Post on 15-Jun-2016
Journal of Memory and Language 44,350–375 (2001)doi:10.1006/jmla.2000.2743, available online at http://www.academicpress.com on
Age of Acquisition, Word Frequency, and the Locus of Repetition Primingof Picture Naming
Christopher Barry and Katherine W. Hirsh
School of Psychology, Cardiff University, Cardiff, United Kingdom
Robert A. Johnston
School of Psychology, University of Birmingham, Birmingham, United Kingdom
and
Catherine L. Williams
School of Psychology, Cardiff University, Cardiff, United Kingdom
We examined the roles of age of acquisition (AoA) and word frequency in picture naming latencies and stud-ied repetition priming to illuminate the locus and mechanism by which the effective variable has its effect. Ex-periment 1 found that AoA affected naming latencies when frequency was controlled, and Experiment 2 foundthat frequency had no effect when AoA was controlled. Experiment 3 found no effects of either AoA or frequencyin delayed picture naming. Picture naming was facilitated by the prior naming of identical pictures and, to a lesserextent, by the prior reading aloud of the names. Repetition priming interacted with AoA but did not interactwith frequency. We conclude that both AoA and long-lasting repetition priming operate at the level of lexical-phonological retrieval and that repetition interacts with AoA because it facilitates the retrieval of lexical-phono-logical elements required for naming, which benefits late-acquired words differentially.© 2001 Academic Press
Key Words:picture naming; repetition priming; age of acquisition; word frequency.
A commonly accepted framework for under-spoi
r
i
t
sion. Recognized and comprehended items then
ssione-ut
of aal
cu-sesicral
islly
standing the naming of a pictured object propothat it comprises three major processing comnents: object recognition, semantic activatiand lexical access (e.g., Ellis, Kay, & Frankl1992; Morton, 1985; Snodgrass & McCulloug1986; Warren & Morton, 1982). Following peceptual analysis, a familiar object activatesstored visual (or structural) representation, whconstitutes the level of object recognition. Ob-jects recognized as familiar then activate thcorresponding stored functional and associaknowledge; that is, they access their semanticrepresentations, which permits their comprehen
0749-596X/01 $35.00Copyright © 2001 by Academic PressAll rights of reproduction in any form reserved.
35
reeoraltis,-te;atxi-
The work of Chris Barry and Bob Johnston was supporby the UK’s Economic and Social Research Council (GraNo. R000235995). We thank Andrew Ellis and StephLupker for comments on an earlier version of this article.
Address correspondence and reprint requests to CBarry, Department of Psychology, University of Kent Canterbury, Canterbury, Kent CT2 7NP, United KingdomFax: +44 1227 - 827030. E-mail: C.Barry@ukc.ac.uk.
0
eso-n,n,h,- ach
eirive
-
activate the stored lexical phonologyof theirnames. Most current models of lexical accefor speech production recognize a separatbetween lexical identification—a process of slecting which item(s) best matches some inpspecification—and lexical retrieval—a processretrieving for a particular lexical representationspecification of its properties in the phonologicoutput domain. Finally, the names must be artilated. Figure 1 shows the three major procesunderlying picture naming, along with the basprocesses proposed to be involved in the oreading of words.
One aspect of Fig. 1 worth emphasizing that picture naming is necessarily semanticamediated, whereas word naming is not. Throutes have been argued to be available for reading (e.g., Coltheart, 1985; Coltheart, CurAtkins, & Haller, 1993; Ellis, 1993): (a) a sublexical, assembled phonological recoding rou(b) a “direct” lexical but nonsemantic route thconnects visual word recognition to stored le
tednt
en
hrisat
.
REPETITION PRIMING OF PICTURE NAMING 351
i
a
vat-tedr-na-nti-thecti-ndills thens ofn-on”
of
FIG. 1. A framework for understanding the component processes underlying picture naming and word reading.
cal phonology (although this has been dispuby Hillis & Caramazza, 1991); and (c) a sematically mediated route. Reading words aloudroutes (b) or (c) would involve activation of thsame lexical phonological representationsare used for picture naming (lexical retrievaHowever, only route (c) has both the lexicidentification and the lexical retrieval stagescommon with picture naming.
Repetition Priming of Picture Naming
The time to name the picture of an object mbe reduced by the prior presentation of the pture. Two classes of explanations of repetiteffects may be distinguished. The first assumthat episodic instancesare created for each stimulus event, along with records of how it w
processed (e.g., Jacoby, 1983; Logan, 1990).tedn-byeasl).alin
ayic-ones-s
peated events enjoy the advantage of reactiing their earlier episodic traces and associaprocessing histories, which facilitates their curent processing. The second class of explation, or what Ellis, Flude, Young, and Burto(1996) refer to as “structural” theories of repetion priming, assumes that the processing of first stimulus event results in changes to the avation levels of permanent representations atheir interconnections. A repeated stimulus wbe responded to faster because it also engagesame, but now “primed,” stored representatioand processing pathways. From their studiesface recognition and naming, Ellis et al. cocluded that there are two distinct loci of repetitipriming effects in face processing: “locus 1priming involves the perceptual recognition
Re-a familiar face (which is specific to having seen
aatigotet
h
orbsenosu
he
pth7-
s
ic
oo
s
n
ten
tsn-nl-
&g-ef-foree-
aks togg-tesner,r-eryll,rg
o-e-minn se-ningeencal ef-im-ndt a
ornn-
ettoffa
on-si-t-d
352 BARRY
the face before, i.e., it is domain-specific) a“locus 2” priming involves the name retriev(which can show both within- and cross-domeffects). Ellis et al. speculate as to whether same framework may be applied to repetitpriming of object recognition and naminLocus 1 priming of object recognition is nstudied directly in the experiments reporhere, as this would necessitate a task thaquires (and is assumed to be restricted to) obrecognition; one possible candidate for suctask is the object decision task introducedKroll and Potter (1984), although it is notewothy that latencies in this task are often as las those for picture naming itself (e.g., BarJohnston, & Scanlan, 1998). However, it will possible to estimate the contribution of locupriming by a comparison of the repetition fects on picture naming from the prior namiof identical pictures of objects with the prireading aloud of the object names. The preexperiments were designed to investigate locrepetition priming of object naming and to futher examine whether this interacts with eitthe age of acquisition or the frequency of objnames.
A number of researchers have found that ture naming is primed by the prior naming of same picture (e.g., Durso & Johnson, 19Mitchell & Brown, 1988). It is logically possible for the repetition effect for an identical itemto occur at any or all of the three main stageprocessing shown in Fig. 1. When an item ispeated its visual, semantic, lexical, and form(phonological and articulatory) characteristare repeated and thus facilitatory repetition fects could occur at any of these levels. In view the lexical level is the dominant locus frepetition priming in speech production tassuch as picture naming.
The argument against a semantic locus repetition priming is based in part on differencin the persistence of repetition priming and mantic priming. That presentation of an itesuch as breadfacilitates subsequent respondito a semantically related item such as butter(e.g., Meyer & Schvaneveldt, 1971) indicathat the repetition of some of an item’s sema
features without repetition of the item itself caET AL.
ndlinheon.tdre-
ject abyr-ngy,e 1f-grents 2r-erct
ic-e9;
ofre-als
ef-urr
ks
forese-mg
stic
facilitate responding. Semantic priming effecdissipate rapidly, however (Meyer, Schvaeveldt, & Ruddy, 1972). In contrast, repetitiopriming typically lasts for minutes (e.g., Wheedon & Monsell, 1992) or hours (e.g., JacobyDallas, 1981). This differential persistence sugests that semantic and repetition priming fects do not have a common locus and thererepetition priming effects are not (wholly) smantically mediated.
A second source of evidence that speagainst a strong semantic-level contributionrepetition priming comes from studies utilizinbilingual participants. Studies of word reconition indicate that processing of noncognadoes not transfer across languages (e.g., KirsSmith, Lockhart, King, & Jain, 1984; Scaboough, Gerard, & Cortese, 1984) except at vshort delays (Kirsner et al., 1984). MonseMatthews, and Miller (1992) found that foEnglish–Welsh bilinguals there was no priminof naming a picture in English by the prior prduction of the (noncognate) word in Welsh in rsponse to a Welsh definition presented 8–20 earlier. The lack of long-lived priming betweeitems such as noncognates that share onlymantic features suggests that shared meacannot be the source of the repetition effect swhen both the semantic and the phonologiform are shared. (Short-lived cross-languagefects can be explained in terms of semantic pring.) Taken together, differential persistence athe lack of cross-language facilitation suppornonsemantic locus for repetition priming.
There is also evidence that neither lexical nsublexical phonological overlap on their oware sufficient to produce long-term repetitiopriming. First, as in the case of semantic priming, phonological/formal priming effects artypically very short-lived. For example, Eveand Humphreys (1981) found that repetitiongraphemic information resulted in facilitation othe processing of both words and nonwords inshort-lag masked priming paradigm. Repetitiof nonwords also results in facilitation (Monsell, 1985); however, both of these effects dispate at long lags. Given this differential persisence, it seems unlikely that form priming an
nrepetition priming have a common locus andl
la
e
i
erinoid
o
e
t
c
o
orui
raln-thery,eb-ofch
e-edasee-
s,todayry
r ail--
i.balu-andla-ast);d-
reicaltynts5).
xi-lecus.ral ingn
m-s,e
REPETITION PRIMING
therefore it would appear that repetition primineffects are not attributable to postlexical phonlogical processing. In addition Brown et a(1991) found no difference in the level of priming of picture naming when overt and covenaming of the prime stimulus was compareThey concluded that the physical act of articution does not contribute to long-term repetitiopriming. Further evidence for this claim comefrom work utilizing homophones, which havidentical phonology (e.g.,hareandhair) but donot share semantic characteristics. Wheeldand Monsell (1992) used homophonprime–probe pairs to show that repetition thinvolves phonological but not semantic overlais not sufficient to produce priming. In the priming phase they required participants to eithread aloud printed words or to produce wordsresponse to printed definitions; 6–12 min latparticipants produced the probe names insponse to pictures. There was repetition primfor their nonhomophonic pairs: however, prducing a homophonic word such as “flour” dnot prime the naming of a picture of aflower.These data complement those obtained wbilinguals: it appears that repetition of phonlogical form without repetition of semantic formis no more conducive to repetition priming ospeech production than is repetition of semanform without concomitant repeti-tion of phonological form. We conclude, therefore, that reptition priming can be localized to the level awhich there is a transition from semanticphonological form—the lexical level. It is onlyat the lexical level that there is a confluencesemantic and phonological influences.
Psycholinguistic Determinants of PictureNaming Latency
Having established that repetition priming efects are most likely to be located at the lexilevel of processing, we can use this to examother claims with regard to localization. The ntion to be exploited in this article is that an inteaction with repetition priming may allow us tdetermine which of the psycholinguistic vaables that have been found to influence pictnaming latencies are candidates for localizat
at the lexical level.OF PICTURE NAMING 353
go-.-rtd.-
ns
oncatp-erinre-g-
ith-
ftic--
to
of
f-aline-
r-
i-re
on
Object recognition. Returning to Fig. 1 andstarting with input processing, there are sevevariables that influence picture naming latecies that are thought to have their effects atlevel of accessing structural descriptions. BarMorrison, and Ellis (1997) found that imagagreement—the extent to which a pictured oject matches the participant’s stored imagethat object—affected picture naming speed suthat objects with a high level of image agrement were named more rapidly. They arguthat high image agreement made for greater ein activating the object’s stored structural dscription. Visual complexity (Ellis & Morrison,1998) and structural similarity (HumphreyRiddoch, & Quinlan, 1988) are also thoughtexert their effects mainly at the level of storestructural descriptions; name agreement malso have some of its effects at this level (Baret al., 1997).
Semantics. Turning next to the conceptual osemantic level, imageability is believed to bekey variable operating at this locus. Imageabity values are by definition difficult to manipulate in experiments utilizing pictorial stimulHowever, the results from studies using verstimuli suggest that imageability has an inflence on production times. Strain, Patterson, Seidenberg (1995) found that word naming tencies were affected by imageability (at lefor low-frequency, irregularly spelled wordsV. Coltheart, Laxon, and Keating (1988) founthat word naming accuracy was affected by imageability (but only in poor readers), and theare numerous reports in the neuropsychologliterature detailing the influence of imageabilion the naming accuracy of dyslexic participa(e.g., Newton & Barry, 1997; Richardson, 197
Phonological assembly. Before moving on toconsider two variables thought to affect the lecal level, we mention word length, a variabthat has been argued to have a postlexical loWord length may be measured in seveunits—phonemes, syllables, and letters bethe most common. Morrison, Ellis, and Quinla(1992) found that phoneme length was an iportant predictor of picture naming latenciebut Barry et al. (1997) found that phonem
length had no significant independent effectu
s
io
e
-
y
fd
lel
r
oxmt
ona-os
m-a-ll,gg-edrets
--
sss-is- in-theis-upr,si--
forc
re-onf-ess,ononell
ngeredd,
e-p-on
-ass
354 BARRY
when other variables were included in their mtiple-regression analysis.
