INTRODUCTION

In recent years, much effort has been made todiscover the neural correlates underlying languageprocessing. With regard to the morphosyntacticfeatures of words, neuroscientific research hasmainly focussed on inflectional morphology. Ullmanand colleagues (Ullman et al., 1997; Ullman, 2001)have proposed the declarative/procedural model,where regularly inflected forms (e.g., walked) areparsed in constituent morphemes via a rule systemwhich activates frontal regions, while irregular forms(e.g., bought) are stored as a whole in a temporo-parietal circuit. Neuropsychological observationsand behavioural studies confirm the role of the leftfrontal areas, and in particular of the left inferiorfrontal gyrus (LIFG), in the processing ofmorphologically inflected words (Novoa and Ardila,1987; Miceli et al., 1989; Tyler et al., 2002; Shapiroand Caramazza, 2003). Neuroimaging studies withunimpaired adults have further supported this claim(Laine et al., 1999; Moro et al., 2001; Shapiro et al.,2001; Miceli et al., 2002; Tyler et al., 2004).

Experimental evidence both in healthy subjectsand in aphasic patients has focussed on theindependent processing of inflectional andderivational morphology. Dissociations have beenreported where the processing of inflectional, butnot derivational, morphology was selectivelyimpaired (Garrett, 1982; Tyler and Cobb, 1987;Miceli and Caramazza, 1988; Caplan, 1992; Tyler

and Marslen-Wilson, 1997). Most of the errorsproduced by Italian aphasic patients aremorphological substitutions of inflectional suffixesin repetition, writing to dictation, reading aloud,and speech production [e.g., vestire (to wear) –vestivi (you were wearing); vivo (I live) – vive (helives)] (Miceli and Caramazza, 1988). Morerecently, the specular dissociation of preservedinflectional morphology but impaired derivationalmorphology has also been described: two Italianright-brain damaged patients have been reported,who showed a selective deficit in the production ofderivational morphology in the absence of a deficitin inflectional processing and of aphasic disorders(Marangolo et al., 2003).

Linguists have traditionally made a distinctionbetween inflection and derivation (Allen andBadecker, 2001). There is general agreement that,whereas inflectional affixes have mostly a syntacticfunction, derivational morphology has mainly alexical-semantic function (e.g., Scalise, 1984;Badecker and Caramazza, 1989; Di Sciullo andWilliams, 1989). For Italian nouns and adjectives,inflectional suffixes specify information, such asgender and number [e.g., tavol-o nuov-o, cas-avecchi-a, tavol-i nuov-i, cas-e vecchi-e (new tableand old house, masculine and feminine singular;new tables and old houses, masculine and feminineplural, respectively)]. Similarly, for Italian verbs,inflectional suffixes mark features such as tense,mood, person, number, etc. [e.g., osserv-o (I

Cortex, (2006) 42, 1093-1106

RESEARCH REPORT

FUNCTIONAL ANATOMY OF DERIVATIONAL MORPHOLOGY

Paola Marangolo1, Fabrizio Piras2, Gaspare Galati3,1 and Cristina Burani4

(1Neuropsychology Research Centre, IRCCS St. Lucia Foundation, Rome, Italy; 2Department of Psychology, “La Sapienza” University, Rome, Italy; 3Department of Clinical Sciences and Bioimaging, “G. D’Annunzio” University,

Chieti, Italy; 4Institute of Cognitive Sciences and Technologies, National Research Council – CNR, Rome, Italy)

ABSTRACT

Lexical morphology involves two types of suffixes: inflectional suffixes, which have a grammatical function, andderivational suffixes with a word formation function. In this study, functional magnetic resonance imaging (fMRI) wasused during processing of Italian derived and inflected words. In the derivational task, subjects were asked to producenouns derived from verbs and from adjectives (e.g., to observe – observation; kind – kindness). After the presentation ofthe derived noun, they had to generate the corresponding verb (e.g., failure – to fail: generation task). In the inflectionaltask, subjects had to produce the past participle of the verb or the plural form of the adjective or the derived noun.Behavioural data were collected in separate sessions in two different conditions. In the first experiment, as in the fMRIstudy, vocal reaction times (RTs) were measured from the offset of the auditory stimulus to the onset of the participant’sresponse. In the second experiment, run with a different group of participants, RTs were recorded starting from the onset ofthe auditory stimulus to the onset of the response. The fMRI results showed that, relative to the inflectional task and to arepetition task, the derivational task, but not the verb generation task, brought about an activation of left fronto-parietalregions, documenting a specific involvement of these areas in the processing of derived words. Although less extended,similar activation was found for verb inflection but was absent for noun and adjective plural forms. Analysis of behavioraldata indicated that an explanation in terms of task difficulty was unlikely related to the imaging results.

Key words: derivational morphology, left inferior frontal gyrus, language processing, neuroimaging

1094 Paola Marangolo and Others

observe, present indicative, 1st person, singular);osserv-ato (observed, past participle)]. Hence,inflectional suffixes can be characterised asproducing different forms of the same word orlexical entry. By contrast, derivational morphologyhas a word formation function: originating from agiven base word, different but related words can becreated through processes such as prefixing,suffixing and infixing. Compared to inflectionalsuffixes, derivational suffixes do create a new wordor lexical entry. In Italian, when combined with theroot of the base word, derivational suffixesgenerate a distinct lexical entry, which can be of adifferent grammatical class than the base form,although it is usually related in meaning [e.g.,gentil-e (kind, adjective) – gentil-ezza (kindness,noun); osserv-are (to observe, verb) – osserv-azione (observation, noun)]1.

In addition to the distinction between a syntacticversus a word formation (lexical/semantic)function, inflectional and derivational morphologyalso differ on the type of processing required toactivate the corresponding suffixes. For Italianregularly inflected words, the process of wordinflection relies on the word “inflectionalparadigm”. Thus the inflectional suffix for the pastparticiple of a regular verb is dictated by theinflectional paradigm to which the verb belongs:within the closed set of the inflectional suffixespertaining to the verb, only one is appropriate forthe past participle. Similarly, the appropriate pluralinflection for nouns and adjectives will beunivocally given within the inflectional paradigm ofeach given noun or adjective. Consequently,inflecting a regular word entails activating, withinthe paradigm of a given lexical entry, the only oneinflectional variant which is appropriate for theintended word form.

In contrast, the process of deriving a new wordfrom a given base word does not rely on a pre-existing paradigm. The choice of the derivationalsuffix that creates a noun from a given verb oradjective usually involves a selection processamong different competitor suffixes (see Burani etal., 1999). For instance, both the nouns osserv-azione (observation) and osservatore (observer) aregood word formations that are derived from theverb osservare (to observe). However, neither*osserv-amento (*observement) or *osservista(*observist), although possible noun derivations,are existing Italian derived words. It should benoticed that for a different verb, like cambiare (tochange), the opposite holds, in that *cambi-azione(*changetion) does not exist, while cambi-amento(changement) is the appropriate derived word.Analogously, the noun piccol-ezza (smallness), but

not the noun *piccol-ità (*smallity) can be derivedfrom the adjective piccolo (small); while the nounsever-ità (severity), but not *sever-ezza(*severeness) is the appropriate derivation fromsevero (severe). The production of a noun derivedfrom a verb entails, with respect to the productionof a noun derived from an adjective, some furtherselection: for a verb, but not for an adjective, thereare often more than one existing derived nouns[e.g., both osserv-azione (observation) andosservatore (observer) can be derived from theverb osservare (to observe), whereas only piccol-ezza (smallness) can be derived from the adjectivepiccolo (small)]. In very much synthesis, theproduction of a suffixed derived word impliesselecting the correct suffix among a set ofcompeting suffixes (see Burani et al., 1999).

