological processing is to consider a patient’s performance in single-word processing. If the patient were to make morphological errors in single-word processing, we would be able to conclude that at least one factor contributing to the patient’s impairment is a deficit of the lexical processing system. In this paper we report a detailed analysis of a patient who makes morphological errors in both sentence and singleword processing. This pattern of results and the presence of a marked dissociation between inflectional and derivational errors in the patient’s single-word repetition performance are consistent with the hypothesis of a selective deficit to an autonomous Inflectional Processing Component of the lexicon. CASE HISTORY F.S. is a 60-year-old, right-handed man. Prior to his illness he was a well-known lawyer. In March, 1978, he underwent surgery for the evacuation of an acute intracerebral hematoma. After surgery, he was severely aphasic and had marked neurological deficits. The patient was referred to one of the authors (G.M.) approximately 2 years postonset, and has been seen regularly as an outpatient since then. The study reported here was carried out between January 1984 and February 1985. The patient’s condition remained stable throughout this period. Neurological Exam The patient has a right hemiplegia, with a severe spastic hypertonus. He shows extinction phenomena on the right in all modalities, but has no clinical signs of neglect. Visual extinctions are more frequent in the right inferior than in the right superior quadrant. A CT-scan, performed in 1982 (4 years postonset), revealed a large hypodense area in the left hemisphere. The lesion involved the temporal lobe in its anterior and middle portions, the claustrum, the internal capsule and (probably) the 28 MICELI AND CARAMAZZA insula, and also extended to the periventricular lobe (see Fig. 2). Neuropsychological white matter of the parietal Examination The patient was submitted to an extensive battery of clinical tests. The results demonstrated normal visuoperceptual and visuospatial abilities, and normal constructional praxis. There were very mild signs of ideomotor apraxia on verbal command, but they disappeared when the patient was asked to imitate gestures made by the examiner. A very mild buccofacial apraxia on verbal command was also present. It also disappeared on imitation, but the movements were executed slowly and with hesitations. The ability to perform sequences of meaningless movements with the buccofacial apparatus was severely impaired (F.S. scored 5/12 on the Nonverbal Agility subtest of the BDAE (Goodglass & Kaplan, 1972)). Visual memory functions were well within normal limits. On Corsi’s blocks, F.S.‘s span was 6 forward, 4 backward. Verbal memory functions, by contrast, were impaired. On the Rey’s 15word list, the patient performed very poorly. The test consists of the auditory presentation of a list of 15 words. After each presentation the patient is asked to report as many words as possible. F.S. reported only 2/15, 5/15, 6/15, 4/15, and 5/15 items after each presentation and, 15 min later, he was able to recall only 5/15. Digit span was 3 forward and 3 backward. Pointing span for objects was 3. The patient scored 24/36 on Raven’s Colored Matrices. A moderate acalculia was also present. Language Examination The motor aspects of F.S.‘s verbal output were peculiar in many respects and could probably be best described as “foreign accent” production. Articulation had an “explosive” character, with abrupt changes in voice pitch, often within the same word. The patient showed some difficulties in producing complex consonant clusters, but no systematic simplifications were noted. Prosody was severely disturbed: Far from being flat, the poor control of pitch made it sound awkward to the listener. F.S. himself reported that people who did not know him mistook him for a foreigner (German, to be specific). Verbal agility was reduced (4/14 on the Verbal Agility subtest of the BDAE). Qualitative and quantitative aspects of F.S.‘s verbal production will be discussed under the Results. Oral naming was mildly impaired for objects (92% correct) and moderately for actions (75% correct). Most of the errors were anomias (inability to provide a response) or phonemic paraphasias, with very few semantic substitutions. Written naming displayed essentially the same features FIG. 2. INFLECTIONAL AND DERIVATIONAL CT-scan F.S. showing of patient a large MORPHOLOGY lesion in the left hemisphere. 29 30 MICELI AND CARAMAZZA (with spelling errors-letter substitutions-instead of phonemic paraphasias). The performance obtained by F.S. in repeating single words and nonwords will be described in detail below. In sentence repetition, the patient demonstrated di&ulties with free-standing grammatical markers (omissions and substitutions) and with bound grammatical markers (substitutions). In addition, he also produced some semantic substitution errors on majorclass lexical items (for example, “Giovanni ha comprato due macchine” (John bought two cars) --, “Giovanni acquistate (f.pl.) due auto” (John purchased (past part., f.pl.) two autos)). Comprehension of isolated words was “normal” for both auditory and visual presentation. This was true for tests using phonemic foils (l/57 incorrect responses) and for tests employing semantic foils (3/60 incorrect responses). Auditory phoneme discrimination based on meaningless CCVC syllables was poor (41/120 incorrect responses) as was auditory-visual matching using the same material (24/120 incorrect responses). Auditory sentence comprehension was also severely impaired. On the syntax comprehension test by Parisi and Pizzamiglio (1970), F.S. responded incorrectly to 14/80 stimuli. He made errors on declarative (l/4) and embedded active reversible sentences (2/4), on declarative passive (3/4) and locative sentences (6/36), and on sentences requiring the comprehension of verb morphology (2/8). In another test of sentence comprehension F.S. produced 21/50 incorrect responses. He made errors on declarative reversible sentences (l/10 active, 6/10 passive), on embedded reversible sentences (3/10 active, 9/10 passive), and on sentences expressing reversible temporal relations of the type before/after (3/10). The patient also had some minor difficulty in acting out sentences which required comprehension of the weak (clitic) vs. strong form of the personal pronoun, of the type: “Dammi/Dagli la penna” (Give me/Give him the pen) vs. “Dai la penna a me/a lui” (Give the pen to me/to him). He made 2/20 errors with sentences of the first type, but performed flawlessly with sentences of the second type. In an auditory sentence-to-picture matching task, F.S. also showed impaired comprehension of sentences containing the reflexive vs. the anaphoric forms of the clitic pronoun (e.g., “La donna si lava” (The woman washes herself), and “L’uomo li ferisce” (The man hurts them), respectively). The patient had to choose from two pictures, one corresponding to the stimulus sentence, the other representing the contrasting clitic personal pronoun (e.g., for the stimulus sentence “La donna si lava” (the woman washes herself), F.S. was shown a picture representing the correct response, and a picture depicting “La donna li lava” (The woman washes them); for the stimulus “L’uomo li ferisce” (The man wounds them) the incorrect alternative portrayed “L’uomo si ferisce” (The man wounds himself); and so on). F.S. made 8/48 errors on this test. INFLECTIONAL AND DERIVATIONAL TABLE 1 ritomo a casa mia 2 Poi ritoma [a] !zj mia casa. 3 Then returns the my house [f.sg.] 4 Then returns my house. MORPHOLOGY 31 1 la Poi io ascolto jj televisione Then I hear the [msg.] television [f.sg.] Then I listen to the television 1 faccio vivo perche’, care dottore, io e solo! 2 0 poi fare il pranzo 3 or then to make the lunch because, dear doctor [m.sg.] I lives alone [m.sg.] 4 or then make lunch because, dear doctor, I lives alone! lunghe 1 ricevo faccio le 2 Poi telefono, riceve, fare, perche’ fl giornate [sono] lunge, 3 Then [I] telephone, receives, to make because the [m.sg.] days [f.pl.] long [m.sg.] 4 Then I telephone, receives, make [phone calls], because the days long, i Parioli [In] via C. (a) il Pariolo street C. il Pariolo 1 vado 2 poi ancora spesso andare [allo] studio. 3 then still often go my office 4 then still often go my office. C. Street, Pariolo, 1 ii mio studio 2 c’e la mia studia ancora aperto. 3 there is the [f.sg.] my [f.sg.] office [f.sg.] still open [m.sg.] 4 there is my office, still open. Note. In this sample the patient is describing what he habitually does in the morning. The patient’s speech is reported in line 2, which reproduces verbatim his spontaneous speech. Words omitted in obligatory context are in square parentheses. Words produced incorrectly are underlined. Line I gives the correct target words. Line 3 is the literal, word-by-word English translation of line 2. Whenever agreement is ambiguous in English, gender and number information is provided in parentheses (m, masculine; f, feminine; sg, singular; pl, plural). Finally, line 4 is a rough equivalent of line 2, to give the reader an approximate idea of how the patient’s spontaneous speech would sound in English. EXPERIMENTAL STUDY 1. Spontaneous Speech A sample of connected speech consisting of 873 words was obtained by asking F.S. to tell us about his illness, to describe some habitual daily activities, and to recount the story of Little Red Riding Hood. F.S.‘s speech displayed marked grammatical deficits as is obvious from the speech sample shown in Table 1 (see also Appendix I). The patient’s speech is characterized by the omission or misselection of free-standing grammatical markers, reduced phrase length, and violation of gender, number, and tense/aspect agreement. Thus, for example, the patient produced “. . . poi ancora spesso andare [allo] studio.” where “allo” (to the) is omitted and the verb is inappropriately inflected in the infinitive (andare) instead of the correct finite, first person, singular form (vado). We have quantified various grammatical aspects of the patient’s speech and this is shown in Table 2, together with the performance of another 32 MICELI AND CARAMAZZA TABLE FEATURES OF F.S.‘s SWNTANEOUS Percentage of words out of syntactic context Mean phrase length (syntactic) Mean phrase length (morphological) Rate of omission of freestanding grammatical markers Rate of substitution of freestanding grammatical markers Rates of Main to Subordinate clauses NP-PP/Total NP Violations of det/N agreement Violations of N/Adj agreement Violations of Subject/V agreement 2 SENTENCE PRODUCTION PERFORMANCE” F.S. C.D.A. (agr.)b E.B. (control) 4.2% 14.4% 0% 6.11 3.2 10.14 3.16 2.4 10.14 54/242 (22.3) 62/255 (24.3) O/227 (0) 481242 (19.8) 191255 (7.4) O/277 (0) 2 21/184 (11.4) 14.9 1l/175 (6.3) 1.7 25/214 (11.7) 19/138 (13.8) 2191 (2.2) (0) 1 I /55 (20.0) o/11 (0) (0) 45182 (54.9) 18/97 (18.6) (0) ” Percentages are in parentheses. b For comparison we have included the performance of an “agrammatic” subject. and a normal patient whose speech would also be clinically classified as “agrammatic” and a matched normal control subject. There are several features of the patient’s performance that are worth focusing on. First, and most important for our present purposes, it is clear that the patient omits (22.3%) or misselects (19.8%) free-standing grammatical markers and misselects inflectional affixes (20.4%). Especially striking is the large proportion of violations of Subject/Verb agreement (55.9%), but the patient also made a substantial number of Determiner/Noun and Noun/Adjective agreement violations (13.8% and 20%, respectively). These difticulties are reflected directly in two other measures of the grammatical appropriateness of the patient’s speech: mean phrase length of syntactically (MPL-S) and morphologically (MPL-M) appropriate speech. The former measure (MPL-S) captures the length of lexically appropriate strings, whether or not there are violations of agreement (e.g., the string “c’e la da studiu ancora aperto” would be scored as length 7 even though the phrase “la mia studia” is morphologically deviant). The other measure (MPL-M) captures the length of grammatically correct strings including agreement. As can be seen in Table 2, F.S. has a reduced MPL-S (6.1) and even shorter MPL-M (3.2) when compared INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 33 to a matched normal control (10.1 for both measures). However, the reduced MPL is not due to a reduction in availability of syntactic framesthat is, it is not due to an inability to produce subordinate clauses (the ratio of main to subordinate clauses is 2 and 1.7 for F.S. and the normal control, respectively) or prepositional phrases (11.4 and 11.7% of total noun phrases were prepositional phrases for F.S. and the normal control, respectively)-the patient appeared to be quite normal in this regard. The reduced phrase length appears to result from the omission of freestanding grammatical markers or the misselection of both free and bound grammatical markers. It is instructive to compare F.S.