Lexical access. A large amount of data habeen amassed in support of two different vaables, both of which are argued to affect lexicaccess: word frequency and age of acquisit(AoA). Many studies have reported effectsthe word frequency of object names on pictunaming latencies: Pictures whose names ocmore frequently in the language are namfaster than those with less frequent names (eHumphreys et al., 1988; Jescheniak & Leve1994; Oldfield & Wingfield, 1965). Word fre-quency effects are also regularly reported in stuies of word naming (e.g., Forster & Chamber1973; Monsell, Doyle, & Haggard, 1989; Rubin1980). AoA effects have also been reportedboth picture naming (e.g., Carroll & White, 1973Gilhooly & Gilhooly, 1979; Morrison et al.,1992) and word reading (e.g., Coltheart et a1988; Morrison & Ellis, 1995). Treiman, Mullennix, Bijeljac-Babic, and Richmond-Welty (1995found no effect of AoA on word reading latencies when 42 other variables (including famiiarity, frequency, neighborhood size, initiaphoneme,number of letters, bigram frequencand spelling-to-sound consistency) were includin their regression analysis. However, their failuto find an effect was based on an analysis osubset of 247 of the 1327 monosyllabic wortheir participants had to read; this gives a “casto-variable” ratio of only 5.7, which is on theborderline of acceptability (Tabachnick & Fidel1988). Finally, there are studies where bothfects have been reported. Barry et al. (1997), Eand Morrison (1998), Lachman, Shaffer, anHennrikus (1974), and Snodgrass and Yudits(1996) all found effects of both AoA (either objective or subjective) and word frequency (agaboth objective and subjective measures weused) on picture naming. Gerhand and Ba(1998) found effects of both AoA and frequencon word naming latencies.
In the picture naming task, both word frequency and AoA are assumed to have their lat the lexical level, either in the process of lecal identification (i.e., selecting the lexical forthat best matches the conceptual specifica
aroused by the picture) or lexical retrieval (i.eET AL.
l-
ri-alonf
recurd
.g.,lt,
d-s,,in;
l.,
)-l-l,edre
ase-
,f-lisdky-inreryy
-cii-
ion
access to the lexical-phonological specificatiof a selected word). In word naming the sitution is more complex; for word frequency twloci are commonly put forward, the output locumentioned above with reference to picture naing and an input locus at the level of identifiction of the orthographic form (see Monse1991, for a review). In the only word naminstudy to manipulate AoA and frequency orthoonally, Gerhand and Barry (1998) suggestthat the word frequency effects they found wedue to an input locus, whereas the AoA effecarose at output, just as in picture naming.
One way to test this hypothesis is to examine the relationship between repetition priming, word frequency, and AoA. If two factorinteract then they are likely to share a proceing locus/stage. Of course factors affecting dtinct stages that operate in cascade may alsoteract, but the presence of an interaction at lexical level, which has been argued to be dcrete and serial by at least one prominent groof theorists (see Levelt, Roelofs, & Meye1999, for the most recent and detailed expotion of this position), would at least strongly encourage a belief in a common mechanism repetition priming and any psycholinguistivariable that interacts with it.
Previous studies would suggest that word fquency is this variable. For example, Wheeldand Monsell (1992) found that the repetition efects they observed were greater for picturwith low-frequency than high-frequency namea pattern of results consistent with the notithat frequency and repetition share a commmechanism. However, Wheeldon and Monsdid not match their items for AoA. We wouldsuggest that in this as in many picture namiexperiments the frequency effects reported wthe result of a confounding of AoA and worfrequency. Were this confound to be removewe would expect that the AoA effects would rmain but the effects of frequency would disapear. If so, we might also obtain an interactiof AoA and repetition.
While we favor an explanation of previous reports of frequency effects in picture naming being due to a failure to control for AoA, thi
.,does make it difficult to account for the presencenf
a
g
h
da
ihr
ntail
e0t
r4a,l
in
ret in
heich
de-bu-r,n-ac-Ablech-t 1-ri-ilet,
ofnger-ta-)
aysny
lye-r-heinfred
enames to trigger the voice key in immediate pic-
c-e
ofm-ar-
ead. Inresresual
e
REPETITION PRIMING
of frequency effects in picture naming expements in which both variables were includ(nor can we fall back on the input locus explation here as Gerhand & Barry, 1998, did word naming, as picture naming is a taskwhich there is no lexical component in the inpto drive a frequency effect). However, the Baet al. (1997) study was a correlational study Ellis and Morrison (1998) did not include a fatorial manipulation of word frequency, althouthey did manipulate AoA in this fashion.
Explanations of Frequency and AoA Effects
Monsell and colleagues (Monsell, 199Wheeldon & Monsell, 1992) have argued texplanations of long-lasting repetition effectsword recognition and picture naming are liketo be the same as explanations of the claimefect of word frequency in these tasks. After as Wheeldon and Monsell say, “frequent iteare those with a lot of accumulated primin(p. 756). Current models of lexical processaccount for frequency effects by proposing eitthat the thresholds of localized lexical repsentations (such as Morton’s logogens) or connection strengths between more distriburepresentations (as in connectionist modelsproposed, for example, by Seidenberg & MClelland, 1989) are determined by frequencyexposure.
Theorists attempting to account for AoA efects have argued that these effects may equalitatively different explanatory mechanismthan those put forward to explain frequency fects (but see Ellis & Lambon Ralph, 200Brown and Watson (1987) have proposed phonological completenesshypothesis of AoA,in which early-acquired words have “a mocomplete phonological representation” (p. 21whereas later-acquired words, which are sumed to be stored in a segmented fashionquire their stored phonology to be assembfor production (which entails longer processtimes). As repetition priming of picture naminmight be expected to facilitate the retrievallexical phonological representations requifor the purpose of spoken word production, iimperative to determine whether repetition
teracts with either frequency or AoA in orderOF PICTURE NAMING 355
ri-eda-
orinut
rrynd
c-h
1;at
inly ef-ll,
msg”ngere-theted asc- of
f-
inform and constrain theorizing about both tlocus of the effects and the mechanisms by whthese variables operate.
One purpose of these experiments was totermine whether there are independent contritions of AoA and word frequency when othepotentially confounding, variables are held costant. Second, we examined the possible intertions between repetition priming and both Aoand frequency in order to elucidate the probaloci of these effects as well as the possible meanisms underlying their operation. Experimenvaried the AoA of the object names, while holding the frequency of the names constant. Expement 2 varied the frequency of the names, whholding AoA constant. Within each experimenpriming was examined from the prior namingthe same pictures and from the prior readialoud of the names of the objects. In both expiments, the interval between the first presention (of either the picture or its printed nameand the second presentation (which was alwa picture) was between 6 and 10 min, and so apriming effects observed would be relativelong-lasting. Experiment 3 examined the dlayed naming of the pictures used in Expeiments 1 and 2 in order to establish whether teffects found were due to any differencesthe time required to initiate the production othe name once its phonology has been prepaor in the ability of the initial phonemes of th
t
sf-).he
e),s-
re-edg
gofdis-
ture naming.
EXPERIMENT 1: AOA VARIED ANDFREQUENCY CONTROLLED
Experiment 1 investigated the priming of piture naming by both the prior naming of thidentical pictures and the prior reading aloudprinted object names. The experiment coprised two phases. In stage I, one group of pticipants named pictures and another group raloud the printed names of these same itemsstage II, all participants named a set of pictuwhich included repeated pictures or the pictuwhose names were read aloud and an eqnumber of control, nonrepeated items.
AoA effects in word naming latencies hav
obeen found in a number of studies (Brown &on&n
e
.
l
dnt
anga
, ero i
tdt
ingl-a-
c-ours-
r-s,re
orid
0).re-
s
irf
Ase
the
te
d
)
as
356 BARRY
Watson, 1987; Gerhand & Barry, 1998; Gilhoo& Logie, 1981; Morrison & Ellis, 1995). Therexist numerous studies reporting effects of wfrequency on word naming (e.g., FrederickseKroll, 1976; Grainger, 1990; Monsell, Doyle, Haggard, 1989) but typically these have controlled for AoA. Thus, apparent frequeneffects on word naming latencies may resfrom (or be contaminated by) a confoundedfect of AoA. The data from stage I of the presexperiment will therefore permit a direct invesgation of the effects of AoA in reading alouwords that have been matched for frequency
Previous studies of picture naming have foupriming effects from the reading aloud of objenames (Durso & Johnson, 1979; WheeldonMonsell, 1992). Typically these effects are smathan the effects of within-domain (picture-tpicture) priming. One reason for this may be fact that it is possible to read aloud regular wosublexically, which would not necessarily egage, or prime, the same mechanisms or resentations involved in the task of naming ptures. Episodic accounts of repetition primisuggest a second reason why cross-dompriming effects may be smaller than withidomain effects. If repetition priming resufrom the activation of stored episodic instancthen word naming and picture naming episocould conceivably share semantic and phological features but will differ in terms of inpuperceptual features. This additional overlapterms of perceptual features may be the sourcthe increase in priming seen when prime probe are identical. Thus word–picture primimay be reduced due to sublexical processinthe word prime and picture–picture priming mbe enhanced due an overlap in the percepfeatures of the prime and probe. Howeverleast one study showed no within-domain hancement. Brown et al. (1991) compared wopicture and picture–picture long-term repetitipriming in several experiments. There wasinteraction between prime task and repetitwhen the manipulation of prime task was btween participants (the design adopted in present study), although these two factors interact when manipulated within subjec
Brown et al. interpreted their findings as evET AL.
lyerd &
otcyultef-nt
ti-d
ndct &llero-therdsn-pre-ic-ngain
n-tses,
dence against transfer-appropriate processaccounts of repetition priming. Therefore athough we did not set out to investigate the vlidity of transfer-appropriate processing acounts of repetition priming, were we to fail tfind a prime task by repetition interaction, ostudy would provide further evidence inconsitent with such accounts.
Method
Participants. The participants were 48 undegraduate students of the University of WaleCardiff, between ages 19 and 27 years. All wemonolingual English speakers with normal corrected-to-normal vision. Each was either paor received course credit for participating.
Stimuli. The critical stimuli were 48 picturestaken from Snodgrass and Vanderwart (198Statistics on the properties of the item sets agiven in Table 1. There were 24 with earlyacquired names (e.g.,cake, frog, pram, andsnowman) and 24 with later-acquired name(e.g., guitar, screw, vase, and toaster). The“early” items were selected such that thenames had AoA ratings (Barry et al., 1997) oless than 2.25 and the “late” items had Aoratings of greater than 2.90. The AoA ratingwere collected as part of a larger normativ
eso-/in
e ofndg ofy
tualatn-d–n
noone-heid
s.
study in which raters were asked to estimate
TABLE 1
Properties of the Sets of Items Used in the Early and LaAge-of-Acquisition Conditions of Experiment 1
Property Early acquired Late acquire
AoASubjective 1.95(0.19) 3.36(0.35)Objective (months) 35.7(28.5) 61.8(21.5)
Word frequencyCelex written 10.7(11.3) 8.7(11.4)Celex spoken 4.0(3.8) 3.8(4.8)Kucera and Francis 9.0(9.3) 9.9(9.2)
Object familiarity 2.93(0.77) 2.82(0.80)%Name agreement 95.5(8.4) 94.5(6.4)Image agreement 3.71(0.58) 3.68(0.76Visual complexity 3.01(0.92) 2.76(0.94)No. of phonemes 4.2(1.4) 5.0(1.7)
Note.The higher the AoA value the later the name w
i-acquired. Data presented as means with standard deviationsin parentheses.
d7s
r
ad
eet
seii
dthl
tedvr
oinutr
wurd
th
e
di-dudalfndthealf
theced
-i-t a
f atoe-
res,reesar-erg).n-hes. aee
ere
nd”i-
re.estof
msl.
the II,of a
REPETITION PRIMING
age at which they thought they had learneword, either in a written or spoken form, on apoint scale, giving a rating of 1 to word“learned at age 2 or under” and a rating of 7words “learned at age 13 or over” (the same pcedure as used by Gilhooly & Logie, 1980Thus, the mean AoA ratings of the early and litems correspond to their having been learneapproximately 2 to 4 years and 6 to 7 yearsage respectively. There were also clear diffences between the early- and later acquired itin terms of the mean age (in months) whennames were acquired, both for Morrison, Chapell, and Ellis’ (1997) measure ofAoA based onlogistic regression andwhen 75% of the childrenthey tested at any particular age could nameitem. Although the two sets of items hanonoverlapping distributions of AoA ratingthe difference between the two sets was nottreme. (This is not too surprising given that,general, concrete words, that is, those nampictureable objects, tend to be early-acquire
The two sets of pictures were matched formean word frequency of their names, using Cewritten frequency, Celex spoken frequency, aKucera and Francis (1967) and the object’s rafamiliarity and percentage name agreem(Barry et al., 1997) and phoneme length andnot differ on ratings of image agreement orsual complexity (Snodgrass & Vanderwa1980). All these measures were establishedthe names of the objects which were the mcommonly produced by British speakers, usthe normative data provided by Barry et al.; ththe word frequencies, AoA ratings, word lengand name agreements are for “pram” and “lorrather than for “baby carriage” and “truck.” Appendix A shows the items used, along with vaous characteristics of both the pictures and thnames.