Until now, there is scarce neuroimagingevidence regarding the areas involved in theprocessing of morphologically derived words (e.g.,walker). Davis et al. (2004) have explored theneural systems involved in processing differentword classes in a functional magnetic resonanceimaging (fMRI) study, contrasting four groups ofwords which differed for morphological complexity(simple, monomorphemic words vs. complexderived or inflected words) and syntactic class(verbs vs. nouns/adjectives). To ensure semanticprocessing of each word, the authors used a one-back synonym-monitoring task in whichparticipants read single words presented on thescreen and pressed a button if the current word wasrelated in meaning to the immediately precedingword. Synonym pairs included monomorphemicwords, inflected verbs and derived words. Noactivation differences were observed formorphologically complex and simple words.However, since all the morphologically complexwords had a transparent semantic relation withtheir stems, it could be the case that the absence ofany difference in activation for morphologicallysimple and complex words was due to the taskused: to judge the presence or absence of asemantic relation between the presented words(e.g., kindness, goodness), subjects could haverelied on the stems only, while ignoring thesuffixes.

A recent unpublished study by Vannest et al.(2003) has used fMRI to compare the silentreading of derivationally suffixed English words tothe reading of inflected and monomorphemicwords. According to the authors, an “increasedactivity in Broca’s area and the basal ganglia” wasfound not only for regular inflected words, but alsofor those derived words “that show evidence ofdecompositional processing in behavioural studies”,namely words that include productive suffixes(e.g., -ness, -less, -able) and are bothphonologically and semantically transparent (e.g.,darkness) with respect to the word they derivefrom (e.g., dark). By contrast, derivationally

1Italian derivational suffixes can additionally be inflected for number andfor gender [e.g., osserv-azion-e (observation) – osserv-azion-i(observations); oper-ai-o (worker, masculine) – oper-ai-a (worker,feminine)].

suffixed words that are not fully transparent andusually do not show behavioural decompositioneffects (e.g., -ity words like serenity) did not showincreased neural activity in the same areas.According to the authors, a morphological rule-based process is used to access derivationallysuffixed words of the first type, while whole-wordprocessing occurs for suffixed derived words thatare not phonologically or semantically transparent.

The data reported by Vannest et al. (2003) areconsistent with several experimental results drawnfrom visual or auditory lexical decision tasks thathave been conducted on unimpaired adults indifferent languages, including English and Italian(Burani and Caramazza, 1987; Marslen-Wilson etal., 1994; Wurm, 1997; Bertram et al., 2000;Burani and Thornton, 2003). These studies haveshown that, in recognition tasks, not all derivedwords are equally subject to being morphologicallydecomposed, but different properties affect theprobability that a derived word will be processedeither decompositionally or as a stored whole-word. Some of these properties are thephonological and semantic transparency of thederived word with respect to the base word, withphonologically and semantically transparent wordsbeing more subject to decompositional processing.Another property that can affect the probability ofmorphological parsing is the frequency of thewhole-word relative to the frequency of itsconstituents, with low-frequency words with high-frequency and productive constituents (roots andaffixes) being more subject to be parsed.

In the present study, fMRI was used duringmorphological processing of derived and inflectedwords. We focussed on those derived words thatare more subject to be produced compositionallyby root-suffix combination, for being transparentand including frequent and productive suffixes (seealso Panzeri et al., 1990; Badecker and Caramazza,1991). Considering that the inflected words wereall regular and transparent, our study aimed atassessing whether similarly transparent suffixedderived words activate the same frontal regionswhich are responsible for morphological processingin the production of regular inflected words. Theprediction could be made that, along with aninvolvement of left frontal areas which are sharedwith the production of transparent regularly

Derivational processing 1095

inflected words, the production of a transparentderived word might involve additional brain areasthat are traditionally associated to lexical-semanticprocessing (see, e.g., Cappa et al., 1998; Perani etal., 1999) and might govern response selection (seealso Marangolo et al., 2003).

A word generation task was employed (for adiscussion of this task, see Tyler et al., 2004, p.520). Participants were auditorily presented withlists composed of either verbs, or adjectives, ormorphologically derived nouns; they performedthree different tasks for each word class (see TableI). In the experimental tasks, after the presentationof either a verb or an adjective, they were asked tosilently produce the corresponding derived noun(derivational task), and, after the presentation of aderived noun, the infinitive form of thecorresponding verb (generation task). Note thatverb-from-(derived) noun generation involves theretrieval of the base word from a derived form anddoes not imply the production of a new lexicalform. In the inflectional task, participants wereasked to silently produce the past participle ofverbs, or the plural form of adjectives and nouns.This task controlled for the morphological parsingcomponent shared with the derivational task. In therepetition task, participants were asked to silentlyrepeat the presented verb, adjective or noun. Thistask controlled for auditory, phonological andautomatic lexical-semantic processing of presentedwords, and for silent word production.

MATERIALS AND METHODS

Participants

Ten healthy right-handed (based on theEdinburgh Handedness Inventory; Oldfield, 1971)participants, aged 21-29, five males and fivefemales, Italian native speakers, gave their informedconsent to participate in the experiment, whoseprocedures had received local ethical approval.

Stimuli

The experimental stimuli (reported in theAppendix) consisted of 90 words belonging tothree grammatical categories: 30 verbs in the

TABLE I

Summary of the experimental conditions

Derivation/generation tasks Inflectional task Repetition task

Verb Derivation of noun from verb: Production of the verb past participle: Repetition of the verbosserv-are (to observe) – osserv-are (to observe) –osserv-azione (observation) osserv-ato (observed)

Adjective Derivation of noun from adjective: Production of the adjective plural form: Repetition of the adjectivegentil-e (kind) – gentil-e (kind, singular) –gentil-ezza (kindness) gentil-i (kind, plural)

Noun Generation of verb from noun: Production of the noun plural form: Repetition of the nounfall-imento (failure) – falliment- o (failure) –fall-ire (to fail) falliment—i (failures)

inflected infinitive form, 30 adjectives in theinflected singular form, and 30 singular derivednouns, which however were not derived from anyof the used verbs. The three classes of words weretaken from a low-to-medium frequency range andwere matched across sets for frequency of usage(Istituto di Linguistica Computazionale del CNR diPisa, 1989). The derived nouns were semanticallytransparent with respect to their base word. The 90words were read by a professional speaker,digitally recorded, normalised for volume, andsaved in digital sound files.

Cognitive Tasks

The experiment consisted of three sessions inwhich participants listened to lists of words fromthe three grammatical classes, respectively. In eachsession, participants alternated three different tasksin which they had to produce a word in response toeach stimulus they had listened to (see Table I). Inthe verb session, they had to produce either thecorresponding derived noun [noun-from-verbderivation task, e.g., osserv-are (to observe) –osserv-azione (observation)], or the verb pastparticiple [verb inflectional task, e.g., osserv-are (toobserve) – osserv-ato (observed)], or they had torepeat the verb (verb repetition task). In theadjective session, they had to produce either thecorresponding derived noun [noun-from-adjectivederivation task, e.g., gentil-e (kind) – gentil-ezza(kindness)], or the adjective plural form [adjectiveinflectional task, e.g., gentil-e (kind, singular) –gentil-i (kind, plural)], or they had to repeat theadjective (adjective repetition task). In the nounsession, participants had to produce either theinfinitive form of the verb the noun was derivedfrom [verb-from-noun generation task, e.g., fall-imento (failure) – fall-ire (to fail)], or the pluralform of the derived noun [noun inflectional task,e.g., fall-imento (failure) – falliment-i (failures)], orthey had to repeat the noun (noun repetition task)(see Table I for reference). The nouns that had to bederived from verbs or adjectives were transparent inmeaning with respect to the base verb or adjective.Thus we made sure that there was an analogouslytransparent semantic relation between stimulus andresponse across tasks (inflectional and derivational).

Sessions lasted 6 minutes each and were run ina fixed sequence: derived nouns first, thenadjectives, and finally verbs, with a few minutes ofrest between sessions. In each session, participantsalternated the three tasks in 15.5 sec blocks, in arandomised sequence. Each block consisted of a 3sec instruction phase, with written instructionsappearing on the centre of the screen, followed by5 words auditorily presented every 2.5 sec, whileparticipants fixated a central cross. Six blocks of 5words were administered for each task, so that all30 words of a given grammatical class werepresented once per task in a randomised sequence.

