The Concreteness Effect in Implicit and Explicit Memory Tests · concreteness effect (better memory for concrete than abstract nouns) in free recall and the explicit general knowledge

Journal of Memory and Language44, 96–117 (2001)doi:10.1006/jmla.2000.2749, available online at http://www.idealibrary.com on

The Concreteness Effect in Implicit and Explicit Memory Tests

Maryellen Hamilton

Baruch College, City University of New York

and

Suparna Rajaram

State University of New York at Stony Brook

Four experiments were conducted to examine the concreteness effect in implicit and explicitmemory measures. Experiment 1 replicated prior reports of an imagery effect on an implicitconceptual memory test. In Experiment 2, we confirmed our prediction of conceptual sensitivity offree recall, explicit general knowledge, explicit word fragment completion, and the implicit generalknowledge tests with a levels of processing manipulation. Furthermore, although we obtained theconcreteness effect (better memory for concrete than abstract nouns) in free recall and the explicitgeneral knowledge test, we failed to find this effect in the implicit general knowledge test. Experi-ment 3 revealed that the failure to find the concreteness effect on the implicit general knowledge testwas not attributable to combining two encoding manipulations in Experiment 2. In Experiment 4, weruled out the possibility that the failure to find the concreteness effect in conceptual implicit memorymay be related to the number of meaningful associates for targets. We discuss the implications ofthese findings within the context of the transfer appropriate processing framework (Roediger,Weldon, & Challis, 1989) and the dual-code hypothesis (Paivio, 1971, 1991) of memory.© 2001

Academic Press

ofer

eo

wncep-hreerds,ingtractare a91)ectsex-

altract

inntsks

tici-deon-nce

re-

rt-LedaVe

ate-Jaananrte

f thPAg o

This1 t

parony76.elle

ol-ail

In this article, we focus on the naturepriming in conceptual memory tests. The expiments we report were motivated by the th

We thank Richard Delaney, Kevin Felker, William Galand, Mana Heydarpour, Laurance Lemmerling, KathyEllen Martin, Tara Price, Maura Rzeznikiewicz, AmanSalomon, Jennifer Sanders, Sandra Siegel, Christinedino, and Carmine Vozzolo for their help in creating mrials and testing participants. We are grateful to Larrycoby, Roddy Roediger, Mary Susan Weldon, andanonymous reviewer for providing helpful comments onearlier version of this article. Parts of the research repoin this article were presented at the annual meeting oEastern Psychological Association in Philadelphia,1996; Washington, DC, 1997; and at the annual meetinthe Psychonomic Society in Philadelphia, PA, 1997.research was supported by NIH Grant R29MH57345-0Suparna Rajaram.

Address correspondence and reprint requests to SuRajaram, Department of Psychology, SUNY at StBrook, Stony Brook, NY 11794-2500. Fax: (631) 632-78E-mail: [email protected], or to MaryHamilton, Department of Psychology, Saint Peter’s Clege, 2641 Kennedy Blvd., Jersey City, NJ 07306. E-m
[email protected].
960749-596X/01 $35.00Copyright © 2001 by Academic PressAll rights of reproduction in any form reserved.

--

retical conceptualization of three well-knovariables that are presumed to engender contual or meaning-based processing. These tvariables are encoding pictures versus woimaging an item’s referent versus readwords, and encoding concrete versus absnouns. These variables are assumed to shcommon mechanism in Paivio’s (1971, 19dual-code hypothesis. We considered the effof these variables on conceptual forms ofplicit and implicit memory with an empiricfocus on the effects of concrete versus absnouns.

Explicit and implicit memory tasks differthe test instructions given to the participa(Graf & Schacter, 1985). Explicit memory tassuch as recall and recognition require parpants to think back to the prior study episoand perform the memory task at hand. In ctrast, implicit memory tasks make no refereto the study episode and participants are

,

r-

-

de,f

o

na

n

:
quired to complete the task at hand with the first

theos

geryonpi

sulipug o93er

s olicionrdtictheer-in-93nt

onmeicitlap, i

ci-ath

le-o-th

indeximesnsrovks).enag

anhavcerdifitte

vio,ionIm-tteref-ered

n ex-all,(see

s ofap-pant6;e,e-d a

icit6;

avemels-5).sscep-eo-icitory

rityess

insis.hishichd atns)d toic-ns,andg are-msetens

o

97CONCRETENESS EFFECT

answer that comes to mind. According totransfer appropriate processing framework ptulated by Roediger and colleagues (RoediWeldon, & Challis, 1989), explicit memotasks typically entail considerable use of cceptual processing for retrieving the study esode (but also see Blaxton, 1989). As a rethese tasks benefit from experimental manlations which enhance conceptual processininformation (see Roediger & McDermott, 19for a review). Implicit memory tasks can eithrely predominantly on perceptual processeon conceptual processes. For instance, imptasks such as word fragment complet(_ l e _ h _ n _; solution: “elephant”) or wostem completion (ele_______) require paripants to complete the word puzzle withfirst solution that comes to mind and prefentially rely on perceptual processing offormation (see Roediger & McDermott, 19for a review). Here, we focus on a differeclass of implicit memory tests that relyconceptual processes. In these tests, notion is made of the study episode but implmemory is aided by the conceptual overbetween study and test items. For instancethe implicit category production test, partipants are provided with a conceptual cuetest that bears no physical resemblance tostudy item (e.g., for the studied item “ephant,” the category label “animal” is prvided at test) and are asked to write downexemplars of that category that come to mThe increase in the generation of studiedemplars over nonstudied exemplars, or pring, provides the measure of memory on thimplicit tests. Experimental manipulatiothat enhance conceptual processing imppriming on these conceptual implicit tas(Blaxton, 1989, Srinivas & Roediger, 1990

The effects of the three variables of presinterest, picture versus word processing, iming the referent versus reading words,studying concrete versus abstract nouns,been well established on measures of contual explicit memory and are straightforwa(see Paivio, 1971, 1983 for reviews). Speccally, pictures are recalled or recognized be
than words (picture superiority effect) on ex-
-r,

--t,-f

rit

-

n-

n

te

e.--e

e

t-

de

p-

-r

plicit memory tests such as free recall (PaiRogers, & Smythe, 1968) and recognit(Shepard, 1967; see also Madigan, 1983).aging the referent of a word produces beretention than nonimaged words (imageryfect), and concrete nouns are better remembthan abstract nouns (concreteness effect) oplicit memory measures including free recrecognition, and paired associate learningPaivio, 1971, 1983, for reviews).

In contrast to these outcomes, the effectthese variables on conceptual implicit taskspear to be less clear and sometimes discre(Blaxton, 1989; McDermott & Roediger, 199Nyberg & Nilsson, 1995; Weldon & Coyot1996; Wippich & Mecklenbrauker, 1995). Spcifically, some researchers have failed to finpicture superiority effect on conceptual implmemory tests (McDermott & Roediger, 199Weldon & Coyote, 1996), and yet, others hreported a positive imagery effect on this saclass of tasks (Blaxton, 1989; Nyberg & Nison, 1995; Wippich & Mecklenbrauker, 199Our aim in this article is to empirically addrethese discrepancies in the results with contual implicit tests and provide a coherent thretical description of these effects on expltasks as well as conceptual implicit memtasks.

As mentioned earlier, the picture superioeffect, the imagery effect, and the concreteneffect are assumed to be related effectsPaivio’s (1971, 1991) dual-code hypotheAccording to the simplest interpretation of thypothesis, there are two possible codes wcan be activated when an item is presentestudy, verbal (logogens) or imaginal (image(Paivio, 1978). These two codes are assumebe additive in nature. It is postulated that ptures, items studied with imagery instructioand concrete words activate both the verbalthe imaginal code, whereas simply readinword only activates one code (verbal). Thefore, the superior memory for pictures, itestudied with imagery instructions, and concritems is the result of the additive contributioof two codes to memory.

Under this possibility, the activation of tw
codes requires cross-system, or referential, pro-

t isis

thegeenriat iss tal-crenda

sinof

refeesturofro

hatyptua

ouleridetheep

d tanle ootheryconemergcee

er-ventsm’singen

tedif-

cep-sferr ettual(i.e.,rd,ere

mictestallyrth

icitpro-

ofto

heicit

ithwasorerdsthese

ro-ler,-

licitlicitlledith

ud-di-theall)nter,of

im-withon,

ctionp heyw rlya thatc e isp t cue

98 HAMILTON AND RAJARAM

cessing (Paivio, 1971, 1991). Specifically, iassumed that if the encoding of a pictureaccompanied with invoking its name, bothverbal and imaginal systems become engathrough referential processing. Similarly, whparticipants are required to image the pictorepresentation of a word (e.g., sailboat), iassumed that referential processing enableactivation of both codes. Finally, in the ducode hypothesis, it is also assumed that conwords facilitate the activation of logogens aimagens, whereas abstract words can onlycess logogens, and thus, referential procesis likely for the former and not the latter setitems (see Paivio, 1983, for a review).

Thus, all three effects of interest are psumed to be guided by cross-system, or reential, processing. Because referential procing is assumed to be conceptual in na(Paivio, 1991, pp. 280–281), the activationdual codes implies one type of conceptual pcessing of information. It should be noted treferential processing represents only oneof conceptual processing and that concepprocesses during encoding and retrieval cbe of different types depending on the expmental manipulation employed to engenconceptual processing. Within the context ofresearch reported here, our definition of conctual processes at encoding will be restrictereferential processing across the verbalimaginal codes. Based on this assumed roreferential processing in the dual-code hypesis, the picture superiority effect, the imageffect and the concreteness effect may besidered as conceptually driven effects on mory. As such, these three effects should emon memory tests that are classified as contual tests, both explicit and implicit, within thtransfer appropriate processing framework.

To examine this idea, Blaxton (1989, Expiment 3) examined the imagery effect in fimemory tasks. In this experiment, participawere asked to form mental images of the itereferent for half the items presented durstudy. No specific study instructions were givfor the remaining items. Blaxton construcfive different memory tests by combining d
ferent test instructions (explicit and implicit) a
d

l

he

te

c-g

-r-s-e

-tel

d-r

-odf

-

--e

p-

and types of processing (perceptual and contual) so as to test the predictions of the tranappropriate processing framework (Roedigeal., 1989). Of these five tests, two conceptests, free recall and semantic cued recalltarget is cued with a semantically related woe.g., study: universe; test: cosmos____ ), wexplicit memory tests. The third test, graphecued recall, was an explicit perceptualwhere the target was cued with a perceptusimilar word (e.g., unversed______). The foutest, word fragment completion, was an implperceptual test where participants werevided with a perceptually degraded versionthe target ( _ _ i v _ _ s e) and were askedprovide the first solution that came to mind. Tfifth test, general knowledge, was an implconceptual test. Participants were provided wconceptually related test questions (“Whatthe Big Bang said to have created?) which bno physical resemblance to the studied wo(universe) and were asked to answer withfirst solution they could think of to solve thequestions.