‘s performance to that of another patient who would also be clinically classified as “agrammatic.” For this purpose we compared his performance to patient C.D.A. The two patients differ on a number of measures as can be seen in Table 2: They differ in the mean phrase length of their speech (both MPL-S and MPL-M), in the proportion of words produced out of grammatical context, in their use of subordinate clauses and prepositional phrases, and in the proportion of agreement violations-whereas F.S. is less impaired than C.D.A. in the first three measures of grammaticality of speech, there is a clear reversal in severity of impairment for the last mentioned measure. It is unlikely that the underlying deficit for these two patients is the same. Thus, the mere classification of these two patients as “agrammatic” would be misleading, at best. In any case, the point here is that F.S.‘s spontaneous speech clearly reflects an impairment in morphological processing. What is not clear is whether the disorder arises from damage to the lexical or to the syntactic component of the sentence processing system. Two hypotheses (at least) may be entertained about the possible locus of functional damage to the sentence production system which would give rise to the pattern of omission and misselection errors produced by F.S. in spontaneous speech. The Lexical Deficit Hypothesis assumes that the locus of damage is to some component of the lexical system. For this hypothesis to be entertained seriously, however, the structure of the lexicon must be articulated in such a fashion that function words and inflectional affixes may be damaged selectively. That is, function words must form an autonomous subcomponent of the lexicon, and the inflectional component of the lexicon must similarly be autonomously represented within the lexicon. (Of course, function words and inflectional affixes could functionally constitute a single processing component.) The other hypothesis, the Syntactic Deficit Hypothesis, assumes that the locus of damage is to a level of sentential processing which distinguishes grammatical features (such as free and bound grammatical markers) from the major lexical items of a sentence-perhaps the Positional Level in Garrett’s model of sentence production (Garrett, 1980, 1982; see also 34 MICELI AND CARAMAZZA Lapointe, 1985). Of course, it is entirely possible that F.S. may have damage to both components of processing. While it is difficult to provide evidence which would rule out the contribution of damage to the syntactic processing component to the overall pattern of speech production impairment in our patient, it is a relatively simple matter to state the conditions under which we would accept the hypothesis that damage to the lexical system is a major contributor to the patient’s impairment. Thus, if it could be demonstrated that F.S. has difficulties in single-word processing that are explicable by appeal to the morphological structure of words, we would be inclined to accept the hypothesis that at least part (if not all) of the spontaneous production disorder is due to damage to the lexical system. Of course, this latter inference is only justified if the single-word processing impairment can be shown to involve the relevant subcomponent of the lexical systemthe inflectional component in this case. To address this issue we assessed F.S.‘s ability to process single words with a variety of tasks. 2. Repetition Tests The patient was asked to repeat, in several sessions, 1832 words and 283 nonwords. The word sample contained several sublists, controlled for form class, frequency, length, morphological complexity, type of suffix (inflectional vs. derivational), and presence of prefix vs. pseudoprefix. The stimuli were presented in random order, at a normal rate of articulation. Some sublists only contained words, whereas others contained both words and nonwords intermixed in random order. The examiner pronounced each stimulus aloud, and the patient was asked to repeat it. a. Word Repetition F.S.‘s ability to repeat single words is severely impaired: He repeated incorrectly 919/1832 (50.2%) stimuli. A breakdown of repetition performance as a function of the form class of the stimulus word (without controlling factors such as word length, frequency, and abstractness/concreteness; but see below) reveals major effects of form class (see Table 3). The patient performed equally poorly on function words and nouns (31.0 and 29.6% errors, respectively) and even worse on adjectives (49.2% errors) and verbs (64.3% errors). A quantitative analysis of the errors produced by F.S. (a total of 919 errors) is shown in Table 4. Responses were scored as correct or as one of a number of incorrect response types. A description, with examples, of each of the error categories used in this analysis is shown in Appendix I. F.S. produced 663/919 (72.1%) incorrect word responses. If we restrict the analysis of errors to polymorphemic words in the sample (i.e., words which permit morphologically legal errors-nouns, adjectives, and verbs), INFLECTIONAL AND DERIVATIONAL TABLE PERFORMANCE OBTAINED REPETITION MORPHOLOGY 35 3 BY F.S. IN WORD TASKS Part of speech Errors % Nouns Adjectives Verbs Function words Total 121/409 2901589 482/150 26184 919/1832 29.6 49.2 64.3 31.0 50.2 659/893 (73.8%) errors resulted in word responses and, most importantly, they were almost all (636; 96.5% of the word responses or 71.2% of the total corpus of errors) morphologically based. These latter errors consist of the correct repetition of the root of the stimulus word and a substitution of the affixal part of the word (e.g., settimana (week) + settimane (weeks); povera (poor; f.sg.) + povero (poor; m.sg.); vestire (to wear) + vestivi (you were wearing)). A striking feature of these morphologically based errors is that they are essentially all (615/637 or 96.7%) inflectional errors. Admittedly the sample of words we are considering here is not a random sample of the words in the language-verbs and adjectives are overrepresented-but, the dissociation between inflectional and derivational errors cannot be attributed to this sampling bias (see below for F.S.‘s performance on controlled sublists). Of the remaining, nonmorphological errors, very few word responses consisted of semantic paraphasias (4; 0.4%) or unrelated word responses (19; 2.1%; e.g., reca (he brings) + credu (that he believes)). Clearly, the primary dimension along which we can characterize the word-error responses is a morphological one. The pattern of nonword responses (186/893; 20.8%) is also informative. A nonnegligible proportion of these errors (55/186; 29.6%) could be considered to be “morphologically based.” This latter type of error involves a substitution of inflectional or derivational affixes. Some of them consist of a correct root plus a suffix not permissible for that root (22/55; 40.0%; e.g., resisteva (he was resisting) -+ resistive (nonword)the affix -ire is the infinitival form for verbs of the 3rd conjugation and is not appropriate for the root resist- which requires the affix -ere for verbs of the 2nd conjugation); The other errors in this category consist of phonologically similar pseudoroots with substituted suffixes (33/55; 60.0%; e.g., jkztavi (you were breathing) -+ &z&e-a pseudoroot jiaf+ the “appropriate” (forfiat-) infinitival affix -are). The remaining nonword errors (13 1/ 186; 70.4%) are morphologically nondecomposable responses which result from phonemic substitutions of one or more phonemes in the stimulus word. The error corpus also contained a small proportion of “fragment” 201 (69.4) 16 (5.5) 1 (0.4) 3 (1.0) 3 (1.0) 40 (13.8) 2 (0.7) 3 (1.0) 12 (4.1) 9 (3.1) 55 (45.5) 6 (5.0) 5 (4.1) 1 (0.8) 14 (11.6) 4 (3.3) 27 (22.3) - 1 (0.8) 6 (5.0) 2 (1.6) 2 (7.7) 3 (11.5) 11 (42.3) 3 (11.5) 1 (3.9) 2 (7.7) 4 (15.4) 45 (5.0) 15 (1.7) 2 (0.2) 21 (2.4) 7 (0.8) 131 (14.7) 4 (.W 19 (2.1) 25 (2.8) 23 (2.6) 15 7 12 (3.1) (1.4) (2.5) 64 (13.3) 2 (0.4) (4.8) (1.9) (0.2) (0.8) 23’ 9 1 4 - 601 (67.3) Function words 345 (71.6) TASKS” Total REPETITION Verbs TABLE 4 BY F.S. IN WORD 20 (2.2) 26 (2.8) 27 (2.9) 47 (5.1) 18 (2.0) 2 (0.2) 21 (2.3) 7 (0.8) 142 (15.5) 7 (0.8) 601 (65.4) Total ’ Percentages are in parentheses. b Twelve errors are incorrect word responses consisting of the substitution of the verb stem with a phonologically related stem and of the verb inflection (e.g., RIDARO’ (I will give again) --j RIDIRA’ (he will say again)); 11 errors are nonword responses consisting of the substitution of the verb stem with a phonologically related pseudostem and of the verb inflection (e.g., SUDAVA (he was sweating) + SUGIARE (nonword)). Inflectional Phonemic subst. on root and inflect. subst. Illegal inflections Prefix omission Derivational Illegal derivations Phonemic substitution Semantic substitution Other word response (phonologically related) Fragments No response MADE Adjectives OF ERRORS Nouns DISTRIBUTION INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 37 responses (29/893; 2.8%; e.g., “unirono” (they united) -+ “uni” . . .) and omissions (23/893; 2.6%). These errors will not be discussed further. F.S.‘s repetition performance for function words contrasts markedly with that for other word classes. Specifically, the majority of errors for function words consist of phonemic substitutions.2 This result reflects the fact that the great majority of function words used in the repetition task are uninflected and, therefore, there is no possibility for making morphological errors. We wish to stress, however, that overall performance for this class of words is not worse than that for other word classes. In fact, as already indicated, the level of performance for function words was comparable to that for nouns, which is better than that for adjectives and verbs. The results we have reported thus far are striking in at least two respects. First, the patient’s single-word repetition difficulty is characterized by morphological errors; of the total storable errors produced, 84.5% were morphologically based and only 15.5% were phonemic paraphasias. This distribution of errors suggests that while the patient does have a phonological processing deficit (at some level of processing), it cannot account (in any simple way) for the massive presence of morphological errors. Second, and perhaps more importantly, the great majority of morphological errors were inflectional (96.7% vs. 3.3% for inflectional and derivational errors, respectively). On the face of it, this latter result rather compellingly suggests a selective deficit to the inflectional component of the lexicon and, therefore, implies functional autonomy for this subcomponent of the lexicon. However, before drawing such a strong conclusion with its associated inferences for the structure of the lexical system, there are several issues that should be addressed: We should investigate the parameters that determine the probability of making an error in repetition and, specifically, a morphological error; we should more carefully evaluate the dissociation between inflectional and derivational errors; we should evaluate possible differences in processing prefixes and suffixes; we should characterize the pattern of morphological errors-that is, the distribution of responses as a function of stimulus characteristics; and, we should attempt to determine whether functional damage is restricted to the output lexicon, the input lexicon, or both. We next turn to these issues through the analysis of various sublists included in the total corpus of words we have just discussed. The effects of word length, word frequency, and form class on repetition performance. In order to assess the effects of word length, word frequency, ’ This result is somewhat surprising given the commonly accepted position that errors on this word class tend not to be phonemic substitutions. However, it must be noted that the particular sample of words used in our task consisted of bi- and polysyllabic words, unlike most of the high frequency function words for which the claim on nonoccurrence of phonemic substitutions is made. 38 MICELI AND CARAMAZZA TABLE 5 F.S.