1096 Paola Marangolo and Others

During fMRI acquisition, participants wererequested to wait for the offset of each presentedword and then subvocally produce their responsesto avoid unnecessary head movements. Behaviouraldata were collected in separate sessions in twodifferent conditions. In the first condition,experimental procedures and participants were thesame as in the fMRI study. Subjects were instructedto wait for the offset of each presented word beforegiving their responses, and vocal reaction times(RTs) were accordingly measured from the offset ofthe auditory stimulus to the onset of theparticipant’s response. Note, however, that weasked participants to wait until the offset of theauditory stimulus before giving their responses, sothey could have prepared their responses inadvance, resulting in the absence of any differencein RTs, as measured from the offset of the stimulus.For this reason, we ran a second experiment with adifferent group of ten participants. RTs wererecorded starting from the onset of the auditorystimulus to the onset of the participant’s response.

fMRI Procedures

Images were acquired using a 1.5 T SiemensVision Magnetom MR system (Siemens MedicalSystems, Erlangen, Germany) equipped for echo-planar imaging, with a quadrature volume head coilfor radio frequency transmission and reception. Foreach participant, we acquired a T1 weightedvolumetric image set [1 mm isotropic voxels, 220coronal slices, Siemens multiplanar rapidacquisition gradient echo sequence, repetition time(TR) = 11.4 msec, echo time (TE) = 4.4 msec], andthree time series of 127 fMR images using blood-oxygenation-level-dependent imaging (TR = 3 sec,TE = 60 msec, 27 axial slices, 64 × 64 imagematrix, 3 × 3 mm in-plane resolution, 3.3 mmthickness with a .7 mm gap between slices,sequential excitation order). Each time seriescorresponded to one experimental session describedabove. The first four volumes of each series werediscarded to achieve steady-state transversemagnetization, and the experimental tasks started atthe beginning of the fifth volume.

Stimuli were generated by a control computer(Power Macintosh G3, Apple Computers,Cupertino, CA, USA) located outside the MRroom, running in-house software (LabScript),implemented in MATrix LABoratory (MATLAB)(The MathWorks Inc., Natick, MA, USA) using thePsychophysics Toolbox extensions (Brainard, 1997;Pelli, 1997), and allowing time-locked presentationof written instructions and auditory stimuli, whilemaintaining millisecond timing accuracy over aperiod of minutes and triggering the acquisition offMR images. An liquid crystal display (LCD) videoprojector with a customised lens was used toproject written instructions to a back projectionscreen mounted above the head coil and visible

through a mirror. Auditory stimuli were presentedthrough a pneumatic headphone device.

Image Processing and Analysis

Images were preprocessed and analysed usingSPM99 (Wellcome Department of CognitiveNeurology, London, UK). First, functional timeseries from each participant were corrected forslice timing and for head movement, manuallycoregistered to the participant’s anatomical image,spatially normalised using transformationparameters obtained from the normalization of theanatomical image (final voxel size: 3 × 3 × 3 mm),and smoothed (6 mm full-width half-maximum –FWHM). Then, images were analyzed using arandom-effect methodology (Friston et al., 1999).A general linear model was fit to the individualdata, with the experimental blocks modeled as box-car functions, convolved with a synthetichemodynamic response function (to account for thedelay of the hemodynamic response). Additionalmodeled explanatory variables included thetemporal derivatives of the convolved box-carfunctions and the head movement parametersestimated during the preprocessing stage. Thesemodels were used to compute subject-specificeffect size images for each effect of interest, whichwere then entered into group one-sample t-tests,testing the null hypothesis that the mean effect sizewas equal to zero in the whole population ourparticipants were extracted from. Significance wasjudged at the voxel level (p < .01) and by clustersize (p < .05 corrected for multiple comparisonsusing gaussian field theory).

Localisation and visualisation of activationswere achieved by using in-house software(BrainShow, written in MATLAB), which allowssuperposing statistical maps over anatomical brainimages as well as over folded, inflated, andflattened representations of the cortical surface. Forvisualisation of group activations, we used thecortical surface of a “standard” brain (the single-subject Montreal Neurological Institute – MNI –brain), reconstructed using the FreeSurfer software(http://surfer.nmr.mgh.harvard.edu). Anatomicallabels were automatically assigned to activatedareas by the BrainShow software, based on themacroscopical anatomical parcellation of the MNIsingle-subject brain described by Tzourio-Mazoyeret al. (2002), which was further refined by splittingthe largest gyri into anterior/posterior orsuperior/inferior portions.

RESULTS

Behavioral Results

As previously reported, behavioural data wereobtained, in a separate session, in two different

Derivational processing 1097

conditions. In both experiments, the number oferrors was negligible: no participant made morethan one error per condition. In the firstexperiment, in which subjects waited until theoffset of the auditory stimulus before giving theirresponses, a two-way analysis of variance(ANOVA) with task and grammatical category aswithin-participant factors did not reveal anysignificant difference in vocal RTs. Thus, thedifferent tasks and word classes posed equaldifficulty for the participants.

In the second experiment, RTs were recordedstarting from the onset of the auditory stimulus tothe onset of the participant’s response. A two-wayANOVA with task and grammatical category aswithin-participant factors revealed a significanteffect of task [F (2, 18) = 9.31, p < .005], with therepetition task significantly faster than thederivational and inflectional tasks, which did notdiffer from each other. The grammatical categoryeffect was also significant [F (2, 18) = 27.98, p <.0001], with faster responses to adjectives than toverbs, which in turn gave faster responses thannouns. The task by grammatical categoryinteraction was also significant [F (4, 36) = 8.16, p< .001]. For the verb class, deriving a noun fromthe verb and producing the verb past participleresulted in the same RTs, and both tasks resulted inlonger latencies than the repetition task (p < .002).For the adjective class, deriving a noun from theadjective was slower (p < .002) than producing theadjective plural form or repeating the adjective;responses to these last two conditions did not differfrom each other. Finally, for the noun class,generating a verb from the derived noun andrepeating the noun did not differ from each other,and both tasks were significantly faster thanproducing the noun plural form (p < .002) (seeFigure 1).

fMRI Results

Our main purpose was to investigate the neuralcorrelates of derivational processing. Thus, as afirst step, we compared the derivational task, i.e.,the noun-from-verb and noun-from-adjectivederivation tasks, with the corresponding repetitiontasks, i.e., the verb and adjective repetition tasks,respectively. We also compared the verb-from-noungeneration task with the noun repetition tasks. Aspreviously stated, the verb-from-noun generationcondition does not involve derivational processing.Therefore, we expected that it would not activate(or would activate to a lesser extent) the brainareas for derivational morphology.

Significant activations are shown in Figure 2and listed in Table II. Noun-from-verb and noun-from-adjective conditions activated a commonnetwork, including the ventrolateral frontal cortexand the bilateral inferior parietal lobule (IPL). Thefrontal activation was bilateral for noun-from-verb

derivation and limited to the left hemisphere fornoun-from-adjective derivation. Most of the frontalactivation was located in the pars triangularis andpars opercularis of the IFG. The parietal activationwas bilateral and included the dorsal part of theIPL and the intraparietal sulcus and extended intothe angular gyrus. The extent of the activation wasmuch larger in the left hemisphere for both frontaland parietal areas. Additional foci of activationwere found in the left supplementary motor areaand middle temporal gyrus for noun-from-adjectivederivation, and in the basal ganglia for bothconditions.

A conjunction analysis (Price and Friston, 1997)was also performed, to show regions which werecommonly activated by derivation versus repetitionacross the two grammatical classes. This analysisconfirmed that the left frontal and the bilateralparietal regions described above were commonlyactivated by noun-from-verb and noun-from-adjective derivation. Conversely, when we lookedat the task by grammatical class interaction, wefound no significant region showing a differentialactivation between noun-from-verb and noun-from-adjective derivation.

A further expectation was that verb-from-noungeneration did not activate the same network as thetwo derivational tasks. Indeed, verb-from-noungeneration only activated the left IFG, and theextent of the activation (972 mm3) was strikinglysmaller than for derivation tasks (20.088 and17.054 mm3 for nouns derived from verbs andadjectives, respectively).