As predicted by the transfer appropriate pcessing framework (Roediger, Weldon, Stad& Riegler, 1989), Blaxton found a typical imagery effect on all conceptual tests, exp(free recall and semantic cued recall) or imp(general knowledge). More items were recaor questions answered with words studied wthe imagery instructions than with words stied without the imagery instructions. In adtion, she found no imagery effect in eitherexplicit perceptual task (graphemic cued recor the implicit perceptual task (word fragmecompletion; see also Durgunoglu & Roedig1987, for converging evidence). This findinga positive imagery effect on a conceptualplicit test has been subsequently replicateda category association task (Nyberg & Nilss1995) and a verb association task1 (Wippich &

1 In the verb association task, participants studied ahrases (i.e., holding the tire) within a text. During test tere provided with verbs (half of which were from eaction phrases) and asked to fill in the first associationsame to mind (e.g., holding _______). Note that thererceptual overlap between the study item and the tes
nd therefore this task does not meet the criterion suggested

ndailstharfor.ct,

omenntlyo

ionaningro-asstoreexsa

erctro-5)an

n).edn-96to

steten

toryerrentdicro

uredes1)imn,

p-theButrityottis

edudy-cep-a-hecifi-ndn,

&ior-Mc-te,onsinetheheseit is

y forkelyvan-ren-ac-

singab-on

ion,983,cit

geineOurdiesre-theep-x-

t onrstectur

verde


Mecklenbrauker, 1995). Together, these fiings support the notion that imagery entconceptual processing (Paivio, 1991) andthis conceptual manipulation enhances pemance on conceptual implicit memory tests

With respect to the picture superiority effethe reported findings are quite different frthose found with the imagery effect. In a recstudy, Weldon and Coyote (1996) consistefailed to find a picture superiority effect on twimplicit conceptual tests (category productand word association) such that pictureswords produced equivalent amounts of primAt the same time, a significant levels of pcessing effect (Craik & Lockhart, 1972) wfound on both tests, indicating that the tewere indeed conceptual in nature. Furthermwhen the same two tests were given withplicit instructions, recall was better for itemstudied as pictures than for items studiedwords. Similarly, McDermott and Roedig(1996) failed to find a picture superiority effeon the implicit conceptual test of category pduction. In fact, only one study (Nicolas, 199has reported a picture superiority effect onimplicit conceptual test (category productioIt is not clear why Nicolas (1995) has obtaina picture superiority effect on an implicit coceptual test while Weldon and Coyote (19and McDermott and Roediger (1996) failedfind one. Weldon and Coyote have suggethat one possibility may be that he used intional study instructions, which may have ledexplicit contamination on the implicit memotest. However, Weldon and Coyote and McDmott and Roediger’s failure to find the pictusuperiority effect is the focus of the curreresearch since this finding seems to contrathe predictions of the transfer appropriate pcessing framework.

Thus, if we assume that imaging and pictprocessing both entail activation of two cothrough referential processing (Paivio, 199then the imagery effect on the conceptualplicit tasks (Blaxton, 1989; Nyberg & Nilsso

earlier for conceptual implicit tests (i.e., no perceptual olap between study and test cues). However, we inclu
here since the authors classify it a conceptual implicit task
-

t-

t

d.

s,

-

s

)

d-

-

t-

,-

1995; Wippich & Mecklenbrauker, 1995) suports both the dual-code hypothesis andtransfer appropriate processing framework.on this view, the absence of a picture superioeffect on conceptual implicit tasks (McDerm& Roediger, 1996; Weldon & Coyote, 1996)problematic for both accounts.

To our knowledge, there are no publishstudies that have examined the effects of sting concrete versus abstract nouns on contual implicit memory. This lacuna in the literture becomes particularly striking in light of tdiscrepancy we have just discussed. Specally, positive imagery effects have been fouon implicit conceptual memory tests (Blaxto1989; Nyberg & Nilsson, 1995; WippichMecklenbrauker, 1995) but no picture superity effect has been found on these tasks (Dermott & Roediger, 1996; Weldon & Coyo1996). Therefore, in order to explore the reasfor this discrepancy it is essential to examwhether a theoretically related effect (i.e.,concreteness effect) would be obtained on ttasks. Based on the dual-code hypothesis,reasonable to assume that superior memorconcrete nouns relative to abstract nouns liarises from the conceptual processing adtage (arising from the cross-system or refetial processing) for concrete nouns. Thus,cording to the transfer appropriate procesframework, an advantage for concrete overstract nouns should be obtained not onlyexplicit tasks such as free recall, recognitand paired associate learning (see Paivio, 1for a review), but also on conceptual implitasks. We set out to test this prediction.

We selected the implicit general knowledtask as our conceptual implicit task to examthe effects of the concrete/abstract variable.selection of this task was guided by the stuthat have tested the effects of theoreticallylated variables. In particular, we selectedimplicit general knowledge test as our conctual implicit task because Blaxton (1989, Eperiment 3) reported a robust imagery effecthis task. Having selected this task, we fisought to replicate the positive imagery eff(Blaxton, 1989) in our laboratory and with o-it
subject pool. To this end, we used Blaxton’s.

liciul

es-agim-

M

esorrkthooudoothe

i-ionourereus

im-raereinnc

ta

nef ofrucplygeionldd)imfoludy

pseoro).

40theuc-tednt.allyd aereenhatin

re-ici-ainkedrd.r orheyonimetheeryke

ion.rent

antsPar-

eri-es-of

ere-rs.

ues-hattionar-r tosted

ereons,hadalfn-es-ues-the


(1989) general knowledge task as our impconceptual test. It was expected that we wofind a typical imagery effect; i.e., more qutions answered with items studied under imery instructions than items studied withoutagery instructions.

EXPERIMENT 1

ethod

Participants. Eighty-three undergraduatfrom SUNY at Stony Brook participated fcredit as partial requirement for course woThe undergraduates were recruited fromPsychology Department human subject pand informed consent to participate in the stwas obtained. Of these 83 participants, 35 tpart in a norming study and 48 took part inexperiment.

Materials and norming.We used 35 particpants to norm 132 general knowledge questselected from Blaxton’s (1989) materials forsubject population. Blaxton’s materials wused because these materials had previoproduced a positive imagery effect on anplicit conceptual test. From the 132 geneknowledge questions, a pool of 80 items wselected based on the results of our normstudy. The overall mean base-rate performafor these items was 19%.

Design. We used a 23 2 within-subjecdesign. A given participant saw 40 wordsstudy and the other 40 words served asitems at test. For each list of 40 words, halthe items were studied under imagery insttions and the other half of the items were simread (nonimaged). The imaged and nonimaconditions were presented in a blocked fashCounterbalancing was done for item type (onew), study instructions (imaged/nonimageand the presentation order of study blocks (aged followed by nonimaged vs nonimagedlowed by imaged), thereby creating eight stlists.

Procedure.Participants were tested in grouof one to three. Items were presented on Z486 computers using Schneider’s (1990) MicExperimental Laboratory (MEL; Version 1.0
Participants took part in both a study and a tes
td

-

.elyk

s

ly

l

ge

tw

-

d.

/,--

s-

phase. All participants were presented withwords during study for 5 s each. Twenty ofwords were presented with imagery instrtions. Specifically, participants were instructo form a mental image of the item’s refereThe participants were asked to hit a specimarked “Y” key as soon as they had formemental representation of the item. They wtold that the word may still remain on the screfor a predetermined amount of time and tduring this time they should keep the imagetheir mind. The remaining 20 items were psented with nonimaged instructions. Partpants were told to simply read the words. Agthey were instructed to hit the specially mar“Y” key as soon as they had read the woParticipants were asked to read no slowefaster than they normally do. Once again twere instructed that the word would remainthe screen for a predetermined amount of tduring which they should continue to readword. Participants in conditions where imaginstructions were given first were told to masure they did not image in the “read” conditas this would slow down their reading times

The test phase was presented as a diffeexperiment to help ensure that the participdid not associate the study and test phase.ticipants were told that for the second expment they would have to answer trivia qutions. They were instructed that the majoritythe questions had multiple solutions, and thfore, there were no right or wrong answeParticipants were instructed to answer the qtions with the first word that comes to mind tanswers the question appropriately. No menof the prior study episode was made. All pticipants took part in a practice session priothe test. Test instructions and practice laapproximately 10 min.

During the test phase the participants wpresented with 80 general knowledge questione question at a time. Half of the questionstarget answers from the earlier study list (hfrom the imaged condition and half from noimaged condition). The other half of the qutions had nonstudied target answers. Each qtion appeared on the computer screen, with
tprompt, “If you have a solution, Press the Y

heeyhitpesenap

ticithetud

tionasesiedan

phae

no sisw ent asf ,M

arpor

redingin

-

eecttud-ms

ery

’sthry

t of

thege,tionetelicitves ef-askon-askr a

im-


Key.” Participants were instructed to hit tspecially marked “Y” key as soon as ththought of a one word solution. Once theythe “Y” key, participants were instructed to tyin their response. After typing in the responor if the participants did not hit the “Y” key ithe allotted time (20 s), the next questionpeared automatically.

At the end of the testing phase, each parpant was debriefed and questioned aboutawareness of any relationship between the sand test phase.