‘s WORD REPETITION PERFORMANCE” Total errors Sublist 1: Form class x Frequency x Length Form class Nouns 15/52 (28.8) Adjectives 20152 (38.5) Verbs 26/52 (50.0) Function words 21/52 (40.4) Length Short (4-6 phonemes) 29/104 (27.9) Long (7-9 phonemes) 53/104 (51.0) Frequency High (a lOO/million 29/104 (27.9) Low (I lO/million) 53/104 (51.0) Sublist 2: Root length x Suffix length Root Short (2 phonemes) 1l/24 (45.8) Long (4-6 phonemes) 15/24 (62.5) suffix Short (1 phoneme) 7/16 (63.7) Medium (3 phonemes) 9/16 (56.2) Long (4-5 phonemes) IO/l6 (62.5) Morphological errors (N.A.V.) 11/52 (21.2) 16/52 (30.8) 21/52 (40.4) 16/78 (20.5) 29178 (37.2) 17178 (21.8) 28/78 (35.9) 5/24 (20.8) lo/24 (41.7) 4/16 (25.0) 6/16 (37.5) 5/16 (31.2) a Percentages are in parentheses. and form class on F.S.‘s repetition performance, we asked him to repeat a list of 156 words controlled for these factors. The list consisted of 52 nouns, 52 adjectives, and 52 verbs. Half of the total set of stimuli (78) had a mean frequency of 150/million, range IOO-300/million; the other half had a mean frequency of 20/million, range lO-30/million (frequencies from Bortolini, Tagliavini, & Zampolli, 1971); half of the total set of stimuli ranged from 4 to 6 phonemes, the other half ranged in length from 7 to 9 phonemes. The high and low frequency and the short and long words were distributed evenly among the three classes of words. We also included in the list 52 function words. However, for these words we could not control the factors of length and frequency. Repetition results for this list of words are shown in Table 5. There are clear effects of length (x2 = 10.649; p < .Ol) and frequency (x2 = 10.649; p < .Ol). The form class condition did not reach an acceptable level of significance (x2 = 4.912; p = n.s) but there is a trend for verbs to be repeated more poorly than adjectives, which in turn are repeated more poorly than nouns. It is worth stressing that the proportion of morphological errors remains roughly invariant as a function of the factors manipulated in the experimental list. INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 39 The results in this section confirm conclusions reached earlier in the paper. Specifically, the effect of word length-a nonlexical factor-suggests that a factor contributing to the patient’s performance is the phonological “complexity” of the stimulus; that is, the sheer number of phonological segments to be processed affects the probability of a correct response. The effect of word frequency points instead to a deficit in the lexicon. The assumption here is that the availability of the whole-word phonological representation of a lexical entry is a direct function of the frequency of usage of that phonological representation-frequently used words are relatively more accessible than infrequently used words (e.g., Morton, 1979). If we further assume that damage to the lexicon accentuates the relative accessibility of phonological representation, then we would expect that words of lower frequency of occurrence in the language would be the ones most difficult to produce. In short, F.S.‘s repetition difficulty arises from damage to independent components of the word production system-damage to the lexicon and damage to nonlexical phonological processes, at either the input or the output level. The lexical deficit does not appear to be form-class-specific (there is only a nonsignificant trend in this direction). Repetition of prefixed words. F.S. was asked to repeat 85 prefixed words, each containing one of eight different prefixes, interspersed in various sublists. The following prefixes were included in the experimental list: ri-, as in rifare (to redo); dis-, as in disgrazia (disgrace); s-, as in scoperto (uncovered); in-, as in inutili (useless); con-, as in congiungono (they conjoin); re-, as in respinto (pushed); di-, as in distacco (detachment); and, pre-, as in pregiudizio (prejudgment). Thirty-nine of these prefixed words, which were part of a large list including items belonging to all word classes, were matched for part of speech, frequency (approximately 20/million), length (mean length: 8.3 letters), and initial letter cluster to 39 pseudoprefixed words (e.g., prefixed word: rifare (to redo); pseudoprefixed word: ricevo (I receive)). The results obtained by F.S. in repeating the entire corpus of prefixed words, and the subsets of matched prefixed and pseudoprefixed words, are shown in Table 6. The most relevant aspect of F.S.‘s performance is his proficiency in repeating prefixes as opposed to suffixes: Of the 44 errors he made on the prefixed words, only two were prefix errors, both omissions (“stendono” (they stretch) + “tendono” (they tend, or they stretch); “slancio” (thrust) --+ “lancio” (throw)), but he reproduced incorrectly 25 suffixes, 20 inflectional and 5 derivational errors. It is worth noting that both prefix omissions occurred with consonant clusters (/s/ + consonant) that are both articulatorily complex and relatively infrequent. This consonant cluster proved difficult for F.S. also when it occurred in wordinitial position of nonprefixed words (consider the following errors in repetition found in the larger error corpus: “stivali” (boots) -+ “tivali” 17 (38.6) 11 (25.0) 5 (11.3) 1 (2.3) 3 (16.7) 3 (16.7) - 15 (83.3) 2 (11.1) 13 (72.2) 27 (61.4) 2 (4.6) 25 (56.8) 10 (55.5) 3 (16.7) 7 (38.9) 2 (11.1) 1 (5.6) 4 (22.2) 11 (61.1) 11 (61.1) 21 (53.8) (n = 39) (n = 39) 21 (53.8) Matched pseudoprefixed Matched prefixed WORDS” 41 (48.2) 20 (45.5) 5 (11.3) 6 OF PREFIXED TABLE ” Performance obtained by F.S. in repeating the whole sample of prefixed words and matched subsets of prefixed and pseudoprefixed Correct responses and error distributions (percentages are in parentheses). 1. Correct responses 2. Error distribution a. Morphological Prefix suffix Inflection Derivation b. Nonmorphological Phonosubsterrors Fragments No response Overall prefixed (n = 85) REPETITION words: 11 (61.1) - K F $ 5 5 z w INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 41 (nonword); “strana” (strange) + “trana” (nonword); “spina” (thorn) + “pina”). F.S.‘s repetition performance for prefixed words leaves no doubt that he does not have particular difficulties in processing prefixes. Instead, it appears that his morphological processing impairment is restricted to suffixes and, in particular, inflectional suffixes. Znflectional vs. derivational suffixes. In an earlier section of this paper we reported that for the full corpus of morphological errors there is a striking dissociation between inflectional and derivational errors (96.7% vs. 3.3%, respectively). As we emphasized in that section, however, the stimulus words included in the total corpus were not controlled for potentially relevant factors that may determine the probability of making an inflectional or derivational error. Thus, for example, one of the factors not controlled for in the total stimulus set is the proportion of derived vs. inflected words. Since in Italian all words except a subset of the free-standing grammatical markers are inflected, there is a very large discrepancy between the number of inflected versus derived words in the total stimulus set. If the probability of making a derivational error is a function of whether a word is derived, then the paucity of such words in the stimulus set would underestimate the extent of processing difficulties of derivational morphology in our patient. To more accurately assess whether inflectional morphology is impaired while derivational morphology is normal in our patient, we analyzed selected subsets of the total stimulus set. In the first analysis we considered the distribution of errors for suffixderived words. Our stimulus set contained 30.5 words which could unambiguously be considered to be derived words. F.S. successfully repeated 13.5 (44.3%) words from this set. Of the errors 109 (64.1%) were morphologically based; of these errors 90 (52.9%) were inflectional, 12 (7.0%) were derivational, and 7 (4.1%) could not be unambiguously scored as either derivational or inflectional (e.g., ballavi (you were dancing + “ballo” (I dance-inflectional-or the dance-derivational). The second analysis considered the repetition performance for a specially designed list of words. The 106-word list consists of 53 inflected words (all verbs) and 53 derived words (nouns and adjectives). The two subsets of words were exactly matched for length and frequency. Since the frequency of derived words tends to be low, more than two-thirds of the stimuli had a frequency of lO/million or less. F.S. repeated correctly 18 (33.4%) inflected and 18 (33.4%) derived words. For the inflected words all the morphological errors were inflectional (25; 71.4%); for the derived words he produced mostly inflectional errors (18; 51.4%), but he also produced a few derivational errors (3; 8.6%). The results reported in this section confirm the conclusion stated earlier in the paper that the patient has a deficit in processing inflectional mor- 42 MICELI AND CARAMAZZA phology. However, there is also some indication that the processing of derivational morphology is not entirely normal in our patient: For the full corpus of morphological errors (136) produced for derived words (either prefixed or saxed), F.S. made a nonnegligible proportion (10.3%) of derivational errors. We will discuss the discrepancy between the proportion of inflectional and derivational errors under the Discussion. Patterns of inflectional errors. We have argued that F.S.‘s repetition impairment results from damage to two independent processing mechanisms: a deficit at some level of phonological processing and a deficit to the morphological processing component of the lexical system. Various arguments may be given in support of this conclusion. The most direct, and the one we will focus on here, is based on the analysis of distributional characteristics of repetition errors. Specifically, we contend that one may plausibly conclude that the functional locus of damage is at some level of phonological processing if the distribution of errors is explicable by appeal to phonological principles. Similarly, one may plausibly conclude that the functional locus of damage is at some level of morphological processing if the distribution of errors is explicable by appeal to morphological principles. There is already clear evidence for supposing that F.S. has a phonological processing impairment: A substantial proportion of repetition errors consisted of phonemic paraphasias. The issue now is whether the putative “morphological” errors might not also merely result from damage to some phonological processing mechanism. This issue was addressed by considering the distribution of inflectional errors made for nouns, adjectives, and verbs. Since the clearest case can be made for adjectives, we will consider this form class first. F.S. was asked to repeat a set of adjectives which could take one of four alternative inflectional affixes. The four inflectional affixes mark gender and number (e.g., care (dear; m.sg.), cari (m.pl.), cara (f.sg.), care (f.pl,)). This experimental list included 30 m.sg. adjectives, 23 m.pl. adjectives, 30 f.sg. adjectives, and 30 f.pl. adjectives, matched for length (6.9 to 7.2 letters), root frequency (27.4/million to 29.6/million), and surface form frequency (6.4/million to 7/million). F.S. repeated