To verify the extent to which this fronto-parietal

1098 Paola Marangolo and Others

network could be attributed to the derivationalprocess, the verb, adjective and noun inflectionaltasks were compared with the verb, adjective and

Fig. 1 – Data from the second behavioral experiment: mean (SD) vocal reaction times (RTs) for the three tasks and the threegrammatical classes.

Fig. 2 – Brain regions activated by derivation/generationand inflection tasks versus repetition tasks, for each grammaticalclass, superimposed to lateral views of a standard brain. Leftcolumn: regions activated by derivation/generation. Rightcolumn: regions activated by inflection. Top row: noun from verbderivation and verb inflection versus verb repetition. Middlerow: noun from adjective derivation and adjective inflectionversus adjective repetition. Bottom row: verb from noungeneration and noun inflection versus noun repetition. A: leftfrontal; B: left parietal; C: right frontal; D: right parietal; E:left temporal; F: left temporo-parieto-occipital; G: rightoccipital; H: right temporal. The same cluster labels are used inTables II and III.

Res

pons

e tim

es (

mse

c)

Grammatical class

Derivation/Generation tasks

Inflectional taskRepetition task

1500

Verb

131212971254

Adjective

124511971202

Noun

130713531308

1400

1300

1200

1100

1000

900

800

noun repetition. Here, we expected that areasselective for derivational morphology would not beactivated, or would be activated to a lesser extent.Significant activations are shown in Figure 2 andlisted in Table III. The pattern of activation wasstrikingly different than for derivation tasks. Verbinflection activated the IFG and the IPL, but onlyin the left hemisphere, and the extent of theactivated clusters was much smaller than for verbderivation (IFG: 4.401 vs. 20.088 mm3; IPL: 1.998vs. 6.102 mm3). Adjective inflection did notactivate the IFG and the IPL at all, and nouninflection activated only a small region in the leftinsula. Additional inflection-related activationswere all located in regions clearly distinct fromthose activated during derivation: medial regions inthe paracentral lobule and the anterior and middlecingulum, inferior temporal, and parieto-occipitalregions.

Derivational processing 1099

In summary, when using repetition as abaseline, the patterns of activation induced by thederivational and the inflectional tasks were verydifferent. The derivational tasks activated abilateral, although left-hemisphere based, fronto-parietal network, while only the verb inflectionaltask showed a limited activation of this network inthe left hemisphere.

To further confirm the involvement of the leftfronto-parietal regions in the derivational task, wedirectly looked at similarities and differencesbetween the derivation and generation tasks and thecorresponding inflectional tasks. Similarities wereassessed through a conjunction analysis aimed atidentifying regions of common activation inderivation and inflection tasks, relative torepetition. This analysis, when performed acrossgrammatical classes (verbs and adjectives), failedto show any region commonly activated by

TABLE II

Brain regions activated by derivation/generation tasks versus repetition. Activated clusters are labelled with the same letters as in Figure 2. For each activated cluster, the table lists the cluster extent (mm3), the maximum voxel Z value in the cluster,

and the anatomical localizations. Stereotactic coordinates (x, y, z in mm) in MNI space (Mazziotta et al., 1995) are also given for representative local maxima in each anatomical subdivision

Noun from Verb Derivation vs. Verb Repetition

Regions Subregions Coordinates

Left frontal (A) Left inferior frontal gyrus (triangular) BA 44 – 53, 28, 2420.088 mm3 (Z max = 4.26) Left inferior frontal gyrus (opercular) BA 45 – 59, 13, 21

Left middle frontal gyrus (anterior) BA 10 – 44, 55, 30Left precentral gyrus (inferior) BA 9 – 50, 13, 30Left insula – 32, 22, 30

Left parietal (B) Left inferior parietal lobule BA 40 – 41, – 48, 426.102 mm3 (Z max = 4.18) Left angular gyrus BA 39 – 38, – 63, 39

Left superior parietal lobule (posterior) BA 7 – 35, – 69, 54Left superior parietal lobule (anterior) BA 7 – 29, – 63, 51

Right frontal (C) Right inferior frontal gyrus (triangular) BA 10 66, 22, 182.538 mm3 (Z max = 4) Right inferior frontal gyrus (opercular) BA 9 50, 13, 24

Right middle frontal gyrus (anterior) BA 46 41, 34, 18Right parietal (D) Right angular gyrus BA 39 35, – 72, 45567 mm3 (Z max = 3.57) Right superior parietal lobule (posterior) BA 7 35, – 72, 51Basal ganglia Right caudate 20, 4, 24432 mm3 (Z max = 3.06)

Noun from Adjective Derivation vs. Adjective Repetition

Regions Subregions Coordinates

Left frontal (A) Left inferior frontal gyrus (triangular) BA 47 – 44, 28, – 117.064 mm3 (Z max = 5.1) Left inferior frontal gyrus (opercular) BA 45 – 53, 13, 18

Left insula – 35, 25, – 1Left middle frontal gyrus (posterior) BA 9 – 38, 10, 36

Left parietal (B) Left inferior parietal lobule BA 40 – 38, – 54, 512.565 mm3 (Z max = 3.58) Left superior parietal lobule (posterior) BA 7 – 26, – 69, 54

Left angular gyrus BA 7 – 35, – 63, 39Left superior parietal lobule (anterior) BA 7 – 32, – 60, 45

Right parietal (D) Right angular gyrus BA 39 35, – 57, 48999 mm3 (Z max = 4.31) Right inferior parietal lobule BA 7 32, – 54, 45Medial frontal Left supplementary motor area BA 6 – 5, 13, 60918 mm3 (Z max = 3.37)Left temporal (E) Left middle temporal gyrus (posterior) BA 22 – 59, – 39, 3594 mm3 (Z max = 3.51) Left middle temporal gyrus (anterior) BA 22 – 56, – 33, 3Basal ganglia Right caudate 20, 4, 24648 mm3 (Z max = 3.64) Left pallidum – 20, 7, 30

Verb from Noun Generation vs. Noun Repetition

Regions Subregions Coordinates

Left frontal (A) Left inferior frontal gyrus (triangular) BA 44 – 53, 28, 30972 mm3 (Z max = 3.33) Left inferior frontal gyrus (opercular) BA 45 – 53, 16, 21

Left precentral gyrus (inferior) BA 44 – 41, 10, 30

derivation and inflection. For verbs only, a regionof 1.863 mm3 in the left IFG appeared to becommonly activated by noun-from-verb derivationand verb inflection, as expected from the resultsreported above.

Differences were assessed by directlycomparing derivation versus inflection. In thiscomparison (Figure 3 and Table IV), the noun-from-verb and noun-from-adjective derivationalconditions significantly activated the IFG and theIPL, while, as expected, the verb-from-noungeneration task did not show any significantactivation. The left frontal and parietal regionswere comparably activated for nouns derived fromverbs and from adjectives, while the rightactivation was significant only for nouns derivedfrom verbs in the frontal lobe and for nounsderived from adjectives in the parietal lobe, andwas much smaller than in the left hemisphere.

We also tested the reverse comparison, i.e., wesearched for regions more activated by inflectionalthan by derivational tasks (Figure 3, and Table V):a region around the left temporo-parieto-occipitaljunction was activated for verbs and adjectives, anda region in the middle cingulum for adjectives andnouns.