Results and Discussion

None of the participants noticed a connecbetween the study phase and the test phFigure 1 illustrates the mean proportion of qutions answered with target items for the studimaged items, the studied nonimaged items,for new items. For all results reported, the allevel was set atp , .05. Unless otherwisoted, the first statistical value (e.g., theF valuer the t value) represents the subject analyhereas the second statistical value repres

he item analysis. A significant difference wound among the means [F1(2,141)5 18.84Se 5 .03;F2(2,237)5 18.93,MSe 5 .04]. To

determine if priming occurred, planned compison t tests were performed between the pro

FIG. 1. Mean proportion of questions answereaged, and new items in Experiment 1.

tion of questions answered with target studied

,

-

-iry

e.-

d

,ts

--

items and the proportion of questions answewith target nonstudied items, for each encodcondition. Significant priming was obtainedboth the imaged condition [t1(47)5 8.37,SE5.02; t2(79) 5 9.36, SE 5 .02] and the nonimaged condition [t1(47)5 5.52, SE 5 .02;t2(79)5 4.66,SE5 .02]. In order to determinif there was an imagery effect, a within subjt test was performed between the priming (sied–nonstudied) proportions of imaged ite(M 5 .21) and nonimaged items (M 5 .09).This comparison revealed a significant imageffect [t1(47)5 4.65,SE5 .02; t2(79)5 4.81,SE 5 .02]. Thus, we replicated Blaxton(1989) findings. Specifically, we found bosignificant priming and a significant imageeffect on our implicit conceptual memory tesgeneral knowledge.

Having replicated the imagery effect onimplicit conceptual test of general knowledwe proceeded to examine our central quesof interest, i.e., the effect of studying concrversus abstract nouns on conceptual impmemory. In Experiment 2, we included fimemory tasks to examine the concretenesfect on memory. We included a free recall tin order to replicate the well-established ccreteness effect on an explicit memory t(Paivio & Csapo, 1969; see Paivio, 1983 fo

nd standard errors) with studied imaged, studied non
d (a
review). Our critical measure was the concep-

itsevis oicitthetasoniseteicactsuransker-m-nt

anamot

hisnded

m-tion.elsip-al

reteskep

nn,er,995)fecer-t isthex

ro-dgee

thewaally

licitif

con-ing,thisthis

esisll-108nddi-ookionn,ral.

atusct),ereestentitbe-

tedon-alfere

andithnta-se-er,to

hus,stud-

andforbal-

t


tual implicit task of general knowledge andselection was based on the findings from prous studies (Blaxton, 1989) and the resultour Experiment 1. We also included an explversion of this task in order to examineconcreteness effect on a conceptual explicitthat met the retrieval intentionality criteri(Schacter, Bowers, & Booker, 1989). Thatwe sought to confirm the presence of a concrness effect in an explicit task that was identto the conceptual implicit task in all respeexcept for the retrieval instructions. To oknowledge, there is no published report ofexplicit version of the general knowledge taFinally, we also included two widely used pceptual tasks, the implicit word fragment copletion task and its explicit version of fragmecued recall. The effect of encoding concreteabstract items has not been previously exined for this category of tasks. We did nexpect any effect on the implicit version of ttask because this task is at the perceptual ethe processing continuum. While we includan explicit version of the word fragment copletion task we did not have a strong predicfor this task (we return to this issue shortly)

In Experiment 2, we also included the levof processing (Craik & Lockhart, 1972) manulation in our experiment to confirm the typicassumptions regarding the processing natuthese five tasks. Past research has documenlevels of processing effect on conceptual tabut not on tasks that are predominantly perctual in nature (Challis & Sidhu, 1993; Hama1990; Srinivas & Roediger, 1990). Howevbased on past research (Roediger, et al., 1Weldon, Roediger, Beitel, & Johnston, 199we also predicted a levels of processing effor fragment cued recall (i.e., the explicit pceptual task). A levels of processing effecbelieved to occur on this test, even thoughcue is perceptual (word fragment), becauseplicit retrieval in itself demands conceptual pcessing (the participant needs to make a jument about whether an item has been searlier). Therefore, this contamination ofperceptual test with conceptual processingexpected to aid performance for semantic
encoded words. On this logic, a small concrete
-f

k

,-l

.

d-

of

ofd as-

2;,t

e-

-n

s

ness effect may also be expected on the expword fragment completion task. However,concreteness of words is not as strong aceptual manipulation as levels of processthe concreteness effect may not occur onpredominately perceptual task (we return toissue under General Discussion).

EXPERIMENT 2

Method

Participants.A total of 324 undergraduatfrom SUNY at Stony Brook participated in thexperiment for credit needed for partial fulfiment of a course requirement. Of these,participants took part in the norming study a216 in one of the five experimental test contions. Of these 216 participants, 40 each tpart in four different tasks, free recall, questcued recall, implicit word fragment completioand fragment cued recall. In the implicit geneknowledge task, 56 participants were tested

Design.In this experiment a 23 2 3 2 3 5mixed-factorial design was used. Studied st(old vs new), item type (concrete vs abstraand levels of processing (deep vs shallow) wmanipulated as within-subject variables. Ttype (free recall, question cued recall, fragmcued recall, implicit word fragment, or implicgeneral knowledge) was manipulated as atween-subject variable.

From the pool of 80 critical items, we creatwo lists of 40 items (20 abstract and 20 ccrete). Each list was further divided so that hthe items (10 abstract and 10 concrete) wshown with semantic encoding instructionsthe remaining half of the items were shown wgraphemic encoding instructions. The presetion of items was blocked with respect tomantic and graphemic encoding. Altogetheight possible study lists were createdachieve appropriate counterbalancing. Tacross participants each item appeared as aied or nonstudied item and for semanticgraphemic encoding. In addition, the orderthese two encoding conditions was alsoanced across participants.

Materials and norming.All items, both a
-study and at test, were presented and data were

ingo-e-

eted tt 2vioofret

nestoth

onaTheaeratendte

weimsser

a bili

10ngere

dif-s o

taS b-

s o-

qd ate

t—n—?)dgepre

m-

rateeral6%.ical) didor-

oas

b idhu( ingc ainl rdc keda ist.A s”a thes ici-p item( onsw 8 oft g., Ist usedo on-c emsw tracti pers d a“ ireda

nta-t n-c durew hei hep tiga-t y-o ringe sedo achi fours odingc od-i e-s timet on( ndi this


collected on Zeos 486 computers by usSchneider’s (1990) Micro-Experimental Labratory (MEL; Version 1.0) software system. Bcause our manipulation of interest (concrness) was inherent in our items, we needeselect a new set of materials for ExperimenA set of 80 nouns was selected from the PaiYuille, and Madigan (1968) norms. Fortythese items were chosen to be highly conc(scores above 6 on the 7-point concretescale,M 5 6.74) and 40 items were chosenbe abstract (scores between 1.5 and 4.5 onconcreteness scale,M 5 2.69) in nature. Paiviet al. also normed items on two additioscales; imageability and meaningfulness.imageability scale is a 7-point scale, wherthe meaningfulness scale reflects the numbmeaningful associates a participant generfor a given word. Since the imageability ameaningfulness scales are highly correlawith the concreteness scale, concrete itemsaccompanied with high scores on both theageability (M5 6.39) and the meaningfulne(M 5 6.44) scales, and the abstract items wccompanied with low scores on the imagea

ty (M 5 3.62) and the meaningfulness (M55.42) scales. All items were between 5 toletters in length and of medium frequency raing from 10 to 49 (M5 25) occurrences pmillion words in print according to Thorndikand Lorge (1944). There was a significantference between concrete and abstract itemthe concreteness scale,t(78)5 26.50,SE5 .15,he imageability scale,t(78)5 31.21,SE5 .09,nd the meaningfulness scale,t(78) 5 8.28,E 5 .12. No significant differences were o

served between concrete and abstract itemboth word length,t(78) 5 1.77, and word freuency,t(78) 5 1.03.For each item, both a word fragment an

general knowledge question were crea(e.g., target—strawberry; word fragmens t _ a _ _ e _ _ y; general knowledge questioWhat fruit wears its seeds inside its skinThese word fragments and general knowlequestions were normed to establish an accable baseline rate of accuracy. One hundeight participants took part in one of four nor
ing studies. Based on the four norming studiesg
-o.,

es

e

lesofs

dre-

e-

-

n

n

d

.et-d

we ensured that the nonstudied baseline hitfor both the word fragments and the genknowledge questions was, on average, 2Furthermore, we ensured that our two critsets of items (abstract and concrete nounsnot statistically differ on their baseline perfmance on word fragment completion,t(78), 1,r general knowledge questions,t(78) 5 1.32.Our levels of processing manipulation w

ased on a method used by Challis and S1993). For the graphemic (shallow) encodondition, participants were asked if a certetter was in the word (e.g., Does this woontain a “c”?). Participants were never asbout the same letter twice within a given lpproximately half of the answers were “yend the other half were “no” responses. Foremantic (deep) encoding condition, partants were asked about the meaning of ane.g., Can you buy this?). A pool of 77 questias created. Normative data showed that 3

he questions focused on concrete ideas (e.his a person?), whereas 39 questions focn abstract ideas (e.g., Is this desired?). Crete questions were used for concrete ithile abstract questions were used for abs

tems. Each question appeared only oncetudy list and half of the questions requireyes” response, whereas the other half requ

“no” response.Our pilot data showed that single prese

ion yielded weak priming on the implicit coeptual test. Therefore, the encoding proceas slightly different for the participants in t

mplicit general knowledge test. Following trecedence established in previous inves

ions (Challis & Sidhu, 1993; Weldon & Cote, 1996), we used massed repetition duncoding in the implicit conceptual tests. Ban Challis and Sidhu’s (1993) parameters, e

tem appeared four times consecutively andeparate questions were asked in each encondition. For example, in the semantic encng condition the word “strawberry” was prented four times consecutively, and eachhis word was followed by a different questii.e., Is this made of wood?; Can this be foun a garden?; Does this involve music?; Can
,row?). An encoding question was used from

noamhadion

ll,r th40

4icitlicitpreforexereonin

psin

eecitrdici-thionn frednhi

sey

onordanprotethent

cittiontterfolticither

pre-

aa

andms

tsto

uldre

herici-

ap-timessy,

ressssfor

nttici-mserebe

eentic-for

lierntsd into

ici-ns.ok

rossthe

ctsofdi-ing

ard


one to four times per study list. However,question was used more than once for the sword. For each word, two of the questions“yes” responses, whereas two of the questhad “no” responses.

During retrieval, for all tests but free recaparticipants were presented with the cues fooriginal 80 items. For a given participant,items were studied, whereas the remainingwere new (nonstudied) items. In the implword fragment completion test and the expfragment cued recall test, participants weresented with 80 word fragments. Similarly,the implicit general knowledge test and theplicit question cued recall test, participants wpresented with 80 general knowledge questiThe implicit and explicit tests differed onlythe instructions given.

