Journal of Memory and Language 46,178–198 (2002)doi:10.1006/jmla.2001.2804, available online at http://www.academicpress.com on

0749© 200All rig

The Production of Noun Phrases in English and Spanish: Implications forthe Scope of Phonological Encoding in Speech Production

Albert Costa and Alfonso Caramazza

Harvard University

The role of the phonological word as a planning unit in the production of noun phrases (NPs) was investigatedin three picture–word interference experiments. We addressed this issue cross-linguistically by asking Spanish

dlogicallyming la-ardless

work of

para-

S

by NmenEduCosSpagrateFritzsistaGallSchn

AdCaraHarvFax:

and English speakers to produce simple (determiner 1 noun [in English]) and complex (determiner 1 adjective1 noun [in English] or determiner 1 noun 1 adjective [in Spanish]) NPs while ignoring phonologically relateor unrelated distractors. The results showed that naming latencies are faster when the distractor is phonorelated to the noun or to the adjective irrespective of the type of NP tested. The results suggest that NP natencies are affected by the level of activation of the phonological content of the lexical nodes of the NP, regof whether they belong to the first or second phonological word. The results are interpreted in the frametheories of phonological encoding.© 2001 Elsevier Science

Key Words:phonological encoding; speech production; phonological word; picture–word interferencedigm

peech production involves several process-

slttn

a

or tte

words. The processes engaged at each level ofanheed,ati-n-s atrs to thege.gi-calur-esr- by

ishp

ing stages that work concurrently over differetypes of representations in an incremental faion (e.g., Dell, 1986; Garrett, 1975; Leve1989). The two main levels of processing are grammatical and the phonological encodistages. At the grammatical encoding stage,speaker selects the appropriate words for thetended message and assigns them to a syntframe that specifies the grammatical relatioamong them. The construction of the syntacframe is itself driven in part by the selection the head lexical nodes of each phrase. Duphonological encoding the speaker retrievesphonological content of the previously selec

The research reported in this article was supported in

-596X/01 $35.001 Elsevier Sciencehts reserved.

17

ounce

ntheno-exi-usare en-oc-

IH Grant DC 04542, a grant from the Catalan Governt (1999SGR00083), and a grant from the Ministerio dcación y Cultura (DGES Project PB97-0977). Albeta was supported by a Postdoctoral Fellowship from tnish government (Fulbright program). The authors aful to Xavier Mayoral, Delia Kong, Ulrich Weger, Tom

sche, Angels Colome, and Mar Rodríguez, for their ance in running the experiments and to Nuria Sebasties, Kimiko Domoto-Reilly, Xavier Alario, and Tatianaur for their helpful suggestions on this work.dress correspondence and reprint requests to Alfonmazza, Department of Psychology, William James Haard University, 33 Kirkland St., Cambridge, MA 02138(617) 496-6262. E-mail: caram@wjh.harvard.edu.

8

nth-,heg

the in-ctic

nsticf

inghed

representation overlap in time, occurring in incremental fashion. That is, while parts of tutterance are being grammatically encodother parts, which have already been grammcally encoded, are being phonologically ecoded. The size or scope of the planning uniteach of these levels of representation appeadiffer. Here, we present evidence concerningscope of the phonological encoding staSpecifically, we study the role of the phonolocal word as the unit that governs phonologiencoding by exploring whether there is concrent phonological activation of lexical nodthat belong to different phonological words duing speech production. We address this issueanalyzing the naming performance of England Spanish speakers in the production of nphrases (NPs) in the picture–word interferenparadigm.

Until recently, the main source of informatioused to inform claims about the sizes of planning units at the grammatical and phological encoding stages was the analysis of lcal and phonological errors in spontaneospeech (e.g., Garrett, 1988). Lexical errors assumed to arise at the level of grammaticalcoding where the selection of lexical nodes curs. For example, in the lexical error “give the

art-e

rthere

s-an-

soll,.

F

-tinthtta

esle rt gi e

rd

h

ot

ein(oegicerh u

n

nden a

ben,ish

g-

otlex-l

g-fi-9,trdsear.-r-rts

anttos ofdgi-u-gi-tge

elt,c-alalve

ee9).ng

inhichlog-tal

ots

PRODUCTION O

baby to the banana” (from Meyer, 1996), thewords “baby” and “banana” have been transposed. Their misordering suggests that the lexical items are concurrently active at the poin time when the first of the two words is iserted into the syntactic frame. And, since two words belong to different phrases in tsame clause, it is reasonable to conclude thascope of grammatical encoding is, at least,clause and not smaller units such as the phrIn contrast, phonological errors such as “heftlemisphere” (from Fromkin, 1973), where thsounds /l/ and /h/ have been transposed, arithe level of phonological encoding. The ements that interact in this type of error tendinvolve sounds relatively close to one anothethe same phrase. Therefore, it appears thascope of processing of grammatical encodinlarger than that of phonological encoding. Whgrammatical encoding seems to encompassphrases included in a clause, phonological coding seems restricted to a smaller unit, phaps no more than two phonological wowithin a phrase.

Linguists have proposed several types prosodic units in order to account for tprosodic regularities observed in speech prodtion. In line with the hierarchical organization syntactic structures, it has been assumed these are also organized hierarchically. Amothe several phonological units that have bproposed in order to account for the productof prosody (phonological phrases, intonatiophrases, etc.), the phonological word prosodic word) has been considered as onthe minimal prosodic constituents. Althouthere are several definitions of the phonologword (e.g., Ferreira, 1993; Nespor & Vog1986; Selkirk, 1984, 1986; Wheeldon & Lahi1997) all of them agree in considering tphonological word as a stressed word plusthe unstressed words that cliticize to it. Becaof this property, some scholars, including Hay(1989) and Nespor and Vogel (1986), have aused the term “clitic group.” Since clitizatiomost often involves function words, one coualso define the phonological word in terms

grammatical classes, as any content word pother function words next to it. However, in

o-

NOUN PHRASES 179

wont-hee

thehese.

e at-

to inthe islethen-

er-s

ofeuc-fhatngenonalr ofhall,i,eallse

eslso

ldoflus

some contexts a function word may correspoto a whole phonological word by itself, as whthe function word is topicalized, and thereforedefinition in terms of stress values seem tomore appropriate. According to this definitiothe prosodic structure of the complex EnglNP “the red car” (determiner 1 adjective 1noun) consists of two clitic groups or phonoloical words (“The red” /də ɹεd/ and “car”/kaɹ/).This is because the lexical word “the” does nbear any stress and therefore clitizises to the ical word “red,” resulting in the phonologicaword /də ɹεd/.

From the processing point of view, phonoloical words are the domain over which syllabication rules are applied (e.g., Levelt, 1981992; Levelt & Wheeldon, 1994). In fluenspeech, the phonological properties of the woare adapted to the contexts in which they appConsider the indefinite determiner “a” in English. The phonological realization of this deteminer depends on whether the next word stawith a vowel (e.g., an elephant) or a conson(e.g., a tiger). The syllabification rules take inaccount the metrical and segmental propertiethe words that form the phonological word ancombine them to produce a series of phonolocal syllables that are then delivered to the articlator. It has also been argued that the phonolocal word is the minimal phonological unit thaneeds to be fully computed before the next staof processing can be engaged (e.g., Lev1989). Thus, in principle, it is possible that artiulation of the elements of the first phonologicword could proceed before the phonologicproperties of the second phonological word habeen retrieved (Wheeldon & Lahiri, 1997; salso Schriefers, de Ruiter, & Steigerwald, 199

However, the fact that segments that beloto two different phonological words interact speech errors suggests that, at the level at wthese errors occur, there is concurrent phonoical activation of lexical nodes from differenphonological words. For example, the lexicitems involved in the error “flow snurries” (for“snow flurries”; from Dell, 1986) belong to twdifferent phonological words, which suggesthat the phonological properties of both phon

logical words were concurrently active.

C

boh

m

yero.,v

ohn

tlusda

s

9 &

rr

e

ctingord

p-d to

staer).on-

andoc-ticing

calt-od-g,rsis-

ili-

ndn-ro-s;

ian-ij,

0;hecalo-cew

ri-b-ento

fro-

duced by the participants. This result was inter-preted as an indication that the second noun of

1 Throughout the paper we use the following notation:double quotation marks for responses,italics for stimuli

180 COSTA AND

Although the study of the regularities oserved in spontaneous speech errors has prto be a valuable tool for understanding tprocesses and representations involvedspeech production, it is not beyond criticisSome researchers have argued that speech ereveal how the normal speech production stem can occasionally get derailed and that infences to the structure of normal processes ferror patterns is not always transparent (eBock & Levelt, 1994; Levelt, Roelofs, & Meyer1999; Meyer, 1992). For example, the obsertion that sound errors involve segmental infomation of words belonging to different phonlogical words may reflect the outcome of tunusual situation in which more than ophonological word is encoded at the same timThat is, it could be argued that under normal ccumstances the scope of phonological encodis one phonological word and that errors arisethose nonnormal situations in which the prodution system is derailed and two phonologicwords are encoded at the same time. Onview, the data from speech errors would not ilminate the normal speech production procesis important, therefore, to complement the stuof slips of the tongue with methodologies thfocus on error-free speech production in ordto obtain converging evidence for specific theretical claims.