The pictures were high-contrast line draings. Pictures were centred on the compscreen and subtended a visual angle of appmately 6 degrees. The words were presenteblack lowercase (on a white background)Geneva bold 18-point font in the center of computer screen.
Each set of 24 items was divided into two s
of 12, which were matched as closely as posOF PICTURE NAMING 357
a-
too-).teat
ofr-mshep-a
thed,x-
nng.)e
exndedntidi-t,forstgs,h,y”-ri-eir
-teroxi- inine
ts
ble on all variables. The 48 participants were vided into two groups of 24: group A name|pictures in stage I and group B read aloprinted names in stage I. Within each group, hof the participants received half of the early- ahalf of the late-acquired items as primed and remainder as unprimed, whereas the other hreceived the opposite arrangement. Thus, items presented in stage I were counterbalanover subgroups of participants.
Procedure. All participants were tested individually in a session, which lasted approxmately 20 min. Each participant was seated acomfortable distance in front of the screen oMacintosh II computer and was first asked say a few words while the sensitivity of a voickey was adjusted.
In stage I, the participants in group A wepresented with 24 critical and 48 filler picturerandomly intermixed, for naming. These wepreceded by six practice trials, using picturnot presented in the main experiment. The pticipants in group B were presented with thprinted names of the 24 critical and 48 filleobjects, randomly intermixed, for oral readin(which were preceded by six practice trialsBoth picture and word presentation was cotrolled by a specially written program whicalso timed the latencies of vocal responsPresentation of a picture or a word startedmillisecond timer that was terminated by thinitiation of the oral naming response madinto a high-sensitivity voice-key microphonconnected to the computer. Participants weinstructed to name each item “as distinctly aquickly as possible” and to avoid saying “umor “er.” Responses stopped the timer and termnated the presentation of the word or pictuNaming latencies were recorded to the nearmillisecond. The experimenter made a note all responses produced, noting any problewith the voice-key, and initiated the next triaThe typical intertrial interval was 2 to 3 s.
After a 5-min unfilled interval while the pic-ture naming experiment was being set up on computer participants began stage II. In stagethe procedure was identical for both groups participants. Participants were presented with
si-randomized order of the 48 experimental pic-th
e
g-
mv,
e
e
nseth
gedi-ininggses
ve.ced
ingn-ngvsts,nd
heac-r.
icleso-
-
ded.-
terwo
358 BARRY
tures, which were preceded by six practice als. The participants were not informed tsome of the items presented in stage I wouldrepeated in stage II. All other aspects of the pcedure were identical to stage I.
Results
The picture naming and word reading timfor four categories of response were exclud(i) picture misnaming “errors” that involved thproduction of alternative names (e.g., callinpicture of a spannera “wrench”), semantic errors (e.g., calling a picture of a screwdrivera“spanner”), and incorrect responses (there wno such word reading errors); (ii) various dysflent responses, such as saying “oh er” or “uwhich stopped the voice-key, or name retriedifficulties for pictures (such as saying “ohknow, it’s a . . .”); (iii) responses below 200 mwhich were either anticipations or the voice-kbeing activated by heavy breathing, and so and (iv) responses above 2 s, most of whwere either voice-key failures or hesitations.
Harmonic means of the latencies of corr(and nonexcluded) naming responses were culated, both for each participant (in each contion) and for each item. We calculated harmomeans rather than arithmetic means, medianmeans after variously defined outliers had bexcluded because the reciprocal transformathey entail has the effect of correcting for t
skew toward slower responses within latendistributions (Ratcliff, 1993).error and excluded responses in each condition are prestage for early-acquired over later-acquired items).
ET AL.
ri-at bero-
esd:
e a
ereu-
,”al
Is,y
on;ich
ctcal-di-ic, orenione
cy
Stage I: First presentation (picture naminand word reading). The arithmetic means (of thharmonic mean) of naming times in each contion of stage I of Experiment 1 are shownTable 2. There were no substantive word readerrors and, overall, only 1.6% of word readinresponses and 3.2% of picture naming responwere excluded using the criteria identified abo
Separate two-factor analyses of varian(ANOVAs) were performed by participants anby items on the harmonic mean correct namlatencies in each condition. Each ANOVA cotained the variables of priming task (namipictures vs naming words) and AoA (early- late-acquired). In the analysis by participanpriming task was a between-subject factor aAoA was a within-subject factor, whereas in tanalysis by items, AoA was a between-item ftor and priming task was a within-item factoUnless otherwise stated, throughout this artall differences referred to as significant are asciated with two-tailed ps < .01.
The main effect of priming task was significant [Fs(1, 46)= 41.78,MSe = 14905.0;Fi(1,46) = 149.37,MSe = 4307.71]; words were reaaloud 161 ms faster than pictures were namThe main effect of AoA was also highly significant [Fs(1, 46)= 58.02,MSe = 1574.06;Fi(1,46) = 18.55, MSe = 5600.17]; early-acquireditems were named 61 ms faster than were laacquired items. The interaction between the tvariables was also significant [Fs(1, 46)= 13.70,
MSe = 1574.06;Fi(1, 46)= 5.69,MSe = 4307.71,ct
ch par-ntage of
TABLE 2
Results of Stage I of Experiment 1
Picture naming Word reading
Early Late AoA Early Late AoAacquired acquired effect acquired acquired effe
Mean 686 778 92 555 587 32SE 17.1 22.6 14.6 15.9 17.7 7.1%Error 1.4 4.9 3.5 2.1 1.0 −1.1
Note.The SEs of AoA effects for latencies cited in Tables 2 and 3 were obtained by calculating AoA effects for eaticipant. Arithmetic means of harmonic mean picture naming and word reading latencies (in milliseconds) and perce
ented. Also shown are the magnitudes of the AoA effects (the advan-
e
he
e
r
nini-vsdbyb--Askl,II-
edg-
ain
REPETITION PRIMING
p < .05]. The AoA effect was larger for picturnaming (92 ms) than for word reading (32 msbut analyses of simple main effects showed tthe AoA effect was reliable for both picturnaming [Fs(1, 46)= 64.06,MSe = 1574.06;Fi(1,46) = 23.15,MSe = 4953.94] and word naming[Fs(1, 46)= 7.67, MSe = 1574.06;Fi(1, 46) =2.77,MSe = 4953.937,p < .05, one-tailed test].
The analysis of errors and excluded responshowed a significant interaction between Aoand priming task [Fs(1, 46) = 5.69, MSe = 0.309,p < .05; Fi(1, 46) = 6.28, MSe = 0.280, p < .05],which reflected the fact that, for picture naminmore errors were made to late- than to early-quired items [Fs(1, 46) = 6.74, MSe = 0.309, p <.05; Fi(1, 46) = 8.60, MSe = 0.242, p < .05],whereas for word reading there was no AoA fect (Fs < 1; Fi < 1).
Stage II: Second presentation (picture naming). The arithmetic means of the harmonmeans of the correct picture naming timeseach condition of stage II of Experiment 1 a
shown in Table 3. Separate three-factor ANOVAcluded responses in each condition for each priming taskadvantage of the repeated over the control conditions), a
OF PICTURE NAMING 359
),at
sesA
g,ac-
f-
-icine
were performed by participants and by items othe harmonic mean correct naming latencieseach condition. Each ANOVA contained the varables of priming task (phase I picture namingphase I word naming), AoA (early vs late), anpriming (repeated vs controls). In the analysesparticipants, priming task was a between-suject factor and AoA and priming were withinsubject factors. In the analyses by items, Aowas a between-item factor and both priming taand priming were within-item factors. Overal3.6% of picture naming responses in stagewere excluded following application of the criteria identified above.
The main effect of AoA was highly signifi-cant [Fs(1, 46) = 94.95, MSe = 3171.43; Fi(1,46) = 20.63, MSe = 18146.40]; early-acquireditems were named faster than later acquiritems. The main effect of priming was also sinificant [Fs(1, 46) = 51.15, MSe = 4673.07; Fi(1,46) = 44.50, MSe = 5861.39]; repeated itemswere named faster than control items. The m
seffect of the prime type was also significanting
and ex-
TABLE 3
Results of Stage II of Experiment 1
Early acquired Late acquired
Priming task Repeated Control Priming Repeated Control Prim
Picture namingMean 611 676 65 662 768 106SE 16.0 17.4 14.2 17.7 19.7 20.4%Error 2.8 4.2 1.4 2.4 4.9 2.5
AoA effect Repeated Control Overall
51 92 7111.0 14.1 10.1
Word namingMean 660 691 31 723 802 79SE 18.7 21.6 12.7 14.8 24.2 16.8%Error 2.1 1.7 −0.4 4.9 4.9 0
AoA effect Repeated Control Overall
63 111 8713.0 18.0 12.7
Note.Arithmetic means of the harmonic mean picture naming latencies (in milliseconds) and percentage of error
group. Also shown are the magnitudes of the repetition effects (thend the AoA effects.fi
o
i
1.ngre
ife.g
for-e aoftsf-dss
hem-theon-red
ac-
inf-heededr-
e ofar-e-d,
ect
tontl.,,heup-ith-eri-
360 BARRY
[Fs(1, 46) = 3.03, MSe = 24979.11, p < .05, one-tailed test; Fi(1, 46) = 24.30, MSe = 3986.130]:Overall, the participants who named picturesstage I were faster (by some 40 ms) to namepictures in stage II than those who read alowords in stage I. The interaction between priing task and priming did not achieve signicance in the analysis by participants [Fs(1, 46) =2.40, MSe = 4673.07, p = .13] but was reliable inthe analysis by items [Fi(1, 46) = 6.69, MSe =2311.45]; the repetition priming effect walarger for the group who named pictures in staI than for those who read aloud words (86 55 ms). The analyses of simple main effeshowed that there were significant repetition fects for both groups (all F ratios > 15.0). Thecritical interaction between AoA and priminwas significant [Fs(1, 46) = 13.90, MSe =1690.44; Fi(1, 46) = 4.58, MSe = 5861.39, p <.05]: The repetition priming effect was larger fpictures with late-acquired than with early-aquired names (93 vs 49 ms), although the pring effect was significant for both (all F ratios >10.0 in the analyses of simple main effect(Alternatively, this interaction could be described as a reduced AoA effect for the repeacompared to the control items; 57 vs 101 ms.
There was a small difference in the meword length of the early- and late-acquireitems. Although Barry et al. (1997) found thword length (in terms of number of phonemehad no independent significant effect on pictunaming latencies, an analysis of covariance items) was performed on the picture namidata from stage II with name length in phonemas a covariate. This revealed no significaeffect of length (t < 1), and the AoA effect re-mained significant [Fi (1, 45) = 21.31, MSe =18176.38].
In the analysis of the errors, the main effectAoA just reached significance in the analysis participants [Fs(1, 46) = 7.07, MSe = 0.239, p <.05], but did not achieve significance in thanalysis by items [Fi(1, 46) = 2.47, MSe = 0.684,p = .12]: There was a trend for more errors to
made to the late-acquired than to the earacquired items (2.7% vs 4.3%). No other effewere significant.in
ET AL.
in theud
m--
sgevsctsef-
g
rc-m-
s).-ted)and
ats)re
(byngesnt
ofby
e
bely-
Discussion
There were four main results of Experiment(1) There was an AoA effect on picture namilatencies; names acquired earlier in life weproduced faster than those acquired later in l(2) There was an AoA effect on word naminlatencies, although this was smaller than picture naming. (3) There were repetition priming effects on picture naming, both from thprior naming of the same pictures and, tolesser extent, from the prior reading aloud the objects’ names. (4) The repetition effecobserved interacted with AoA: The priming efects were reliably greater for late-acquirethan for earlier-acquired names. We discueach of these findings in turn.