Procedure.Participants were tested in grouof one to three. A given participant took partonly one of five test conditions. In the frrecall, explicit question cued recall, impliword fragment completion, or the explicit wofragment cued recall test conditions, partpants saw single presentations of items instudy phase. During the single presentateach word appeared on the computer scree3 s. Following the word, a question that requia yes/no response appeared on the scree2 s. Participants were asked to respond byting the specially marked “Y” or “N” key(which were adjacent to each other on the kboard). In the shallow processing conditiparticipants were instructed to read the wcarefully, pay attention to the letters, andswer the graphemic question. In the deepcessing condition, participants were instructo read the word carefully, pay attention tomeaning of the word, and answer the semaquestion.

In the encoding procedure for the impligeneral knowledge test, the same manipulawas used except that each word was presefour times consecutively on the compuscreen with each presentation lasting 1 slowed by a question presented for 2 s. Parpants were given the same instructions assingle presentation groups except that they w
informed that the same item may appear morg
e

s

e

0

-

-

s.

e,or

fort-

-,

--

d

ic

ned

--ee

than once and that each time an item wassented a new question would be asked.

After the study condition, participants had10-min retention interval where they playedgame of cards on the computer for 6 minwere given test instructions and practice itefor the remaining 4 min.2

Following the retention interval, participanin the free recall condition were instructedwrite down in any order all the words they corecall from the earlier study lists. They weinformed that the words could come from eitof the two lists that they saw earlier. Partpants were given 7 min for the task.

For the remaining four tests, test itemspeared on the computer screen one at awith the prompt, “If you have a solution, pre“Y.” When the participants pressed the “Y” kethey were asked to type in the solution and pthe enter key. If the participant failed to prethe “Y” key, the item remained on the screen20 s and then timed out.

For the explicit tests (i.e., word fragmecued recall and question cued recall), parpants were instructed to try to solve the itepresented in the allotted time. Participants walso informed that some of the cues couldsolved with words that they had previously sin either of the two encoding conditions. Paripants were urged to solve only those cueswhich their solution was a word from the earstudied lists. For the implicit tests participawere instructed to solve the items presentethe allotted time with the first word that camemind. In the general knowledge tasks partpants were urged to give one word solutioFor all five groups, the entire procedure toless than 1 h.

Results

In Fig. 2, we have collapsed the results acthe concrete/abstract variable to presentpriming and the levels of processing effefirst. Figure 2 illustrates the mean proportionitems retrieved for the nonstudied (new) contion and both levels of processing encod

2 Participants in the free recall condition played the c
eame for 7 min because there was no practice phase.

tsne

or-die

e.,an

diefo

eS tgt f

p ex-p2e waso m-p

andc tractn lus-t or-t fore butf dw pe( per

skssk;

st.

17)18)17)5)


conditions for each of the five memory tesThe levels of processing analyses and plancomparisont tests were performed on the crected measures of studied minus nonstuitems.

A typical levels of processing effect (i.greater retention for items studied under semtic processing when compared to items stuunder graphemic instructions) was observedall our conceptual tests [free recall,t1(39) 55.95, SE 5 .02, t2(79) 5 4.95, SE 5 .02;xplicit general knowledge test,t1(39) 5 7.6,E5 .03,t2(79)5 8.66,SE5 .02; and implicieneral knowledge test,t1(55)5 2.2,SE5 .02,

2(79)5 2.40,SE5 .02]. A significant level o

FIG. 2. Mean proportion of correct responses (aused in Experiment 2. GK5 general knowledge ta

TAB

Mean Proportion of Studied and Nonsfor Each Item Type in A

Task

Concrete

Studied Nonstudied

Free Recall .23 (.12) 0Explicit GK .35 (.17) .03 (.05)Explicit WFC .42 (.13) .06 (.07)Implicit WFC .48 (.13) .26 (.09)Implicit GK .32 (.15) .27 (.15)

Note.GK 5 general knowledge task; WFC5 word fragme

.d

d

-dr

rocessing effect was also observed on thelicit word fragment completion test [t1(39) 5.75,SE5 .03, t2(79) 5 3.06,SE5 .03]. Asxpected, no levels of processing effectbserved for the implicit word fragment coletion task [t1(39), 1, t2(79) , 1].Table 1 presents the studied, nonstudied,

orrected scores for both concrete and absouns on each of the five tests. Figure 3 il

rates the corrected recall and priming propions for both concrete and abstract nounsach of the five memory tests. For every test

ree recall, a 23 2 ANOVA was performeith study status (old vs new) and item ty

concrete vs abstract) as factors. In addition,

tandard errors) for each item type for each of the five taWFC5 word fragment completion task.

1

ied Items Retrieved and Standard Deviationsive Tasks in Experiment 2

Abstract

t.-Nst. Studied Nonstudied St.-N

(.12) .17 (.10) 0 .17 (.10)(.18) .29 (.15) .04 (.06) .25 (.

6 (.15) .41 (.16) .08 (.08) .33 (.(.14) .41 (.14) .24 (.12) .17 (.(.16) .37 (.15) .30 (.14) .07 (.1

nd s

LE

tudll F

S

.23.32.3.22.05

nt completion task.

eete

true aallbemrac

atus

d

ect

[ffe

ojec

er-ionson

onud

F

edst ob-our

sk

ntus

m

b tiFr n-c66o ob-s icitwt alt k isu mee ska .

eo tus[1 m

theled


our hypothesis, planned comparisont tests werperformed to detect priming and the concrness effect on our two implicit measures.

Since there were no nonstudied items (insions) reported in free recall we did not us2 3 2 ANOVA on these data. For the free rectask, significant differences were observedtween the proportion of studied concrete iterecalled and the proportion of studied abstitems recalled (the concreteness effect) [t1(39)52.71,SE5 .02, t2(78)5 2.22,SE5 .03].

In the general knowledge explicit task,main effect was found for study sta[F1(1,39) 5 140.63,MSe 5 .02, F2(1,78) 5147.94,MSe 5 .02]. No main effect was founfor item type [F1(1,39)5 2.27,F2(1,78), 1].A significant interaction was found by subj[F1(1,39)5 7.01,MSe 5 .05], but not by itemF2(1,78) 5 1.27]. Planned comparisont testsrevealed that a significant concreteness ewas found on the corrected priming scoresthe general knowledge explicit test by sub[t1(39) 5 2.64, SE 5 .05], but not by item[t2(78) 5 1.12], although note that the numical difference was in the expected direct(concreteM 5 .32; abstractM 5 .26). Thus, aexpected, the overall pattern produced a ccreteness effect in this task.

For the explicit word fragment completitask, there was a significant main effect of ststatus [F1(1,39) 5 352.36, MSe 5 .01;

FIG. 3. Recall and priming proportions (and stafive tasks used in Experiment 2. GK5 general know

2(1,78) 5 279.50,MSe 5 .02] and no main

-

-

-st

ctft

-

y

effect of item type [F1(1,39), 1; F2(1,78),1]. No significant interaction was observ[F1(1,39) , 1; F2(1,78) , 1]. Therefore, aexpected, there was no concreteness effecserved on the corrected priming scores inexplicit word fragment completion ta[t1(39) , 1; t2(78) , 1].

In the implicit word fragment completiotest, there was a main effect of study sta[F1(1,39) 5 152.71,MSe 5 .01; F2(1,78) 593.35, MSe 5 .02] and a main effect of itetype by subject [F1(1,39)5 5.23,MSe 5 .01],ut not by item [F2(1,78), 1]. No significan

nteraction was found [F1(1,39)5 1.08,2(1,78) 5 1.05]. Planned comparisont tests

evealed significant priming for both the corete items [t1(39)5 9.89,SE5 .02; t2(39)5.79,SE5 .03] and for abstract items [t1(39)5.81,SE5 .02; t2(39) 5 6.99,SE5 .03]. Perur predictions, no concreteness effect waserved on the priming scores for the implord fragment completion task [t1(39)5 1.04;

2(78) 5 1.03]. Note that the small numericrend for the concreteness effect in this tasnlikely to be real because this trend becaven smaller in the explicit version of the tand was not significant, as described aboveIn the implicit general knowledge task, w

btained a main effect of study staF1(1,55) 5 14.16, MSe 5 .01; F2(1,78) 54.21, MSe 5 .01] and a main effect of ite

rd errors) for concrete and abstract nouns for each ofge task; WFC5 word fragment completion task.

nda

type by subject [F1(1,55)5 10.81,MSe 5 .01],

t

-for

t s[

ri-dgecemou

te-ory

alicitvelzedowthee

parof

rd-b-tthraler-on

rolevtuaex-

p ef-f nd-i tionso me-w ntso byt wed ip-u ourp tualt uedr

ri-m esse st.A ss-i ions ask.W thei s-t tualm n-c on-c testc ns( il-u licitg e tot forew ca-t woe ter-a con-c set .

ac tualt ofp in-c ats ouri atedt cha-n asf d tob on-c ems( ng

ls op tiont . Wd r the re-v thac ptup


but not by item [F2(1,78), 1]. No significaninteraction was observed [F1(1,55), 1;F2(1,78), 1]. Planned comparisont tests confirmed that there was significant primingboth concrete items [t1(55)5 1.93,SE5 .02;2(39) 5 1.99, SE 5 .02] and abstract itemt1(55) 5 3.36,SE5 .02; t2(39)5 3.56,SE5.02]. The most critical finding in this expement concerned the implicit general knowletest; we did not find a significant differenbetween the priming scores for abstract iteand the priming scores of concrete items onimplicit general knowledge test [t1(55) , 1;t2(78), 1]. Thus, we failed to find a concreness effect in our conceptual implicit memtest.

To help ensure that this failure to findsignificant concreteness effect on our impconceptual test was not due to the small leof priming observed in this task, we reanalythe data by dividing the data for high and lresponders. Specifically, we examined whewe would see a concreteness effect emergwe only analyzed the performance of thoseticipants who showed relatively high levelspriming. The data were divided in half accoing to the proportion of overall priming oserved for each participant. A within subjecttest was then performed on the data forhigh-scoring participants in the implicit geneknowledge test. However, no significant diffence was found for the priming scores of ccrete (M5 .13) and abstract nouns (M5 .16),t(27) , 1, of these high responders.

Discussion

As predicted by the transfer appropriate pcessing framework, there was a significantels of processing effect on all the conceptests (both explicit and implicit) and as wepected on our explicit perceptual task.3 Also as

3 We note here that even though we did obtain a leverocessing effect on the explicit word fragment comple

ask we failed to find a concreteness effect on this tasko not have a ready explanation for this discrepancy foxplicit word fragment completion task. Although, as piously mentioned, we believe this finding may suggestoncreteness is not as strong a manipulation of conce

processing as is levels of processing.

sr

s

rif-

e

-

--l

redicted there was no levels of processingect on the implicit perceptual test. These fings suggest that, according to the assumpf the transfer appropriate processing fraork, we identified the processing requiremef our tests correctly. Also, as predicted

ransfer appropriate processing framework,id not find an effect of the conceptual manlation of concrete and abstract nouns onerceptual tests. Furthermore, our concep

ests of free recall and the explicit question cecall did produce the concreteness effect.