In recent years, the picture–word interferenparadigm has been used to investigate issuelated to the functional architecture and proceing dynamics of lexical access in speech pduction (e.g., Jescheniak & Schriefers, 19Schriefers, Meyer, & Levelt, 1990; StarreveldLa Heij, 1995, 1996). In this paradigm, particpants are asked to name a picture while ignoa distractor word (see MacLeod, 1991, for a view of Stroop-like tasks). By manipulating threlationship between the picture’s name anddistractor word, different effects are observThe two best-studied effects are the semantic in-terferenceeffect and the phonological facilita-tion effect. The semantic interference effestands for the observation that picture namlatencies are longer when the distractor w

and the picture are semantically (categoricalrelated than when they are not (e.g., Glaser

ARAMAZZA

-venein.

rrorss-r-m

g.,

a-r--eee.

ir-inginc-alhis-. Ityt

ero-

ce re-

ss-ro-8;

i-inge-ethed.

Glaser, 1989; Glaser & Düngelhoff, 1984; Luker, 1979). By contrast, phonologically relatedistractors speed up naming latencies relativephonologically unrelated distractors (e.g., Co& Sebastian-Galles, 1998; Lupker, 1982; Mey& Schriefers, 1991; Rayner & Springer, 1986Researchers have argued that these two ctrasting phenomena reflect the mechanisms representations at two different stages of pressing. The dominant view is that the semaninterference effect is the product of a delay the selection of the target lexical item duringrammatical encoding and that the phonologifacilitation effect is due to facilitation in selecing representations at the phonological encing stage (e.g., Levelt, Schriefers, VorberMeyer, Pechman, & Havinga, 1991; Schriefeet al., 1990; see also Starreveld, 2000, for a dcussion of the nature of the phonological factation effect in this paradigm).

Schriefers (1992, 1993) was the first to extethe picture–word interference paradigm to ivestigate lexical access in the context of NP pduction (see also Alario & Caramazza, in presCaramazza et al., in press; Costa, SebastGalles, Miozzo, & Caramazza, 1999; La HeMark, Sander, & Willeboordse, 1998; Miozzo &Caramazza, 1999; Schriefers & Teruel, 200van Berkum, 1997). Meyer (1996) extended tparadigm to address the size of the grammatiand phonological planning units in speech prduction. She asked Dutch speakers to produeither a conjunction noun phrase (“the arroand the bag”)1 or a simple sentence (“the arrowis next to the bag”), while ignoring a distractoword. The distractor word could be related ether to the first or to the second noun. Meyer oserved that naming latencies were slowed whthe distractor word was semantically relatedthe first (bombfor arrow) or to the second (sackfor bag) noun of the utterance, irrespective othe type of utterance (NPs or sentences) p

ly) &(pictures or words), and single quotation marks for lexicalrepresentations.

F

-—

el

ode

llthen

eh

vo

tnrsatnesnr

naoa.obth

-ndro-h”

ele

ase-ate-rdo-

-edsr,

hatesla--asof

call

ersealizer-s),

se oflo-NPra-ar-

hethnehee

PRODUCTION O

the utterance (bag) was concurrently active withthe first noun (arrow) at the level where the semantic interference effect is thought to occurthe lexical node selection stage. Interestingthe results observed with phonological distrators contrasted sharply: naming latencies waffected by distractors that were phonologicarelated to the first noun of the utterance (art forarrow) (see Miozzo & Caramazza, 1999, Expeiment 4, for similar results in Italian) but nowhen the distractors were related to the secnoun (ball for bag). This result was interpreteto indicate that the level of activation of thphonological properties of the second nounthe utterance (bag) does not affect the onset ospeaker’s articulation. The contrasting resuobtained with semantically and phonologicarelated distractors were taken as evidencethe scope of the grammatical encoding procis larger than that of the phonological encodiprocess.

The results of Meyer (1996) are consistwith the notion that the phonological word is tbasic unit of phonological encoding. This is bcause phonological facilitation arises whenethe primed word is included in the first phonlogical word (e.g.,arrow is included in the firsphonological word “the arrow”), but not whethe primed word is located outside the fiphonological word. This observation suggethat the level of activation of the phonologicproperties of the words that are not part of first phonological word does not affect namilatencies (e.g.,bag in the NP “the arrow and thbag”). These results fit well with the hypothethat the retrieval of the phonological segmeof the lexical nodes that comprise the fiphonological word must be completed befoarticulation of the phrase is initiated. They aalso consistent with the proposal that the phological properties of the lexical nodes that not part of the first phonological word do nneed to be selected before articulation st(e.g., Levelt & Maassen, 1981; Meyer, 1996)

However, recent results have questioned bthe notion that the phonological word has tofully processed before articulation starts and

assumption that the phonological propertieslexical items that do not belong to the firs

NOUN PHRASES 181

ly,c-rely

r-tnd

offltsyatssg

ntee-er-

sttsl

heg

istsstrereo-retrts

thee

of

phonological word do not affect naming latencies. Regarding the first issue, Schriefers aTeruel (1999) asked German speakers to pduce no-determiner noun phrases (“rosa Tisc[pink table]) while ignoring distractors that werphonologically related either to the first syllabof the adjective (ro for rosa) or to the first sylla-ble of the noun (tis for Tisch). Naming latencieswere speeded only when the distractor word wrelated to the first word of the utterance. This rsult by itself seems to support the notion thonly the first phonological word needs to bcomputed for articulation to start. This is because in the NPs used in this study the first woof the utterance corresponds to the first phonlogical word (e.g., “rosa”, [pink]), while the second word, the noun (e.g., “Tisch”), falls outsidof the first phonological word and corresponto the second phonological word. HoweveSchriefers and Teruel also reported a result tsuggests that the first phonological word donot need to be fully processed before articution starts. In their study, phonological facilitation was not observed when the distractor wphonologically related to the second syllablethe adjective (e.g.,safor rosa[pink]). This resultled the authors to conclude that the phonologiword may not be the lower limit of phonologicaencoding and that it is possible that speakstart articulation without having completed thphonological encoding of the first phonologicword. These authors conclude then that the sof phonological encoding may be, in certain cicumstances (not specified by the authorsmaller than the first phonological word.

More importantly for our purposes are thoresults that suggest that the level of activationthe phonological elements of lexical nodes cated in the second phonological word of the may affect naming latencies. Miozzo and Camazza (1999, Experiment 5) asked Italian pticipants to produce determiner 1 adjective 1noun NPs in a simple picture naming task. Tdefinite determiner in Italian depends on bothe gender of the noun and, for masculinouns, on the phonological properties of tword that follows it. Masculine nouns take thdeterminer “lo” when the following word’s

tonset consists of a vowel, a consonant cluster of

C

n

e

o

f

h

ahnes

ains

diobe

indtr

heohrth

jeo

itb-ednteste.aimthe

d(inetscti-lals-chednt

cechre

nollr,hatn-ds- atire

ar-r

toehet--

ishon--

in tode-de

182 COSTA AND

the form “s1 consonant” or “gn,” or an af-fricate; in all other contexts masculine noutake the determiner “il.” Thus, for example, fothe masculine word “scoiattolo” [squirrel] thdeterminer is “lo” (“lo scoiattolo” [the squirrel])but it is “il” if the adjective “grande” [big] is in-terposed between the determiner and the n(“il grande scoiattolo” [the big squirrel]). Therefore, in order to select the proper determiner masculine nouns, the phonology of the neword must be available to the selection mecnism. Miozzo and Caramazza took advantagethis property to investigate what phonologicinformation is available at the point at which tform of a determiner is selected. They reasothat if the only phonological information activat the moment at which the determiner is lected were that of the word immediately folowing it (the adjective), then the phonologicproperties of the noun should not affect namtimes since this information could not serve abasis for interference in determiner selectioHowever, if the phonology of the noun were ative at the point at which determiner forms aselected, there should be a cost in naming timwhen the phonology of the noun provides infomation conflicting with that provided by the ajective. Thus, in the latter case, the productof phrases such as “il grande scoiattolo” [the squirrel] should be slower relative to phrassuch as “il grande tavolo” [the big table]. Thisbecause the initial segments of “grande” a“scoiattolo” require the determiners “il” an“lo,” respectively, whereas the initial segmenof “grande” and “tavolo” both require the deteminer “il.” The results were clear: slower naming latencies were observed when the phonogy of the adjective and the phonology of tnoun in the NP required different determinforms (“il grande scoiatolo”) in comparison tthe condition where the two words require tsame determiner (“il grande tavolo”). These sults suggest that the phonological form of noun is active at the point at which determinform selection takes place. Similar results habeen observed in French with possessive adtives and demonstrative pronouns whose gder-marked forms also depend on the phonol

of the following word (Alario & Caramazza, in

ARAMAZZA

sr

un-orxta- ofl

eed

e-l-lg an.c-reesr--nigs

sd

s--ol-er

ee-e

erveec-n-gy

press). Also, Roelofs (1998), using the implicpriming paradigm (see Meyer, 1990, 1991), oserved phonological activation of words locatin the second phonological word. Finally, receresults by Ferreira and Swets (in press) suggthat processing difficulty of material beyond thfirst phonological word affect initiation timesOn the basis of these results, the authors clthat the system requires planning beyond initial phonological word.

Taken together, the results from Miozzo anCaramazza (1999), Alario and Caramazzapress), Roelofs (1998), and Ferreira and Sw(in press) suggest that there is concurrent avation of phonological information of lexicanodes that belong to different phonologicwords. Interestingly, this conclusion is consitent with the data from spontaneous speeerrors, where quite often the elements involvin sound exchange errors belong to differephonological words.