correctly 28 m.sg. adjectives (93.3%), 10 m.pl. adjectives (43.5%), 16 f.sg. adjectives (53.3%), and 8 f.pl. adjectives (26.7%). The patient displayed a marked tendency to produce the m.sg. as the incorrect response: of the 45 inflectional errors, 36 were substitutions of the m.sg. inflection for other inflectional endings (80.0%). Since the m.sg. form of the adjective is usually the most frequent, as well as the citation form,3 F.S. was asked to repeat another list of 3 By citation form of a word, we mean that form which is considered to be basic in respect to other inflected forms. It is the form which is listed in dictionary entries and is INFLECTIONAL AND DERIVATIONAL TABLE NUMBER OF CORRECT WITH FOUR ENDINGS, SURFACE RESWNSES MATCHED PROVIDED IN LENGTH 43 MORPHOLOGY I BY F.S. IN REPEATING A SUBLIST OF ADJECTIVES AND CONTROLLED FOR ROOT FREQUENCY AND FORM FREQUENCY Form administered Surface form frequency m.s.g. > non-m.sg. m.sg. < non-m.sg. m.sg. non-m.sg. 41150 40150 5150 11/50 adjectives to determine whether the very high percentage of incorrect m.sg. responses merely reflected a tendency to produce the most frequent form of the adjective. This list included 200 adjectives equally divided among the following four types: (i) adjectives whose m.sg. form is more frequent than non-m.sg. forms, presented in the m.sg. form (root frequency: 40/million; form frequency of the m.sg. form: 20.9/million; form frequency of non-m.sg. forms: 3/million); (ii) a non-m.sg. form of the adjectives in (i); (iii) adjectives whose m.sg. form is very infrequent, presented in the m.sg. form (root frequency: 40/million; form frequency of the m.sg. form: S/million; form frequency of the non-m.sg. form: 22S/million; (iv) a nonm.sg. form of the adjectives in (iii). The incidence of the three non-m.sg. inflections was balanced across subsets (i) and (iv). Length was balanced across subsets. Once again, the results are clear-cut (see Table 7): F.S. repeated correctly m.sg. adjectives a high proportion of the time (41/50 from subset (i) and 40/50 adjectives from subset (iii)), but non-m.sg. forms were reproduced correctly much less often (5/50 from subset (ii), and 11/50 from subset (iv)). Of the inflectional errors made in this task (71/113; 62.8%), the vast majority (56/71; 78.3%) consisted of the substitution of the m.sg. inflection for a non-m.sg. suffix. If we consider the distribution of correct responses and inflectional errors for the entire set of adjectives that take four inflectional affixes, there is a very strong effect of inflectional form on performance. Table 8a summarizes these data. As is immediately apparent, the m.sg. form is repeated correctly with very high probability (94.9%). The probability of correctly repeating the other inflectional forms is quite low, by contrast, ranging from 25.5 to 34.2 to 39.8% for f.pl., m.pl., and f. sg., respectively. Furthermore, there is an overwhelming tendency to produce the m.sg. form of the adjective independently of the inflectional form of the stimulus. usually given in response to questions such as “How language?” do we say x in such and such a 44 MICELI AND CARAMAZZA TABLE CONFUSION MATRIX FOR INFLECTIONAL ERRORS Z-ENDING M.SG. (a) Four-ending adjectives MSG. 149 (94.9) M.PL. 40 (52.6) FSG. 43 (48.9) F.PL. 34 (61.8) Total 266 (70.7) (b) Two-ending adjectives SG. PL. 8 BY F.S. IN REPEATING MADE ~-ENDING AND ADJECTIVES M.PL. F.SG. F.PL. Total 8 (5.1) 26 (34.2) 1 (1.1) 2 (3.6) 37 (9.8) 5 (6.6) 35 (39.8) 5 (9.1) 45 (12.0) 5 (6.6) 9 (10.2) 157 76 88 55 376 SG. PL. 56 (81.2) 36 (65.5) 13 (18.8) 19 (34.5) a Percentages are in parentheses. A similar analysis for adjectives that only take two inflectional endings (e.g., forte (strong, m.sg. and f.sg.), forti (strong, m.pl. and f.pl.)) gives comparable results. As is shown in Table 8b, whereas the singular form is very often repeated correctly (81.2%), the plural form is only infrequently repeated correctly (34.5%). And, again here, there is a strong tendency to incorrectly produce the singular form when repeating an adjective marked for plural. These results leave little doubt that the major determinant of morphological errors in our patient is a deficit to the inflectional processing component of the lexical system. The distribution of repetition errors is explicable by appeal to morphological principles: The most often produced form of an adjective, both correctly and incorrectly, is the citation form. That is, the patient’s tendency to produce the m.sg. form for the fourending adjectives does not merely reflect a frequency effect-the citation form was given both when it is the relatively high and when it is the relatively low frequency form of an adjective-nor, does it reflect a bias to produce the phonological form /o/, which is associated with the m.sg. also form of four-ending adjectives (e.g., care, cara, care, cari)-F.S. overwhelmingly produced the singular form for two-ending adjectives which take the phonological form /e/, the phonological form associated with f.pl. inflections for four-ending adjectives. In short, F.S.‘s pattern of inflectional errors for adjectives cannot be explicated by appealing to the relative frequencies of alternative inflectional forms of adjectives nor by appealing to potential preferences for alternative phonological forms. Instead, there is strong support for the view that the primary determinant of the probability of correctly repeating an adjective is whether the INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 45 stimulus is given in the citation form or in another inflected form-a morphological principle. Although we don’t have nearly as strong a data base for nouns as we have for adjectives and, therefore, we cannot evaluate the effects of such factors as frequency and phonological form on repetition, what little data we have on F.S.‘s performance in repeating nouns essentially mirrors that obtained with adjectives. F.S. repeated correctly 300/323 (92.9%) of the nouns when these were given as stimuli in the singular form and only 56/86 (65.1%) when given in the plural form. This result is consistent with the claim that the citation form of a word is relatively spared in comparison to other inflected forms. The analysis of F.S.‘s verb repetition performance provides a further opportunity for evaluating the hypothesis that his deficit is functionally localized in the morphological processing component of the lexicon and, more specifically, in the inflectional processing component. Over a number of sessions, F.S. was asked to repeat 757 verbs. He repeated correctly 268 (35.7%) of these stimuli. Of the 482 errors, the great majority 345 (71.6%) were inflectional errors. The remaining errors were divided among phonemic substitutions, phonemic substitutions + substitution of the inflection, morphologically illegal substitutions of the inflection, unrelated word responses, fragments, and omissions. Only 4 unambiguously derivational errors were produced by F.S. in response to a verb stimulus (0.8% of total errors). A breakdown of verb stimuli by mood, aspect, and tense reveals that some forms were more difficult to repeat than others (see Table 9). Thus, for example, the imperfect form was more difficult to repeat than the present form in the indicative mood or the past participle form. However, as can be seen in Table 9, inflectional errors always (except for the future and conditional forms) constituted the most frequent type of error. A stimulus-response matrix for those verb forms with a reasonably large representation in the stimulus set is shown in Table 10. This table shows the probability of producing a response (whether or not correct) of the same form as the stimulus (i.e., given a verb in the present form (e.g., “Temo” (I fear)) the response is in the present form (e.g., “temo,” correct, or “temi” (you fear), “teme” (he fears), “temiamo” (we fear), “temete” (you fear; pl.) or “temono” (they fear)). The present and past participle are the best preserved while the imperfect and (simple) past tense are the most impaired verb forms. It is worth noting that there is no indication in the data for preservation of the mood feature of a verb; for example, a verb in the indicative voice, say the imperfect, is not more likely to be repeated as another indicative form, say the (simple) past tense, than a verb form in another mood, say in the infinitival form. It is also worth stressing that unlike nouns and adjectives where the inflectional ending is just a single phoneme (e.g., mel-a (apple, f.sg.)), (n = 211) (n = 153) (II = 64) (?I = 31) (n = 459) (n = 29) (n = 6) (n = 2) (n = 125) (n = 12) (n = 83) (n = 35) Participle Present Past Gerund Infinitive Ambiguous 52 (41.6) 9 (75.0) 40 (48.2) 9 (25.7) (0) 101 (47.9) 34 (22.2) I2 (18.7) 9 (29.0) I56 (34.0) 2 (6.9) (0) Correct OF THE ERRORS 9 2(100.0) 54 (43.2) (0) 29 (34.9) 23 (65.7) 72 (34.1) 100 (65.4) 41 (64.1) 3 (9.7) 216 (47.0) 21 (72.4) 2 (33.3) Inflectional VERB FORMS 6 (4.8) (0) 1 (1.2) 2 (5.7) (0) 10 (4.7) 6 (3.9) 3 (4.7) 2 (6.4) 21 (4.6) 1 (3.4) 1 (16.7) Inflectional pW BY F.S. IN REPEATING TABLE MADE ” 2(100.0) 58 (48.0) (0) 30 (36.1) 25 (71.4) 82 (38.9) 106 (69.3) 44 (68.6) 5 (16.1) 237 (51.6) 22 (75.8) 3 (50.0) Total inflectional 5N 9 2 $ z 3 (50.0) (0) 13 (10.4) 3 (25.0) 13 (15.7) 1 (2.9) ,% z g 28 (13.2) 13 (8.5) 8 (12.5) I7 (54.9) 66 (14.4) 5 (17.3) Others“ * Percentages are in parentheses. ’ Incorrec;responses that contain an inflection substitution error, resulting in a nonpermissible combination of root + inflection (e.g., resistere (to resist) + resistivo (nonword)), and incorrect responses that contain a substituted inflection + a substituted root (e.g., testare (to test) + distate (to be distant, ind. pres. 2nd pl., or past participle, f.pl.)), or a substituted inflection + phonologically related pseudoroot (e.g., fiatavi (you were breathing) -+ Mare (nonword)). ’ Verb forms that cannot be unambiguously identified (e.g., tit-ate (to pull), ind. pres. 2nd pl., or past part. f.pl.; esclusi (to exclude), ind. past. 1st sg., or part. past., m.pl. d Includes all the remaining incorrect (word and nonword) response types listed in Table 3. Present Imperfect Past tense Future Total Subjunctive Indicative DISTRIBUTION 1 (0.6) 50. (37.3) 2 (3.8) 6 (8.7) 634 (11.8) (3.9) 152 (87.9) 11 (8.2) 5 (9.4) 6 (8.7) 1795 (33.6) (4.8) Indicative imperfect MATRIX MADE 20 (37.8) (0) 251 (4.7) (1.0) 1 (0.7) 3 (1.7) Indicative past tense OF THE ERRORS IN REPEATING 10 8 (15.1) 5% 101 (18.9) 37 (27.6) 6 (3.4) Infinitive BY F.S. TABLE 9 (17.0) 12 (17.4) 127 77 (23.8) (74.0) 20 (14.9) 9 (5.2) Past participle VERB FORMS VERB 5 (9.4) 5 (7.2) 317 (5.8) (6.7) 13 (9.7) 1 (0.6) Ambiguous” OF SOME 4 (7.5) (0) 7 (1.3) (0) 2 (1.5) 1 (0.6) Other verb forms TENSES” 53 (loo) 69 (100) 533 104 (100) (loo) 134 (100) 173 (100) Total * Percentages are in parentheses. b This category includes incorrect verb responses that cannot be unambiguously assigned to one of the other response categories (e.g., lavavi (YOU were washing) + lavate (you wash, indicative present, 2nd plural, or washed, past participle, f.pl.)). Indicative present Indicative imperfect Indicative past tense Infinitive Past Total participle Indicative present STIMULUS-RESPONSE 4 h E 2 s 2 $ r 8 z 3 2 u 2 r P 5 8 7 48 MICELI AND CARAMAZZA verbal inflections can consist of a single or a series of phonemes (e.g., am-a (he loves; present), am-are (to love; infinitival), am-uto (loved; past participle), am-ussero (that they loved; subjunctive), am-erebbero (they would love; conditional), and so forth). Thus, inflectional errors for verb forms with the longer inflections will necessarily involve multiplephoneme changes. As can be seen from Table 10, F.S. produced very many such multiple-phoneme substitutions. This result is important, because it conclusively demonstrates that inflectional substitutions cannot merely be attributed to chance phoneme-substitution errors (as opposed to inflectional errors). Thus, consider as an example the following error: spingu (that he pulls) + spingono (they pull). Here the change involves substituting one phoneme and adding two more. Phonological errors involving multiple phoneme substitutions, additions, or deletions which resulted in the production of nonwords (e.g., tern-o (I fear) -+ tern-elo or tern-uro, where elo and uro are not legal afhxes) as opposed to inflectional errors were not found in the corpus of errors. b. Nonword Repetition In order to assess nonlexical processing mechanisms in repetition, we asked F.S. to repeat 283 nonwords, randomly interspersed with words. F.S. performed very poorly in this task. He repeated correctly only 28/283 (9.9%) nonwords. There was no clear effect of stimulus length on performance: ll/lOl (10.9%) correct versus 17/182 (9.3%) correct, for short (four to six phonemes) and long (seven to nine phonemes) nonwords, respectively. The absence of a length effect should not be given undue weight, however, since performance may have been too poor (floor effect) for drawing strong implications from these data. F.S.