The finding of major interest in this expeent is that we failed to obtain a concretenffect on the implicit general knowledge teccording to the transfer appropriate proce

ng framework, a conceptual manipulathould affect performance on a conceptual te did find a levels of processing effect on

mplicit general knowledge test, thereby illurating that the test is sensitive to a concepanipulation. However, we failed to find a co

reteness effect on this task. In addition, a creteness effect was found when the sameues were given with explicit instructioquestion cued recall), illustrating that our fare to find a concreteness effect on our impeneral knowledge test was not attributabl

he nature of the test cues themselves. Bee set out to describe the theoretical ramifi

ions of our findings, we sought to rule out tmpirical possibilities that may have councted a potential concreteness effect in oureptual implicit memory test. The first of thewo possibilities was tested in Experiment 3

One potential reason why we did not findoncreteness effect on our implicit concepest might be that we also included a levelsrocessing manipulation at study. That is,luding a levels of processing manipulationtudy may have changed the way in whichtems were encoded and, therefore, eliminhe concreteness effect. The possible meism by which such elimination may occur is

ollows. The concreteness effect is presumee due to a concrete items entailing more ceptual encoding compared to abstract itPaivio, 1971). However, specifically requiri

f

ee

tal
articipants to encode both concrete and ab-

aon

licietenvath

s a92

tagth

er-esska

neso

tuaipuweceof

mi-

M

-ar-ditre-

enthe

on-nceuseeadingthatheyan

achions

ecton-cant,

M

r-

ete

,t thepa. ,w hef ouri ofp thed hos nt

st.

.14)

.16)


stract nouns for meaning in our experiment mhave reduced the default differences in the cceptual encoding of these items. The expconceptual tests may have been less pronthis problem because these tests utilize intional retrieval. The use of a deliberate retriestrategy may permit the subtle differences inconceptual and distinctive attributes of itemencoding to emerge (see Roediger et al., 19thereby giving the concrete words an advanover the abstract words despite the use oflevels of processing encoding.

To test this possibility, we conducted Expiment 3 to examine whether the concreteneffect would arise on a conceptual implicit taif we did not use the levels of processing mnipulation at study. Because the concreteeffect is attributed to conceptual encodingconcrete items (Paivio, 1971) and conceptests should be sensitive to conceptual manlations at encoding (Roediger et al., 1992),predicted a concreteness effect on our contual implicit test once the use of the levelsprocessing manipulation at study was elinated.

EXPERIMENT 3

ethod

Participants.An additional group of 56 undergraduates from SUNY at Stony Brook pticipated in the experiment to obtain creneeded for partial fulfillment of a coursequirement.

Materials and procedure.All materials andprocedures were identical to that of Experim2 with the following changes. We used only

TAB

Mean Proportion of Studied and Nonsfor Each Item Type

Concrete

Studied Nonstudied

Experiment 3 .34 (.16) .29 (.12)Experiment 4 .28 (.18) .23 (.12)

implicit general knowledge test because wed

y-tto-let),ee

s

-s

fl-

p-

t

were interested in examining whether a ccreteness effect would arise on this test ostudy instructions were changed. We did notthe levels of processing manipulation; instparticipants were told that they were betested for their reading reaction times andthey should read no faster or slower than tnormally do. Each item was preceded byasterisk for 1 s, and, as in Experiment 1, eitem was presented for four massed repetitof 1 s each.

Results and Discussion

Table 2 contains the proportion of corrstudied and nonstudied questions for both ccrete and abstract items. There was a signifimain effect of study status [F1(1,55)5 13.28

Se 5 .01; F2(1,78)5 14.65,MSe 5 .01] andno main effect of item type [F1(1,55), 1;F2(1,78), 1]. In addition, no significant inteaction was observed [F1(1,55), 1; F2(1,78)51.63]. As expected, planned comparisont testsrevealed significant priming for both concritems [t1(55)5 2.44,SE5 .02; t2(39)5 1.82,SE 5 .02] and abstract items [t1(55) 5 3.14,SE5 .02; t2(39)5 3.44,SE5 .03]. Howeverhere was no significant difference betweenriming scores for concrete words (M 5 .05)nd the priming scores for abstract words (M 5

06) [t1(55), 1; t2(78) 5 1.77]. Once againe tried to eliminate the possibility that t

ailure to observe the concreteness effect onmplicit measure was due to the low levelsriming observed. Therefore, we reanalyzedata only for that half of the participants whowed higher levels of priming. No significa

2

ied Items Retrieved and Standard DeviationsExperiments 3 and 4

Abstract

t.-Nst. Studied Nonstudied St.-N

5 (.15) .35 (.15) .29 (.12) .06 (5 (.14) .30 (.14) .26 (.12) .04 (

LE

tudin

S

.0

.0

ifference was observed between the priming

te

en-not 3ofabng-conl.’

thededge

ntul-ss

eminth

hanbilales

tteriavcteiallter

ado

tothetioheoseyion

aouto

methe, th

larci-orees-isusetion

ot belicit

so-notan-re-

s ineenrete

e isantshatr ofici-ingful-o’sful-nses,sso-nly

sin-oney a

ractw-tionossi-ur-lsoredseds ofFors toatedd theen-dge


scores for these participants on concrete (M5.14) and abstract items (M 5 .14) [t(27) , 1].Thus, once again we failed to find a concreness effect on a conceptual implicit test.

Given that the default differences in thecoding of concrete and abstract nouns didproduce a concreteness effect in Experimenwe considered the following possibility. Onethe important ways in which concrete andstract items differ is in the number of meaniful associations that can be generated forcrete and abstract nouns (see Paivio et a1968 norms). This difference may impairproduction of concrete words on open-entests such as the implicit general knowletest.

As mentioned under “Method” in Experime2, in Paivio et al.’s (1968) norms, meaningfness is measured by the mean number of aciations participants generated for a given itThe original pool of items that we usedExperiments 2 and 3 were chosen so thatconcrete items scored significantly higher tthe abstract items on Paivio et al.’s imageaity, concreteness, and meaningfulness scWe used these criteria because these scalehighly correlated and because we wantedcreate a large item set. However, these criled to the selection of concrete words that hmore meaningful associations than the seleabstract words and, therefore, could potentprovide more incorrect solutions. This latpoint is elaborated below.

All of our general knowledge questions hmany possible solutions. Due to the naturethe questions, all of the possible solutionsgiven question were associated with each oFor example, our general knowledge ques“What animal is a political symbol?” had ttarget response “elephant.” However other psible solutions to that question were donkeagle, and dove. Note that all possible solutare associated with each other since theybelong to the category “animals.” Becausetest was implicit (participants were instructedcomplete the item with the first word that cato mind), and many possible solutions toquestions were associates of one another
intrusions may have been affected by the num
-

t,

-

-s,

d

o-.

e

-s.areoaed

y

far.n

-,sllr

e

ber of meaningful associations for a particuitem. Specifically, the more meaningful assoations a particular item can produce, the mpotential incorrect solutions exist to our qution for that particular item. This problemunique to implicit open-ended tests becathese tests can be completed with any soluand, therefore, competing responses may nsuppressed as would be done in the expversions of the same tests.

It should be noted that the meaningful asciates for concrete and abstract nouns dodiffer on at least one other measure of meingfulness. Nelson and Schreiber (1992)cently provided the meaning set size normwhich little differences were observed betwthe meaning set size for abstract and concnouns. In this approach, meaning set sizmeasured by having large groups of participprovide the first meaningful associate tcomes to mind for a given item. The numbedifferent items generated by a pool of partpants for a given item constitutes its meanset size. Note that this measure of meaningness differs from Paivio’s measure. In Paivimultiple association measure of meaningness, each participant provides many respowhereas in Nelson and Schreiber’s single aciation measure, each participant provides oone response.

Based on Nelson and Schreiber’s (1992)gle association measure of meaningfulness,would not expect response competition to plarole in the production of concrete and abstnouns on our conceptual implicit task. Hoever, based on Paivio’s multiple associameasure, response competition remains a pble confound in our previous experiments. Fthermore, Nelson and Schreiber (1992) afound that concrete and abstract nouns diffeon meaningfulness scales similar to those uby Paivio even though the meaning set sizeconcrete and abstract nouns did not differ.these reasons, the goal of Experiment 4 wadetermine whether the larger size of associresponses to concrete nouns had diminisheproduction of target concrete words in an opended task such as implicit general knowle
-in our prior experiments. We tested this idea by

foal.’s eri-ne

M

-ar-nttici56

frol.’screon-lne

wl-d sccullnt

aem

dy

9 ofiT nd[ in

for

oect

inig-resg

heb-ete-nlytedr-for

on-for

ivios.

ex-susry.

latedeetheto

n-res

dingoryte,

heone

vem-on-thertual

then,oryentin-ese

wl-ex-m-hes of

of


holding constant the meaningfulness scoresconcrete and abstract nouns from Paivio et(1968) norms. We predicted a concretenesfect on our conceptual implicit task in Expement 4 because this measure of meaningfulwas controlled.

EXPERIMENT 4

ethod

Participants.An additional group of 86 undergraduates from SUNY at Stony Brook pticipated for credit needed for partial fulfillmeof a course requirement. Of these 86 parpants, 30 took part in a norming study andtook part in the experiment.

Materials and procedure.A pool of 60 items(30 concrete and 30 abstract) was selectedboth Experiment 2 and from Paivio et a(1968) norms such that the abstract and conitems only varied on the imageability and ccreteness scales and not the meaningfuscales [concreteM 5 5.82, abstractM 5 5.76;t(58), 1]. For each new item, a general knoedge question was constructed and normethat the mean baseline rate of completion aracy for this pool of 60 items was 28%. Aprocedures were the same as in Experimeexcept that the participants saw 30 itemsstudy (15 concrete and 15 abstract) and 60 itat test (30 old and 30 new).