As reviewed here, the chronometric evidenon the role of the phonological word in speeproduction is somewhat contradictory. There athree main sets of results. One set showsphonological activation of elements that faoutside the first phonological word (Meye1996). Another set of results demonstrates tthere is concurrent phonological activation ecompassing at least two phonological wor(Alario & Caramazza, in press; Miozzo & Caramazza, 1999; Roelofs, 1998). Finally, there isdata set that suggests that encoding of the enphonological word is not necessary to begin ticulation (Schriefers & Teruel, 1999). Furthediscussion of this third possibility is deferred the General Discussion. In the following, wfocus on the extent to which the activation of tphonological form of lexical nodes that fall ouside of the first phonological word affect naming latencies.

We address this question by asking Spanand English speakers to produce NPs in the ctext of a picture–word interference task. An important difference in the structure of NPs these two languages affords the opportunityaddress the role of grammatical structure in termining whether lexical nodes that fall outsi

the first phonological word are concurrently ac-

F

l

t

vef

fr-

PRODUCTION O

tive with it. In the experiments we report, walso included semantically related distractoThis distractor condition was included primarito assess the sensitivity of our task to lexical terference. Semantic interference is a highly liable effect in the picture–word interferencparadigm and its presence indicates that the

has the sensitivity to reveal effects of the mag

v

nr,

egrlaaohu

hnen

),r

c””a.gy

nicalr-96;

theea-P-

in-g- &a,

o-o--

g.,

ematli-itsver

-byzanc-

ftsreetal

groups.

2 Starreveld (2000) has recently argued that the phonolog-ical facilitation effect may reveal the ease with which theword-form representation of the target is retrieved ratherthan the ease with which the segmental information of thatword is retrieved (Meyer & Schriefers, 1991). Our predic-

tude of the semantic interference effect.

EXPERIMENT 1: SEMANTIC ANDPHONOLOGICAL EFFECTS FOR THENOUN IN ENGLISH NP PRODUCTION

In this experiment, three groups of natiEnglish speakers were asked to name a seriecolored pictures using one of three differetypes of utterances: (a) bare nouns (e.g., “ca(b) determiner1 noun NPs (e.g., “the car”)and (c) determiner1 adjective 1 noun NPs(e.g., “the red car”). The pictures were prsented with semantically related, phonolocally related, and unrelated control distractoOur main goal was to test whether namingtencies are affected by distractor words thatphonologically related to the noun when the psition of the noun varies between being in tfirst, second, or third position in the NP. Crcially, and given the standard definition of thphonological word given above, whether tnoun belongs to the first or to the second phological word varies across these different typof utterances. In the bare noun (e.g., “car”) adeterminer1 noun conditions (e.g., “the car”the noun belongs to the first phonological wo(“car”/kaɹ/ and “the car” /də kaɹ/ respectively),while in the determiner1 adjective 1 nouncondition it belongs to the second phonologiword (e.g., first phonological word: “The red/də ɹεd/; second phonological word: “car/kaɹ/). The question is whether phonologicdistractors related to the noun of the NP (ecap) will affect naming latencies differentlthan unrelated distractors (e.g.,pen) in all threeNP conditions.

We expect to find phonological facilitatiowhen the noun is placed in the first phonologword even if it is not the first word in the utteance, replicating previous results (Meyer, 19

Miozzo & Caramazza, 1999; but see Schriefe

NOUN PHRASES 183

ers.yin-re-easkni-

es oft

”),

-i-s.-re-e-eeo-sd

d

al

l.,

& Teruel, 1999). Therefore, we should obserphonological facilitation in the first two types oNPs (e.g., bare noun, “car”, and determiner 1noun, “the car”).

The prediction regarding the effects ophonologically related distractors for the deteminer 1 adjective 1 noun NPs is more compli-cated. If naming latencies are independent of level of activation of the words included in thsecond phonological word, the phonological fcilitation effect should not arise in complex Nproduction. In contrast, if phonological encoding comprises more elements than those cluded in the first phonological word, as sugested by the results of several studies (AlarioCaramazza, in press; Miozzo & Caramazz1999; Roelofs, 1998), it is possible for phonlogically related distractors to produce phonlogical facilitation even when the noun is located outside the first phonological word (e.determiner 1 adjective 1 noun; “the red car”/də ɹεd/ /kaɹ/). And, on the assumption that thphonological effects observed in this paradigindicate phonological activation of the word this related to the distractor, a phonological facitation effect for the noun would suggest that phonological properties are concurrently actiwith the phonological encoding of the otheparts of the utterance.2 We also included a semantically related condition, as was done Meyer (1996) and Miozzo and Caramaz(1999), in order to provide more informatioabout the effects of semantically related distrators in the production of NPs.

Method

Participants. Fifty-one native speakers oEnglish took part in the experiment. Participanwere students at Harvard University and wepaid for their participation. Participants werrandomly assigned to the three experimen

rstions do not depend on which of these two explanations iscorrect, and therefore we will not pursue this issue.

C

th

lo ae

gtrtt

rehulao

t

alat

eortwsarti

en voe

lsreuliualn-of nontm-.

u-in-

as ei-

in-irs

thetedirterndadt

theofaspro-ti-se

thee- byey,eresion

rethatri-

er-bal(c)out-via-re-rried

184 COSTA AND

Material. We selected 28 pictures wimonomorphemic names (see Appendix A). Tpictures were presented in three different co(red, green, or blue), but each picture alwayspeared in the same color. The pictures wpaired with five different distractors: (a) a smantically related distractor, (b) a phonolocally related distractor, (c) an unrelated condistractor matched on various variables with semantically related distractors, (d) an unreladistractor matched with the phonologically lated distractors, or (e) a string of XXXs. Tpictures and the distractors were paired sthat each distractor appeared once in the recondition and once in the unrelated conditiFor example, the picture of a horse appearedwith the semantically related distractor donkeyand with the unrelated distractor spoon. Thesame distractors appeared with the picture fork.However, for this picture the word donkeyserved as the unrelated distractor and the wspoonas the semantically related distractor. Tsame matching procedure was used for phonologically related distractors and their urelated controls. This design allows us to confor possible effects of individual distractors thmight affect the results in any systematic wother than their semantic or phonological reedness with the pictures’ names. The distracpaired with a given picture had similar frequecies (Francis & Kuc˘era, 1982). Care was taketo avoid any phonological or semantic relatness between the distractor words and the cname of each picture. The phonologically lated distractors shared at least their first segments with the pictures’ names. Twenty-pictures were selected as fillers and were presented five times each with unrelated disttors. The fillers were also used as warmup suli in the first three trials of each block.

The distractor words were shown in uppcase letters (Helvetica font, bold, 27 point) awere superimposed on the pictures. Picturespeared in the center of the screen. To preparticipants from anticipating a distractor’s psition, word position varied randomly in the rgion around fixation with a maximum variatioof 1.5 cm. For a given picture, however, the d

tractors always appeared in the same locati

ARAMAZZA

hersp-re

e-i-olheed-

echtedn.

ordhethen-rolaty,t-

orsn-nd-lor

e-o

ixlsoac-m-

r-d

ap-ent--

nis-

Stimuli were presented in 5 blocks of 54 tria(28 experimental plus 26 fillers). Each pictuappeared once per block. In each block, stimof the various conditions appeared an eqnumber of times (5 or 6). Block trials were radomized with the restriction that distractors the same experimental condition appeared inmore than two consecutive trials. Five differeblock orders were constructed, and similar nuber of participants randomly assigned to each

Procedure. Participants were tested individally in a sound-attenuated room. They were structed to name the picture and its colorquickly and as accurately as possible usingther a bare noun, a determiner 1 noun NP or adeterminer 1 adjective 1 noun NP, dependingon the group of participants. They were formed that they would see picture–word paand were asked to ignore the words. Beforeexperiment proper, participants were presenwith the entire set of pictures along with thedesignated names. Next, the experimenshowed all the pictures without their names aparticipants named the stimuli. Each trial hthe following structure. First, a fixation poin(an asterisk) was shown in the center of screen for 1 s, followed by a blank interval 500 ms. Then, the picture–word pairing wpresented for 600 ms. If a response was not vided within 3 s, the next trial started automacally. The intertrial interval was 1.5 s. Responlatencies were measured from the onset ofstimulus to the beginning of the naming rsponse. Stimulus presentation was controlledthe program Psyscope (Cohen, MacWhinnFlatt & Provost, 1993). Response latencies wmeasured by means of a voice key. The seslasted approximately 45 min.

Analyses. Three types of responses wescored as errors: (a) production of names differed from those designated by the expementer, (b) verbal dysfluencies (stuttering, uttance repairs, and production of nonversounds that triggered the voice key), and recording failures. Erroneous responses and liers (i.e., responses exceeding 3 standard detions) were excluded from the analyses of sponse latencies. Separate analyses were ca

on.out with subjects and items as dependent vari-

F

.e--tinuarahar

.5om bo

n

llhif

r-teder-we

tedse-

-erllsoors

er-

nd

-

ithc-as

-

PRODUCTION O

ables, yielding F1 and F2 statistics, respectivelyThree variables were analyzed: Type of Uttance (Bare Noun, determiner 1 noun NP, determiner 1 adjective 1 noun NP), Semantic Relationship (Semantically Related vs SemanUnrelated), and Phonological Relationsh(Phonologically Related vs Phonologically Urelated). The first variable was a between-sjects variable and the others within-subject vables. The results of the error analyses reported only if significant. Table 1 shows tmean response latencies and error rates function of type of distractor and type of utteance.

Results

Erroneous responses were observed on 7of the trials. The naming latencies derived frone picture were excluded from the analysescause it elicited a high percentage of err(more than 14%).