‘s performance in nonword repetition clearly reflects major damage to phonological processes, either at the input or output level, but most likely at both levels. Interim Summary The results we have reported to this point paint a coherent picture. F.S. presents with morphological processing difficulties both in spontaneous speech and single-word repetition. These difficulties are certainly compounded by a moderate deficit in phonological processing (as is indicated by nonword repetition performance) but are not reducible to a phonological processing disorder. The strongest evidence against such a possibility is the patterned nature of the inflectional errors produced by F.S. in the repetition task. The pattern of inflectional errors is explicable by appeal to morphological and not phonological principles of the language. On these grounds it is reasonable to conclude that the morphological errors produced by F.S. in spontaneous speech and in repetition are properly characterized as resulting from damage to a morphological processing INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 49 mechanism. Furthermore, since F.S. produced inflectional errors not only in spontaneous speech but also in single-word repetition, we are justified in entertaining the parsimonious hypothesis that damage to the inflectional component of the lexical system is the common source of damage for both forms of impairment. The preliminary conclusion we have reached here identifies as the most probable locus of functional damage the inflectional component of the output lexicon. However, as already noted, F.S. also presents with phonological processing difficulties. Some of these difficulties clearly involve speech production mechanisms as indicated by his performance in spontaneous speech. We have argued that such a deficit cannot be the basis for the morphological errors in spontaneous speech or repetition. By the same token neither could a deficit in processing the auditory input serve as the basis for the morphological errors in repetition. But, rather than relying just on argument to draw this latter conclusion, we assessed F.S.‘s ability to process auditory inputs. Auditory Processing F.S.‘s auditory processing ability was first evaluated through a lexical decision task. The lexical decision test consisted of 560 items, equally divided between words and nonwords. Half of the words (N = 140) were verbs, the remaining 140 words belonged to the other grammatical classes (nouns, N = 78; adjectives, N = 33; functors, N = 29). Two sets of nonwords, exactly matched in length to the word stimuli were used for this task. The first set consisted of 140 stimuli. Half of the nonwords in this set (N = 70) were “morphologically legal,” that is, they could be morphologically decomposed into a real root and a real inflection; the remaining 70 nonword stimuli in this set were exactly matched to the other nonword stimuli in length and in degree of similarity to real words, but could not be morphologically parsed into a verb root and a verb inflection. Nonwords in the second set (N = 140) were matched in length to the nonwords in the first set, and were all of the “morphologically nondecomposable” type. An example of the “morphologically decomposable” nonword stimuli is veneva, resulting from the root vey1- (“venire,” to come), with the nonpermissible inflection -eva (legal for verbs like volere (to want)-voleva, he was wanting). The corresponding nondecomposable stimulus is voredu, which contains neither a verb root nor a verbal inflection. F.S. performed relatively well on this task: He produced 1 l/280 (3.9%) false rejections on words, and 7/280 (2.5%) false alarms on nonwords. F.S. incorrectly rejected as nonwords l/78 (1.3%) nouns, l/33 (3.0%) adjectives, O/29 functors, and 9/140 (6.4%) verbs, and, he incorrectly accepted as words 2/70 (2.9%) decomposable nonwords, O/70 matched nondecomposable nonwords, and 5/140 (3.6%) “other” nonwords. 50 MICELI AND CARAMAZZA F.S.‘s performance on this test demonstrates that he can reliably (though not flawlessly) distinguish words from nonwords. However, his good performance in a lexical decision test does not say much about the nature of the information that is available to him upon the presentation of a word stimulus. For example, even though he might be able to say that chiumavo (I was calling) is a word, he may do so on the basis of incomplete information about the input string (i.e., based on the recognition of the root morpheme and the presence of an unspecified, but permissible suffix). In order to obtain more information on this issue, an auditory, same-different judgment task was administered. Same-Different Auditory Judgment Task F.S. was presented with 428 pairs of verbs: Half of the pairs consisted of the same stimulus; the remaining pairs consisted of different verb forms. Two lists of “different” pairs were constructed. In one list (N = 107), verbs differed by one phoneme in the root morpheme part of the word; in the other list (N = 107) they differed by one phoneme in the suffix part. Different verb pairs were matched across lists, so as to include the same phoneme contrasts. So, for example, the first list included items like spat-a (he shoots)-spira (he dies), or like compativo (I was pitying)-comparivo (I was appearing). The corresponding stimuli in the second list would be speri (you hope)-spera (he hopes), or spezzare (to break)-spezzate (broken; f.sg.). One of the two verbs used in each different pair was used to construct “same” pairs. The stimuli included in the two sets of “semi” and “different” pairs were matched for length and frequency. The 428 pairs were presented in random order. The examiner read aloud the first verb. The patient was then invited to count aloud backward from 4 to 0 (which usually took F.S. more than 5 set). Subsequently, the examiner pronounced the second stimulus and the patient was invited to say whether the two stimuli he had heard were “same” or “different.” F.S. made 12/214 (5.6%) errors on “same” pairs and 14/214 (6.5%) errors on “different” pairs. He made the same number of errors on pairs that differed in the root morpheme and on pairs that differed in the suffix (7/107, 6.5%). Although not perfect, F.S.‘s performance on this test is quite good. Some of the verbs used in the “same-different” auditory judgment task were included in a repetition test (N = 200). Compared to his good performance in the former task, the performance obtained by F.S. in repetition was extremely poor: He repeated correctly only 74 (37.0%) verb stimuli. Out of 126 errors, he produced, as usual, a very high number of substitutions of inflections (94/126; 74.6%). The remaining errors were phonologically similar nonword responses (21/126; 16.7%), or substitutions INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 51 of a phonologically related word for the correct target (e.g., rimango (I remain) + rimundo (I postpone)) (9/126; 7.0%). F.S. also produced 1 semantic substitution (giurero (I will swear) + giudice (judge)), and failed to produce any response in one case. Taken together, the results obtained by F.S. in the auditory lexical decision task and in the auditory “same-different” judgment task (especially when considered in light of performance on the repetition task in which some of the stimuli used in the “same-different” judgment task were included) rule out the possibility that the patient’s repetition disorder is merely the consequence of a reduced ability to process morphologically complex input strings under auditory presentation. DISCUSSION Although the pattern of language dysfunction in our patient is complex and, therefore, unlikely to be the result of damage to a single processing mechanism, there are aspects of his performance which can sustain conclusions specifically about the structure of the lexical system. Our discussion will focus on this latter issue. But, first, a brief summary of the major features of F.S.‘s language and cognitive performance that are relevant to questions about lexical processing. There are four major features of F.S.‘s performance that are relevant to our discussion: First, there is the fact that F.S.‘s spontaneous speech is characterized by morphological errors which involve the substitution of inflectional affixes and the substitution or omission of free-standing grammatical markers; second, the patient’s spontaneous speech and repetition of sentences and single words contain phonological errors (phonemic paraphasias); third, his markedly impaired single-word repetition performance results most frequently in inflectional errors; and, fourth, even though F.S.‘s auditory/perceptual processing ability is not intact, the nature of this particular impairment does not appear to be quantitatively nor qualitatively of the sort that could account for the qualitative and quantitative features of his repetition disorder. The overall pattern of results we have described suggests as the functional locus of damage the morphological processing component of the lexicon. We elaborate on this conclusion below. There are three issues we wish to take up here: We will argue that damage to nonlexical phonological processes cannot account for the pattern of morphological errors produced by our patient; we will consider the possibility that the morphological errors in spontaneous speech and in single-word repetition have a common source; and, we will discuss the pattern of repetition errors in relation to claims about the organization of the lexicon and, in particular, the morphological processing component. 52 MICELI AND CARAMAZZA Phonological Dejicit and Morphological Errors There is no doubt that F.S. has damage at some level of phonological processing. His almost total inability to repeat nonwords indicates a deficit in processing segmental phonology. Furthermore, a nonnegligible proportion of repetition errors for word stimuli can confidently be attributed to a phonological impairment since these errors involved phonemic transformations of a to-be-repeated word (e.g., “pagata” (paid, past part., f.sg.) --f “pagara” (nonword); “avverte” (he warns) + “averre” (nonword)). A similar conclusion can be reached for some of the errors in spontaneous production (e.g., “divertimento” (amusement) * “diritimento” (nonword)). An important characteristic of these errors is that the locus of phonological transformation was as likely to be the root as the affixal part of a word, as might be expected if the determining factor for these errors were to be nonmorphologically based. One issue that cannot be resolved with the available data on this patient is the extent to which an auditory/perceptual processing deficit contributes to phonological errors. Although it is entirely possible that a proportion of the phonological errors in repetition reflect misperception of the stimuli, it must, nonetheless, be the case that the patient has a phonological disorder at the production level since phonological errors also occurred in spontaneous speech. Thus, the most conservative conclusion we can reach is that F.S. may have a processing deficit for segmental phonology both at the perceptual and production levels. The critical issue here, however, is whether such a deficit could account for the putative morphological errors produced by our patient. Under the Results we argued against this possibility. There are several arguments that can be offered for this position. First, the nature of the perceptual deficit in our patient is not commensurate with the extent and type of morphological errors produced by F.S. in the repetition task. And, in any case such a deficit is irrelevant to the production of morphological errors in spontaneous speech. Second, if we were to adopt the position that it is a phonological production deficit that is responsible for the putative morphological errors, we would be hard pressed to explain the patterned nature of the inflectional errors obtained with nouns and adjectives. Recall that the inflectional errors produced by F.S. were not distributed randomly among the various inflectional endings but appear to follow a specific pattern-the production of the singular form of nouns and the singular masculine form of adjectives (the citation forms for these two word-classes, respectively) were much better preserved than other inflectional forms. And, finally, the great majority of morphological errors were inflectional, even when a stimulus was unambiguously a derived word. It is unclear how a nonlexical deficit hypothesis could account for this pattern of morphological errors. We must conclude, INFLECTIONAL therefore, that the putative phological errors. AND DERIVATIONAL morphological MORPHOLOGY 53 errors are, afterall, really mor- On the Possible Relationship between Morphological Spontaneous Speech and Single-Word Repetition Errors in The second issue we would briefly like to consider is the possibility that the morphological errors in F.S.