Results

There was a significant main effect of stustatus [F1(1,55)5 9.76,MSe 5 .01;F2(1,58)5

.42,MSe 5 .01]. There was no main effecttem type [F1(1,55)5 2.99; F2(1,58) , 1].here was also no interaction fou

F1(1,55), 1; F2(1,58), 1]. As can be seenTable 2, we obtained significant primingconcrete items [t1(55)5 2.85, SE 5 .02;t2(29)5 2.91,SE5 .02]. The priming was alssignificant for abstract items by subj[t1(55) 5 1.92, SE 5 .02], but not by item[t2(29)5 1.64] (but note that the means werethe right direction). However, there was no snificant difference between the priming scofor concrete words (M 5 .05) and the primin
scores for abstract words (M 5 .04) [t1(55), 1;
rsf-

ss

-

m

te

ss

o-

3ts

t2(58) , 1]. Once again, to help eliminate tpossibility that the low levels of priming oserved on this task were masking the concrness effect, we reanalyzed the data for othose half of the participants who demonstrahigher levels of priming. No significant diffeence was found between priming scoresconcrete (M5 .15) and abstract items (M 5.13) for these participants [t(27), 1]. Thus, wedid not find a concreteness effect on the cceptual implicit test even when we equatedthe number of meaningful associations (Paet al., 1968) for concrete and abstract word

GENERAL DISCUSSION

The goal of this research project was toamine the effect of studying concrete verabstract words on conceptual implicit memoPrior research has demonstrated that a revariable of imaging the referent of a word (sPaivio, 1983) produces greater priming onimplicit general knowledge test comparedsimply reading words (Blaxton, 1989). In cotrast, a related manipulation of studying pictudoes not produce greater priming than reawords on another conceptual implicit memtask of category association (Weldon & Coyo1996). Within this context, we explored teffects of concrete versus abstract wordsdifferent forms of memory. In Experiment 1 wreplicated Blaxton’s (1989) finding of a positiimagery effect on the conceptual implicit meory test of general knowledge. We then cducted three experiments to examine whethe concreteness effect occurs on concepimplicit memory tests. To test this, we usedimplicit general knowledge test (Blaxto1989). In Experiment 2, we compared memfor concrete and abstract nouns in five differmemory tasks that a priori make differentstructional and processing demands. Thtasks were free recall, explicit general knoedge test, implicit general knowledge test,plicit word fragment completion test, and iplicit word fragment completion test. Tconceptual/perceptual processing demandthese tasks were confirmed with a levelsprocessing manipulation at study.

Specifically, free recall and explicit and im-

esefnt

levder

thedidesepu-icingal.

er-prhe

ffeks

geob

sks-

ecfecen-

ex. Inro-th

anveulanthathab

s eas

entina

no

ien, who

in-weeen

8tedract

19for-

toer-tract-

inataralingight

-

wes. Ifete-a-onourim-

e at onhenofctts.

oa

s,ere-Co-ournt

t intope-ryon


plicit versions of the general knowledge tproduced significant levels of processingfects. The explicit version of the word fragmecompletion test also produced a significantels of processing effect, as has been reporteother studies (Roediger et al., 1992). Furthmore, as expected, the implicit version ofperceptual word fragment completion tasknot exhibit a levels of processing effect. Thfindings with the levels of processing manilation are completely consistent with the predtions of the transfer appropriate processframework (Roediger, 1990; Roediger et1989, 1992).

With respect to our central question of intest, the concrete/abstract variable produceddicted results with one important exception. Tpredicted results included a concreteness eon the explicit versions of the conceptual tasi.e., free recall and explicit general knowledFurthermore, no concreteness effect wastained on the predominantly perceptual tai.e., explicit and implicit word fragment completion. However, contrary to the general exptations, we failed to find a concreteness efon the implicit conceptual memory task of geral knowledge.

We conducted two more experiments toamine two possible causes of this null effectExperiment 3, we removed the levels of pcessing manipulation on the reasoning thatinstructions to process both the concreteabstract nouns at a deep (or a shallow lemay have inadvertently eliminated the defaprocessing differences between concreteabstract nouns. However, our hypothesisthe emergence of the default differences inconceptual processing of concrete versusstract nouns would produce a concretenesfect on the implicit general knowledge test, wnot supported in Experiment 3. In Experim4, we equated the number of possible comperesponses for concrete nouns compared tostract nouns. However, this control also didgive rise to a concreteness effect.

In order to help ensure that we had sufficpower for observing the concreteness effectanalyzed the data for all 168 participants w
took part in the implicit general knowledge test
t-

-in-

-

,

e-

ct,.-,

-t

-

edl)tdt

e-f-

tgb-t

te

across Experiments 2, 3, and 4 in one withsubject t test. Even in this meta-analysis,failed to detect a significant difference betwconcrete (M5 .05) and abstract items (M 5.06) [t (167) , 1]. In addition, of the 16participants, only 72 (or 43%) demonstrabetter performance for concrete over abstnouns, 77 (or 46%) showed a reversal, and(or 11%) participants showed equal permance for the two types of items. Accordinga Wilcoxon matched pairs sign test, the diffence in performance for concrete and absnouns was not significant (p 5 .66). Furthermore, similar to the analysis we performedExperiments 2, 3, and 4, we divided the dfrom all 168 participants in the implicit geneknowledge test in half and analyzed the primscores of those participants who showed hpriming. We still failed to find a significandifference between concrete (M 5 .14) and abstract items (M5 .15), t (83) , 1.

To ensure that power was not an issue,also used Cohen’s (1988) power estimatewe assume a medium size effect of concrness, then in a one-tailedt test conducted withwithin-subject design where then is 56, according to Cohen (1988; also see Aron & Ar1999, p. 278), we had 96% power in each ofanalyses of the concreteness effect in theplicit general knowledge test. If we assumsmall effect size for the concreteness effecthe implicit measure, then according to Co(1988; and Aron & Aron, 1999) the numbersubjects needed for a one-tailed within-subjettest with 80% power would be 172 participanOur collapsed data had ann extremely close tthat (n 5 168) but there was no sign ofconcreteness effect (M5 .05 for concrete itemM 5 .06 for abstract items) in these data. Thfore, based on our collapsed data, and onhen’s power estimates, we conclude thatnull effects are not a result of insufficiepower.

Our failure to obtain a concreteness effeca conceptual implicit memory task is similarthe reports of a failure to obtain a picture suriority effect in conceptual implicit memotasks (McDermott & Roediger, 1996; Weld
& Coyote, 1996). However, our findings stand

iveged inetwoap, eivio

ca-o

theicitn rn titreielila,le,b

oryth

edenas

aneltioThfo

onmatuace

on-poor

rlaprfo

teeluscalei

tiontes

ofate-sifi-nceor

icitingndnot

theanddi-ef-rd

ourt onn-ask

athel-

find-

edsralar-t isusem-

thisan

sk,n’sl-

d aothralis-nlyon-ingsthee of

essthe

eper


in sharp contrast to the report of a positimagery effect in an implicit general knowledtest reported by Blaxton (1989) and observeExperiment 1 in the present series. Combitogether, these findings are problematic forinfluential theories of memory, the transferpropriate processing framework (Roedigeral., 1989) and the dual-code hypothesis (Pa1971, 1991).

Before we consider the theoretical implitions and possible theoretical resolutionsthese findings, it is important to reiteratesignificance of the specific conceptual implmemory tasks used across these studies. Icent reports, researchers have speculated opossibility that different conceptual implictasks may vary subtly in their processingquirements (Cabeza, 1994; Vaidya, GabrKeane, Monti, Gutierrez-Rivas, & Zarel1997; Weldon & Coyote, 1996). For exampCabeza (1994) demonstrated a dissociationtween two implicit conceptual tasks, categassociation and free association (similar toword association task previously discusswith two conceptual encoding conditions, gerating category labels and generating wordsociations. Cabeza found that when participencoded items by generating category labpriming was better on the category associatask compared to the word association task.opposite pattern of priming was observedthe word association encoding conditiCabeza’s findings suggest that differencesexist both between the two types of concepprocessing at encoding and between the contual processing used on the two implicit cceptual tests. Note that his findings also supthe transfer appropriate processing framewin that the more similar the processing ovewas between study and test, the better pemance was at test.

Vaidya et al. (1997) have also demonstrathat a conceptual manipulation at study (levof processing) can differentially affect varioimplicit conceptual tests. Specifically, a typilevels of processing effect was found on thimplicit conceptual task of category associaand word association with weak associa
However, no levels of processing effect was
nd

-t,

f

e-he

-,

e-

e),--

tss,ner.ylp-

rtk

r-

ds

r

.

observed on their implicit conceptual tasksword association with strong associates, cgory verification, and abstract/concrete clascation (i.e., memory measured by performalatencies for classifying words as concreteabstract).

Therefore, both studies illustrate that implconceptual tasks may vary in their processrequirements. Although this is an important anecessary consideration, this reasoning issufficient to account for the discrepancies indata we have discussed here. In WeldonCoyote’s (1996) and McDermott and Roeger’s (1996) studies, no picture superiorityfect was obtained on the implicit tasks of woassociation or category production. Instudy, we failed to find a concreteness effeca different conceptual implicit test, that of geeral knowledge. We took care to select this tbecause Blaxton (1989) originally reportedpositive effect of a related variable, i.e.,imagery effect, on the implicit general knowedge test and we have since replicated thising with this task.

In this context, another important point neto be made regarding the implicit geneknowledge task. Vaidya et al. (1997) havegued that the implicit general knowledge tesnot a good measure of implicit memory becathis task may be contaminated by explicit meory strategies. Our data do not supportclaim. We conducted both an explicit andimplicit version of the general knowledge taand together, the findings support Blaxto(1989) original claim that the general knowedge test can tap implicit memory. We founsignificant levels of processing effect on bthe implicit and explicit versions of the geneknowledge tests. Yet, the implicit version dsociated from the explicit version such that othe explicit version of the task produced a ccreteness effect. Taken together, our findprovide strong support for the proposal thatgeneral knowledge test is a useful measurconceptual implicit memory.

In sum, our failure to find the concreteneffect cannot be attributed to the selection ofparticular task. Instead, a broader and de
theoretical analysis is necessary to account for

thebe

m-theesbaasn

tioptutuaasoidesinss

rityes

wade

ectasltsingl inmam

romf thsinvate

odepose) ththet ire

nichnemeriopntefes,

ren-is-

thero-

ionsndsu-

mtoc-

el-79;totionlexarehis

vesonture

oryish

tioniedkewithere-

aa-

ul-il-

onhisportpe-

eenffectc-tedoreespre-vernoer-sts,


our results in conjunction with the results ofother studies cited here. The discrepancytween finding a positive imagery effect on iplicit conceptual tasks and not finding eitherpicture superiority effect or the concreteneffect raises questions about the underlyingsis of these three effects. Specifically, twopects of this issue deserve consideration. Owe need to consider whether the assumpthat these three effects are based on conceprocessing is a valid one. Two, if the concepbasis of these effects does seem to be a reable assumption, then we need to conswhether the underlying conceptual procesvaries quantitatively or qualitatively acrothese three effects.