Semantically related distractors. The maineffect of semantic relationship was significa[F1(1,48) 5 26.3, MSE 5 579.2, p , .001;F2(1,26) 5 8.7, MSE 5 3288.0, p , .007]:naming latencies were slower with semanticarelated than with unrelated distractors. Tmain effect of type of utterance was also signicant [F1(2,48) 5 5.1, MSE 5 23094.8,p ,.009;F2(2,52) 80.0,MSE5 2405.5,p , .001].

Importantly, the interaction between the two p-

Note. Noun phrase production in English; distractors rel

NOUN PHRASES 185

r-

icp-b-i-rees a-

%

e-rs

t

ye-

Fs , 1), suggesting that the semantic interfeence effect created by the semantically reladistractors was independent of the type of uttance produced by the participants. Whenconsider determiner1 adjective1 noun utter-ances alone, we find that semantically reladistractors produce more interference thatmantically unrelated distractors [F1(1,16) 56.7, MSE 5 519.5,p , .019; F2(1,26)5 3.5,MSE5 1532.6,p , .07]. Both semantically related and unrelated distractors led to slownaming latencies than the XXX condition (aps, .014). Semantically related distractors alproduced more errors than unrelated distract[F1(1,48) 5 5.9, MSE 5 2.6, p , .018;F2(1,26)5 3.05,MSE5 3.2,p , .09]. This ef-fect was also independent of the type of uttance (Fs , 1). Finally, the XXX condition ledto fewer errors than the semantically related aunrelated conditions (allps , .012).

Phonologically related distractors. The maineffect of phonological relationship was significant [F1(1,48)5 32.5,MSE5 707.5,p , .001;F2(1,26)5 11.2,MSE5 3310.6,p , .002], in-dicating that naming latencies were faster wphonologically related than unrelated distrators. The main effect of type of utterance walso significant [F1(2,48) 5 3.9, MSE 524180.2,p , .02; F2(2,52) 5 44.0, MSE 53263.1,p , .001]. More importantly, the interaction between the two factors did not even a

0.27.36.48.6

factors did not even approach significance (bothproach significance (bothFs , 1), suggesting

TABLE 1

Naming latencies (in Milliseconds) by type of Distractor and Type of Utterance in Experiment 1

Type of utterance

Bare noun Det 1 noun Det 1 adj 1 noun Total

Type of distractors Mean SD E% Mean SD E% Mean SD E% Mean E%

Semantically related 788 100 10.2 669 101 10.9 715 138 9.5 724 1Semantically unrelated 761 95 6.7 644 83 6.3 694 126 8.9 699Phonologically related 729 104 5.0 629 102 6.8 680 119 7.4 679Phonologically unrelated 768 111 6.3 657 102 9.2 704 128 10.2 709XXXXs 704 105 4.3 604 81 4.4 668 111 5.6 658 4.8Semantic effects 227 225 221 225

(unrelated-related)Phonological effects 39 28 24 30

(unrelated-related)

ated to the Noun.

C

ise

oteot

e

ion

a

wnir

ta

h

teah

e

rg&ocet

he

-r-

sttscting,nine-

zza-

here

cd

flo-

dg-e-y

to sno-the

eart. 1.thhers:

186 COSTA AND

that the effects of phonologically related dtractors are similar in the three types of uttances. When we consider determiner1 adjec-tive 1 noun utterances alone, we find somevidence that phonologically related distractled to faster naming latencies than unreladistractors. The 24-ms difference between cditions was only significant in the subjecanalyses [F1(1,16)5 6.0, MSE 5 802.1,p ,.026;F2(1,26)5 2317.6,MSE5 2.7,p , .11].Naming latencies in the XXX condition werfaster than in the phonologically related and urelated conditions (allps , .011). Participantsmade more errors in the unrelated conditthan in the phonologically related conditio[F1(1,48) 5 4.9, MSE 5 1.8, p , .032;F2(1,26)5 2.9, MSE5 1.52,p , .091]. Thiseffect was independent of the type of utteran(Fs , 1). Error rates in the XXX conditionwere lower than in the unrelated condition (ps , .010).

Discussion

The results of this experiment showed tmain findings. First, distractors that are semacally related to the noun slow naming latencmore than unrelated distractors. Second, disttors that are phonologically related to the nolead to faster naming latencies than unreladistractors. Furthermore, these two effects independent of the type of NP that speakersto produce.

The semantic interference effect and phonological facilitation effect were expectwhen participants named the pictures with bnouns (e.g., Lupker, 1979, 1982). Similarly, tsemantic interference effect observed in theterminer 1 noun and determiner 1 adjective 1noun conditions was expected since the scopgrammatical encoding has been shown to beclause (Meyer, 1996).

The phonological facilitation effect for deteminer 1 noun NPs replicates previous findinin Dutch (Meyer, 1996) and Italian (Miozzo Caramazza, 1999; Experiment 4). Furthermit extends the observation of phonological faitation effects for the second word of the uttance to a language in which the first elemen

the NP (the determiner) is independent of t

ARAMAZZA

-r-

ersd

n-s

n-

n

ce

ll

oti-esac-uned

grammatical and phonological properties of tnoun, as opposed to Dutch and Italian.

The most critical result for us is the phonological facilitation effect observed in the deteminer 1 adjective 1 noun NP condition, inwhich the noun was located outside the firphonological word. Assuming that the effecof phonologically related distractors refleprocesses at the stage of phonological encodthe results imply that the phonology of the nouin an NP is activated even when it is locatedthe second phonological word. The results rported here and those of Miozzo and Carama(1999, Experiment 5; see also Alario & Caramazza, in press) invite the inference that tscope of phonological encoding extends mothan the first phonological word.

However, before going into more specificlaims about the implications of the observeresults it is important to test the reliability othe phonological facilitation effect for lexicanodes that are located outside the first phonlogical word. In the following experiment wereplicate the conditions where we obtainephonological facilitation in Experiment 1, usindifferent materials and a slightly different design. This is especially relevant given that thphonological facilitation effect for the determiner 1 adjective1 noun utterances was onl

read

hedree

de-

ofthe

-s

re,il-r- of

significant in the analysis by subjects.

EXPERIMENT 2: PHONOLOGICALEFFECTS FOR THE NOUN IN ENGLISH NP

PRODUCTION

In this experiment participants were askedproduce complex NPs of the form determiner1adjective 1 noun in English. As in the previouexperiment, we analyzed the effects of phologically related distractors to the noun on onset of naming latencies.

Method

Participants. Twenty participants from thsame population as in Experiment 1 took pNone of them had participated in Experiment

Material. We selected 40 pictures wimonomorphemic names (see Appendix B). Tpictures were presented in four different colo

hegreen, red, blue, and purple. The pictures were

F

ra

oau

eortw

nc

u

o

aai ikiaa

n

ho

r

e

ted

is-d

sePndeion

gestnds

a-nts

ed of

al-ical sec-erystheriesail-rase

eri-log-m- inmi- isEx- the

u-tionofisti-

meitsheti-of

er-

PRODUCTION O

paired with three different distractors: (a)phonologically related distractor, (b) an unlated control distractor matched on various vables with the phonologically related distrators, or (c) an unrelated filler distractor. As Experiment 1, the pictures and the distractwere paired such that each distractor appeonce in the related condition and once in the related condition. The distractors paired withgiven picture had similar frequencies. Care wtaken to avoid phonological or semantic relatness between the distractor words and the cname of each picture. The phonologically lated distractors shared at least their first segments with the pictures’ names. Four ptures were selected as fillers and were preseas warmup stimuli in the first four trials of eablock along with unrelated distractors. All thpictures were presented before the experimproper along with unrelated distractors. Theach picture appeared four times (once in training phase and three times in the experimproper). Each picture appeared in each of four colors included in the experiment. The cors were distributed within the different condtions a similar number of times.

Stimuli were presented in 4 blocks of 44 tri(40 experimental plus 4 fillers). Each picture peared once per block. In each block, stimulthe three different conditions appeared a simnumber of times (13 or 14). Trials within blocwere randomized with the restriction that dtractors of the same experimental condition peared in no more than two consecutive triTen different block orders were constructed atwo participants were randomly assigned each order. The other details of the experimedesign and procedure were similar to thoseExperiment 1.

Results and Discussion

Following the same criteria as in Experime1, 12.6% of the trials were excluded from tanalyses. The reaction times derived from participant and from one picture were discardfrom the analyses due to their large numbeerrors (more than 16%).

Naming latencies were significantly fast

when the distractors were phonologically re

NOUN PHRASES 187

ae-ri-c-inrsredn-

aasd-lor

e-o

ic-tedheents,

theentthel-i-

lsp-oflarss-p-ls.ndtot’s

of

nteneed of

r

lated (865 ms) than when they were unrela[887 ms;F1(1,18)5 5.8 , MSE5 820.8,p ,.027;F2(1,38)5 4.2,MSE5 1979.7,p , .046].The error rates in the two conditions were stattically identical (13.2 and 12.1 for related anunrelated distractors respectively, bothFs , 1).

The results of this experiment replicate thoobserved in Experiment 1 for the complex Nproduction group. In both studies, a reliable arobust phonological facilitation effect for thnoun of the NP was observed in the productof complex NPs (determiner 1 adjective 1noun). As discussed above, these results sugthat the scope of phonological encoding extebeyond the first phonological word.