1100 Paola Marangolo and Others

TABLE III

Brain regions activated by inflection tasks versus repetition. Details as for Table II

Verb Inflection vs. Verb Repetition

Regions Subregions Coordinates

Left frontal (A) Left inferior frontal gyrus (triangular) BA 45 – 50, 43, 604.401 mm3 (Z max = 3.88) Left precentral gyrus (inferior) BA 6 – 53, 4, 18

Left inferior frontal gyrus (opercular) BA 45 – 59, 13, 21Left middle frontal gyrus (anterior) BA 10 – 44, 55, 30Left middle frontal gyrus (posterior) BA 9 – 35, 10, 36Left precentral gyrus (superior) BA 44 – 50, – 3, 48

Left parietal (B) Left inferior parietal lobule BA 40 – 44, – 42, 481.998 mm3 (Z max = 4.28) Left superior parietal lobule (posterior) BA 7 – 29, – 69, 51

Left angular gyrus BA 7 – 38, – 72, 45Medial frontal Left paracentral lobule BA 6 – 11, – 18, 69675 mm3 (Z max = 3.27) Left supplementary motor area – 11, – 9, 78Left temporal (E) Left inferior temporal gyrus (posterior) BA 37 – 47, – 57, – 13324 mm3 (Z max = 3.23)

Adjective Inflection vs. Adjective Repetition

Regions Subregions Coordinates

Right frontal (C) Right middle frontal gyrus (posterior) BA 6 47, – 6, 57297 mm3 (Z max = 3.48)Medial frontal Left paracentral lobule BA 7 – 5, – 27, 57324 mm3 (Z max = 3.52)Left temporo-parieto-occipital (F) Left middle occipital gyrus BA 19 – 38, – 72, 241.539 mm3 (Z max = 3.83) Left angular gyrus – 53, – 69, 27Right occipital (G) Right middle occipital gyrus BA 19 32, – 84, 15405 mm3 (Z max = 2.63) Right superior occipital gyrus BA 19 26, – 96, 15

Noun Inflection vs. Noun Repetition

Regions Subregions Coordinates

Left frontal (A) Left insula – 38, 19, – 1297 mm3 (Z max = 3.86)Medial frontal Right anterior cingulum BA 32 8, 25, 27702 mm3 (Z max = 3.34) Right middle cingulum BA 32 2, 7, 42Right temporal (H) Right superior temporal gyrus (polar) BA 38 52, 4, – 12459 mm3 (Z max = 3.99)

Fig. 3 – Brain regions activated by derivation/generationtasks versus inflection tasks, for each grammatical class,superimposed to lateral views of a standard brain. Left column:regions activated by derivation/generation versus inflection.Right column: regions activated by inflection versusderivation/generation. Top row: noun from verb derivation andverb inflection. Middle row: noun from adjective derivation andadjective inflection. Bottom row: verb from noun generation andnoun inflection. A: left frontal; B: left parietal; C: right frontal;D: right parietal; F: left occipito-temporal; I: medial frontal.The same cluster labels are used in Tables IV and V.

DISCUSSION

The goal of the present study was to investigateif derivational processing shares the same frontalregions shown to be involved in inflectionalmorphology, or whether the two morphologicalprocesses would involve different neural structures.The core of our results is the demonstration that,even though derivational processing shared someregions of activation with inflectional morphology,it showed specific activation of left fronto-parietalregions. Interestingly, adjective and noun

Derivational processing 1101

inflections did not activate this network but showeda common activation in the medial frontal regionswhich seems specific to this latter process.

Since we were mainly interested ininvestigating the cerebral areas involved inderivational processing, we will first discuss theactivations specifically related to this process, i.e.,the left ventrolateral frontal cortex, the left inferiorparietal lobule and, to a lesser extent, the rightventrolateral frontal cortex and the right parietalregions. Activations related to inflectionalprocesses will be also commented.

TABLE IV

Brain regions activated by derivation tasks versus inflection. The verb from noun generation versus noun inflection comparison is notshown, because it did not yield any significant activation. Details as for Table II

Noun from Verb Derivation vs. Verb Inflection

Regions Subregions Coordinates

Left frontal (A) Left inferior frontal gyrus (triangular) BA 45 – 41, 34, 309.666 mm3 (Z max = 3.97) Left inferior frontal gyrus (opercular) BA 44 – 56, 19, 33

Left precentral gyrus (inferior) BA 44 – 50, 10, 30Left middle frontal gyrus (anterior) BA 46 – 47, 37, 21

Left parietal (B) Left inferior parietal lobule BA 40 – 32, – 57, 391.836 mm3 (Z max = 3.81) Left superior parietal lobule (posterior) BA 7 – 32, – 66, 48Right frontal (C) Right inferior frontal gyrus (triangular) BA 44 56, 19, 24189 mm3 (Z max = 2.61) Right inferior frontal gyrus (opercular) BA 9 47, 19, 27Basal ganglia Right caudate 20, 4, 18432 mm3 (Z max = 3.06)

Noun from Adjective Derivation vs. Adjective Inflection

Regions Subregions Coordinates

Left frontal (A) Left inferior frontal gyrus (triangular) BA 46 – 47, 31, 219.342 mm3 (Z max = 4.08) Left inferior frontal gyrus (opercular) BA 44 – 47, 10, 18

Left precentral gyrus (inferior) BA 6 – 38, 4, 30Left middle frontal gyrus (posterior) BA 9 – 38, 13, 36Left middle frontal gyrus (anterior) BA 46 – 44, 43, 18

Left parietal (B) Left inferior parietal lobule BA 7 – 38, – 54, 541.782 mm3 (Z max = 3.54) Left angular gyrus BA 39 – 35, – 63, 39

Left superior parietal lobule (anterior) BA 7 – 32, – 63, 45Right parietal (D) Right inferior parietal lobule BA 7 38, – 54, 48270 mm3 (Z max = 2.81)Medial frontal Left supplementary motor area BA 6 – 5, 13, 60297 mm3 (Z max = 3.28)

TABLE V

Brain regions activated by inflection versus derivation tasks. Details as for Table II

Verb Inflection vs. Verb Derivation

Regions Subregions Coordinates

Left temporo-parieto-occipital(F) Left middle occipital gyrus BA 39 – 44, – 72, 24108 mm3 (Z max = 2.84) Left angular gyrus BA 39 – 50, – 72, 26

Left middle temporal gyrus (posterior)

Adjective Inflection vs. Adjective Derivation

Regions Subregions Coordinates

Medial frontal Middle cingulum BA 24 – 2, – 2, 42108 mm3 (Z max = 3.61)Left temporo-parieto-occipital (F) Left middle occipital gyrus BA 19 – 38, – 72, 20567 mm3 (Z max = 3.41) Left middle temporal gyrus BA 19 – 46, – 74, 20

Left angular gyrus

Noun Inflection vs. Verb from Noun Generation

Regions Subregions Coordinates

Medial frontal (I) Middle cingulum BA 32 2, 10, 42216 mm3 (Z max = 3.24)

Left Fronto-Parietal Regions

As stated in the introduction, two separatesystems seem to underlie the processing ofinflected words: the left frontal system responsiblefor morphological decomposition of regularinflected forms, and the left temporo-parietalsystem which processes irregular forms as wholeunits (Ullman et al., 1997; Ullman, 2001). Bothneuropsychological observations and neuroimagingdata with unimpaired participants support this view(Miceli et al., 1989, 2002; Laine et al., 1999; Moroet al., 2001; Shapiro et al., 2001; Tyler et al., 2002,2004). Frontal regions and specifically the IFGwere found to be active in morphological parsingtasks. This led some authors to suggest that theIFG is responsible for morphologicaldecomposition which could be in common to bothinflected and phonologically transparent derivedwords (Vannest et al., 2003; Tyler et al., 2004).

In the present study, similar regions andspecifically the pars triangularis (BA 44, 45) andopercularis (BA 45) of the IFG were activatedwhen contrasting both the derivational and theinflectional tasks with the repetition task. Since thederived words that we used were very likely to beproduced compositionally, these results seem toconfirm that those regions were engaged inmorphological decomposition. Interestingly, whenthe derivational and the inflectional tasks weredirectly contrasted, different portions of the leftIFG (BA 45 pars triangularis, BA 44 parsopercularis) still remained active. This supports thehypothesis that additional components areresponsible for the derivational task (Badecker andCaramazza, 1989; Burani et al., 1999).