The assumption that the picture superioeffect, the imagery effect, and the concreteneffect arise from conceptual processingbased on the following logic. In the dual-cohypothesis, it is assumed that all three effarise from the activation of verbal as wellimaginal codes. This dual activation resufrom cross-system or referential processwhich in turn is assumed to be conceptuanature. However, an alternate mechanismalso explain these three effects in explicit meory. These three effects may indeed arise fcross-system processing but not because oconceptual nature of such referential procesbut because cross-system processing actithe imaginal code. Because the imaginal codassumed to be a “mnemonically superior” cin the dual-code hypothesis (Paivio & Csa1973), it is the superiority of this code per(subsequent to the cross-system processinggives rise to the picture superiority effect,imagery effect, and the concreteness effecmemory. A quick clarification is needed hewith respect to the nature of the “mnemosuperiority” of the imaginal code. AlthougPaivio considered the image code to be a mmonically superior code, it is not clear froPaivio’s description whether mnemonic supority arises from conceptual or perceptual prerties of the stimulus. The reasoning presehere assumes that mnemonic superiority reto perceptually distinctive information. Thu
the three said effects may arise not from the
-

s--e,nalln-r

g

ss

s

y-

egesis

,

at

n

-

--drs

conceptual processing afforded by the refetial processing but rather from the visual dtinctiveness afforded by the activation ofimaginal code secondary to the referential pcessing.

Others have also considered some variatof this possibility. For example, Weldon aCoyote (1996) have argued that the pictureperiority effect in explicit memory arises frothe visual distinctiveness of pictures relativewords. This interpretation is similar to the acount of the picture superiority effect in Nson’s sensory-semantic model (Nelson, 19Nelson, Reed, & McEvoy, 1977). AccordingNelson’s model, the perceptual representaof pictures involves more unique and compvisual features compared to words thatbased on only 26 letters of the alphabet. Tdifference, according to Nelson’s model, girise to the picture superiority effect. Weldand Coyote (1996) have suggested that a picsuperiority effect emerges on explicit memtests because intentional efforts to distingubetween studied and nonstudied informabenefit from the distinctive aspects of studinformation. However, the implicit tasks mano reference to the past and are completedthe first response that comes to mind and, thfore, distinctive information does not haveparticular advantage relative to other informtion on implicit tests of memory (see also Mligan, 1996). Weldon and Coyote’s (1996) faure to find a picture superiority effectimplicit conceptual memory tasks confirm tprediction. Thus, there is considerable supfor the possibility that at least the picture suriority effect is not conceptually based.

A distinctiveness hypothesis has also bconsidered to explain the concreteness e(Marschark & Hunt, 1989). Specifically, acording to this claim the imagery associawith concrete items causes them to be mdistinct and it is this distinctiveness that drivthe concreteness effect. This assumptiondicts better memory for concrete items oabstract items in explicit tasks. Althoughpredictions were made in this context for pformance on implicit conceptual memory te
straightforward predictions can now be derived

ndincbutn-tive

bp-odobbutan-wlthe-tooteth

riouctem

inathebeo

esit

notweask

inls-5).9)orthativeimypean

m-theetansta-to

n-sualpro-d byssn-

pro-and

al-ro-thes forver,tualifysesay

thects

andpe-

heu-insrdith

inalringe ofms,

pro-s tobuttialag-

rds.1)tacts or

ex-te

ab-lsoitter

or ad


based on Mulligan’s (1996) and Weldon aCoyote’s (1996) ideas that perceptually disttive information enhances explicit memorynot implicit memory. Specifically, if the cocreteness effect arises from more distincencoding of concrete items (as suggestedMarschark & Hunt, 1989) and implicit concetual tests are not sensitive to distinctive encing, then a concreteness effect should betained on the explicit general knowledge testnot on the implicit version of this test. Our daconfirm this prediction. We found a typical cocreteness effect on our explicit general knoedge test and we failed to find this effect onimplicit version of this task. This “distinctiveness” explanation is not in oppositionPaivio’s dual-code hypothesis because, as nearlier, the dual-code hypothesis assumesthe pictorial or imaginal code is a more supemnemonic code and could, on its own, prodbetter memory. To the extent that cross-sysprocessing activates the pictorial or imagcode, both the picture superiority effect andconcreteness effect in explicit memory couldthe result of distinctiveness or the superioritythe pictorial code. For this very reason, thtwo effects may fail to occur on the implicversions of these tasks.

Note, however, that this explanation issufficient to account for the imagery effectobtained on the implicit general knowledge tin Experiment 1 and others have reportedprior research (Blaxton, 1989; Nyberg & Nison, 1995; Wippich & Mecklenbrauker, 199Specifically, if Marschark and Hunt (198claim that imagery makes concrete items mdistinct, then it is reasonable to assumeimaged items should also be more distincthan read words. Furthermore, if conceptualplicit memory tests are not sensitive to this tof distinctiveness, then we should not findimagery effect on these tasks.

Thus, the imagery effect in conceptual iplicit tasks cannot be easily explained byvisual or perceptual distinctiveness interprtion without some additional consideratioDespite this complication with the interpretion of the imagery effect, it is not necessary
entirely abandon the possible interpretation tha
-

y

--t

-

datre

l

fe

et

-

-.

all three effects (i.e., picture superiority, cocreteness, and imagery) are the result of vior perceptual distinctiveness. Instead, wepose that these three effects can be explainecombining the effects of visual distinctiveneof the imaginal code with the operation of coceptual processes entailed by referentialcessing in the processing of concrete wordsimaged items.

Our proposal is derived from Paivio’s ducode hypothesis in which both referential pcessing as well as mnemonic superiority ofimaginal code are considered to be the basethe three effects under consideration. Howewe argue that the data from the concepimplicit memory tasks systematically specthe conditions under which referential procesversus the superiority of the imaginal code mdifferentially aid memory performance andmechanism by which each of the three effeemerge. Specifically, we agree with WeldonCoyote’s (1996) analysis that the picture suriority effect results from the activation of tmore visually distinctive (or “mnemonically sperior”) imaginal code. This proposal also gasupport from the fact that in the picture/womanipulation, participants are presented wpictures that can directly contact the imagcode (see also Nelson, 1979) without requiextensive referential processing. In the casprocessing of imaged items and concrete itewe argue that the conceptual or referentialcessing plays a larger role. One possibility awhy imaging enhances conceptual primingconcrete words do not may be that referenprocessing is greater in magnitude during iming than during the encoding of concrete woThis possibility is supported by Paivio’s (199suggestion that concrete words may conimagens to a lesser extent than either picturethe imagery instructions. For example, inplicit memory tasks, memory for concrewords is found to be greater than that forstract words (Paivio & Csapo, 1969, see aPaivio, 1983 for a review). In turn, explicmemory for imaged words is found to be greathan for concrete words (see Paivio, 1983 freview). Finally, explicit memory for image
twords is equivalent to that for named pictures

they iu

epthaingin

rory

rdstintu

imaner

onetehaenimtheheumm-

asprocodi-tt,91zed

oweeses

ro-b

ptum-a

t.”laituaatiomeer

91;c-

out,forses.heip-tual

fars inlana-on-

ofow-n-ely.heicitone

of(thetheefin-tualula-istial

ita-re-alsofer-ren-Ca-ourcon-ual-za,dif-eseinge-sere

ly,


(Paivio & Csapo, 1973). Thus, even thoughconcreteness effect is mediated by imagerthe dual-code hypothesis and, therefore, mentail referential processing (which is conctual in nature), it is reasonable to proposethe referential processing involved in encodconcrete words is less than that involvedimagery. In sum, the extent of conceptual pcessing involved for different items may vaalong a continuum such that abstract woconcrete words, and imagery span the conuum from lesser to greater degree of concepprocessing.

On this reasoning, although conceptual pring is likely to be greater for imaged words thconcrete words (as in Blaxton, 1989 and Expiment 1 in the current article), enhanced cceptual priming should occur for concrwords relative to abstract words as well. Tthis “weaker” conceptual processing duringcoding was not picked up on a conceptualplicit task suggests that the dual-code hyposis should be modified to explicitly include tnotion of a referential processing continuwith respect to explicit as well as implicit meory tasks.

A similar idea of a processing continuum halready been adopted within the transfer appriate processing framework. Roediger andleagues (McDermott & Roediger, 1996; Roeger & Blaxton, 1987; Roediger & McDermo1993; Roediger et al., 1989; Weldon, 19Weldon & Roediger, 1987) have emphasithat conceptual and perceptual propertiestasks are not dichotomies but should be vieas endpoints on a continuum. Therefore, thresearchers suggest that there may be degregradations of “conceptual” or “perceptual” pcessing. In fact two separate continua maynecessary, i.e., a perceptual and a concecontinuum (Weldon, 1991) to account for meory performance, where each continuum mhave its own endpoints of “least” and “mosThis idea has been previously used to expsome of the inconsistent findings in perceptests and has become an accepted modificof the transfer appropriate processing frawork (McDermott & Roediger, 1996; Roedig
& Blaxton, 1987; Roediger & McDermott,
nst-t

-

,-

al

-

--

t---

-l-

;

fdeor

eal

y

nln

-

1993; Roediger et al., 1989; Weldon, 19Weldon & Roediger, 1987). However, as MDermott and Roediger (1996) have pointedthe same notion has not been widely utilizedconceptual implicit memory tests and procesOur findings in conjunction with those on tpicture/word manipulation and imagery manulation support the notion of such a concepprocessing continuum.