However, such a conclusion may be premture. The type of NPs used in these experimeallows two interpretations of the observphonological effects with respect to the scopephonological encoding. One possibility, as ready noted, is that the scope of phonologprocessing encompasses at least part of theond phonological word of the NP. The othpossibility is that phonological encoding alwaincludes the head of the phrase plus any owords located before it. These two possibilitare indistinguishable on the basis of the avable data. This is because the head of the phin the English NPs used in the previous expments always belonged to the second phonoical word and therefore word position and gramatical class are conflated. That is, sinceEnglish the adjective is always located prenonally, the third element of a three-word NPnecessarily a noun. Therefore the results of periments 1 and 2 are consistent with both ofalternatives proposed above.

Note that the latter hypothesis is not implasible. It may be argued that, because selecof the noun is necessary for the constructionan NP frame, the lexical node of the nounthe first element selected during the grammacal encoding of a phrase. If we further assuthat after the selection of a lexical nodephonology is activated automatically, then tphonological properties of the noun are acvated before the selection of other elementsthe NP (determiner and adjective, in our exp

-iments). According to this idea, even though

C

ehiea

-

leehb

lyhe-s

ei

is ir mg

onho

-wen

h,idso

head of the NP and that are located outside the

to

h innishsi-i- is

exi-o-

(in doforthat re-on-alg- in-ono-al

n-rt

dit.ofin-ic-se-ntedtorlor

—

or for

ac-Thepre-

188 COSTA AND

the phonological content of the noun will bproduced after that of the determiner and tadjective, it actually becomes activated earlthan that of the other two items. If this explantion is correct, then our results would be completely compatible with those obtained bMeyer (1996) in which no phonological facilitation was observed when the distractor wowas related to the second noun in a compNP (e.g., “the arrow and the car”). In this casthe first noun of the NP may be acting as thead of the NP while the second noun couldacting as the head of a second NP. Thereforeis possible that phonological facilitation is onobserved for the head of the first but not for thead of the second NP, and this is why Mey(1996) did not find any phonological facilitation for the words placed outside the firphonological word.

In Experiment 3 we try to adjudicate betwethe two explanations described above by askSpanish participants to produce determiner1noun 1 adjective NPs. Unlike English, color adjectives are usually placed postnominally Spanish (e.g., “La casa azul” [literally, the houblue]). Therefore, a comparison of the effectsphonological distractors related to the thword of Spanish and English NPs allows usdistinguish between word position and gramatical class—the third word is a noun in Enlish NPs but is an adjective in Spanish NPswhile keeping constant their position in termsphonological words (e.g., always the secophonological word). This is because tprosodic structure of these Spanish NPs ctains two phonological words (“la casa”: /lakasa/ and “azul”: /aɵul/). If the scope of phonological encoding encompasses the first tphonological words of the NP regardless of thgrammatical class, we should observe a phological facilitation effect when the distractoword is related to the lexical node corresponing to the second phonological word (in tSpanish case, the adjective, “azul”). Howeverthe phonological facilitation effect observed English reflects the special status of the heathe NP (the noun) during phonological proceing, phonological facilitation effects should n

be observed for lexical items that are not t

ARAMAZZA

h

er--

y

rdx,ee

, it

er

t

nng

-neofdto--

—fden-

oiro-

rd-eifn ofs-t

first phonological word.

EXPERIMENT 3: SEMANTIC ANDPHONOLOGICAL EFFECTS IN THE

PRODUCTION OF SPANISH NPS

In this experiment we asked participantsproduce determiner 1 noun 1 adjective NPs inSpanish. As described above, and althougsome cases (mostly in poetic contexts) Spacolor adjectives may occupy a prenominal potion, they typically occur in postnominal postion and therefore the third element of the NPan adjective and not a noun. Therefore the lcal item that now falls outside the first phonlogical word is the adjective, whereas beforeExperiments 1 and 2) it was the noun. If wenot observe a phonological facilitation effect the adjective in Spanish, we can conclude the results obtained in Experiments 1 and 2flect the special status of the noun in the cstruction of the NP. By contrast, phonologicfacilitation effects for the adjective would sugest that the scope of phonological encodingcludes the element located in the second phlogical word regardless of its grammaticproperties.

Method

Participants. Thirty native speakers of Spaish from the University of Barcelona took pain the experiment in exchange for course cre

Materials. The design was similar to that Experiment 1. Two sets of materials were cluded (see Appendix C). In the first set, 26 ptures with monomorphemic names were lected. Each of these pictures was presewith three different distractors: (a) a distracsemantically related to the adjective—a coname (e.g.,gris [gray] for azul[blue]), (b) a dis-tractor semantically unrelated to the adjectivean unrelated adjective (e.g.,ideal [ideal] for azul[blue]), or (c) a string of XXXs. Each distractin this set was presented four times (exceptthe distractors rosa [pink] and eficaz[efficient],which were presented six times). All the distrtors in these two conditions were adjectives. 26 pictures in the second set of items were

esented with three different distractors each: (a) a

F

iv

iv

y

m

,m

a

alerec

pectern

wwe

s a

nt

llye-

ro-s-

thed

-

iths-hes-

r-e-toofd

ar-

r

PRODUCTION O

distractor phonologically related to the adject(e.g.,azucar[sugar] for azul [blue]), (b) a dis-tractor phonologically unrelated to the adject(e.g., drama [drama] for azul [blue]), or (c) astring of XXXs. These distractors were alwanouns and were presented only once.

The distractors included in each set had silar frequencies [Set 1: semantically related5161, semantically unrelated5 158 (F , 1); Set2: phonologically related5 152, phonologicallyunrelated5 155 (F , 1)] (Sebastian, MartiCuetos, & Carreiras, 1996). They also had coparable numbers of letters [Set 1: semanticrelated 5 5.8, semantically unrelated5 6.5(F , 1); Set 2: phonologically related5 5.5,phonologically unrelated5 5.2; (F , 1)]. Thir-teen pictures were selected as filler pictureswere presented three times each. The filwere also used as warmup stimuli in the fithree trials of each block. Stimuli were prsented in 3 blocks of 52 experimental trials ea(26 pictures from Set 1 and 26 pictures from S2). Each picture appeared once per block.each block, stimuli of the various conditions apeared a similar number of times (eight or ninBlock trials were randomized with the restrition that distractors of the same experimencondition appeared in no more than two consutive trials. Three different block orders weconstructed, and similar numbers of participawere assigned to each of them.

The pictures were presented in one of thdifferent colors: blue (azul), brown (marronand green (verde). However, each picture always presented in the same color. Care taken to avoid phonological or semantic relatness between the distractor words and the ture’s name. Approximately half of the picturenames were masculine; the remainder w

feminine. The other details of the design a

hnt

procedure are similar to those in Experiment

RESULTS AND DISCUSSION

The scoring criteria used in Experimentwere also used in this experiment, leading toexclusion of 9.6% of the data points. In tanalysis we considered the effect of semaand phonological relatedness of the distrac

to the adjective in the NP. Table 2 shows t

NOUN PHRASES 189

e

e

s

i-

-lly

ndrs

st-h

etIn-).-alc-e

mean response latencies and error rates afunction of type of distractor.

Semantically related distractors. The maineffect of semantic relationship was significa[F1(1,29) 5 10.7, MSE 5 994.6, p , .003;F2(1,25) 5 7.1, MSE 5 1292.2, p , .013]:naming latencies were slower with semanticarelated than with unrelated distractors. Both smantically related and unrelated distractors pduced longer naming latencies than XXX ditractors (all ps , .001). The only significantdifference in the error analyses was between XXX condition and the semantically relatecondition (all ps , .01)

Phonologically related distractors. The maineffect of phonological relationship was significant [F1(1,29) 5 4.5,MSE5 871.5,p , .041;F2(1,25) 5 6.1,MSE5 728.9,p , .020], indi-cating that naming latencies were faster wphonologically related than with unrelated ditractors. Naming latencies were faster in tXXX condition than in the other two condition(all ps , .001). No significant effects were observed in the error analysis.

Two main results were obtained in this expeiment. First, semantically related distractors dlayed the production of the NP in comparisonunrelated distractors, replicating the resultsExperiment 1. Second, phonologically relate

ts

ree),asasd-

pic-’serend1.

1theetic

ors

distractors sped up naming latencies in comp

TABLE 2

Naming Latencies (in Milliseconds) by Type of Distractoand Set of Pictures in Experiment 3

Type of distractors Mean SD E%

SET 1Semantically related 755 124 12.7Semantically unrelated 728 112 10.0XXXXs 703 106 7.6

SET 2Phonologically related 735 109 10.1Phonologically unrelated 752 123 9.4XXXXs 694 109 8.2

Semantic effects 227(unrelated-related)

Phonological effects 17(unrelated-related)

Note. Noun phrase production in Spanish, with distrac-

hetors related to the Adjective.

C

ua

)o

at

d.ultsno-e-P

ectherd

am-si-ns

190 COSTA AND

ison to unrelated distractors. These results sgest that naming latencies are affected by thetivation level of the phonological properties othe adjective in the production of complex NPwith the structure “determiner1 noun1 adjec-tive” (la casa azul [literally, the house blue]This result indicates that by altering the levelactivation of the phonological properties of thlexical node that belongs to the second phonlogical word (see footnote 2), even when thlexical node is not the head of the phrase (but

.

th oeaend

haorexnP

snnhnt

reNr

calo

-e-nts&veif-n

m-atr--rsg

aes-

ofstheto-dn

hedlecenallynceo-

nds),ofsnt

adjective), naming latencies are also affected

GENERAL DISCUSSION

The main goal of this study was to explore role of the phonological word as a basic unitphonological encoding in speech productiand the extent to which there is concurrphonological activation of lexical nodes that foutside the phonological word being encodWe carried out three experiments in which Elish and Spanish speakers were asked to proadjectival NPs (English: determiner 1 adjective1 noun; Spanish: determiner 1 noun 1 adjec-tive) while ignoring the distractor words. On tassumption that phonologically related distrtors affect the level of activation of the phonlogical properties of the words (or the woforms) with which they are related, we usphonological facilitation effects to reveal the etent to which the phonological representatioof the different phonological words of the Nare activated.