‘s spontaneous production and in single-word repetition have a common basis. This issue cannot be resolved unambiguously given the available data, but its consideration is instructive. There are at least three possible accounts we can give for the cooccurrence of morphological errors in spontaneous production and single-word repetition. We briefly discuss these below. One possibility is that these two types of errors arise from damage to distinct functional mechanisms: Damage to a syntactic processing mechanism accounts for the impairment in spontaneous production and damage to the morphological component of the lexicon accounts for the singleword repetition impairment. If we were to adopt this position, we would have to hypothesize a fairly specific form of damage to the syntactic component of the language production system (the S-structure level in current syntactic (transformational) theory (e.g., Chomsky, 1981), or the Positional level in Garrett’s (1980) model of sentence production). That is, the hypothesized deficit would have to be restricted to just those morphosyntactic features of S-structure that govern selection of inflectional affixes, determiners, prepositions, and auxiliaries which appear to be most severely impaired in our patient. However, subordinating conjunctions, which presumably are also selected at the S-structure level, are relatively unaffected. Indeed, the syntactic complexity of our patient’s speech as indexed by the ratio of main to subordinate clauses does not differ from that of a normal control subject (see Table 2). Thus, F.S.‘s speech contained constructions such as the following: la faccio Poi io ascolto il televisione 0 poi fare il pranzo Then I listen the (m.sg.) TV (f.sg.) or then to make (infin.) the lunch Then I watch TV or then make lunch vivo perche’, care dottore, io vive solo! because, dear doctor, I lives alone! because, dear doctor, I lives alone! C’era un process0 the c’era quest0 professore There was a trial that there was this professor There was a trial where this professor was 54 MICELI AND CARAMAZZA interessava perche’ [era] una perizia per un omicidio the interessa me. because [it was] a consult for a murder that interests (pres., sg.) me because a consult for a murder that interests me. Although this Dual Deficit Hypothesis is logically possible, it is unparsimonious and fails to capture an important similarity in the types of morphological errors produced by the patient in spontaneous sentence production and single-word repetition. Another possibility is that damage to only the morphological component of the lexicon is responsible both for the morphological errors in spontaneous sentence production and for the morphological errors in singleword repetition. This hypothesis can readily account for the presence of inflectional errors in both types of tasks. However, we are forced to make a strong assumption about the kind of information represented in the morphological component of the lexicon in order to account for the substitution and omission of free-standing grammatical markers. Since, F.S. omitted or substituted these latter items in spontaneous production, a unitary hypothesis of the locus of functional damage for spontaneous production and single-word repetition would have to assume that freestanding grammatical markers and inflectional affixes are represented in a common subcomponent of the lexicon as proposed by Caramazza et al. (1985). However, whether this assumption can be given independent linguistic motivation is unclear (but see Kean, 1977).4 The third possibility is that the morphological errors in spontaneous sentence production result from damage to both a syntactic processing mechanism and the morphological component of the lexicon, while the morphological errors in single-word repetition result from damage to the lexicon alone. This hypothesis could easily be the correct one but it is also the least interesting because it is the most difficult to disconfirm. Unless there is a principled way for distinguishing the relative contribution, both quantitatively and qualitatively, of damage to the syntactic and morphological processing components, this complex hypothesis about the underlying cause of the patient’s impairment in spontaneous speech is not especially illuminating. Despite our inability to decide which of the three hypotheses of the locus (loci) of functional damage best accounts for our patient’s morphological errors in spontaneous sentence production and single-word repetition, we can, at least, draw a firm conclusion about one issuewe must postulate a deficit to the lexicon in order to account for the morphological errors produced by the patient in the single-word repetition tasks. The difficulty we encountered in our effort to provide a general account 4 Alternatively, we could assume that function words and inflectional aflixes form distinct subcomponents of the lexicon and that both are damaged in our patient. INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 55 of all the morphological errors (i.e., both in spontaneous production and repetition) produced by our patient illustrates the futility of certain efforts to provide a motivated explanation of clinically defined disorders such as agrammatism, paragrarnmatism, deep dyslexia, and the like. The correct explanation of the basis for a pattern of language dysfunction depends crucially on a reasonable characterization of the structure of the impaired performance (see Badecker & Caramazza, 1985; Caramazza, 1986, for discussion). This task is hard enough for extensively studied, individual patients (such as the present case, for example); it requires, at the very least, a proper characterization of a patient’s relevant performance. When one seeks to provide an explanation of ill-defined categories such as “agrammatism,” it is not clear whether such an effort is meaningful: Categories of this latter sort reflect groupings based on either theoretically arbitrary features of language or empirically indefensible abstractions; in either case, the sought-for explanation of the clinically defined disorder contributes neither to the development of a theory of normal language processing nor to an understanding of the nature of language disorders. The Dissociation of Inflectional and Derivational Morphology F.S.‘s performance in the single-word repetition task is unambiguous in at least two respects: First, the great majority of errors were morphologically based and, second, there is a striking dissociation between inflectional and derivational morphology in repetition errors. Of the total corpus of repetition errors (N = 893) for which a morphological error was possible, the great majority (637; 71.2%) were morphologically based. Crucially, however, these errors were almost all (96.7%) inflectional. It is our contention that these results support a model of the lexicon in which lexical items are represented in morphologically decomposed form. However, before discussing this possibility in any detail we wish to discuss several alternative accounts, of varying degrees of plausibility, which may be offered as explanations for the results we have reported. One rather implausible account, as a full explanation of the reported results, is based on the assumption of a phonological processing deficitboth at the input and output levels. On this account, the patient misperceives the stimulus and produces a phonologically similar response. This explanation will simply not do on several counts. First, it is completely silent on the presence of morphological errors in spontaneous production where an input phonological disorder is irrelevant. Second, it fails to account for the distribution of errors-f the total corpus of repetition errors that resulted in word responses (N = 659), the vast majority were morphological errors (N = 636; 96.5%). It is difficult to imagine how a strictly perceptual/phonological deficit (or, output phonological deficit, for that matter) could explain this distribution of errors. And, third, a direct test of the patient’s ability to perceptually discriminate phonologically similar word stimuli failed to reveal a significant impairment in input 56 MICELI AND CARAMAZZA phonological processing. This is not to say, of course, that F.S. does not have a phonological processing impairment-vi& his phonemic paraphasic errors in repetition. However, this deficit fails to account for the patterned nature of the repetition errors where inflectional errors were the most conspicuous. And, within this class of errors there was further structure explicable along morphological and not phonological principles-e.g., the distribution of errors for adjectives. In sum, although F.S. most likely has a deficit in phonological processing, this deficit is in addition to some other deficit that is responsible for the morphological errors (most of them, at least) in repetition and spontaneous production. A second explanation, not based on the assumption of morphological decomposition, which may be offered as an account of the reported pattern of results assumes an impairment in lexical retrieval from the phonological output lexicon. On this account, the patient computes the correct meaning of the stimulus word but this semantic representation fails to activate the correct phonological representation in the output lexicon and instead activates a semantically related word. This unadorned explanation is completely inadequate on the face of it-it fails to account for the distribution of inflectional versus derivational errors and, within the class of inflectional errors, it fails to explain the preference for the citation form of nouns and adjectives. Furthermore, this hypothesis of the nature of the impairment in F.S. would predict the massive presence of semantic paraphasic errors, errors that were virtually nonexistent in the corpus collected. A more plausible version of this h,>othesis may be offered. This hypothesis--the Satellite Hypothesis-will have to assume a richer structure for the phonological output lexicon, although still not assuming morphological decomposition. This hypothesis is based on a related proposal by Lukatela, Gligorijevic, Kostic, and Turvey, 1980) for word recognition. The basic proposal is that lexical entries are organized about a central lexical form (the nominative case in Serbo-Croatian). It is not clear what would be the corresponding organization in Italian where nouns are not case-inflected. Nonetheless, we could make one of two assumptions: either that all morphologically (derivationally and inflectionally) related words cluster about a major lexical form (Satellite Hypothesis I) or, alternatively, that there is a family of clusters where each cluster consists of inflectionally related forms (Satellite Hypothesis II). In the former case we would cluster together nouns, adjectives, verbs, and adverbs with the same root form (e.g., blame: coZpa (N); colpevole (Adj.); colpevolizzare (V); colpevolizzazione (N); colpevolmente (Adv.); etc.; and related inflected forms); in the latter case, each syntactic category of a lexical family would be represented independently in a subcluster (e.g., blame: colpa (N.sing.) and cofpe (N.pl.) would be represented separately INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 57 colpevolizzo dlscolpare fncol~are\ I incolpo ) / dl:colpo , colpe;olme"te colpevollzzazfone colpevollzzare \ colpevolisti / i\ colpevollsta colpevolezza FIG. 3. Schematic representation of Satellite Hypothesis I. from colpevole (Adj.sing.) and colpevoli (Adj.pl.)). Figures 3 and 4 graphically represent these alternative Satellite Hypotheses. It is immediately apparent that Satellite Hypothesis I cannot provide a motivated account for the results we have reported. In this hypothesis no distinction is made among various inflectional and derivational forms of a word and, therefore, errors should not reflect morphological structure, contrary to the results we have obtained. Satellite Hypothesis II is superhcially more interesting. This hypothesis colpevollrro colpevollsta M b colpevole FIG. 4. Schematic representation of Satellite Hypothesis II. (Verb paradigms are abbreviated by reporting only the Indicative Present 1st person and the infinitive, e.g., incolpo, incolpare.) 