It is widely accepted that the left frontal regionsand in particular the LIFG are involved inlexical/semantic retrieval process (see e.g.,Warburton et al., 1996; Cappa et al., 1998; Peraniet al., 1999). Recent imaging studies furthersuggest that the activity of the left IFG is notrelated to semantic retrieval per se but rather todemands for the selection of information fromcompeting alternatives (Thompson-Schill et al.,1997, 1999; Desmond et al., 1998; Nathaniel-Jamesand Frith, 2002). A number of studies areconsistent with this hypothesis: for example,lesions in the IFG have been shown to impairsentence completion and word association tasks(Robinson et al., 1998), but only when there areseveral possible responses. Also, IFG lesions affectpatients’ ability to select the appropriate meaningof ambiguous words (Swaab et al., 1998). Selectioneffects in the IFG have also been shown inparadigms such as verb generation tasks in highversus low response competition conditions(Thompson-Schill et al., 1997). As stated in theintroduction, different from inflectionalmorphology, derivational processing implies notonly lexical/semantic retrieval (word formation

1102 Paola Marangolo and Others

function) but also the selection of the appropriatederivational suffix among a set of alternatives(Badecker and Caramazza, 1989; Burani et al.,1999). Thus, we suggest that the more extensiveinvolvement of the left frontal regions found forderivational processing reflects both lexicalretrieval and response selection processes.

The activation of parietal regions was alsointeresting. In the neuropsychological literature,there are reports of patients with parietal damagewho make morphological errors (Silveri and DiBetta, 1997; Marangolo et al., 2003), although it iswidely accepted that left parietal regions aremainly involved in the processing of the lexical-semantic components of words (Alexander et al.,1989). Results from neuroimaging studies supportthis claim. Parietal areas are active in semanticprocessing and word generation tasks (Frith et al.,1991; Demonet et al., 1992; Warburton et al., 1996;Binder et al., 1997). Since the derivational processimplies the generation of a lexical form with adistinct meaning relative to the base form, onereasonable explanation of our results is that parietalareas were activated because these regions arespecialised in the retrieval of lexical-semanticinformation.

To further confirm the specificity of theactivation for the retrieval of derived nouns, ourexperimental design included a task whereparticipants had to produce the base verb infinitivethe noun was derived from (verb-from-noungeneration). Although this task implies a wordretrieval process and a change of grammaticalcategory, it does not imply retrieving a derivedword (i.e., it does not imply a word formationprocess). Hence, we predicted that it would notactivate (or would activate to a lesser extent) thebrain network for derivational morphology. Theresults confirmed this prediction.

A final point regards the partial activationobserved in the left fronto-parietal regions in theverb inflection condition (but only when comparedto the repetition task), which was absent in thenoun and adjective inflectional tasks. It has beenshown that frontal regions, and specifically the leftIFG, are responsible for the processing of thegrammatical properties of verbs. Bothneuropsychological and neuroimaging studiesconfirmed this issue (Marslen-Wilson and Tyler,1997; Xiong et al., 1998; Shapiro et al., 2001; Tyleret al., 2002, 2004). Left frontal hemisphere patientsfrequently show agrammatic speech with difficultyin verb retrieval and morphological deficits whichaffect verb inflection (Miceli et al., 1989; Marslen-Wilson and Tyler, 1997; Tyler et al., 2002). In ourstudy, the activation of the left IFG found in theverb inflection condition is in line with thesefindings. Recently, Tyler et al. (2004) presenteddata consistent with our results. In a neuroimagingstudy, the authors contrasted the processing ofregularly inflected nouns (e.g., dogs) with regularly

inflected verbs (e.g., hitting) and found that the IFGwas more strongly activated for verb processingcompared to nouns. The authors advanced thehypothesis that the greater involvement of the leftIFG activity during the processing of inflectedverbs could reflect processes sensitive to themorphological structure of verbs. In other words,the authors suggested that verb processing activatesprocesses of decomposition of its constituentmorphemes in which the left IFG is specificallyengaged. In their opinion, these processes are lessactive in the elaboration of noun inflection since“noun plurals have a primarily semantic role insentence interpretation” (Tyler et al., 2004, p. 521)and therefore, they are subject to a lessersegmentation (for similar conclusions, see alsoColombo and Burani, 2002). Following this line ofreasoning, this differential processing whichunderlies verb and noun inflections, could probablyexplain why, in our study, we found a significantactivation of the left IFG in the verb inflectioncondition which was absent for noun and adjectiveplurals. Similarly to nouns, adjective inflectionshave a primarily semantic role in sentenceinterpretation, and show to be less subject tomorphological decomposition relative to verbs(Kostic and Katz, 1987; Traficante and Burani,2003). Therefore, it could be the case that theirprocessing, like for nouns, is subserved by regionsdifferent from the left IFG.

Right now, the relatively small number ofneuropsychological and neuroimaging studies whichdirectly compared the morpho-syntactic functions ofnouns and verbs confirmed that differentmechanisms may underlie the processing of theinflected forms of nouns and verbs. Patients havebeen described with greater difficulties either in theprocessing of nouns or in the processing of verbsinflections (Shapiro et al., 2000; Shapiro andCaramazza, 2003). Along the same lines, Shapiro etal. (2001) found that transcranial magneticstimulation of the left frontal cortex in healthysubjects differentially affected the processing ofverb inflection compared to the plural and singularform of regular nouns. Some neuroimaging studiesspecifically aimed at studying the grammaticalproperties of nouns found the involvement of frontalregions but only in the processing of grammaticalgender (Heim et al., 2002; Miceli et al., 2002). Inour study, the absence of activation of the left IFGin noun and adjective inflection conditions wasaccompanied by an activation of the middle frontalregions common to both processes. This activationmight reflect the areas that subserve the processingof the grammatical properties of nouns andadjectives and specifically of their plural forms.

As previously stated, only for verb inflection,we also found a partial activation of parietalregions. A plausible explanation for this resultcould be related to the fact that in Italian the pastparticiple of some verbs can sometimes also be an

Derivational processing 1103

adjective [e.g., pul-ito (cleaned)]. In this last case,the transformation of the verb infinitive into theadjective might be considered a derivationalprocess. Therefore, it cannot be excluded that thepast participle was sometimes processed as aderived form, leading to some partial activation ofthe same areas involved in derivational processing.

Right Fronto-Parietal Regions

In the derivation of nouns from verbs, we foundactivation of the right ventrolateral frontal andparietal regions as well as of the left fronto-parietalregions. These latter activations were absent in thecomparison with the inflectional task. In thederivation of nouns from adjectives only theactivation of the right parietal regions wassignificant.

Although we can not draw any definiteconclusion from the present data, it is interesting tonote the relation that exists between the observedactivations and the pattern of linguistic errorsdescribed in two recently reported right-braindamaged patients (Marangolo et al., 2003). In theabsence of any aphasic symptom, these patientspresented with a selective difficulty in derivationalmorphology. They were selectively unable toproduce nouns but only when they were derivedfrom verbs. Interestingly, these patients showed aright temporo-parietal lesion, which extendedsubcortically and frontally, and in one patient alsoinvolved the right superior frontal gyrus. In thestudy by Marangolo et al. (2003), we argued for apossible right hemisphere contribution to thederivation of nouns from verbs. We assumed thatthe activation of fronto-parietal circuits of bothhemispheres was involved in the response selectionprocesses necessary for deriving nouns from verbs.The present data seem to support that prediction.

Task Difficulty

It could be argued that the conditions whichactivated the left fronto-parietal cortex were, insome sense, more difficult relative to the others.An explanation in terms of task difficulty wouldpredict a different level of accuracy during theperformance of the three tasks; however, accuracydata from both behavioural experiments showedthat, for all participants, the number of errors wasnegligible. Secondly, in the first behaviouralexperiment, in which instructions were the same asfor the fMRI sessions, analysis of RTs alsoindicated that the three tasks were equally difficultfor the participants. Finally, in the secondbehavioural experiment, RTs were different amongword classes but the trend was not completelyconsistent with the pattern of the observedactivations. In fact, if the pattern of activationsreflected the pattern of RTs and, therefore, an effectof task difficulty, we should have found a

correspondence between fMRI data and RTs.However, while the fronto-parietal activation wasobserved for both noun-from-adjective and noun-from-verb derivation, both relative to inflection andto repetition, derivation RTs were longer thaninflection RTs but for the class of verb thisdifference was not significant. Although we cannottotally exclude that task difficulty didn’t play anyrole, the observed fronto-parietal activation seemednot to depend on the effect of this variable (in thiscase we would expect a difference between verbsand adjectives activations).