In Experiment 1 and 2 we tested the effectphonologically related distractors to the nouof English NPs. The results of both experimeshowed that naming latencies were faster wthe pictures were presented along with phologically related distractors than when presenwith unrelated distractors. Importantly, thphonological facilitation effect was observed gardless of the position of the noun in the (first, second, or third position), and therefothe effect is independent of whether the nobelongs to the first or second phonologiword. The results of Experiment 3 demonstrthat when the distractor word is phonologicarelated to an adjective placed in the sec

phonological word of a Spanish NP a reliab

ARAMAZZA

g-c-

fs

.feo-t

he

eofn,ntlld.g-uce

ec--

dd-s

ofstseno-ede-Peunaltelynd

phonological facilitation effect is obtaineTaken together, as shown in Fig. 1, the ressuggest that the level of activation of the phological information of the lexical nodes that blong to the second phonological word in an Naffects naming latencies. Furthermore, the effof the phonologically related distractors on twords located in the second phonological woof the NP seems to be independent of the grmatical status of the word placed in that potion, since they were observed for both nou(in English) and adjectives (in Spanish).

Our results fit well with other evidence suggesting that the phonology of the elements blonging to the second phonological word of aNP is concurrently active with earlier elemenof the phrase (e.g., Ferreira, 1991; DellO’Seaghdha, 1992). Importantly, we now haconvergent evidence for this proposal with dferent experimental paradigms relying ochronometric analyses of behavior. For exaple, Miozzo and Caramazza (1999) found ththe phonology of the noun in Italian deteminer 1 adjective1 noun NPs affects the selection of the phonological form of determine(the first word in the NP) in a simple namintask. Similar results have been obtained inFrench variant of this experiment using posssive 1 adjective1 noun NPs (Alario & Cara-mazza, in press). Along the same lines, Roel(1998) obtained larger facilitation effects, wian implicit priming paradigm, when the targwords share not only parts of the first phonlogical word but also parts of the seconphonological word. The fact that the activatioof phonological elements that are not in tfirst phonological word of the NP is observewith different chronometric paradigms (simppicture naming, the picture–word interferenparadigm, and implicit priming) as well as ithe analyses of spontaneous and experimentinduced speech errors increases our confidein the reliability of the phenomenon and its ptential theoretical importance.

It is important to note that our results athose of Alario and Caramazza (in presMiozzo and Caramazza (1999), and Roel(1998) are not, strictly speaking, inconsiste

lewith those reported by Meyer (1996). Meyer’s

r

d

o

heuh

h

nee’

aldeng

ets,hrdrts.nd--

la-ed-atrdalc-se,he

sec-ond phonological word—the noun in English and the adjective in Spanish.

study differs from the others cited here not onin the types of structures subjects were requito produce (e.g., conjoined NPs and sentencbut also in the number of phonological worinvolved in these structures. In the structurtested in our experiments, the primed word walways located in the second phonologicword. In contrast, in Meyer’s experiment usinsentences such as “the bag is next to the arrthe position of the primed word “arrow” is nothe second phonological word but rather tthird or fourth phonological word. Along thsame lines, it is unclear whether in the struct“the bag and the arrow” arrow is located in tsecond or third phonological word. In sumwhereas in our experiments we tested whetthe second phonological word was concurrenactive with the first phonological word, iMeyer’s study it is arguable whether the primword was located in the second or third (or latphonological word. Thus, taking both Meyerresults (no concurrent activation of first an

third phonological word) and ours (concurre

lyedes)s

esasalgw,”te

ree,er

tly

dr)

sd

activation of first and second phonologicword) into account, we can tentatively concluthat the upper bound of phonological encodiis two phonological words.

Although our results are clear regarding thexistence of concurrent activation of elementhat belong to different phonological wordsthey are silent regarding the extent to whicprocessing of the second phonological woneeds to be completed before articulation staDo our results necessarily mean that the secophonological word must be fully retrieved before the articulatory routines of the first phonological word can be executed? Since namingtencies are affected by manipulating the spewith which the second phonological word is retrieved, one may be tempted to conclude ththe encoding of the second phonological womust be completed before the first phonologicword can be delivered to the next stage of proessing. However, this is not necessarily the caand our results can also be explained in t

PRODUCTION OF NOUN PHRASES 191

FIG. 1. Differential effects of semantically related distractors and phonologically related distractors in NPproduction (English and Spanish). In both cases, the distractors were related to the lexical node located in the

ntframework of an incremental processing archi-

C

hls

eh rdh h9pfohrdeaievn

-r

ugr

televoa ee

e ecttn

otheinnce).d-iter-

blereherefirst

re-othcil-a

thes in re-x-nd

torseyes

endst-

aresedmile

rtedter-er-t inntedrs

notultsac-l fa-st

geith ofhy-

pothesis that the scope of phonological encod-ing may be variable.3 They propose that the

3 Ferreira and Swets (in press) have also suggested that in-crementality may be under strategic control, but that the sizeof the planning unit cannot be smaller than one phonological

192 COSTA AND

tecture in which the phonological word is tbasic unit of phonological encoding (see aRoelofs, 1998).

In the introduction we mentioned that spefluency is explained (in part) by assuming tthe processes engaged at different levels ofresentation are functioning concurrently on ferent parts of the utterance. Some researchave assumed that the phonological word isminimal information required for computing tnext level of representation (e.g., Levelt, 198In such a framework, speakers could be coming a phonological word and passing this inmation onto the next level of analysis as tbegin to work on the next phonological woThus, speakers could be retrieving the phonplan of the first phonological word as they engaged in computing the second phonologword. In this scenario two processes are gaged simultaneously at the phonological lebroadly defined: the formation of the secophonological word (e.g., “car”/kaɹ/) and the retrieval of the articulatory gestures of the fiphonological word (e.g., “The red”/də ɹεd/). Ifwe were to assume that the processing resoimplicated at these two levels of processinphonological word formation and articulatoroutine retrieval—are not independent, thenease with which a representation at one levprocessing is computed may affect the retriof representations at the other level. Therefby facilitating the retrieval of the phonologicproperties of the second phonological wordthe utterance (/kaɹ/) more processing resourcmay be allocated in computing the phonetic ments of the first phonological word (/də ɹεd/).As a consequence, naming latencies may bcelerated. The phonological facilitation effobserved in our experiments could reflect differential use of processing resources in phonologically related and unrelated conditioThis type of explanation is consistent with mels that claim that the phonological word is minimal phonological information required order to start the production of an uttera(e.g., Levelt, 1989; Wheeldon & Lahiri, 1997

Before concluding, it is important that we adress an apparently discordant result in the l

ature. In the introduction, we noted tha

ARAMAZZA

eo

chatep-if-ers

thee).ut-r-ey.ticrecaln-

el,d

st

rces—yhe ofalre,lofsle-

ac-t

hehes.

d-

Schriefers and Teruel (1999) obtained a reliafacilitation effect when the distractors wephonologically related to the first syllable of tNP (ro for “rosa Tisch”), but not when they werelated either to the second syllable of the phonological word (sa for “rosa Tisch”) or tothe second phonological word (ti for “rosaTisch”). This result is at variance with those ported here as well as with Meyer’s (1996), bof which demonstrated clear phonological faitation for the second lexical item within phonological word. What is the source of discrepancy? There are several differencethe procedures of those studies that may besponsible for the conflicting results. In our eperiments and in those of Meyer (1996) aMiozzo and Caramazza (1999), the distracwere words, while in Schriefers and Teruel thwere word fragments (two or three phonemcorresponding either to the beginning or the of the word; see Starreveld, 2000, for contraing effects when word fragments vs words used as distractors). Furthermore, the NPs uby Schriefers and Teruel were different frothose used in the other two studies. WhSchriefers and Teruel used adjective 1 nounNPs, in the other studies the NPs always stawith a determiner (the red car; la casa azul [lially, the house blue]; de kerk en het boek [litally, the church and the book]). Note also thaour experiments the distractors were presevisually instead of auditorially as in Schriefeand Teruel’s study. However, this factor canbe the main reason for the contrasting ressince Meyer (1996) also used auditory distrtors, and she observed reliable phonologicacilitation for the second syllable of the firphonological word (the noun).

Schriefers and Teruel (1999) acknowledimplicitly that their results are inconsistent wother results in the literature on the scopephonological encoding and entertain the

tword.