There was a clear and robust effect of trated AoA of the object names on picture naing latencies when the word frequencies of names (and other variables) were carefully ctrolled. Pictures whose names were acquiearlier in life (such as comb) were named reli-ably faster than those whose names were quired later in life (such as vase). Indeed, theAoA effect was remarkably consistent withthe experiment: The magnitude of the AoA efect was 92 ms for the first presentation of tpictures in stage I and 102 ms for the unprimpresentation of the pictures in stage II (summover the two groups of participants). The intenal reliability of the AoA effect was thereforquite high, as it was found for the two subsetspictures we used and for the two groups of pticipants. Further, the AoA effect was not rducible to an effect of mean word lengths (anindeed, the number of phonemes in the objnames had no effect).
Our finding of an AoA effect lends support the previous reports that AoA is an importadeterminant of picture naming times (Barry et a1997; Ellis & Morrison, 1998; Morrison et al.1992, 1997). However, we maintain that tpresent study provides stronger evidence to sport the AoA case than has been reported herto; the present results come from an expmental study adopting a factorial design
ctswhich the word frequency of the objects’ names(and other important variables that have been
lyelnnn
te
dc
nin
i
ba
d
rh
tcnco be
e
r
oa
e-e
itisatendl.,geng.,ap-t,ted
elyd
ctes-reli-a-ct-
fa-f
notctne,p--nsesr).re-ueden
wengonal.ti-
REPETITION PRIMING
implicated in picture naming) were carefulcontrolled. Experiment 1 has therefore provida direct manipulation of AoA and found a highreliable (and fairly substantial) AoA effect opicture naming latencies. Following Morrisoet al. (1992), Barry et al. (1997), and Ellis aMorrison (1998), we propose that the AoA efect on picture naming latencies operates at level of retrieving a word’s phonology for thpurpose of spoken word production.
We also found a clear AoA effect on wornaming times in stage I when word frequenwas controlled. This replicates the results Morrison and Ellis (1995) and Gerhand aBarry (1998), who also found effects of AoA word reading times when frequency was cotrolled. Whatever the ultimate resolution of thdebate about the role of frequency in lexicprocessing with respect to that of AoA, the preent results confirm that AoA has a clear and dependent effect in word reading latencies,least for the concrete nouns used in Expement 1. The magnitude of the AoA effect oserved for word naming was, however, substtially (and significantly) smaller than that fonaming pictures. However, as has been founprevious studies (e.g., Potter & Faulconer, 1975),word naming times were faster than pictunaming times (532 vs 747 ms in stage I of texperiment). We propose that the AoA effeboth for word and picture naming operates at level of lexical retrieval in the speech prodution system. The process of mapping a semarepresentation to a word’s stored phonologiform is necessarily implicated in the task naming a picture. Although this process mayinvolved in the task of reading aloud a printword, there may also be some contribution sublexical spelling-to-sound phonological rcoding, which would circumvent activation ostored lexical phonology. We therefore interpthe difference in the magnitude of the AoA efect for word and picture naming in terms some contribution of sublexical phonologicrecoding to word reading.
There were clear long-term repetition primineffects on picture naming times from both thprior naming of the same pictures and the pri
r
OF PICTURE NAMING 361
dy
df-he
yofd
n-eals-n-atri--n-
r in
eiscthe-ticalfedof-fetf-fl
g
reading aloud of the objects’ names. The reptition effect was greater from naming the sampictures than from reading their names, butwas reliable in both conditions. This finding consistent with episodic and transfer-appropriprocessing accounts of repetition priming (ainconsistent with the results of Brown et a1991). Repetition priming of picture naminfrom the prior presentation of pictures has befound by a number of previous studies (e.Durso & Johnson, 1979, where the lag was proximately 5 min). In the present experimenthe average time between the first and repeapresentations of the pictures was approximat8 min. The repetition priming effects observehere are therefore quite long-lasting.
The faster naming of a picture of an objefollowing the prior naming of the same picturmay result from facilitation at a number of posible stages of processing underlying pictunaming. It may represent a long-lasting facitation of object recognition, of semantic activtion, or of lexical access. The repetition effefound from the prior reading aloud of the objects’ names must be due to the long-lasting cilitation of either the semantic activation othe named object or to lexical access; it canbe due to priming of the process of objerecognition. On the basis of these results alowe cannot rule out any contribution to the reetition effect from the facilitation of the activation of the objects’ semantic representatio(the semantic information activated by picturand printed names should be quite similaHowever, as argued in the introduction, we ainclined to believe that the long-lasting priming effects we have demonstrated are not dto semantic priming, the effects of which tento be short-lived and easily eliminated by thpresentation of intervening items betweeprime and target (see Neely, 1991). Instead,propose that the repetition effect from readialoud the object names reflects the facilitatiof the name retrieval process that Ellis et (1996) refer to as locus 2 priming: Having acvated a word’s phonological form from eithe
eorreading aloud a word or naming a picture, it iseasier to access it again when required as ai
oi ralo
po
inges.
ter n
0).g.,
tsrdtsne-
ionsin parentheses.
362 BARRY
naming response to a picture. We interpretadditional repetition priming effect from naming the same pictures over that from readaloud the object names as reflecting the conbution of what Ellis et al. refer to as locuspriming in the former condition.
The repetition priming effects we found inteacted with AoA, with the effects being larger fnaming objects whose names were acqulater in life than for those whose names werequired earlier in life. We interpret this inteaction by suggesting that the two effects hthe same locus. We assume that the phonocal representations of late-acquired words harder to access in order to be producedspeech. Words that have been previously duced in stage I are easier to produce again repeated presentation. For words that areready easier to access for speech produc(namely early-acquired words), then there wbe less scope for increasing access speedther, and so there will be less repetition primfor naming pictures with early-acquired namAs we also found a repetition priming effect fboth early- and late-acquired words, there malso be a general facilitation effect operatingthe stage of access to word’s phonology. If interpretation is correct, then we would also pect a priming effect in the following expement, in which we manipulate the frequencythe picture names while holding AoA constaHowever, if previous reports of effects of wofrequency on picture naming are the result ofconfounding of AoA and frequency, then wwould not expect to find any main effect of fr
quency nor any interaction between repetiti -ssi-s
asesal
s.
ible fa-t,r-
and frequency.
EXPERIMENT 2: FREQUENCY VARIEDAND AOA CONTROLLED
Experiment 2 investigated the priming of thnaming of pictures that varied in the word frquency of their names and where AoA (aother variables) were held constant.
Method
Participants. The participants were 48 students from the same population as Experimen
although none had taken part in that experimeorust athisx-
i-oft.
rdthee
e-on
ee-nd
-t 1,
Stimuli. The critical stimuli were 48 picturestaken from Snodgrass and Vanderwart (198There were 24 with high-frequency names (e.bottle, church, horse, and wheel) and 24 withlow-frequency names (e.g., axe, pear, scissors,and zebra). The means of the frequency counof these names, using four measures of wofrequency, are shown in Table 4. The two seof items had nonoverlapping distributions othree of the four frequency measures. We blieve that the difference in the word frequencies of the picture names was as large as poble given the constraints of matching the itemon other dimensions, and the difference waslarge as those typically manipulated in studiof frequency effects in other aspects of lexicprocessing.
As shown in Table 4, the two sets of picturewere very closely matched for ratings of AoAThey were also matched as closely as possfor percentage name agreement, rated objectmiliarity (Barry et al., 1997), image agreemenand visual complexity (Snodgrass & Vande
ET AL.
the-ngtri-1
r-r
redac--vegi-
are inro-n aal-
tionillfur-
TABLE 4
Properties of the Sets of Items Used in the High and Low Frequency Conditions of Experiment 2
High Low Property frequency frequency
Word frequencyCelex written 84.6(46.9) 4.4(2.4)Celex spoken 38.8(31.2) 2.3(2.8)Kucera and
Francis 92.6(67.9) 4.1(2.9)Hofland andJohansson 88.4(77.1) 3.5(3.9)
AoASubjective 2.40(0.42) 2.28(0.30)Objective (months) 35.5(15.0) 36.0(16.5)
Object familiarity 3.58(0.83) 2.82(0.81)%Name agreement 91.4(7.76) 96.0(6.04)Image agreement 3.60(0.65) 3.87(0.50)Visual complexity 2.52(0.88) 3.00(0.96)No. of phonemes 3.79(1.29) 3.79(0.98)No. of letters 5.08(1.67) 4.79(1.25)
Note.Data are presented as means with standard deviat
nt.wart, 1980) and for mean length of the names, in
t
tahs
c
ips
ns
mn
i
i-.ly
ereto
li-
cy
ofce
-dsifi-
re-ct,
2%
m-icchn
REPETITION PRIMING
number of phonemes and number of letters.measures (including frequencies) were eslished for the names of the objects most comonly produced by British speakers; thus, measures are for “pram” and not “baby criage,” for example. Appendix B presents titems used, along with various characteristicboth the pictures and their names.
Both the 24 high- and the 24 low-frequenitems were divided into two sets of 12, whiwere matched as closely as possible on all vables. Half of the participants in each groupand B) received half of the high- and half of tlow-frequency items as primed and the remader as unprimed, whereas the other half recethe opposite arrangement. Thus, the items sented in stage I were counterbalanced over groups of participants.
Procedure. This was the same as in Expement 1.
Results
As in Experiment 1, any naming responthat was incorrect or dysfluent or which failedstop the timer was excluded (as were respobelow 200 ms or above 2 s). Harmonic meancorrect responses were calculated, both for eparticipant in each condition and for each ite
Stage I: First presentation (picture naming aword reading). The arithmetic means of the hamonic mean picture naming and word readtimes in each condition of stage I of Experimen
are shown in Table 5. Separate priming taskcentage of error and excluded responses in each conditiofects (the difference between low- and high-frequency ite
OF PICTURE NAMING 363
Allab-m-her-e
of
cyh
ari-(Ahein-vedre-ub-
ri-
setoses ofach.dr-ngt 2
frequency ANOVAs were performed by particpants and by items on these harmonic means
The main effect of priming task was highsignificant [Fs(1, 46) = 120.97, MSe = 11291.72;Fi(1, 46) = 380.54, MSe = 3963.58]: Words wereread aloud 239 ms faster than pictures wnamed. The main effect of frequency failed reach significance [Fs(1, 46) = 1.47; Fi < 1]. Theinteraction between the two variables was reable by participants [Fs(1, 46) = 6.85, MSe =873.22, p < .05] but not by items [Fi(1, 46) =2.00, MSe = 3963.58, p = .16]. This interaction re-flected the fact that pictures with high-frequennames were actually named more slowly thanthose with low-frequency names; the analysissimple main effects showed that this differenwas significant by subjects [Fs(1, 46) = 7.34, MSe = 873.22] but not by items [Fi (1, 92) = 2.33,MSe = 3961.21, p = .13]. The small (9-ms) advantage of reading aloud high-frequency worfaster than low-frequency words was not signcant (Fs < 1; Fi < 1).
The analysis of the error and excluded sponses produced only one significant effethat of priming task [Fs(1, 46) = 19.46, MSe =0.619; Fi(1, 46) = 20.94, MSe = 0.575]: More er-rors were made to pictures than to words (8.vs 2.4%).
Stage II: Second presentation (picture naing). The arithmetic means of the harmonmeans of correct picture naming times in eacondition of stage II of Experiment 2 are show
byin Table 6. Separate priming task by frequency
fects forand per-
TABLE 5
Results of Stage I of Experiment 2
Picture naming Word reading
High Low Frequency High Low Frequencyfrequency frequency effect frequency frequency effect
Mean 765 742 −23 511 520 9SE 16.0 16.8 11.2 15.2 15.6 4.5%Error 9.4 6.9 −2.5 1.4 3.1 1.7
Note.SEs of frequency effects for latencies cited in Tables 5 and 6 were obtained by calculating the frequency efeach participant. Arithmetic means of harmonic mean picture naming and word reading latencies (in milliseconds)
n are presented. Also shown are the magnitudes of the frequency ef-ms).
ni
y
f
tplo
c
tsnt
e-
-w-ifi-
e-e
n
y-
vs
ing
and ex-cluded responses in each condition for each priming task group. Also shown are the magnitudes of the repetition effects (the
by priming condition ANOVAs were performeby participants and by items on the harmomean correct naming latencies in each condit
There was no main effect of frequency (Fs <0.1; Fi < 0.2); pictures with high-frequencnames were named only 2 ms faster than thwith low-frequency names. The main effect priming was highly significant [Fs(1, 46) =121.83, MSe = 2108.03; Fi (1, 46) = 77.83, MSe = 3622.79]; repeated items were namedms faster than the control items. The main efof priming task was also significant [Fs(1, 46) =11.01, MSe = 24784.70; Fi (1, 46) = 130.75,MSe = 2452.96]; the participants who named pictures in stage I were slower to name the tures in stage II than those who had read awords in stage I (718 vs 642 ms). Furthermopriming task interacted with priming [Fs(1, 46)= 17.60, MSe = 2108.03; Fi (1, 46) = 18.46, MSe = 2713.50]. The repetition priming effewas larger for those participants who namedpictures in stage I (101 ms) than for those w
advantage of the repeated over the control conditions) a
read aloud the object names in stage I (45 m
dic
on.
oseof
73ect
heic-ud
re,
ttheho
although analyses of simple main effecshowed that the repetition effect was significafor both groups (F > 14.0 for all effects). Therewas no interaction between priming and frquency (Fs < 1; Fi < 0.1); the repetition primingeffect was 77 ms for pictures with high-frequency names and 70 ms for pictures with lofrequency names. There were no other signcant effects (Fs < 1; Fi < 1).