The similarity in the activation pattern fornoun-from-adjective and noun-from-verb derivationis also at odds with another possible explanation,i.e., that the activation is related to the number ofalternatives that, only for derivation, can begenerated from the base-word. In fact, since thenumber of plausible derived alternatives is greaterfor nouns derived from verbs [e.g., osserv-azione(observation) and osservatore (observer) fromosservare (to observe)] than from nouns derivedfrom adjectives [e.g., piccol-ezza (smallness) frompiccolo (small)], again, we should have founddifferent areas active for the two conditions, whichwas not the case.

As a further confirmation of the commonprocesses involved in the noun-from-verb andnoun-from-adjective conditions, results of theconjunction analysis confirmed that the left frontaland parietal regions were commonly activated inthe two derivational tasks. Finally, when we lookedat the task by grammatical class interaction, wefound no significant region showing a differentialactivation between noun-from-verb and noun-from-adjective derivation.

CONCLUSION

The derivation of nouns from verbs and fromadjectives activated a neural network that includesthe left frontal and parietal regions. The activationof frontal areas, and specifically of the left IFG,further confirms the importance of these structuresin morphological processing, and extends therelevance of these structures to the processing ofderived words. The new finding is that thederivational task, in addition to the activation ofareas which are common to inflectional processing,also activated frontal regions and parietal areas thatwere not active in the inflectional task. This resultsuggests that these areas may contribute to thelexical-semantic processes and the selectioncomponents involved in derivational morphology.

REFERENCES

ALEXANDER MP, HILTBRUNNER B and FISCHER RS. Distributedanatomy of transcortical sensory aphasia. Archives ofNeurology, 46: 885-892, 1989.

ALLEN M and BADECKER W. Morphology: The internal structure of

1104 Paola Marangolo and Others

words. In Rapp B (Ed), What Deficits Reveal about the HumanMind/Brain: A Handbook of Cognitive Neuropsychology.Philadelphia: Psychology Press, 2001.

BADECKER W and CARAMAZZA A. A lexical distinction betweeninflection and derivation. Linguistic Inquiry, 20: 108-116,1989.

BADECKER W and CARAMAZZA A. Morphological composition inthe lexical output system. Cognitive Neuropsychology, 8: 335-367, 1991.

BERTRAM R, SCHREUDER R and BAAYEN RH. The balance ofstorage and computation in morphological processing: Therole of word formation type, affixal homonymy, andproductivity. Journal of Experimental Psychology: Learning,Memory and Cognition, 26: 489-511, 2000.

BINDER RJ, FROST AJ, HAMMEKE TA, COX RW, RAO SM andPRIETO T. Human brain language areas identified by functionalmagnetic resonance Imaging. The Journal of Neuroscience,17: 353-362, 1997.

BRAINARD DH. The psychophysics toolbox. Spatial Vision, 10:433-436, 1997.

BURANI C and CARAMAZZA A. Representation and processing ofderived words. Language and Cognitive Processes, 2: 217-227, 1987.

BURANI C, DOVETTO FM, SPUNTARELLI A and THORNTON AM.Morpho-lexical access and naming: The semanticinterpretability of new root-suffix combinations. Brain andLanguage, 68: 333-339, 1999.

BURANI C and THORNTON AM. The interplay of root, suffix andwhole-word frequency in processing derived words. InBaayen H and Schreuder R (Eds), Morphological Structure inLanguage Processing. Berlin: Mouton de Gruyter, 2003.

CAPLAN D. Structure, Processing, and Disorders. Cambridge, MA:MIT Press, 1992.

CAPPA SF, PERANI D, SCHNUR T, TETTAMANTI M and FAZIO F. Theeffects of semantic category and knowledge-type on lexical-semantic access: A PET study. NeuroImage, 8: 350-359, 1998.

COLOMBO L and BURANI C. The influence of age of acquisition,root frequency and context availability in processing nounsand verbs. Brain and Language, 81: 398-411, 2002.

DAVIS MH, MEUNIER F and MARSLEN-WILSON WD. Neuralresponses to morphological, syntactic, and semantic propertiesof single words: An fMRI study. Brain and Language, 89:439-449, 2004.

DEMONET JF, CHOLLET F, RAMSAY S, CARDEBAT D, NESPOULOUSJL, WISE R, RASCOL A and FRACKOWIAK R. The anatomy ofphonological and semantic processing in normal subjects.Brain, 115: 1753-1768, 1992.

DESMOND JE, GABRIELI JD and GLOVER GH. Dissociation offrontal and cerebellar activity in a cognitive task: Evidence fora distinction between selection and search. NeuroImage, 7:368-376, 1998.

DI SCIULLO AM and WILLIAMS E. On the Definition of Word.Cambridge, MA: MIT Press, 1989.

FRISTON KJ, HOLMES A and WORSLEY KJ. How many subjectsconstitute a study? NeuroImage, 10: 1-5, 1999.

FRITH CD, FRISTON KJ, LIDDLE PF and FRACKOWIAK RS. A PETstudy of word finding. Neuropsychologia, 29: 1137-1148,1991.

GARRETT M. Production of speech: Observations from normal andpathological language use. In Ellis AW (Ed), Normality andPathology in Cognitive Functions. London: Academic Press,1982.

HEIM ST, OPITZ B and FRIEDERICI AD. Broca’s area in the humanbrain is involved in the selection of grammatical gender forlanguage production: Evidence from event-related functionalmagnetic resonance imaging. Neuroscience Letters, 328: 101-104, 2002.

ISTITUTO DI LINGUISTICA COMPUTAZIONALE DEL CNR DI PISA.Corpus di Italiano Scritto Contemporaneo, 1989.Unpublished manuscript.

KOSTIC A and KATZ L. Processing differences between nouns,adjectives, and verbs. Psychological Research, 49: 229-236,1987.

LAINE M, RINNE OJ, KRAUSE BJ, TERAS M and SIPILA H. Lefthemisphere activation during processing of morphologicallycomplex word forms in adults. Neuroscience Letters, 271: 85-88, 1999.

MARANGOLO P, INCOCCIA C, PIZZAMIGLIO L, SABATINI U,CASTRIOTA-SCANDERBEG A and BURANI C. The righthemisphere involvement in morphologically derived words.Journal of Cognitive Neuroscience, 15: 364-71, 2003.

MARSLEN-WILSON WD, TYLER LK, WAKSLER R and OLDER L.Morphology and meaning in the English mental lexicon.

Derivational processing 1105

Psychological Review, 101: 3-33, 1994.MARSLEN-WILSON WD and TYLER LK. Dissociating types of

mental computation. Nature, 387: 592-594, 1997.MAZZIOTTA JC, TOGA AW, EVANS A, FOX P and LANCASTER J. A

probabilistic atlas of the human brain: Theory and rationalefor its development. The International Consortium for BrainMapping (ICBM). NeuroImage, 2: 89-101, 1995.

MICELI G and CARAMAZZA A. Dissociation of inflectional andderivational morphology. Brain and Language, 35: 24-65,1988.

MICELI G, SILVERI MC, ROMANI C and CARAMAZZA A. Variationsin the pattern of omissions and substitutions of grammaticalmorphemes in the spontaneous speech of so-calledagrammatic patients. Brain and Language, 36: 447-492, 1989.

MICELI G, TURRIZIANI P, CALTAGIRONE C, CAPASSO R, TOMAIOLO Fand CARAMAZZA A. The neural correlates of grammaticalgender: An fMRI investigation. Journal of CognitiveNeuroscience, 15: 618-628, 2002.

MORO A, TETTAMANTI M, PERANI D, DONATI C, CAPPA SF andFAZIO F. Syntax and the brain: Disentangling grammar byselective anomalies. NeuroImage, 13: 110-118, 2001.

NATHANIEL-JAMES DA and FRITH CD. The role of the dorsolateralprefrontal cortex: Evidence from the effects of contextualconstraint in a sentence completion task. NeuroImage, 16:1094-1102, 2002.

NOVOA O and ARDILA A. Linguistic abilities in patients withprefrontal damage. Brain and Language, 30: 206-225, 1987.