F

du

edt

ngatubccsm

seaerthti,

r

oeee

urt

n

r

e

theely

ebyen

inlay

is- inua-c-eof

our inrilye-

leara-

hehecal the

ted ofesf-stsnde-ntnms, incit- ofre-

ofgi--

ed ofer

PRODUCTION O

amount of phonological information that neeto be computed before speakers start artiction may not be fixed, but instead may depeon the speaking demands of a particular expmental (or natural) context. However, they not identify the variables in their experimenthat could have led to such a drastic reductiothe scope of phonological encoding—a sinsyllable. A tentative answer to this question mbe found when we analyze two properties of study conducted by Schriefers and Ter(1999): the number of pictures and the numof repetitions. In the experiments in whiSchriefers and Teruel did not find phonologifacilitation for the second syllable of the firword of the NP, participants were asked to naa small set of pictures (6 pictures) many tim(14 times) in four different SOA conditionThus, participants were repeating each of thtarget words a total of 56 times (14 times in eof the 4 SOAs). It is possible that the use of vsmall phonological planning units (e.g., the fisyllable of the utterance) depends, among othings, on how easy it is to retrieve the targelexical nodes. If, furthermore, the ease wwhich name retrieval occurs depends, in partthe number of possible responses in the expment plus the number of times they are peated, then Schriefers and Teruel’s descould have favored the use of such a smphonological planning unit. Note that in thother studies in which the phonological planing unit appears to be larger, the numberitems was larger and the number of repetitiwas smaller than in Schriefers and Terustudy, presumably preventing the use of vsmall phonological planning unit. Future rsearch will have to determine which of theproperties (or possibly some other variable sas word frequency, length, etc.) might be sponsible for the seemingly contrasting resul

The results of the present experiments areplicate and extend the observation of semainterference effects when producing NPs English. Critically, English differs in one impotant dimension from the languages (Dutch aItalian) in which the semantic interference fects have already been observed in NP prod

tion. In those languages the determiner of t

NOUN PHRASES 193

sla-ndri-os ofley

heelerhalte

es. 6chrysters’

NP agrees with the grammatical gender of noun (het boek [the book]; de kerk [thchurch]), and therefore its selection can ontake place after the selection of the noun. If thselection of the noun is delayed, for example,presenting a semantically related distractor, ththe selection of the determiner (the first wordthe utterance) is also delayed, leading to a dein naming latencies. Therefore, it is not surpring that a semantic interference effect ariseslanguages such as Dutch and Italian. The sittion in English is different, however. The seletion of the determiner in English could bachieved independently from the selection the noun (at least in the type of NPs used in experiments). Therefore, in principle, a delaythe selection of the noun need not necessaresult in a delay in naming latencies. Neverthless, the results of our experiments show a csemantic interference effect even in this sitution. These results, along with those from tphonologically related distractors, support thypothesis that the scope of the grammatistage is, at least, as large as the scope of

thoneri-e-ignallen-ofnsl’sry-

seche-s.lsoticin-ndf-uc-

phonological encoding unit.

CONCLUSION

The results of the three experiments reporhere demonstrate that the level of activationthe phonological properties of the lexical nodthat fall outside the first phonological word afect naming latencies. This observation suggethat the phonological elements of the secophonological word of the NP are activated bfore articulation begins. This is an importafinding that fits quite well with the informatiogathered with other chronometric paradigsuch as picture naming (Alario & Caramazzapress; Miozzo & Caramazza, 1999) and implipriming (Roelofs, 1998). Importantly, our results are also in agreement with analysesspontaneous speech errors, in which, as viewed above, the phonological information lexical nodes that belong to different phonolocal words do interact. Although the full implications of this observation still need to be workout, we can at least conclude that the windowphonologically active representations is larg

hethan a single phonological word.

194 COSTA AND CARAMAZZA

APPENDIX A

Materials Employed in Experiment 1

Picture Semantically related Unrelated Phonologically related Unrelated

Hat (R) Scarf Mosquito Hag WandPigeon (R) Sparrow Flute Pimple OrchidLamp (R) Torch Fox Latch DockHorse (R) Donkey Spoon Horde ForeHammer (R) Shovel Skirt Hammock VentRabbit (G) Mouse Onion Racquet CarriageCorn (G) Wheat Cannon Cord TackLeg (G) Elbow Shark Lemon DollLettuce (G) Tomato Elephant Lens CamperBench (G) Stool Cow Bell PillPeach (B) Apple Rose Peak TumorBottle (B) Pitcher Lion Boss TileMug (B) Bowl Piano Mud GuiltArm (B) Chest Truck Arc CardWasp (R) Mosquito Scarf Wand HagOrgan (R) Flute Sparrow Orchid PimpleDog (R) Fox Torch Dock LatchFork (R) Spoon Donkey Fore HordeVest (R) Skirt Shovel Vent HammocCarrot (G) Onion Mouse Carriage RacquetTank (G) Cannon Wheat Tack CordDolphin (G) Shark Elbow Doll LemonCamel (G) Elephant Tomato Camper LensPig (G) Cow Stool Pill BellTulip (B) Rose Apple Tumor PeakTiger (B) Lion Pitcher Tile BossGuitar (B) Piano Bowl Guilt MudCar (B) Truck Chest Card Arc

Note. The color of the pictures is shown in parentheses (R 5 Red; B 5 Blue; G 5 Green).

APPENDIX B

Materials Employed in Experiment 2

Picture Phonologically related Unrelated

Basket Batch HammockCarpet Card LackCarrot Carriage SafeCastle Cast HarvestCherry Check VesselCradle Crane HedgeCrown Crowd DrugDrum Drug CrowdEagle Eager CarriageFlag Flash EagerFork Form StandGlobe Glow PeakHammer Hammock BatchHarp Harvest FlashHelmet Hedge Cast

Lamp Lack Form

PRODUCTION OF NOUN PHRASES 195

APPENDIX B—Continued

Picture Phonologically related Unrelated

Peach Peak GlowSail Safe CraneStamp Stand CheckVest Vessel CardBed Bell GutBottle Bottom WinterBridge Brick DollCoat Coal MouseDog Doll BrickDress Dread WheatGun Gut BellHat Hatchet ShinMoon Moose TrumpetMountain Mouse RaceNose Node TreatPlane Play CoalRain Race PlayShip Shin HatchetSun Sum TrailTrain Trail SumTree Treat NodeTruck Trumpet MooseWheel Wheat Dread

Window Winter Bottom

APPENDIX C

Materials Employed in Experiment 3

Picture Phonologically related Unrelated

SET 1

Manzana (BL) (apple) Gris (gray) Ideal (ideal)Cuchillo (BL) (knife) Gris (gray) Ideal (ideal)Cama (BL) (bed) Violeta (violet) Egoista (egoist)Falda (BL) (skirt) Naranja (orange) Vertical (vertical)Cebolla (BL) (onion) Lila (purple) Ruin (contemptible)Coche (BR) (car) Violeta (violet) Egoista (egoist)Silla (BR) (chair) Granate (garnet) Crujiente (cruchy)Jarra (BR) (pitcher) Granate (garnet) Crujiente (crunchy)Ventana (BR) (window) Rosa (pink) Eficaz (effective)Collar (GR) (necklace) Naranja (orange) Vertical (vertical)Fresa (GR) (strawberry) Lila (purple) Ruin (comtemptible)Aranya (GR) (spider) Rosa (pink) Eficaz (effective)Barco (GR) (boat) Rosa (pink) Eficaz (effective)Trompeta (BL) (trumpet) Gris (grey) Ideal (ideal)Zanahoria (BL) (carrot) Gris (grey) Ideal (ideal)Tanque (BL) (tank) Violeta (violet) Egoista (egoist)Tiburon (BL) (shark) Naranja (orange) Vertical (vertical)Castillo (BL) (castle) Lila (purple) Ruin (contemptible)Brazo (BR) (arm) Violeta (violet) Egoista (egoist)Bicicleta (BR) (bike) Granate (garnet) CrujienteNariz (BR) (nose) Granate (garnet) Crujiente

Conejo (BR) (rabbit) Rosa (Pink) Eficaz (effective)

196 COSTA AND CARAMAZZA

APPENDIX C—Continued

Picture Phonologically related Unrelated

Leon (GR) (lion) Naranja (orange) Vertical (vertical)Guitarra (GR) (guitar) Lila (purple) Ruin (contemptible)Pistola (GR) (gun) Rosa (pink) Eficaz (effective)Luna (GR) (moon) Rosa (pink) Eficaz (effective)

SET 2

Barba (BL) (beard) Azote (stroke) Fianza (surety)Casa (BL) (house) Azufre (sulfur) Jasmin (jasmine)Gato (BL) (cat) Azurcar (sugar) Drama (drama)Botella (BL) (bottle) Azotea (roof) Trenza (plait)Arbol (BR) (tree) Marfil (ivory) Pincel (brush)Bomba (BR) (bomb) Margen (margin) Rabia (rage)Avion (BR) (plane) Marca (mark) Techo (ceiling)Cangrejo (BR) (crab) Marco (frame) Leche (milk)Calcetin (BR) (sock) Marcha (march) Golpe (blow)Hombre (GR) (man) Verbo (verb) Farol (lantern)Casco (GR) (helmet) Verja (fence) Bruja (witch)Gorra (GR) (cap) Vera (edge) Tubo (tube)Camello (GR) (camel) Verso (verse) Tropa (troop)Mono (BL) (monkey) Azote (stroke) Fianza (surety)Piano (BL) (piano) Azufre (sulfur) Jazmin (jasmine)Raqueta (BL) (racquet) Azucar (sugar) Drama (drama)Payaso (BL) (clown) Azotea (root) Trenza (braid)Vaso (BR) (glass) Marfil (ivory) Pincel (brush)Paloma (BR) (pigeon) Margen (margin) Rabia (rage)Pelota (BR) (ball) Marca (mark) Techo (ceiling)Perro (BR) (dog) Marco (frame) Leche (milk)Pera (BR) (pear) Marcha (march) Golpe (blow)Mano (GR) (hand) Verbo (verb) Farol (lantern)Pierna (GR) (leg) Verja (fence) Bruja (witch)Reloj (GR) (watch) Vera (edge) Tubo (tube)Martillo (GR) (hammer) Verso (verse) Tropa (troop)

Note. SET 1 5 Semantically Related/Unrelated Distractors to the Adjective; SET 2 5 Phonologically Related/Unrelatedh

:

&

Jfo

o-e

a,

l

al

x-

Distractors to the Adjective. The color of the pictures is s

REFERENCES

Alario, F. X., & Caramazza, A. (in press). The productionNoun Phrases in French. Cognition.