The analysis of the error and excluded rsponses produced the following results. Thmain effect of frequency was just significant ithe analysis by participants [Fs(1, 46)= 5.05,MSe = 0.546,p < .05], but did not reach signifi-cance in the analysis by items [Fi(1, 46)= 1.47,MSe = 1.871,p = .23]; there were slightly fewererrors to pictures with high- than low-frequencnames (3.6% vs 5.6%). The main effect of priming hovered around significance [Fs(1, 46) =6.34,MSe = 0.435,p < .05;Fi(1, 46)= 3.60,MSe
= 0.766,p = .064]; there were fewer errors tothe repeated than to the control items (3.6%
nd the frequency effects.
364 BARRY ET AL.
TABLE 6
Results of Stage II of Experiment 2
High frequency Low frequency
Priming task Repeated Control Priming Repeated Control Prim
Picture namingMean 662 771 109 672 766 94SE 16.1 14.9 15.7 10.6 16.6 12.8%Error 5.6 4.5 −1.1 3.5 8.3 4.8
Frequency effect Repeated Control Overall
10 −5 211.0 16.9 9.5
Word namingMean 620 664 44 619 666 47SE 17.5 20.9 11.0 21.6 22.4 13.9%Error 1.7 2.4 0.7 3.5 6.9 3.4
Frequency effect Repeated Control Overall
−1 2 110.3 12.9 8.0
Note.Arithmetic means of the harmonic mean picture naming latencies (in milliseconds) and percentage of error
s),5.6%). The main effect of priming task was sig-
rege
eew
e
taoe
e
n
t
e
ae
u
ngioref-ndheWegesas-idctctef-ith
s,d-o-nd
ch-g of
sk,-
g.,7).
rdadthghegedcec-,ep-onin
eirxt,de-ds-
REPETITION PRIMING
nificant [Fs(1, 46)= 4.06,MSe = 0.566,p < .05;Fi(1, 46)= 6.19,MSe = 0.371,p < .05]; the par-ticipants who named pictures in stage I pduced more errors on picture naming in stagthan those who had read aloud words in sta(5.4% vs 3.7%). The interaction betwepriming and frequency was also just significa[Fs(1, 46)= 6.22, MSe = 0.523,p < .05; Fi(1,46) = 4.25,MSe = 0.766,p < .05]; for pictureswith high-frequency names, there was no diffence in errors between repeated and control it(3.7% vs 3.5%), whereas for pictures with lofrequency names, there were fewer errors topeated than control items (3.5% vs 7.6%). Thwere no other significant effects (Fs< 1; Fi < 1).
Discussion
There was no effect of the word frequency othe objects’ names on picture naming latencwhen the AoA of the names was held consand when other pertinent variables were ctrolled. No frequency effect was found in stagfor the group who named pictures or in stagewhere all participants named pictures. (Thwas a very small frequency effect on pictunaming accuracy in stage II only, but this wnot reliable in the analysis by items.) This ressupports most previous studies of picture naing (e.g., Morrison et al., 1992, 1997), but nothers; for example, Barry et al. (1997) fouthat frequency did have an effect but only signicantly for those items with later acquirenames. The present study represents whabelieve to be the first using a factorial expemental design to examine frequency effectspicture naming latencies while holding all othvariables—and in particular AoA—constanWhen AoA was controlled, there was no effof the frequency of the object names, evthough the difference in the frequency countsthe names of the pictures was quite substanThe absence of a frequency effect supportsresults reported by Morrison and Ellis (199but is not consistent with those of Gerhand Barry (1998), who found effects of both frquency and AoA on word naming times instudy that varied both variables factorially. Wshall consider the implications of this res
under General Discussion.OF PICTURE NAMING 365
o-II
e Innt
r-ms-re-re
fiesntn- I
II,rereasultm-otdif-d weri-onert.cten oftial.the5)nd-aelt
In Experiment 2 we found clear repetitiopriming effects of 102 ms from the prior naminof the same pictures and 46 ms from the prreading aloud of the object names. (These fects are of a similar magnitude to those fouin Experiment 1, although in that experiment teffects were larger for later acquired names.) interpret the “cross-domain” repetition priminfrom the prior reading aloud of the object namas reflecting what Ellis et al. (1996) refer to locus 2 priming, that is, to the facilitation of lexical access. The repetition priming effects dnot interact with the frequency of the objenames: when the AoA ratings of the objenames are carefully controlled, the repetition fects were essentially equal for pictures whigh- and low-frequency names (77 and 70 mrespectively). This result contradicts the finings of Wheeldon and Monsell (1992), whfound greater priming for pictures with lowfrequency names (from both word reading anaming to definitions), but they did not mattheir stimuli for AoA. The failure to find a repetition by frequency interaction in picture naminalso stands in marked contrast to the findingsprevious studies that show that, for word recog-nition as assessed by the lexical decision tarepetition effects are reliably larger for lowfrequency than for high-frequency words (e.Scarborough, Cortese, & Scarborough, 197However, Scarborough et al. found no interac-tion between frequency and repetition in a wonaming task (in their Experiment 3); instethey found an equal repetition effect for bohigh- and low-frequency words. Scarborouet al. included nonwords intermixed with thwords to name, and this may have encouraparticipants to read aloud all items by relianon sublexical phonological recoding, thus effetively “bypassing” lexical involvement whichwe have argued, is likely to be the source of retition effects. Interactions between repetitiand word frequency have not been found other reading paradigms; for example, in thstudy of eye movements while reading teRaney and Rayner (1995) found that the crease in fixation duration for repeated worwas similar in magnitude for low- and high
frequency words.resc-
neimrgoane
ad
e
n
kctee
n
t
oreri- totherac-sere-om
ereus
Aheyap- tri-ing to
ex-iestedre)gers ofachm.ach
e-res
e-
ed
366 BARRY
EXPERIMENT 3: DELAYED NAMING OF PICTURES
In both Experiments 1 and 2, we measuimmediate picture naming latencies. It is posble that the results could have been produartifactually by differential detectability of response initiation times by the voice-key duedifferences in the initial phonemes of the namof the objects we used in these two experimeor by the time to initiate the articulation of thnames. In order to test for (and, we hope, to elinate) these possibilities, we presented the ccal pictures in a delayed naming task. The loof measuring delayed naming times is as flows: Immediate picture naming latencies mreflect processes involved in picture recogtion, activation of the object’s semantic reprsentation, accessing of thephonology of thename, and processes involved in preparinginitiating articulation of the word. Delayepicture naming is assumed to assess onlycontribution of articulatory processes involvin producing the name. There is evidencesuggest that AoA affects word articulatiodurations: Gerhand and Barry (1998) aRoodenrys, Hulme, Alban, Ellis, and Brow(1994) found that later acquired words talonger to actually articulate than earlier aquired words in a task in which participanwere required to repeat words in rapid succsion. The study of delayed naming will, therfore, serve to establish whether the AoA effectobserved in Experiment 1 (and the failure to fia frequency effect in Experiment 2) was dueany differences arising from differential firs
phoneme detectability by the microphone for thh
aereiccisr
at 1intices
re-e
required to initiate the production of an object’s.
A
picture names used.
METHOD
Participants. The 48 participants were drawfrom the same population as before. None taken part in Experiments 1 or 2.
Stimuli. The stimuli were the experimentpictures used in Experiments 1 and 2. Thwere 48 pictures in the “AoA set” and 48 ptures in the “frequency set.” Half of the partipants named the pictures from the AoA (group A) and the remainder named the pictu
from the frequency set (group B).ET AL.
di-ed
toests
-iti-icl-yi--
nd
thedtond
ne-
ss--
dto-e
nad
l
Procedure. Each picture was presented f1.5 s using the same equipment as in Expments 1 and 2. Participants were instructedname the picture when it disappeared from screen (and not before) and were given 15 ptice trials to become familiar with this responrequirement. One set of 48 pictures was psented to each participant in a separate randorder. All other aspects of the procedure widentical to those in stage II of the previoexperiments.
Results and Discussion
The data from three participants, 2 in groupand 1 in group B, were excluded because tproduced the name before the picture dispeared from the screen on more than 15% ofals. These participants were replaced. Namresponses that were incorrect, which failedstop the timer, or which were made beforethepicture disappeared from the screen were cluded. Delayed naming times (i.e., latenccalculated from 1500 ms after the timer starat the onset of the presentation of the pictuthat were either shorter than 150 ms or lonthan 2 s were also excluded. Harmonic meancorrect responses were calculated, both for eparticipant in each condition and for each iteThe mean delayed picture naming times in econdition are shown in Table 7.
There were no reliable differences in the dlayed naming latencies either between pictuwith early- and later acquired names (Fs < 1; Fi< 1) or between pictures with high- and low-frquency names (Fs < 1; Fi < 1). Also, there wereno significant effects in the error and excludresponses (all Fs < 1). These results suggest ththe effect of AoA we observed in Experimentand the failure to find an effect of frequency Experiment 2 were not due to any systemadifferences in the initial phonemes of the namof the objects we used. More importantly, the sults show that there is no AoA effect in the tim
--etes
name once its phonology has been prepared
GENERAL DISCUSSION
Experiments 1 and 2 showed that the Ao
and not the frequency of object names deter-uo
t
hann
h
fh
uo
hc
nn
wr
avetedy
e-eiran
ey ac-ingef-nlythendw).
ine-ssi-l-nd,2)to-h-d,
ofryr,
ture
Note.Arithmetic means of harmonic mean picture naming latencies (in milliseconds) and percentage of error and excludedresponses in each condition. Also shown are the magnitudes of the AoA and frequency effects.
mines picture naming latencies and that pictnaming was primed both by the prior namingthe same picture and, to a lesser extent, byprior reading aloud of the object’s printed namFurthermore, repetition priming interacted wiAoA but not with frequency. Word naming timewere affected by AoA but not reliably by frequency. Experiment 3 found that there wereeffects of either AoA or frequency on delayepicture naming times, which indicates that tfinding of an AoA effect (and the failure to findfrequency effect) in immediate picture namiwas not an artifact. The theoretical implicatioof the results are considered below.
AoA and Frequency Effects in Word Reading
We found that early-acquired words weread aloud faster than later acquired words wfrequency was controlled; however, there wno reliable difference in the naming times high- and low-frequency object names matcfor AoA. The AoA effect on word naming timelends further weight to the AoA case and sports many previous findings. The absence frequency effect supports the results reportedMorrison and Ellis (1995) but is not consistewith those of Gerhand and Barry (1998), wfound independent (and noninteracting) effeof both frequency and AoA on word namitimes in a study that varied both variables ifully factorial design. All the words used in thpresent study were concrete nouns (as they the names of the pictures we used), wheGerhand and Barry’s stimuli sampled acro
the range of concreteness/imageability ratinref
thee.hs-node
gs
reenasoredsp-f a byntotsg aeere
easss
Strain, Patterson, and Seidenberg (1995) hreported that word naming latencies are affecby imageability, at least for low-frequencwords with irregular spelling-to-sound corrspondences; in particular, they found that thparticipants were slower to name abstract thconcrete low-frequency exception words. Thproposed that semantic representations arecessed automatically in the task of readaloud words. However, they argue that the fects of semantic activation are observed ofor low-frequency, exception words because “phonological processing of regular words ahigh-frequency words is too efficient to allomuch impact of the word’s meaning” (p. 1152Developing this argument, it is possible that,the reading of words with “rich” semantic reprsentations (namely concrete nouns), any poble effects of frequency will be overridden. Athough the object names used in this study (ain particular, the words used in Experiment were not selected to be matched for spelling-sound regularity, the majority of both the higand low-frequency words were regular. (Indeethe only irregular words were heart, comb, and,possibly, glove, two of which are actually of lowfrequency.) It would appear from the results Experiment 2 that frequency has only a veweak effect on word naming times for regulahighly concretewords.