OLDFIELD RC. The assessment and analysis of handedness: TheEdinburgh inventory. Neuropsychologia, 9: 97-113, 1971.

PANZERI M, SEMENZA C, FERRERI T and BUTTERWORTH B. Free useof derivational morphology in an Italian jargon aphasic. InNespoulous JL and Villiard P (Eds), Morphology, Phonology,and Aphasia. Berlin: Springer Verlag, 1990.

PELLI DG. The VideoToolbox software for visual psychophysics:Transforming numbers into movies. Spatial Vision, 10: 437-442, 1997.

PERANI D, CAPPA SF, SCHNUR T, TETTAMANTI M, COLLINA S, ROSAMM and FAZIO F. The neural correlates of verb and nounprocessing: A PET study. Brain, 122: 2337-2344, 1999.

PRICE CJ and FRISTON KJ. Cognitive conjunction: A new approachto brain activation experiments. NeuroImage, 5: 261-270, 1997.

ROBINSON G, BLAIR J and CIPOLOTTI L. Dynamic aphasia: Aninability to select among competing verbal responses? Brain,121: 77-89, 1998.

SCALISE S. Generative Morphology. Dordrecht: Foris, 1984.SHAPIRO KA and CARAMAZZA A. Grammatical processing of nouns

and verbs in left frontal cortex. Neuropsychologia, 41: 1189-1198, 2003.

SHAPIRO KA, PASCUAL-LEONE A, MOTTAGHY FM, GANGITANO Mand CARAMAZZA A. Grammatical distinctions in the left frontalcortex. Journal of Cognitive Neuroscience, 15: 713-720, 2001.

SHAPIRO KA, SHELTON J and CARAMAZZA A. Grammatical class inlexical production and morphological processing: Evidencefrom a case of fluent aphasia. Cognitive Neuropsychology, 17:665-682, 2000.

SILVERI MC and DI BETTA AM. Noun-verb dissociations in brain-damaged patients: Further evidence. Neurocase, 3: 477-488,1997.

SWAAB TY, BROWN C and HAGOORT P. Understanding ambiguous

words in sentence contexts: Electrophysiological evidence fordelayed contextual selection in Broca’s aphasia.Neuropsychologia, 36: 737-761, 1998.

THOMPSON-SCHILL SL, D’ESPOSITO M, AGUIRRE GK and FARAHMJ. Role of left inferior frontal cortex in retrieval of semanticknowledge: A reevaluation. Proceedings of the NationalAcademy of Sciences of the USA, 94: 14792-14797, 1997.

THOMPSON-SCHILL SL, D’ESPOSITO M and KAN IP. Effects ofrepetition and competition on activity in left prefrontal cortexduring word generation. Neuron, 23: 513-522, 1999.

TRAFICANTE D and BURANI C. Visual processing of Italian verbsand adjectives: The role of inflectional family size. In BaayenHR and Schreuder R (Eds), Morphological Structure inLanguage Processing. Berlin: Mouton de Gruyter, 2003.

TYLER LK, BRIGHT P, FLETCHER P and STAMATAKIS EA. Neuralprocessing of nouns and verbs: The role of inflectionalmorphology. Neuropsychologia, 42: 512-523, 2004.

TYLER LK and COBB H. Processing bound grammaticalmorphemes in context: The case of an aphasic patient.Language and Cognitive Processes, 2: 245-262, 1987.

TYLER LK, DEMORNAY-DAVIES P, ANOKHINA R, LONGWORTH C,RANDALL B and MARSLEN-WILSON WD. Dissociations inprocessing past tense morphology: Neuropathology andbehavioral studies. Journal of Cognitive Neuroscience, 14: 79-94, 2002.

TYLER LK and MARSLEN-WILSON WD. Disorders of combination:Processing of complex words. Brain and Language, 60: 40-48, 1997.

TZOURIO-MAZOYER N, LANDEAU B, PAPATHANASSIOU D, CRIVELLOF, ETARD O, DELCROIX N, MAZOYER B and JOLIOT M.Automated anatomical labeling of activations in SPM using amacroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage, 15: 273-289, 2002.

ULLMAN M. The declarative/procedural model of lexicon andgrammar. Journal of Psycholinguistic Research, 30: 37-69,2001.

ULLMAN M, CORKIN S, COPPOLA M, HICKOK M, GROWDON ,KOROSHETZ W and PINKER S. A neural dissociation withinlanguage: Evidence that the mental dictionary is part ofdeclarative memory and that grammatical rules are processedby the procedural system. Journal of Cognitive Neuroscience,9: 266-276, 1997.

VANNEST J, POLK TA and LEWIS RL. Decompositional Processingof Derived Words: A fMRI Study. New Evidence for a Dual-Route Model. Poster presented at The Cognitive NeuroscienceSociety Meeting, New York, 2003.

WARBURTON E, WISE RJ, PRICE CJ, WEILLERC, HADAR U, RAMSAYS and FRACKOWIAK R. Noun and verb retrieval by normalsubjects. Studies with PET. Brain, 119: 159-179, 1996.

WURM LH. Auditory processing of prefixed English words is bothcontinuous and decompositional. Journal of Memory andLanguage, 37: 438-461, 1997.

XIONG J, SHOBINI R, GAO JH, WOLDORFF M and FOX PT.Evaluation of hemispheric dominance for language usingfunctional MRI: A comparison with positron emissiontomography. Human Brain Mapping, 6: 42-58, 1998.

Paola Marangolo, Centro Ricerche di Neuropsicologia, IRCCS Fondazione SantaLucia, via Ardeatina 306 - 00179 Roma, Italia. e-mail: p.marangolo@hsantalucia.it

(Received 25 June 2004; reviewed 1 October 2004; revised 12 November 2004; accepted 28 April 2005; action editor:Roberto Cubelli)

1106 Paola Marangolo and Others

APPENDIX

Word stimuli auditorily presented in the fMRI and in the behavioural sessions. English translation in brackets

Verbs Adjectives Derived nouns

Assassinare (to murder) Onesto (honest) Pulizia (cleanliness)Accelerare (to accelerate) Vile (vile) Dominazione (domination)

Fallire (to fail) Saggio (wise) Evasione (evasion)Aprire (to open) Veloce (fast) Esposizione (exhibition)Liberare (to free) Oscuro (dark) Costruzione (construction)

Inaugurare (to inaugurate) Prudente (prudent) Dondolio (swinging)Aggredire (to assault) Indecente (indecent) Finzione (figment)Sfinire (to exhaust) Coerente (coherent) Miagolio (caterwaul)Gracidare (to croak) Frequente (frequent) Lavaggio (washing)

Confondere (to confound) Efficace (effective) Lucidatura (polishing)Imitare (to imitate) Superbo (superb) Navigazione (navigation)Guarire (to heal) Misero (wretched) Abitazione (habitation)Avviare (to start) Perfido (perfidious) Bruciatura (burn)

Organizzare (to organize) Villano (rude) Divertimento (fun)Immaginare (to imagine) Cortese (kind) Meditazione (meditation)Distruggere (to destroy) Avaro (avaricious) Giuramento (oath)

Sciogliere (to melt) Pigro (lazy) Atterraggio (landing)Cinguettare (to twitter) Furbo (cunning) Giustificazione (justification)

Illuminare (to illuminate) Denso (dense) Bombardamento (bombing)Sbalordire (to amaze) Morbido (soft) Flessione (flexion)

Soffrire (to suffer) Gioviale (jovial) Apertura (opening)Accendere (to light) Elegante (elegant) Punizione (punishment)Riparare (to repair) Ruvido (rough) Spiegazione (explanation)Spostare (to move) Debole (weak) Espressione (expression)Riempire (to fill in) Sciocco (silly) Illustrazione (illustration)

Occupare (to occupy) Violento (violent) Esclusione (exclusion)Smarrire (to mislay) Allegro (happy) Soddisfazione (satisfaction)Registrare (to record) Acido (acid) Uccisione (killing)Reclutare (to recruit) Negativo (negative) Irrigazione (irrigation)Resistere (to resist) Sensibile (sensitive) Licenziamento (firing)