The theoretical analysis presented thusfocuses on possible quantitative differencethe conceptual processes because an exption based on differences in the amount of cceptual processing fit the data on effectsimagery and concreteness quite nicely. Hever, it is important to bear in mind that coceptual processes can also differ qualitativWithin our own study we found that both timplicit general knowledge test and the explfragment cued recall test were sensitive totype of conceptual processing (the levelsprocessing manipulation) but not anotherconcrete/abstract manipulation). Althoughresearch presented in this article does not ditively demonstrate whether the concepanalyses engendered by these two maniptions differ qualitatively or quantitatively, thpattern of results does point to the potendifferences in the sensitivity of tasks to qualtive differences in meaning. Other recentsearch earlier described in our article hasdemonstrated that possible qualitative difences in conceptual encoding produce diffetial effects on conceptual memory tasks (beza, 1994; Vaidya et al., 1997). Thus,findings and those of others suggest thatceptual processing requirements may vary qitatively not only across tasks (as in Cabe1994; Vaidya et al., 1997), but also acrossferent encoding manipulations. Together, thpatterns of data point to the varieties of meanthat guide cognitive, and in particular, mnmonic operations. A fuller exploration of theideas may prove to be a fruitful topic for futuresearch.

CONCLUSION

We conclude by noting that, until recent
our understanding of the relationship between

waofgsas

on&ca

es,nsithth

tuardwe

oteturualhan

rendntoin

eseith

thewenc-n oexrds

ongriaty-

C licittype-

ofs orn-

v-

r--

citesicrn-

con-l

d.),r of

of

nd

ndorythe-

gy,

fer-m-l

rds:makan

7).of

-

te-nts of.ate-

t inh,

andmanal

. Inn:-


pictures, imagery, and concrete wordslargely based on findings from a varietyexplicit memory tasks. However, the findinfrom conceptual priming tasks in our studywell as in some previous studies (Blaxt1989; Roediger & McDermott, 1996, WeldonCoyote, 1996) necessitate important modifitions in our conceptualization of how picturimages, and concrete words are related. Cotent with the proposals of the dual-code hypoesis (Paivio, 1991) these findings supportview that imaging requires stronger concepprocessing than processing of concrete woWith respect to the processing of pictures,agree with the views of Weldon and Coy(1996, also see Nelson, 1979) that the picsuperiority effect arises largely from the visdistinctive properties of the pictures rather tfrom enhanced conceptual processing. Withspect to the processing of concrete words, fiings from the present series of experimeclearly indicate that conceptual processingconcrete words is weaker than that foundimaging the words. As we noted earlier, thspecifications are not entirely inconsistent wthe current versions of the dual-code hyposis. However, unlike the dual-code view,have explicitly emphasized the role of distitiveness in picture processing and the notioa continuum of conceptual processing toplain the effect of imagery and concrete woon memory.

REFERENCES

Aron, A., & Aron, E. (1999).Statistics for psychology(2nded.). New York: Prentice Hall.

Blaxton, T. A. (1989). Investigating dissociations ammemory measure: Support for a transfer—appropprocessing framework.Journal of Experimental Pschology: Learning, Memory, and Cognition,15, 657–668.

abeza, R. (1994). A dissociation between two impconceptual tests supports the distinction betweenof conceptual processing.Psychonomic Bulletin & Review,1, 505–508.

Challis, B. H., & Sidhu, S. (1993). Dissociative effectmassed repetition on implicit and explicit measurememory.Journal of Experimental Psychology: Leaing, Memory, and Cognition,19, 115–127.

Cohen, J. (1988).Statistical power analysis for the behaioral sciences(2nd ed.). Hillsdale, NJ: Erlbaum.

Craik, F. I. M., & Lockhart, R. S. (1972). Levels of pro-

s

,

-

s--els.

e

--sf

-

f-

e

s

f

cessing: A framework for memory research.Journal ofVerbal Learning and Verbal Behavior,11, 671–684.

Durgunoglu, A. Y., & Roediger, H. L. (1987). Test diffeences in accessing bilingual memory.Journal of Memory and Language,26, 377–391.

Graf, P., & Schacter, D. L. (1985). Implicit and explimemory for new associations in normal and amnsubjects.Journal of Experimental Psychology: Leaing, Memory, and Cognition,11, 501–518.

Hamann, S. B. (1990). Level of processing effects inceptually driven implicit task.Journal of ExperimentaPsychology: Learning, Memory, and Cognition,16,970–977.

Madigan, S. (1983). Picture memory. In J. C. Yuille (EImagery, memory, and cognition: Essays in honoAllan Paivio (pp. 65–89). Hillsdale, NJ: Erlbaum.

Marschark, M., & Hunt, R. R. (1989). A reexaminationthe role of imagery in learning and memory.Journal ofExperimental Psychology: Learning, Memory, aCognition,15, 710–720.

McDermott, K. B., & Roediger, H. L. (1996). Exact aconceptual repetition dissociate conceptual memtests: Problems for transfer appropriate processingory. Canadian Journal of Experimental Psycholo50, 57–71.

Mulligan, N. W. (1996). The effects of perceptual interence at encoding on implicit memory, explicit meory, and memory for source.Journal of ExperimentaPsychology: Learning, Memory, and Cognition,22,1067–1087.

Nelson, D. L. (1979). Remembering pictures and woAppearance, significance, and name. In L. S. Cer& F. I. M. Craik (Eds.),Levels of processing in hummemory(pp. 45–76). Hillsdale, NJ: Erlbaum.

Nelson, D. L., Reed, V. S., & McEvoy, C. L. (197Learning to order pictures and words: A modelsensory and semantic encoding.Journal of Experimental Psychology: Human Learning and Memory,3, 485–497.

Nelson, D. L., & Schreiber, T. A. (1992). Word concreness and word structure as independent determinarecall.Journal of Memory and Language,31,237–260

Nicolas, S. (1995). The picture-superiority effect in cgory-association tests.Psychological Research,58,218–224.

Nyberg, L., & Nilsson, L. (1995). The role of enactmenimplicit and explicit memory.Psychological Researc57, 215–219.

Paivio, A. (1971). Imagery and verbal processes.NewYork: Holt, Rinehart, Winston.

Paivio, A. (1978). The relationship between verbalperceptual codes. In E. C. Caterette & M. P. Fried(Eds.), Handbook of perception. Vol. IX: Perceptuprocessing.New York: Academic Press.

Paivio, A. (1983). The empirical case for dual codingJ. C. Yuille (Ed.), Imagery, memory and cognitioEssays in honor of Allan Paivio(pp. 307–332). Hills
dale, NJ: Erlbaum Associates.

cur

reey,

rbay,

arei-

n-r 92

ngl o

itF.

9).licit. Lylvin

&m-nd

y-

ionten-&

es

yoft-

rds,g

gana-

-.

A.,i-ingl

ony:

them-rn-

al

on,s innd-

m-


Paivio, A. (1991). Dual coding theory: Retrospect andrent status.Canadian Journal of Psychology,45, 255–287.

Paivio, A., & Csapo, K. (1973). Picture superiority in frecall: Imagery or dual coding?Cognitive Psycholog5, 176–206.

Paivio, A., & Csapo, K. (1969). Concrete image and vememory codes.Journal of Experimental Psycholog80, 279 –285.

Paivio, A., Rogers, T. B., & Symthe, P. C. (1968). Whypictures easier to recall than words?Psychonomic Scence,11, 137–138.

Paivio, A., Yuille, J. C., & Madigan, S. A. (1968). Cocreteness, imagery and meaningfulness values fonouns.Journal of Experimental Psychology,76, 1–25.

Roediger, H. L., & Blaxton, T. A. (1987). Effects of varyimodalities, surface features, and retention intervapriming in word fragment completion.Memory andCognition,15, 379–388.

Roediger, H. L., III., & McDermott, K. B. (1993). Implicmemory in normal human subjects. In H. Spinnler &Boller (Eds.),Handbook of neuropsychology(Vol. 8,pp. 63–131). Amsterdam: Elsevier.

Roediger, H. L., Weldon, M. S., & Challis, B. H. (198Explaining dissociations between implicit and expmeasures of retention: A processing account. In HRoediger & F. I. M. Craik (Eds.),Varieties of memorand consciousness: Essays in honour of Endel Tu(pp. 3–14). Hillsdale, NJ: Erlbaum.

Roediger, H. L., III, Weldon, M. S., Stadler, M. L.,Riegler, G. L. (1992). Direct comparison of two iplicit memory tests: Word fragment completion aword stem completion.Journal of Experimental Pschology: Learning, Memory, and Cognition,18,1251–1269.

Schacter, D. L., Bowers, J., & Booker, J. (1989). Intentawareness, and implicit memory: The retrieval intionality criterion. In S. Lewandowsky, J. C. Dunn,K. Kirsner (Eds.),Implicit memory: Theoretical issu
(pp. 47–55). Hillsdale, NJ: Erlbaum.
-

l

5

n

.

g

,

Schneider, W. (1990).Micro-Experimental Laborator[Computer program]. Pittsburgh, PA: Psychology Sware Tools, Inc.

Shepard, R. N. (1967). Recognition memory for wosentences, and pictures.Journal of Verbal Learninand Verbal Behavior,6, 156–163.

Srinivas, K., & Roediger, H. L., III (1990). Classifyinimplicit memory tests: Category association andgram solution.Journal of Memory and Language,29,389–412.

Thorndike, E. L., & Lorge, I. (1944).The teachers’ handbook of 30,000 words.New York: Columbia UnivPress.

Vaidya, C. J., Gabrieli, J. D. E., Keane, M. M., Monti, L.Gutierrez-Rivas, H., & Zarella, M. M. (1997). Evdence for multiple mechanisms of conceptual primon implicit memory tests.Journal of ExperimentaPsychology: Learning, Memory, and Cognition,23,1324–1343.

Weldon, M. S. (1991). Mechanisms underlying primingperceptual tests.Journal of Experimental PsychologLearning, Memory, and Cognition,17, 526–541.

Weldon, M. S., & Coyote, K. C. (1996). Failure to findpicture superiority effect in implicit conceptual meory tests.Journal of Experimental Psychology: Leaing, Memory, and Cognition,22, 670–686.

Weldon, M. S., & Roediger, H. L. (1987). Altering retrievdemands reverses the picture superiority effect,Mem-ory & Cognition,15, 269–280.

Weldon, M. S., Roediger, H. L., Beitel, D. A., & JohnstT. R. (1995). Perceptual and conceptual procesimplicit and explicit tests with picture fragment aword fragment cues.Journal of Memory and Language,34, 268–285.

Wippich, W., & Mecklenbrauker, S. (1995). Implicit meory for textual materials.Psychological Research,57,131–141.

(Received August 18, 1999)
(Revision received June 27, 2000)

Documents

The Concreteness Effect in Implicit and Explicit Memory Tests · concreteness effect (better memory for concrete than abstract nouns) in free recall and the explicit general knowledge