It Is AoA and Not Frequency That DeterminesPicture Naming Latencies
There has been debate as to whether pic
REPETITION PRIMING OF PICTURE NAMING 367
TABLE 7
Results of Experiment 3 (Delayed Picture Naming)
Age of acquisition Word frequency
Early Late AoA High Low Frequencyacquired acquired effect frequency frequency effect
Mean 527 526 −1 530 525 −5SE 29.3 29.4 10.5 29.2 29.0 9.5%Error 6.8 7.6 0.8 8.0 7.6 −0.4
gs.naming is determined by AoA or frequency,
n,a
vr
e
i
le
tb
o
ec
n
fe
ud
ino-st,icetoli;hect-tedlly,ingar-udwn es-er-, it
uldcts orpe-nc- a
testl,nta-e- toion
n-c-neingnotne
we ef- be.
p-elyts.esand
368 BARRY
both in naming accuracy in aphasic patie(e.g., Hirsh & Ellis, 1994; Hirsh & Funnell1995) and in naming latencies in normal speers. So far, all the reported studies of the fects of AoA on picture naming latencies harelied on multiple regression analyses (Baet al., 1997; Carroll & White, 1973; Gilhooly &Gilhooly, 1979; Lachman, 1973; Lachmaet al., 1974; Morrison et al., 1992, 1997; Snograss & Yuditsky, 1996). Experiment 2 usedmore conventional (and arguably more powful) experimental design in which frequencwas varied and seven other variables (includAoA) were controlled and failed to find anyeffect of frequency on picture naming timeMany studies that have included AoA have afound no frequency effect (e.g., Carroll & Whit1973; Gilhooly & Gilhooly, 1979; Morrisonet al., 1992). Barry et al. (1997) found a reliabfrequency effect only for late-acquired itemBoth Ellis and Morrison (1998) and Snodgraand Yuditsky (1996) found a frequency effectaddition to one of AoA; the frequency effecmay have been restricted to late AoA words, neither study evaluated the interaction. The sults of Experiment 2 allow us, with some condence, to achieve closure on the debate concing the major determinant of picture naminlatencies: We have shown that it is the AoA the object’s name and not its frequency that fects the time to name a picture of that objeFurthermore, as the names of the high- and lfrequency items in Experiment 2 differed boon their frequency in print and, more impotantly, on their spokenfrequency, we may also bconfident that our failure to find any frequeneffect on naming cannot be due to the manipution of an inappropriate measure of frequencyis our position that frequency effects in pictunaming are unreliable and may be limited to olate-acquired words (see Barry et al., 1997).
Locus 1 and 2 Repetition Priming of Picture Naming
There were reliable long-term repetition priming effects in both Experiments 1 and 2, whethe prime was presented some 6 to 10 min beits corresponding target picture. The priming
fects were larger from the prior naming of thET AL.
ts
k-ef-ery
nd- ar-yng
s.so,
les.ssinsut
re-fi-ern-gofaf-ct.w-
thr-
yla-. Itrely
-reoref-
same pictures than from the prior reading aloof the object names.
We offer an interpretation of our results terms of “structural” rather than episodic theries of repetition priming for four reasons. Firin stage I of Experiments 1 and 2, we used twas many filler items as critical stimuli in order minimize any strategic learning of the stimuthis should have reduced the likelihood of tutilization of episodic traces when, unexpeedly, the stimuli were subsequently repea6 to 10 min later. Second, and more criticathe form of the stimulus-to-response processassociations were quite different for those pticipants in Experiments 1 and 2 who read aloprinted words in stage I and named line-drapictures in stage II. As the episodic instancestablished in stage I would have been quite diffent from those in stage II for these participantsis difficult to see that an episodic account coeasily explain any cross-domain repetition effeobserved. Third, even advocates of episodictransfer appropriate processing accounts of retition priming have argued that “lexical activatiois a critical component of priming effects and acounts for at least minimal transfer even withphysical mismatch between the prime and stimulus” (Brown, Neblett, Jones, & Mitchel1991, p. 523). Finally, we propose a represetional account of the effects of AoA and word frquency on picture naming and so would wishbe consistent in the nature of the interpretatoffered for repetition priming.
Repetition priming of picture naming latecies from the prior naming of the identical pitures could be due to facilitation from any o(or many) of the stages of processing underlypicture naming. As we argued earlier, we do believe that activation at the semantic level alocan be responsible for the long-lasting effectshave observed in Experiments 1 and 2, as thefects were too durable (and too substantial) tolikely to result solely from semantic primingDifferential persistence of form priming and reetition priming also appears to rule out a purpostlexical phonological locus for these effecThis leaves us with two other major possibilitito consider, the stages of object recognition
elexical access, or what Ellis et al. (1996) refer ton
v
t
e
oe
e
de
h
sd
bn-
rx
s
an
-ofha-ili-thenta- ort ofonlyayghton’sateht of
in-linens 2valat anat
val.
tslifeosere-e-tas
esd toedelyn-uc-edheed ofthe
REPETITION PRIMING
as locus 1 and locus 2 priming. Locus 1 primiinvolves the facilitation of the processes of peceptual recognition of familiar items; this is proposed to be “domain-specific,” where “domainis used to refer to “classes of stimuli that aprocessed by their own specialized recognitisystems” (p. 304). On this account, the prior sual presentation (and recognition) of an objewould prime the subsequent recognition of thobject by the temporary activation of the objecrepresentation in the recognition system (or changing the strength of the appropriate conntions between the recognition system and themantic system for that object). Locus 1 primineffects are the sort that are argued for by ththeorists who put forward episodic or transfappropriate processing accounts of repetitipriming. Locus 2 repetition priming involves thfacilitation of lexical access. Ellis et al. claimthat this does “not depend on repeating a stimlus in the same perceptual domain but rather pends on accessing the same name on two srate occasions” (p. 305). Thus, there can be bwithin- and cross-domain locus 2 priming.
In Experiments 1 and 2 we found botwithin-domain repetition priming (from namingpictures to naming the same pictures) and crodomain repetition priming (from reading printewords to naming pictures with those namesWe suggest that the within-domain effects oserved reflected a combination of locus 1 alocus 2 priming. Although the repetition priming of object recognition was not studied di-rectly in Experiments 1 and 2, we can estimaits contribution to picture naming times. The additional within-domain compared to cross-domain repetition priming may be taken asmeasure of the contribution of locus 1 repetitiopriming. The difference between the magnitudof the repetition priming effects from the prionaming of words and pictures was 31 ms in Eperiment 1 and 56 ms in Experiment 2. However, there is an alternative reason for findinless cross-domain than within-domain repetitiopriming. If there was some contribution of sublexical phonological recoding to the readinaloud of words in stage I of the experimentthen this would not involve the activation o
stored lexical phonology and so would have reOF PICTURE NAMING 369
gr--”reoni-ctat’sbyc-
se-gser
on
u-e-pa-
oth
s-
).-d
te--ane
--gn-g,
f
sulted in generating less locus 2 priming thpicture naming.
Of primary interest, however, is locus 2 priming. Having localized this effect to the level lexical access, we can distinguish two mecnisms through which it might be obtained. Factation could arise as a result of a change in strength of the connections between represetional levels, semantic-to-lemma connections,lemma-to-lexeme connections (enhancementhe processes of lexical identification/selectiand lexical retrieval respectively). Alternativethe availability of particular representations mbe enhanced. In a logogen-type model this micorrespond to an increase in a representatiresting level of activation or a change in the rof activation gain; in an attractor network it migcorrespond to a deepening of an item’s basinattraction. Our results do not allow us to distguish between these accounts. However, in with Wheeldon and Monsell (1994) and Griffiand Bock (1998), our preference is that locurepetition effects operate at the stage of retrieof the lexical-phonological form. We argue threpetition strengthens the association betweenitem’s lemma and its phonological form and ththis strengthening facilitates subsequent retrie
Explanations of AoA and Its Interaction withRepetition Priming
Experiment 1 found that pictures of objecwhose names were words acquired early in were named aloud faster than those whnames were acquired later in life when the fquency of the words was controlled. Also, reptition priming interacted with the AoA but nowith the frequency of picture names: There wa larger repetition effect for naming picturwhose names were acquired later comparethose acquired earlier in life. We have arguthat the locus of both effects is the same, namthe retrieval of lexical phonological represetations for the process of spoken word prodtion. The main effect of AoA may be accountfor by Brown and Watson’s proposal that tphonological representations of early-acquirwords are retrieved more easily than thoselater-acquired words. They suggested that
-“phonological output representations are storedln
ohe
lw
r-d
o eag
n
-deoa
e
the ati
et
oy
edlly
nof
reof
er-ndly,r-o-ac-nhee
ndmee-a-b-
inns)l.,o-
eirhatr toro-edalu-alin
ala-falc-itsala-or
a-
ede
370 BARRY
in a relatively complete form during the earstages of vocabulary acquisition” (p. 214) athat, for later-acquired words, “only minimal information is stored explicitly” (p. 215), due tpresumed storage limitations. Exactly how tstructure (and integrity) of phonological reprsentations is to be understood was not madeplicit by Brown and Watson’s “phonologicacompleteness hypothesis.” In order to devethe phonological completeness hypothesis, suggest that the lexical phonological representions (lexemes) of monosyllabic, monomophemic words are stored in terms of a threemensional matrix whose structure is the worinitial consonant (or consonant cluster) × thevowel × terminal consonant (or consonant cluter). This notion is an elaboration of the prposal advanced by Seymour (1987) that thethographic input lexicon might be constructin such a fashion. We suggest that early-quired words (e.g., frog) could be seen as beinstored as integral entries in the “cells” of thphonologically defined matrix (e.g., as /fr⊃ g/),but that late-acquired words (e.g., flag) arestored as vector points specifying positioalong each axis (e.g., as /fl/ + /æ/ + /g/). Inorder to produce a late-acquired word for naming, individual phonological elements woulneed to be retrieved, which takes more timthan the direct retrieval of the integral form an early-acquired word. (We note that later-quired words tend in general to be longer acertainly include many polysyllabic wordswhich may require further levels of segmentstorage.) Once a late-acquired word has hadelements retrieved (in the priming stage of experiment), it will be easier to retrieve those ements again when a naming response isquired later. These suggestions—and they speculative—enable us to account for both main effects of AoA and repetition and, constent with the interaction, to locate them boththe level of the retrieval of lexical phonology.
Some support for the phonological completness hypothesis may be gleaned from a selecreading of the literature on phonological deveopment in children. For example, in a reviewchildren’s vocabulary development, Walle
(1993) suggested that phonological word formET AL.
yd
-
e-ex-lop
eta--
di-’s
s--
or-d c-
is
s
fc-nd,d
itsel-re-re
hes-at
-ivel-f
“are not, at the outset, segmentally organizand that such structure emerges only graduawith vocabulary growth” (p. 329), and Wijne(1992) suggested that, at the 50-word stagevocabulary acquisition, phonological forms aholistic and unanalyzed. However, the namesthe early- and late-acquired items used in Expiment 1 were learned at approximately 3–4 a6–7 years of age respectively which, arguabare both beyond any such nonsegmentally oganized or holistic stage. Furthermore, the phnological completeness hypothesis does notcord easily with models of speech productio(e.g., Levelt et al., 1999; Roelofs, 1997) in whicword-form encoding is assumed to involve thassignment of a word’s phonemes to a framstructure. Indeed, in such models, the souforms of words are generated anew each tithey must be produced; “word forms are rtrieved from the mental lexicon not as unanlyzed wholes but rather as sublexical and susyllabic units, which are to be positionedstructures (such as word and syllable skeletothat are independently available” (Levelt et ap. 19). In such models, the stored lexical phonlogical representations ofall words need to besegmented and then recombined with thstored frame. Phonological representations tare assumed to be “complete”—and so hardesegment—should therefore not enjoy any pcessing advantage in models of this kind; indethe reverse should be the case. Rather than evate detailed models of how the phonologicforms of words are stored and processedspeech production, we simply wish to affirmthat it is AoA and not word frequency that hasclear and reliable effect upon picture namingtencies. This is the major empirical finding othe work reported in this article, and genermodels of speech production will need to acommodate it. We have argued that AoA haslocus at the stage of the retrieval of lexicphonology. It is likely that the stored representtions that enable the information required f(and so support) word-form encoding havelower threshold of required activation for earlyacquired words than those of late-acquirwords. This locus is entirely consistent with th
sresults of Experiment 1 that AoA interactedrhficafe-
he
0i.g
a
ruo
ln
oo
g
th
be
p
lid
tltst-
ngm-of-tmeere-m-erh
-h ofcte-ef-ndslyghren)gg
tsre ofd
atss-s
)c-
-d
REPETITION PRIMING
with repetition priming (and especially locuspriming). Experiment 2 found no reliable inteaction between repetition priming and tword frequency of picture names matchedAoA. The results of this experiment contradWheeldon and Monsell’s (1992) finding thcross-domain repetition priming was greaterpictures with low-frequency names; howevthey did not match their stimuli for AoA. In addition, the absence of an interaction suggethat word frequency effects—if they can bfound in picture naming—are not located at tlevel of lexical retrieval. Indeed, the results rported here force the conclusion that the crucvariable that needs to be modeled in pictunaming is AoA, not frequency.