58 MICELI AND CARAMAZZA incorporates in the organization of the lexicon the basic distinction between derivational and inflectional morphology: Each subcluster represents inflectional variants of the base and derived forms of a word. To retrieve a specific lexical form (e.g., colpevoli) one must enter a lexical cluster through the designated entry point-colpa in the example considered here. Even a brief consideration of how this retrieval process might work reveals immediately that this Satellite Hypothesis is forced to make assumptions as strong as those made by the Morphological Decomposition Hypothesis without any of the advantages of this latter hypothesis. Lexical retrieval within the Satellite Hypothesis II requires that the semantic representation of a word specify three independent sets of semantic values: one for the root morpheme which specifies the entry point for a lexical cluster (ROOT SEMANTICS); one for the derivational affix which specifies the appropriate subcluster within a lexical cluster (DERIVATIONAL AFFIX SEMANTICS); and, one for the inflectional affix which specifies the particular lexical form to be retrieved (INFLECTIONAL AFFIX SEMANTICS). The semantics of a word would, then, have the following structure: [ROOT SEM.; (DER. AFFIX SEM.) INFL. AFFIX SEM.], where the material in parentheses is assumed to be an optional component. Thus, for example, the word colpevoli should have the semantic representation [(COLPA) (ADJ.) (PL.)], where COLPA is the semantic component for the root morpheme (COLP-), ADJ. is the semantic component for the selection of an adjectival affix (EVOL-), and PL. is the semantic component for the selection of the appropriate inflectional affix plural (I). Each component part of the semantic representation serves a specific role in lexical retrieval. The whole representation is needed to select a particular phonological form. The Satellite Hypothesis II is consistent with the pattern of morphological processing impairment in our patient. To account for the reported results all we need assume is that F.S. has a selective deficit either at the level of the semantic units that specify inflectional forms in a subcluster or at the level of the inflectional forms themselves. Despite the success of this hypothesis in accounting for the reported results we contend that it should not be accorded the status of a serious alternative to the Morphological Decomposition Hypothesis. The reason for taking this position is quite compelling. Note that the Satellite Hypothesis II, in order to provide a principled means for lexical access, de facto includes morphological decomposition at the semantic level. However, it does not include provisions for morphological productivity. This hypothesis only allows access to stored phonological representations in the phonological output lexiconit cannot produce novel, morphologically legal forms. This limitation decisively rules out Satellite Hypothesis II as a viable alternative to the Morphological Decomposition Hypothesis, (For related arguments against a morphologically nonproductive lexicon see Stemberger, 1985.) INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 59 In other papers (Caramazza, Laudanna, & Romani, 1988; Caramazza et al., 1985) we have argued for a morphologically decomposed structure of the lexicon on the basis of experimental results we have obtained with normal subjects, the reading performance of a dyslexic patient, and various other published results with normal subjects (e.g., Burani, Salmaso, & Caramazza, 1984; MacKay , 1979; Taft, 1984) and brain-damaged patients (e.g., Job & Sat-tori, 1984; Patterson, 1980; but see Badecker & Caramazza, 1987). F.S.‘s highly selective disorder of morphological processing reported here provides one of the clearest experimental results yet in favor of the Morphological Decomposition Hypothesis for the organization of the lexicon. An even more important implication of the reported results concerns whether inflectional and derivational morphology constitute independent subcomponents of the morphological processing system. The answer from F.S.‘s performance would appear to be clearly affirmative-F.S. is severely impaired in processing inflectional morphology and only mildly (if at all) impaired in processing derivational morphology. Furthermore, the patient almost never made derivational errors on words that did not contain a derivational aflix, despite the fact that he made many inflectional and other errors on these words. That is, there is literally a categorical dissociation of error types: F.S. made inflectional errors on inflected words (both derived and nonderived) and occasionally derivational errors on derived words. This categorical dissociation could only occur if the two classes of affixes, inflectional and derivational, were functionally distinct. The basis for this conclusion is as follows: Let us suppose that inflectional and derivational affixes are not organized into functionally distinct subcomponents of the lexicon but, instead, constitute a single morphological component that stores together both types of affixes. If this component were to be damaged, the retrieval of morphological affixes, both inflectional and derivational, would be affected. In such a case there is no reason to expect different levels of performance for inflectional and derivational affixes. However, we found a large discrepancy in performance for inflectional and derivational affixes. Furthermore, since, on this account, the two types of affixes are “stored” in the same damaged component of the lexicon we would expect affix misselections not to honor the distinction between inflectional and derivational affixes. In other words, if affix x were, for whatever reason, not accessible, then some other affix, y, might be selected, independently of the morphological status of x and y. Thus, if F.S. were to be asked to repeat the adjective “bella” (beautiful, f.sg.) we would expect him to produce errors not only like “bello” (beautiful, m.sg.; an inflectional error) but also like “bellezza” (beauty; a derivational error). However, cross-category errors almost never occurred-affix misselections respected morphological distinctions. This means that the mechanism for selecting atIix type, inflectional 60 MICELI AND CARAMAZZA or derivational, must be intact in our patient and that damage is restricted to those processes that select specific affixes wirhin the inflectional subcomponent of the lexicon. The view of lexical organization that emerges from these considerations is the following: A lexical semantic representation which may take the form already discussed for the Satellite Hypothesis serves as input to the phonological output lexicon. The semantic representation is articulated into distinct parts, each part specifying different aspects of the lexical form. That is, the semantic representation includes what we have called root semantic features, derivational semantic features (where present), and inflectional semantic features. The root plus derivation semantic features jointly specify a stem plus default inflectional affix in the phonological lexicon. The default inflectional affix is the citation form of a word which we take to be the singular form for nouns, the masculine singular form for adjectives, and the infinitive form for verbs (and possibly the past participle form for verbs since this lexical form behaves differently from other verbal inflections). The critical assumption we have made here is that although the lexicon is derivationally productive, those derived forms that have been experienced are stored in the phonological lexicon and are accessed by the root and derived semantic features directly. Novel derived forms must be computed by the application of the Derivational Processes Component. The other important assumption we have made is that the phonological form of a lexical item stored in the lexicon is the citation form. This form is transformed by the application of appropriate inflectional processes driven by the inflectional semantic features.’ It is important to note that we assume that the phonological lexicon may store well-learned inflected forms, and certainly irregularly inflected forms. In conclusion, we have provided the strongest yet experimental evidence for a dissociation of inflectional and derivational processes in the lexicon. These results are consistent with the model of morphological processing discussed under the Introduction. Specifically, our results are consistent with the view that both derivational and inflectional morphology are located in the lexicon-the Strong Lexicalist Hypothesis (e.g., Lapointe, 1979). However, our results also suggest that inflectional and derivational affixation constitute different processes (e.g., Anderson, 1982; Aronoff, 1976). The view of lexical organization that emerges from our results is well captured by the model presented in Fig. 1, where we distinguish among three subcomponents of the lexicon: the Root Morpheme Com5 It should be noted that, although we have emphasized how inflectional affixes are activated from the semantic component of the lexicon, these phonological forms may also be accessed directly from information specified in a syntactic frame in sentence production. That is, we assume that inflectional aIIixes may be accessed either through semantic or through syntactically specified information depending on whether single-word or sentence production is being considered. INFLECTIONAL ponent, the Derivational cesses Component. AND DERIVATIONAL Processes Component, 61 MORPHOLOGY and the Inflectional Pro- APPENDIX I: LEGEND FOR TABLES InJlectional error: incorrect word response, consisting of the production of the correct stem + incorrect inflection. Examples: finestre (windows, f.pl.) grossa (big, f.sg.) leggeva (he was reading) -+ finestra (window, f.sg.) + gross0 (big, m.sg.) + leggere (to read) Derivational error: incorrect word response, consisting of the production of the correct stem + incorrect derivational suffix, substituting for the correct inflection. Examples: pittura (painting, noun) illusorie (illusory, f.pl.) passando (passing, gerund) -+ pittore (painter) + illusione (illusion) + passaggio W.sage) Prefix omission: incorrect word response, consisting of the production stem and of the correct suffix, with omission of the prefix. of the correct Example: scambio (exchange) + cambio (change) Semantic substitution: incorrect word response, consisting of the production of an item semantically (but not phonologically) related to the target. Examples: Other word geniale (brilliant) giurero’ (I will swear) response: + un tipo originale (a character) -+ giudice (judge, noun) production of a word phonologically (but not semantically) related to the stimulus. Examples: rimango (I remain) vieta (he prohibits) + rimando (I postpone) + dieta (diet, noun) Illegal injection: incorrect nonword response, consisting of the production of the correct stem + incorrect inflection, present in the language, but not permissible for that particular stem. Examples: Stimulus Error Permissible, incorrect Correct use of inflection aspettare (you are waiting) aspettiv2 aspettava (he was waiting) riempiro (filled, m.sg.) riempiereh riempire (to fill) sentiva (he was hearing) cogliere (20 pick) Illegal derivation: incorrect nonword response, consisting of the production of the correct stem + substituted derivation, present in the language but not permissible for that stem. Stimulus autunnale (autumnal) + prontezza (readiness) -+ Error autunnaio prontenza Correct instance of deriv. fioraio (florist) potenza (power) 62 MICELI AND CARAMAZZA Phonological error: incorrect nonword response, phonologically related to the stimulus. It can contain the target stem only, the target inflection only, or neither (see Table footnotes). Examples: Phonological c&don0 (they fall) misi (I put, past) sopravvivere (to survive) error of the production inflection. Examples: on root + substituted of a pseudoroot, inflection: phonologically + c&dote --f mensi + sopravisimo incorrect nonword response, consisting related to the target + an incorrect fiatavi (you were breathing) contato (counted, m.sg.) + Mare + coltare APPENDIX II We briefly present some general principles of Italian inflectional morphology. Nouns usually carry number inflection; gender is lexically determined. The most frequent ending of the masculine singular (m.sg.) is -0, and the corresponding plural form (m.pl.) is -i (e.g., cappotto (coat), cappotti (coats)); the most frequent form of the feminine singular (f.sg.) is -a, and the corresponding plural form (f.sg.) is -e (e.g., pianta (plant), piante (plants)). Since noun gender is lexically determined, however, there are many exceptions to this general case, and the morphology of a noun is not an indicator of its gender (e.g., elefante (elephant, m.sg.), elefanti (elephants, m.pl.); poeta (poet, m.sg.), poeti (poets, m.pl.); artista (artist, m.sg. and f.sg.), artisti (artists, m.pl.), artiste (artists, f.pl.); pelle (skin, f.sg.); pelli (skins, f.pl.), etc.). Some nouns are invariable (e.g., citta’ (city, f.sg., and cities, f.pl.). Adjectives can be broadly divided in two categories. Some carry four different endings, corresponding to m.sg. (-o), m.pl. (-~3, f.sg. (-a), f.pl. (-e )-see for example buono, buoni, buona, buone (good). Other adjectives only carry two inflections, that mark number (sg., -e; pl., -i) and are ambiguous with respect to gender-see for example, forte (msg. and f.sg.), forti (m.pl. and f.pl.) (strong). In connected speech, adjectives always agree with the name that they modify, in gender and number. The gender of the noun also determines the choice of the article (e.g., in the most typical cases, il cappotto nero (the black coat, m.sg.), i cappotti neri (the black coats, m.pl.), la pianta sana (the healthy plant, f.sg.), le piante sane (the healthy plants, f.pI.). Verbs have an inflectional paradigm that is much more complicated than the paradigm of either nouns or adjectives. Italian verbs belong to one of three conjugations. For each verb in each conjugation, Italian has four finite modes (indicative, subjunctive, conditional, imperative) and three nonfinite modes (gerund, participle, and infinitive). Some of the verb forms are generated by adding an inflection to the verb root; others are generated starting from a au + V root + past participle structure, where aax can be either “essere” (to be) or “avere” (to have). For forms of the first type (V root + inflection), within the finite modes, indicative has four tenses (present, imperfect, past and future), subjunctive has two tenses (present and imperfect), conditional and imperative each have one tense (present); each finite mode has six voices, corresponding to the lst, 2nd, and 3rd singular, and to the lst, 2nd, and 3rd plural. Within the nonfinite modes, the gerund and the infinitive have one tense (present), and the participle has two (present and past). These forms are invariable, except for the past participle, which under certain conditions behaves like an adjective with four endings (e.g., “I bambini sono baciati dalla mamma” (The boys are kissed (m.pl.) by their mother; “La mamma li ha baciati” (The mother them has kissed (m.pl.)). The forms of the second type (au + V root + past participle inflection) are much less INFLECTIONAL AND DERIVATIONAL 63 MORPHOLOGY frequent in the language, but are just as numerous as the forms of the type V roof + inflection (four indicative tenses, two subjunctive and participle tenses, one conditional, imperative, gerund, and infinitive tense). The paradigm of the V root + i&ction forms of the verb “amare” (to love), a regular verb of the 1st conjugation, is reported below. INDICATIVE Present Imperfect am-o (I love) -i -a -iamo -ate -ano am-avo (I was loving) -avi -ava -avamo -avate -avano Past Future am-ai (I loved) -asti am-ero’ (I will love) -erai -era’ -eremo -erete -eranno -0’ -ammo -aste -arono SUBJUNCTIVE Present am-i (that I love) -i -i -iamo -iate -imo Imperfect am-assi (that I loved) -assi -asse -assimo -aste -assero CONDITIONAL IMPERATIVE Present Present am-erei (I would love) -eresti -erebbe -eremmo -ereste -erebbero am-a (2nd sg.) -ate (2nd pl.) The remaining forms are borrowed from the Subjunctive PARTICIPLE C~ERUND Present am-ando (loving) INFINITIVE Present Past Present am-ante (loving) am-ato (loved, m.sg.) am-are (to love) REFERENCES Anderson, S. 1982. Where’s morphology? Linguistic Inquiry, 13, 571-612. Aronoff, S. 1976. Word formation in generative grammar. Cambridge, MA: MIT Press. Badecker, W., & Caramazza, A. 1985. On considerations of method and theory governing the use of clinical categories in neurolinguistics and cognitive neuropsychology: The case against agrammatism. Cognition, 20, 97-125. Badecker, W., & Caramazza, A. 1987. The analysis of morphological errors in a case of acquired dyslexia. Brain and Language, 32, 278-305. 64 MlCELl AND CARAMAZZA Bortolini, V., Tagliavini, C., & Zampolli, A. 1971. Lessico difrequenze della lingua italiana contemporaries. Milano: Garzanti. Burani, C., Salmaso, D., & Caramazza, A. 1984. Morphological structure and lexical access. Visible Language, 18, 342-352. Butterworth, B. 1983. Lexical representation. In B. Butterworth (Ed.), Language production, Vol. 2. New York: Academic Press. Caplan, D., Keller, L., & Locke, S. 1972. Inflection of neologisms in aphasia. Brain, 95, 169-172. Caramazza, A. 1986. On drawing inferences about the structure of normal cognitive systems from the analysis of patterns of impaired performance: The case for single-patient studies. Brain and Cognition, 5, 41-66. Caramazza, A., & Bemdt, R. S. 1985. A multicomponent deficit view of agrammatic Broca’s aphasia. In M.-L. Kean (Ed.), Agrammatism. Orlando, FL: Academic Press. Pp. 2163. Caramazza, A., Laudanna, A., & Romani, C. (1988). Lexical access and inflectional morphology. Cognition, 28, 297-332. Caramazza, A., Miceli, G., Silveri, M. C., & Laudanna, A. 1985. Reading mechanisms and the organization of the lexicon: Evidence from acquired dyslexia. Cognitive Neuropsychology, 2, 81-l 14. Chomsky, N. 1981. Lectures on government and binding. Dordrecht, Netherlands: Foris Publications. Coltheart, M. 1985. Cognitive neuropsychology and the study of reading. Attention and performance, XI. Hillsdale, NJ: LEA. Cutler, A. 1983. Lexical complexity and sentence processing. In G. B. Flores d’Arcais & R. J. Jarvella (Eds.), The processes of language understanding. London: Wiley. Dell, G. 1986. A spreading-activation theory of retrieval in sentence production. Psychological Review, 93(3), 283-321. De Villiers, J. G. 1978. Fourteen grammatical morphemes in acquisition and aphasia. In A. Caramazza & E. B. Zurif (Eds.), Language acquisition and language breakdown: Parallels and divergences. Pp. 121-144. Baltimore, MD: The Johns Hopkins Univ. Press. Garrett, M. 1980. Levels of processing in sentence production. In B. Butterworth (Ed.), Language production. New York: Academic Press. Vol. 1. Garrett, M. 1982. Production of speech: Observations from normal and pathological language use. In A. Ellis (Ed.), Normality andpathology in cognitive functions. New York/London: Academic Press. Gleason, J. B. 1978. The acquisition and dissolution of the English Inflectional System. In A. Caramazza & E. B. Zurif (Eds.), Language acquisition and language breakdown: Parallels and divergences. Baltimore, MD: The Johns Hopkins Univ. Press. Goodglass, H. 1976. Agrammatism. In H. Whitaker & H. A. Whitaker (Eds.), Studies in neurolinguistics. New York: Academic Press. Vol. 1. Goodglass, H., & Berko, J. 1960. Agrammatism and inflectional morphology in English. Journal of Speech and Hearing Research, 3, 257-267. Goodglass, H., Gleason, J., Ackerman-Bemholtz, N., & Hyde, M. 1972. Some linguistic structures in the speech of a Broca’s aphasic. Cortex, 8, 191-212. Goodglass, H., & Kaplan, E. 1972. The assessment of aphasia and related disorders. Philadelphia: Lea & Febiger. Henderson, L. 1985. Towards a psychology of morphemes. In A. W. Ellis (Ed.), Progress in the psychology of language. London: LEA Limited. Vol. I. Pp. 15-68. Jackendoff, R. 1975. Morphological and semantic regularities in the lexicon. Language, 51, 639-671. Job, R., & Sartori, G. 1984. Morphological decomposition: Evidence from crossed phonological dyslexia. The Quarterly Journal of Experimental Psychology, 36A, 435-458. INFLECTIONAL AND DERIVATIONAL MORPHOLOGY 65 Kean, M.-L. 1977. The linguistic interpretation of aphasia syndromes: Agrammatism in Broca’s aphasia, an example. Cognition. 5, 9-46. Lapointe, S. 1979. A theory ofgrammatical agreement. Unpublished doctoral dissertation, University of Massachusetts, Amherst, MA. Lapointe, S. 1985. A theory of verb form use in the speech of agrammatic aphasics. Brain and Language, 24, 100-155. Lieber, R. 1980. On the organization of the lexicon. Unpublished MIT Ph.D. thesis. Lukatela, G., Gligorijevic, B., Kostic, A., & Turvey, M. T. 1980. Representation of inflected nouns in the internal lexicon. Memory and Cognition, 8, 415-423. MacKay, D. 1979. Lexical insertion, inflection. and derivation. Journul OfPsycholinguistic Research, 8, 477-498. Matthews, P. 1974. Morphology. London: Cambridge Univ. Press. Morton, J. 1979. Word recognition. In J. Morton & J. Marshall (Eds.), Psycholinguistics 2: Structures and Processes. Cambridge, MA: MIT Press. Parisi, D., & Pizzamiglio, L. 1970. Syntactic comprehension in aphasia. Cortex, 6, 204215. Patterson, K. 1980. Derivational errors. In M. Coltheart, K. Patterson, & J. Marshall (Eds.), Deep dyslexia. London: Routledge & Kegan Paul. Patterson, K. E. 1982. The relation between reading and phonological coding: Further neuropsychological observation. In A. W. Ellis (Ed.), Normality and pathology in cognitive functions. New York/London: Academic Press. Scalise, S. 1980. Towards an “extended” Italian morphology. Journal ofItalian Linguistics, l/2, 197-244. Scalise, S. 1984. Generative morphology. Dordrecht, Netherlands: Foris Publications. Selkirk, L. 1982. The syntax of words. Cambridge, MA: MIT Press. Siegel, D. 1974. Topics in English morphology. Unpublished manuscript, MIT, Cambridge, MA. Stemberger, J. P. 1985. An interactive activation model of language production. In A. W. Ellis (Ed.), Progress in the psychology of language. London: LEA Limited. Vol. 1, pp. 143-183. Taft, M. 1984. Evidence for an abstract representation of word structure. Memory and Cognition, 12, 264-269. Taft, M. 1985. The decoding of words in lexical access: A review of the morphographic approach. In D. Besner, T. Wailer, & G. Mackinnon (Eds.), Reading research: Advances in theory and Factice. New York: Academic Press. Vol. 5. Williams, E. 1981. X features. In S. Tavakolian (Ed.), Language acquisition and linguistic theory. Cambridge, MA: MIT Press.