Bock, K., & Levelt, W. (1994). Language production: Grammatical encoding. In M. A. Gernsbacher (Ed.),Hand-book of psycholinguistics(pp. 945–984). San DiegoAcademic Press.

Caramazza, A., Miozzo, M., Costa, A., Schiller, N.,Alario, X. (in press). Lexical selection: A cross-language investigation of determiner production. In Dupoux (Ed.),Language, brain and cognitive development. Cambridge, MA: MIT Press.

Cohen, J. D., MacWhinney, B., Flatt, M., & Provost,(1993). PsyScope: An interactive graphic system designing and controlling experiments in the psychogy laboratory using Macintosh computers. Behavior

Research Methods, Instruments, and Computers, 25,257–271.

ro-

own in parentheses (GR 5 Green; BL 5 Blue; BR 5 Brown).

of

-

-E.-

.orl-

Costa, A., & Sebastian-Galles, N. (1998). Abstract phonlogical structure in language production: Evidencfrom Spanish. Journal of Experimental Psychology:Learning, Memory, and Cognition, 24,886–903.

Costa, A., Sebastian-Galles, N., Miozzo, M., & CaramazzA. (1999). The gender congruity effect: Evidence fromSpanish and Catalan. Language and CognitiveProcesses, 14,381–391.

Dell, G. S. (1986). A spreading activation theory of retrievain sentence production. Psychological Review, 93,283–321.

Dell, G. S., & O’Seaghdha, P. G. (1992). Stages of lexicaccess in language production. Cognition, 42,287–314.

Ferreira, F. (1991). Effects of length and syntactic compleity on initiation times for prepared utterances. Journalof Memory and Language, 30,210–233.

Ferreira, F. (1993). Creation of prosody during sentence p

duction. Psychological Review, 100,233–253.

F

a um

f

i-

io

.-s.

o

i

ve o:

8k

-

u

n &s

.ri

v

o

h

the

n-c-

n-ng

ngita-

ncex-

l-s:of

e-

i-

an-e

grti--

of

tion:

-n:

9).ri-

n

er ef-:

PRODUCTION O

Ferreira, F., & Swets, B. (in press). How incremental is lguage production? Evidence from the production ofterances requiring the computation of arithmetic suJournal of Memory and Language.

Francis, W. N., & Kuc˘era, H. (1982). Frequency analysis oEnglish usage. Boston: Houghton Mifflin.

Fromkin, V. A. (1973). Speech errors as a linguistic evdence. The Hague: Mouton.

Garrett, M. F. (1975). The analysis of sentence productIn G. Bower (Ed.),The psychology of learning and motivation: Advances in research and theory(Vol. 9, pp.133–177). New York: Academic Press.

Garrett, M. F. (1988). Processes in language productionF. J. Newmeyer (Ed.),Linguistics: The Cambridge survey. Vol 3: Language: Biological and psychological apects (pp. 69–96). Cambridge, MA: Harvard UnivPress.

Glaser, W. R., & Düngelhoff, F. (1984). The time coursepicture–word interference. Journal of ExperimentalPsychology: Human Perception and Performance, 10,640–654.

Glaser, W. R., & Glaser, M. O. (1989). Context effect Stroop-like word and picture processing. Journal ofExperimental Psychology: General, 118,13–42.

Hayes, B. (1989). The prosodic hierarchy in meter. P. Kiparsky & G. Youmans (Eds.),Phonology andmeter: Phonetics and phonology(Vol. 1, pp. 201–260).San Diego: Academic Press.

Jescheniak, J., & Schriefers, H. (1998). Discrete serial sus cascaded processing in lexical access in spproduction: Further evidence from the coactivationnear-synonyms. Journal of Experimental PsychologyLearning, Memory, and Cognition, 24,1256–1274.

La Heij, W., Mak, P., Sander, J., & Willeboordse, E. (199The gender-congruency effect in picture–word tasPsychological Research, 61,209–219.

Levelt, W. J. M. (1989). Speaking: From intention to articulation. Cambridge, MA: MIT Press.

Levelt, W. J. M. (1992). Accessing words in speech prodtion: Stages, processes and representations. Cognition,42,1–22.

Levelt, W. J. M., & Maassen, B. (1981). Lexical search aorder of mention in sentence production. In W. KleinW. Levelt (Eds.),Crossing the boundaries in lingusitic(pp. 221–252). Dordrecht: Reidel.

Levelt, W. J. M., Schriefers, H., Vorberg, D., Meyer, A. SPechmann, T., & Havinga, J. (1991). The time couof lexical access in speech production: A study of pture naming. Psychological Review, 98,122–142.

Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioral& Brain Sciences, 22,1–75.

Levelt, W. J. M., & Wheeldon, L. (1994). Do speakers haaccess to a mental syllabary? Cognition, 50,239–269.

Lupker, S. J. (1979). The semantic nature of response cpetition in the picture–word interference task. Memory& Cognition, 7, 485–495.

Lupker, S. J. (1982). The role of phonetic and orthograpsimilarity in picture–word interference. Canadian

Journal of Psychology, 36,349–367.

NOUN PHRASES 197

n-t-s.

n.-

In

-

f

n

In

er-echf

).s.

c-

d

,sec-

-

e

m-

ic

MacLeod, C. M. (1991). Half a century of research on Stroop effect: An integrative review. PsychologicalBulletin, 109,163–203.

Meyer, A. S. (1990). The time course of phonological ecoding in language production: The encoding of sucessive syllables of a word. Journal of Memory andLanguage, 29,524–545.

Meyer, A. S. (1991). The time course of phonological ecoding in language production: Phonological encodiinside a syllable. Journal of Memory and Language,30,69–89.

Meyer, A. S. (1992). Investigation of phonological encodithrough speech error analyses: Achievements, limtions, and alternatives. Cognition, 42,181–211.

Meyer, A. S. (1996). Lexical access in phrase and senteproduction: Results from picture–word interference eperiments. Journal of Memory and Language, 35,477–496.

Meyer, A. S., & Schriefers, H. (1991). Phonological faciitation in picture–word interference experimentEffects of stimulus onset asynchrony and typesinterfering stimuli. Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 17,1146–1160.

Miozzo, M., & Caramazza, A. (1999). The selection of dterminers in noun phrase production. Journal of Exper-imental Psychology: Learning, Memory, and Cogntion, 25,907–922.

Nespor, M., & Vogel, I. (1986). Prosodic phonology. Dor-drecht: Foris.

Rayner, K., & Springer, C. J. (1986). Graphemic and semtic similarity effects in the picture–word interferenctask. British Journal of Experimental Psychology, 77,207–222.

Roelofs, A. (1998). Rightward incrementality in encodinsimple phrasal forms in speech production: Verb–pacle combinations. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24,904–921.

Schriefers, H. (1992). Lexical access in the productionnoun phrases. Cognition, 45,33–54.

Schriefers, H. (1993). Syntactic processes in the producof noun phrases. Journal of Experimental PsychologyLearning, Memory, and Cognition, 19,841–850.

Schriefers, H., Meyer, A. S., & Levelt, W. J. M. (1990). Exploring the time course of lexical access in productioPicture–word interference studies. Journal of Memoryand Language, 29,86–102.

Schriefers, H., de Ruiter, J. P., & Steigerwald, M. (199Parallelism in the production of noun phrases: Expements and reaction time models. Journal of Experimen-tal Psychology: Learning, Memory, and Cognition, 25,702–720.

Schriefers, H., & Teruel, E. (1999). Phonological facilitatioin the production of two-word utterances. EuropeanJournal of Cognitive Psychology, 11,17–50.

Schriefers, H., & Teruel, E. (2000). Grammatical gendin noun phrase production: the gender interferencefect in German. Journal of Experimental Psychology

Learning, Memory, and Cognition, 26,1368–1377.

198 COSTA AND CARAMAZZA

Sebastian, N., Marti, M. A., Cuetos, F., & Carreiras, M.(1996). LEXESP: Base de datos informatizada de la

se of

e

n

s

e

naming tasks. Journal of Experimental Psychology:Learning, Memory, and Cognition, 21,686–698.

Starreveld, P. A., & La Heij, W. (1996). Time-course analy-ic-:

ech

ch

lengua espanola (LEXESP: Computerized data baSpanish Language). Departamento de Psicologiasica. Universitat de Barcelona.

Selkirk, E. (1984).Phonology and syntax: The relation btween sound and structure. Cambridge, MA: MITPress.

Selkirk, E. (1986). On derived domains in sentence phoogy. Phonology Yearbook, 3, 371–405

Starreveld, P. A. (2000). On the interpretation of onsetauditory context effects in word production. Journal ofMemory and Language, 42,497–525.

Starreveld, P. A., & La Heij, W. (1995). Semantic interfence, orthographic facilitation and their interaction

Ba-

-

ol-

of

r-in

sis of semantic and orthographic context effects in pture naming. Journal of Experimental PsychologyLearning Memory and Cognition, 22,896–918.

Van Berkum, J. J. A. (1997). Syntactic processes in speproduction: The retrieval of grammatical gender. Cog-nition, 64,115–152.

Wheeldon, L., & Lahiri, A. (1997). Prosodic units in speeproduction. Journal of Memory and Language, 37,356–381.

(Received November 10, 2000)(Revision received February 8, 2001)(Published online November 8, 2001)