Recently, Ellis and Lambon Ralph (200have shown how connectionist models can sulate age (or ratherorder) of acquisition effectsThey trained networks using the backpropation algorithm with cumulative (or interleavedpresentation: After early (or first-entered) pterns had been learned, later patterns wadded to the training set but the early pattealso continued to be presented. It was argthat this form of presentation mirrors normal vcabulary acquisition, where the occurrenceearly forms does not cease when later onesencountered. Ellis and Lambon Ralph’s simutions always showed effects of both AoA afrequency, although the AoA effects tendedbe larger in magnitude; this was shown mdramatically in a simulation in which a setearly entered low-frequency patterns had lowerror scores than a set of late entered hifrequency patterns. We foundno effect of wordfrequency on picture naming latencies, buclear effect of AoA. Ellis and Lambon Ralpdemonstrated the important point that orderacquisition (or training exposure) effects caninterpreted in connectionist models, but thsimulations used only abstract patterns. Itpossible that behavioral AoA effects—and escially whether they are accompanied by oserved frequency effects—may be differentdifferent domains such as picture naming, woreading, and word recognition. Further, Eland Lambon Ralph’s simulations did not a
dress repetition effects or how these might moOF PICTURE NAMING 371
2-eor
erate effects due to either AoA or frequency. Athis stage, we shall simply say that our resushow a clear AoA effect on picture naming buno frequency effect and that AoA (but not fre
ttorr,stsee-
ialre
)m-
a-)t-erensed-ofarea-dtostferh-
a
ofeirise-b-inrds-
quency) interacted with repetition.
CONCLUSION
We reported three experiments investigatipicture naming that were designed to (1) exaine the effects on naming latencies of the age-acquisition (AoA) and word frequency of objecnames, (2) contrast the repetition priming frothe prior naming of identical pictures and thprior reading aloud of the object names in ordto investigate what Ellis et al. (1996) have rferred to as locus 1 and locus 2 repetition priing for object naming, and (3) determinwhether repetition priming interacts with eitheAoA or frequency in order to determine whicvariable is the major determinant of picture naming and to inform theories of the locus at whicit has its effect. There were three main resultsthe experiments. (1) AoA had a significant effeon picture naming latencies when name frquency was controlled, but frequency had no fect when AoA was controlled. (2) The repetitioeffect was larger from having previously namethe same pictures compared to having previouread aloud the names of the pictures, althouthis also produced a priming effect. Thus, theare both long-lasting locus 1 (object recognitioand locus 2 (name retrieval) repetition primineffects in picture naming. (3) Repetition primininteracted with AoA but not with frequency. Inaddition, we found that (a) there were no effecof name AoA or frequency on delayed pictunaming latencies and (b) there was an effectAoA but no reliable effect of frequency on worreading latencies for object names.
To account for these results, we proposed ththe locus of the AoA effect is the stage of acceto lexical phonology for the task of spokenword production, which is the same locus alocus 2 repetition priming, which is why the twoeffects interact. Ellis and Lambon Ralph (2000have shown that connectionist models can acount for AoA effects; however, their simulations always produced effects of both AoA an
d-frequency and were not applied to modeling
tr
iu
372 BARRY
repetition priming. Our finding that AoA, bunot word frequency, affects spoken-word pduction presents a major challenge to Ellis aLambon Ralph’s modeling, and future work wbe necessary to specify in adequate comp
tional detail the mechanism by which AoA ha6
9
n-,al.,e
(Kucera & Francis, 1967); AoA = the rated age of acquisi-
asss &nt
its powerful effect.
APPENDIX A
Names of the Stimuli Used in Experiment 1
In Experiment 1, AoA was varied and frequency wheld constant. (Names in parentheses are the SnodgraVanderwart, 1980, names of those items given differnames by British participants.)
Items with early-acquired names
Name of object ImAg Fam %NA Freq AoA
balloon 4.33 2.86 100 10 1.85banana 4.42 3.71 90.9 4 1.80butterfly 3.92 2.73 100 2 2.20cake 3.45 3.32 100 13 1.90carrot 4.50 4.23 100 1 2.05coat 2.59 3.77 100 43 1.85comb 3.78 3.68 100 6 2.10doll 2.28 2.5 72.7 10 1.60drum 3.71 2.41 100 11 2.00duck 3.85 2.59 81.8 9 1.95elephant 3.85 2.18 100 7 2.05frog 3.60 2.38 90.9 1 2.10leaf 3.88 3.41 100 12 2.05lion 3.88 1.91 100 17 1.75lorry
(truck) 2.80 3.41 77.3 0 2.10mouse 4.22 2.59 81.8 10 2.05orange 4.00 3.73 100 23 1.85pear 4.62 3.23 100 6 2.16pig 3.62 2.36 95.5 8 1.80pram (baby
carriage) 3.65 2.27 100 1 1.75shoe 3.02 4.68 100 14 1.50snail 3.33 2.45 100 1 2.20snowman 4.00 2.18 100 0 1.95tiger 3.82 1.77 100 7 2.10
Mean 3.71 2.93 95.4 9.0 1.94SD .58 .77 8.4 9.3 .19
Items with later acquired names
ImAg Fam %NA Freq AoA
ET AL.
o-ndllta-s
asss &ent
anchor 4.32 1.73 100 15 3.35arrow 2.27 3.27 90.9 14 3.25ashtray 3.20 3.5 100 0 3.35belt 4.05 3.81 95.5 29 3.00bow 2.67 2.36 81.8 15 3.25cigar 2.75 2.23 95.5 10 3.90cigarette 4.65 3.86 100 25 3.35envelope 4.70 4.27 95.5 21 3.45flag 3.22 2.22 100 16 3.20flute 3.41 1.91 95.5 1 3.65guitar 4.20 3.00 95.5 19 3.35kangaroo 4.30 1.41 95.5 0 3.05lamp 3.26 3.73 95.5 18 2.95light bulb 4.42 3.41 77.3 0 3.20pineapple 4.60 2.36 86.4 9 3.05pliers 4.22 2.24 86.4 1 4.35ruler 3.98 3.82 100 3 3.20screw 3.67 2.77 95.5 21 3.42screwdriver 4.30 2.73 100 0 3.45spanner
(wrench) 2.51 2.55 86.4 0 4.10toaster 3.92 3.86 100 0 3.45trumpet 2.89 2.05 95.5 7 3.10vase 2.72 2.50 100 4 3.25violin 4.18 2.14 100 11 2.95
Mean 3.68 2.82 94.5 9.9 3.35SD .76 .80 6.4 9.2 .35
Note.ImAg = rated image agreement (Snodgrass & Vaderwart, 1980); Fam = rated object familiarity (Barry et al.1997); %NA = percentage of name agreement (Barry et 1997); Freq = the word frequency of the object’s nam
tion of the object’s name (Barry et al., 1997).
APPENDIX B
Names of the Stimuli Used in Experiment 2
In Experiment 2, frequency was varied and AoA wheld constant. (Names in parentheses are the SnodgraVanderwart, 1980, names of those items given differenames by British participants.)
Items with high-frequency names
Name of object ImAg Fam %NA Freq AoA
bottle 2.85 4.41 95.5 76 2.10box 2.90 3.64 100 70 2.00chain 4.46 2.57 95.5 50 3.25church 2.98 3.09 95.5 348 2.70cigarette 4.65 3.86 100 25 3.35desk 3.18 4.09 90.9 65 2.80
dress 2.30 3.14 100 67 2.05glass 4.40 4.45 95.5 99 2.352
,a
ge of
ces of
des
B.g:
t:pre-y.
d la-y
n-
thein
).el-
of’s
-
nd
i-
i--
-its.
i- ofc-
REPETITION PRIMING
gun 3.82 2.00 90.9 118 2.75hat 3.65 2.59 95.5 56 2.10heart 4.49 3.09 90.9 173 2.45horse 4.20 2.82 100 117 1.85iron 4.08 3.05 100 43 2.90key 4.58 4.68 95.5 88 2.35knife 3.25 4.82 95.5 76 2.10moon 3.15 3.32 90.9 60 2.10mountain 3.52 2.41 86.4 33 2.95ring 3.08 3.82 95.5 47 2.30table 3.42 4.50 100 198 1.85telephone 4.28 4.36 63.6 76 2.20train 3.20 3.64 100 82 2.40watch 3.18 4.27 100 81 2.60wheel 3.48 2.68 86.4 56 2.16window 3.25 4.64 95.5 119 2.00
Mean 3.60 3.58 94.1 92.6 2.40SD .65 .83 7.8 67.9 .42
Items with low-frequency names
ImAg Fam %NA Freq AoA
ant 2.92 2.81 86.4 6 2.55axe 4.50 2.14 100 6 2.95banana 4.42 3.71 90.9 4 1.80carrot 4.50 4.23 100 1 2.05clown 3.25 2.09 100 3 2.35comb 3.78 3.68 100 6 2.10duck 3.85 2.59 81.8 9 1.95frog 3.60 2.38 90.9 1 2.10glove 3.65 2.91 90.9 9 2.25hammer 4.10 2.82 100 9 2.45kite 4.10 2.14 100 1 2.60owl 4.10 2.18 100 2 2.45peach 3.28 3.01 81.8 3 2.45pear 4.62 3.23 100 6 2.16pram (baby
carriage) 3.65 2.27 100 1 1.75scissors 4.40 3.91 100 1 2.65snail 3.33 2.45 100 1 2.20sock 3.72 4.73 100 4 1.80spider 2.95 3.09 95.5 2 2.26stool 4.12 3.50 90.9 8 2.35swan 3.69 2.23 100 3 2.50tiger 3.82 1.77 100 7 2.10whistle 4.55 2.45 100 4 2.45zebra 4.05 1.41 95.5 1 2.50
Mean 3.87 2.82 96.0 4.1 2.282SD .50 .81 6.0 2.9 .30
Note.ImAg = rated image agreement (Snodgrass & Vaderwart, 1980); Fam = rated object familiarity (Barry et al.1997); %NA = percentage of name agreement (Barry et 1997); Freq = the word frequency of the object’s nam
Kucera & Francis, 1967); AoA = the rated age of acquisitionof the object’s name (Barry et al., 1997).OF PICTURE NAMING 373
n-
l.,e
REFERENCES
Barry, C., Morrison, C. M., & Ellis, A. W. (1997). Naminthe Snodgrass and Vanderwart pictures: Effects of agacquisition, frequency and name agreement. QuarterlyJournal of Experimental Psychology, 48A, 560–585.
Barry, C., Johnston, R. A., & Scanlan, L. C. (1998). Are fa“special” objects? Associative and semantic primingface and object recognition and naming. QuarterlyJournal of Experimental Psychology, 51A, 853–882.
Bartram, D. J. (1974). The role of visual and semantic coin object naming. Cognitive Psychology, 6, 325–356.
Brown, A. S., Neblett, D. R., Jones, T. C., & Mitchell, D. (1991). Transfer of processing in repetition priminSome inappropriate findings. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 17,514–525.
Brown, G. D. A., & Watson, F. L. (1987). First in, first ouWord learning age and spoken word frequency as dictors of word familiarity and word naming latencMemory & Cognition, 15, 208–216.
Carroll, J. B., & White, M. N. (1973). Word frequency anage of acquisition as determiners of picture-namingtency. Quarterly Journal of Experimental Psycholog,25, 85–95.
Centre for Lexical Information. (1993). The Celex lexicaldatabase.Nijmegen, The Netherlands: Max Planck Istitute for Psycholinguistics.
Coltheart, M. (1985). Cognitive neuropsychology and study of reading. In M. I. Posner & O. S. M. Mar(Eds.), Attention and performance XI(pp. 3–37). Hills-dale, NJ: Erlbaum.
Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993Models of reading aloud: Dual-route and paralldistributed-processing approaches. Psychological Re-view, 100, 589–608.
Coltheart, V., Laxon, V. J., & Keating, C. (1988). Effects word imageability and age of acquisition on childrenreading. British Journal of Psychology, 79, 1–12.
Durso, F. T., & Johnson, M. K. (1979). Facilitation in naming and categorizing repeated pictures and words. Jour-nal of Experimental Psychology: Human Learning aMemory, 5, 449–459.
Ellis, A. W. (1993). Reading, writing and dyslexia: A cogntive analysis(pp. 24–31). Hillsdale, NJ: Erlbaum.
Ellis, A. W., Flude, B. M., Young, A., & Burton, A. M.(1996). Two loci of repetition priming in the recogntion of familiar faces. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 295–308.
Ellis, A. W., Kay, J., & Franklin, S. (1992). Anomia: Differentiating between semantic and phonological deficIn D. I. Margolin (Ed.), Cognitive neuropsychologyin clinical practice(pp. 207–228). New York: OxfordUniv. Press.
Ellis, A. W., & Lambon Ralph, M. A. (2000). Age of acqusition effects in adult lexical processing reflect lossplasticity in maturing systems: Insights form conne
tionist networks. Journal of Experimental Psychology:Learning, Memory, and Cognition, 26, 1103–1123.e-yn-
d
e-
ic-nd
ess-
n-ty,xi--
-
ty:
innce
2).e
-
i-n
.e
ncyysg-
tsre-
e-st
a
n
-k
a
o
-
d
8
374 BARRY
Ellis, A. W., & Morrison, C. M. (1998). Real age of acquistion effects in lexical retrieval. Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 24,515–523.
Evett, L. J., & Humphreys, G. W. (1981). The use of astract graphemic information in leixcal access. Quar-terly Journal of Experimental Psychology, 33A,325–350.
Forster, K. I., & Chambers, S. M. (1973). Lexical access naming time. Journal of Verbal Learning and VerbaBehavior, 12, 627–635.
Frederiksen, J. R., & Kroll, J. F. (1976). Spelling and souApproaches to the internal lexicon. Journal of Experi-mental Psychology: Human Perception and Perfomance, 2, 361–379.
Gerhand, S., & Barry, C. (1998). Word frequency effectsoral reading are not merely age-of-acquisition effectsdisguise. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 267–283.
Gilhooly, K. J., & Gilhooly, M. L. M. (1979). Age-of-acquisition effects in lexical and episodic memory tasMemory & Cognition, 7, 214–223.
Gilhooly, K. J., & Logie, R. H. (1980). Age-of-acquisitionimagery, concreteness, familiarity and ambiguity meures for 1944 words. Behaviour Research Methods anInstrumentation, 12, 395–427.
Gilhooly, K. J., & Logie, R. H. (1981). Word age-of-acquistion, reading latencies and auditory recognition. Cur-rent Psychological Research, 1, 251–262.
Grainger, J. (1990). Word frequency and the neighborhfrequency effects in lexical decision and word naminJournal of Memory and Language, 29, 228–244.
Griffin, Z. M., & Bock, K. (1998). Constraint, word frequency, and the relationship between lexical processlevels in spoken-word production. Journal of Memoryand Language, 38, 313–338.
Hillis, A. E., & Caramazza, A. (1991). Mechanisms for acessing lexical representations for output: Evidenfrom a category-specific semantic deficit. Brain andLanguage, 40, 106–144.
Hirsh, K. W., & Ellis, A. W. (1994). Age of acquisition anlexical processing in aphasia: A case study. CognitiveNeuropsychology, 11, 435–458.
Hirsh, K. W., & Funnell, E. (1995). Those old, familiathings: Age of acquisition, familiarity and lexical access in progressive aphasia. Journal of Neurolinguis-tics, 9, 23–32.
Hofland, K., & Johansson, S. (1982). Word frequencies inBritish and American English.Bergen: The NorwegianComputing Centre for the Humanities.
Humphreys, G. W., Riddoch, M. J., & Quinlan, P. T. (198Cascade processes in picture identification. CognitiveNeuropsychology, 5, 67–103.
Jacoby, L. L. (1983). Perceptual enhancement: Persiseffects of an experience.Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 9, 21–38.
Jacoby, L. L., & Dallas, M. (1981). On the relationship b
tween autobiographical memory and perceptual leaing. Journal of Experimental Psychology: General, 3,ET AL.
306–340.Jescheniak, J. D., & Levelt, W. J. M. (1994). Word fr
quency effects in speech production: Retrieval of stactic information and of phonological form. Journal ofExperimental Psychology: Learning, Memory, anCognition, 20, 824–843.
Kirsner, K., Smith, M. C., Lockhart, R. S., King, M. C., &Jain, M. (1984). The bilingual lexicon: Language-spcific units in an integrated network. Journal of VerbalLearning and Verbal Behavior, 23, 519–539.
Kroll, J. F., & Potter, M. C. (1984). Recognizing words, ptures and concepts: A comparison of lexical, object areality decision times. Journal of Verbal Learning andVerbal Behavior, 23, 39–66.
Kucera, H., & Francis, W. H. (1967). Computational analy-sis of present-day American English.Providence, RI:Brown Univ. Press.
Lachman, R. (1973). Uncertainty effects on time to accthe internal lexicon. Journal of Experimental Psychology, 99, 199–208.
Lachman, R., Shaffer, J. P., & Hennrikus, D. (1974). Laguage and cognition: Effects of stimulus codabiliname-word frequency, and age-of-acquisition on lecal reaction time. Journal of Verbal Learning and Verbal Behavior, 13, 613–625.
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in speech production. Behavioraland Brain Sciences, 22, 1–75.
Logan, G. D. (1990). Repetition priming and automaticiCommon underlying mechanisms. Cognitive Psychol-ogy, 22, 1–35.
Meyer, D. E., & Schvaneveldt, R. W. (1971). Facilitation recognizing pairs of words: Evidence of a dependebetween retrieval operations. Journal of ExperimentalPsychology, 90, 227–234.
Meyer, D. E., Schvaneveldt, R. W., & Ruddy, M. G. (197Activation of lexical memory.Paper presented at thPsychonomic Society Meeting, St. Louis, MO, November, 1972.
Mitchell, D. B., & Brown, A. S. (1988). Persistent repettion priming in picture naming and its dissociatiofrom recognition memory.Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 14,213–222.
Monsell, S. (1985). Repetition and the lexicon. In A. WEllis (Ed.), Progress in the psychology of languag(Vol. 2, pp. 147–195). London: Erlbaum.
Monsell, S. (1991). The nature and locus of word frequeeffects in reading. In D. Besner & G. W. Humphre(Eds.), Basic processes in reading: Visual word reconition (pp. 148–197). Hillsdale, NJ: Erlbaum.
Monsell, S., Doyle, M. C., & Haggard, P. N. (1989). Effecof frequency on visual word recognition tasks: Wheare they? Journal of Experimental Psychology: General, 118, 43–71.
Monsell, S., Matthews, G. H, & Miller, D. C. (1992). Reptition of lexicalization across languages: A further te
i-
b-
ndl
d:
r-
in in-
s.
,s-
d
i-
odg.
ing
c-ce
r-
).
tent
e-
of the locus of priming. Quarterly Journal of Experi-mental Psychology, 44A, 763–783.rn-
eal
le
em
e
on
o
e
il
nll
e
de-ni-r
ib-nd
ni-ri,
of
Se-
g-
vi-
g-
izedage
nd
for
-e-hy.
t intion
on
of
enor.
in
REPETITION PRIMING
Morrison, C. M., Chappell, T. D., & Ellis, A. W. (1997). Agof acquisition norms for a large set of object names their relation to adult estimates and other variabQuarterly Journal of Experimental Psychology, 50A,528–559.
Morrison, C. M., & Ellis, A. W. (1995). Roles of word frequency and age of acquisition in word naming and ical decision. Journal of Experimental PsychologyLearning, Memory, and Cognition, 21, 116–133.
Morrison, C. M., Ellis, A. W., & Quinlan, P. T. (1992). Agof acquisition, not word frequency, affects object naing, not object recognition. Memory & Cognition, 20,705–714.
Morton, J. (1985). Naming. In S. Newman & R. Epst(Eds.), Current perspectives in dysphasia(pp. 217–230).Edinburgh: Churchill–Livingstone.
Neely, J. H. (1991). Semantic priming effects in visual wrecognition: A selective review of current findings atheories. In D. Besner & G. W. Humphreys (EdsBasic processes in reading: Visual word recogniti(pp. 264–336). Hillsdale, NJ: Erlbaum.
Newton, P. K., & Barry, C. (1997). Concreteness effectsword production but not word comprehension in dedyslexia.Cognitive Neuropsychology, 14,481–509.
Oldfield, R. C., & Wingfield, A. (1965). Response latencin naming objects. Quarterly Journal of ExperimentaPsychology, 17, 273–281.
Potter, M. C., & Faulconer, B. A. (1975). Time to understapictures and words. Nature, 253, 437–438.
Ratcliff, R. (1993). Methods for dealing with reaction timoutliers. Psychological Bulletin, 114, 510–532.
Ratcliff, G., & Newcombe, F. (1982). Object recognitioSome deductions from clinical evidence. In A. W. E(Ed.), Normality and pathology in cognitive function(pp. 147–171). London: Academic Press.
Raney, G. E., & Rayner, K. (1995). Word frequency effeand eye movements during two readings of a tCanadian Journal of Experimental Psychology, 49,151–173.
Richardson, J. T. E. (1975). The effect of word imagability in acquired dyslexia.Neuropsychologia, 13,281–288.
Roelofs, A. (1997). The WEAVER model of word-form encoding in speech production.Cognition, 64, 249–284.
Roodenrys, S., Hulme, C., Alban, J., Ellis, A. W., & BrowG. D. A. (1994). Effects of word frequency and ageacquisition on short-term memory span. Memory &Cognition, 22, 695–701.
Rubin, D. C. (1980). 51 Properties of 125 words: A uanalysis of verbal behavior. Journal of Verbal Learning
and Verbal Behavior, 19, 736–755.Scarborough, D. L., Cortese, C., & Scarborough, H. (1977). Frequency and repetition effects in lexicmemory. Journal of Experimental Psychology: HumaPerception and Performance, 3, 1–17.
OF PICTURE NAMING 375
ndes.
-x-
:
-
in
rdd
.),n
inp
es
nd
e
:iss
ctsxt.
e-
-
n,of
nit
Scarborough, D. L., Gerard, L., & Cortese, C. (1984). Inpendence of lexical access in bilingual word recogtion. Journal of Verbal Learning and Verbal Behavio,23, 84–99.
Seidenberg, M. S., & McClelland, J. L. (1989). A distruted, developmental model of word recognition anaming. Psychological Review, 96, 523–568.
Seymour, P. H. K. (1979). Human visual cognition.London:Collier Macmillan.
Seymour, P. H. K. (1987). Developmental dyslexia: A cogtive experimental analysis. In M. Coltheart, G. Sarto& R. Job, (Eds.), The cognitive neuropsychology language(pp. 351–395). Hillsdale, NJ: Erlbaum.
Strain, E., Patterson, K., & Seidenberg, M. S. (1995). mantic effects in single-word naming. Journal of Ex-perimental Psychology: Learning, Memory, and Conition, 21, 1140–1154.
Snodgrass, J. G., & McCullough, B. (1986). The role ofsual similarity in picture categorization. Journal of Ex-perimental Psychology: Learning, Memory, and Conition, 12, 147–154.
Snodgrass, J. G., & Vanderwart, M. (1980). A standardset of 260 pictures: Norms for name agreement, imagreement, familiarity, and visual complexity. Journalof Experimental Psychology: Human Learning aMemory, 6, 174–215.
Snodgrass, J. G., & Yuditsky, T. (1996). Naming timesthe Snodgrass and Vanderwart pictures.Behavior Re-search Methods, Instruments, and Computers, 28,516–536.
Tabachnick, B. G., & Fidell, L. S. (1988). Using multivari-ate statistics.New York: Harper & Row.
Treiman, R., Mullennix, J., Bijeljac-Babic, R., & RichmondWelty, E. (1995). The special role of rimesin the dscription, use, and acquisition of English orthograpJournal of Experimental Psychology: General, 124,107–136.
Walley, A. C. (1993). The role of vocabulary developmenchildren’s spoken word recognition and segmentaability. Developmental Review, 13, 286–350.
Warren, C., & Morton, J. (1982). The effects of priming picture recognition. British Journal of Psychology, 73,117–129.
Wheeldon, L. R., & Monsell, S. (1992). The locusrepetition priming of spoken word production.QuarterlyJournal of Experimental Psychology, 44A, 723–761.
Wheeldon, L. R., & Monsell, S. (1994). Inhibition of spokword production by priming a semantic competitJournal of Memory and Language, 33, 332–356.
Wijnen, F. (1992). Incidental word and sound errors
S.aln
young speakers. Journal of Memory and Language, 31,734–755.
(Received December 1, 1997)(Revision received June 15, 2000)