Constructing Meaning: The Role of Affordances and ...martin/Csci6402/Papers/kaschak00.pdfthe objects...

Journal of Memory and Language43, 508–529 (2000)doi:10.1006/jmla.2000.2705, available online at http://www.idealibrary.com on

Constructing Meaning: The Role of Affordances and GrammaticalConstructions in Sentence Comprehension

Michael P. Kaschak and Arthur M. Glenberg

University of Wisconsin—Madison

The Indexical Hypothesis describes how sentences become meaningful through grounding theirinterpretation in action. We develop support for the hypothesis by examining how people understandinnovative denominal verbs, that is, verbs made from nouns and first encountered by participantswithin the experiment (e.g., to crutch). Experiments 1 and 2 demonstrated that different syntacticconstructions provide scenes or goals that influence the meaning created for the innovative verbs.Experiment 3 used reading time to demonstrate that people also consider possible interactions withthe objects underlying the verbs (i.e., the affordances of the objects) when creating meaning.Experiment 4 used a property verification procedure to demonstrate that the affordances derived fromthe objects depend on the situation-specific actions needed to complete the goal specified by thesyntactic construction. Thus the evidence supports a specific type of interaction between syntax andsemantics that leads to understanding: The syntax specifies a general scene, and the affordances ofobjects are used to specify the scene in detail sufficient to take action.© 2000 Academic Press

Key Words:embodied cognition; construction grammar; sentence comprehension; denominalverbs.

ntrsio&

ei-re92.g.&rg

.g.,

e aon

Hy-pre-sses:riv-f-f-sic

len-) asthe

sse

eas.g.,nd

In-se

ularin

the-ex-tive

Nafirssinthaney f

anarleonsateousateudesite

eithnsoents

How sentences are understood is a cequestion. Theories of sentence comprehencan inform work on parsing (e.g., FrazierClifton, 1996; MacDonald, Pearlmutter, & Sdenberg, 1994; see Mitchell, 1994, for aview), lexical access (e.g., Andrews, 19Marlsen-Wilson, 1990), mental models (eZwaan & Radvansky, 1998; van DijkKintsch, 1983; Johnson-Laird, 1983; GlenbeMeyer, & Lindem, 1987), and meaning (e

This article is based upon work supported under ational Science Foundation Graduate Fellowship to theauthor and by a grant from the University of WisconGraduate School to the second author. The authorsDavid Robertson, Joshua Alexander, and Jareb Courtnhelp in developing the materials for these experimentsrunning participants. Thanks also to Jenny Saffran, ChSnowdon, and Gabriella Vigliocco for helpful discussiand suggestions on the work presented here. We are grfor the comments of Michael Spivey and two anonymreviewers which helped to improve this article. The mrials used in the experiments reported here (but not inclin the appendices) can be found at the following webhttp://psych.wisc.edu/glenberg/jml_k&g.html.

Address correspondence and reprint requests toauthor at Department of Psychology, 1202 West JohStreet, Madison, WI, 53706. E-mail: mpkaschak@stud
wisc.edu; [email protected].
5080749-596X/00 $35.00Copyright © 2000 by Academic PressAll rights of reproduction in any form reserved.

aln

-;,

,

Glenberg, 1997). In this article, we proposmodel of sentence comprehension basedGlenberg and Robertson’s (1999) Indexicalpothesis. This hypothesis asserts that comhending a sentence requires three proceindexingwords and phrases to referents, deing affordances(Gibson, 1979) from these reerents, andmeshing(Glenberg, 1997) these afordances under the guidance of intrinbiological and physical constraints (see Gberg, 1997; Glenberg & Robertson, in presswell as constraints provided by the syntax ofsentence.

In this article, we focus on the third proce(meshing) and how it is guided by syntax. Wdevelop this account by incorporating idfrom the language development literature (eLandau & Gleitman, 1985; Fisher, 1996) alinguistics (e.g., Goldberg, 1995) into thedexical Hypothesis. In particular, we propothat the meanings associated with particsyntactic forms function to constrain the waywhich sentences are interpreted. This hyposis is investigated in four experiments thatplore how adult readers understand innova

-t

kords

ful

-d:

ern.

denominal verbs (Clark & Clark, 1979).

anderre

rmsedeeamV–Xnca

9;rguenclextu

llt thes

ingh enti &F &G 6)T .g“ kss ass hil-d air-i ant

it-

ex-ernce

enttedbe-ticsine.g.,re is

theningela-ishertheentf anwl-andnal

age

cticagent

e,eee-

T

D

C

R

WOBL”

509CONSTRUCTING MEANING

SYNTAX AND MEANING

The relationship between syntactic formsmeaning has been explored in terms of vargument structures, or, roughly, the structuof simple sentences. Several sentence fosuch as the double-object, transitive, caumotion, and others (see Table 1) have bassociated with particular meanings. For exple, the transitive sentence structure (N–OBJ) is linked with a putative meaning of “acts on Y,” whereas the double-object sentestructure (N–V–OBJ1–OBJ2) is linked withmeaning of “X transfers Y to Z” (Pinker, 198Goldberg, 1992, 1995). Whereas these ament structures correspond to simple sentforms, they are also found in more compsentences. For instance, the transitive strucis found in the sentence, “Mike kicked the bahat was laying in the yard,” even thoughentence structure is complex.The relationship between form and mean

as been explored in a number of experimnvolving children (e.g., Pinker, Lebeaux,rost, 1987; Gropen, Pinker, Hollander,oldberg, 1991; Naigles, 1990; Fisher, 199he results from a series of paradigms (eacting out” tasks and preference looking tasuggests that particular syntactic forms areociated with particular meanings. In fact, cren’s sensitivity to these form—meaning p

ngs has become a significant aspect of m

TAB

Syntactic For

Form Example

ransitiveN–V–OBJ “Mike kicked the toy”

ouble object:N–V–OBJ1–OBJ2 “Mike gave David a toaused motionN–V–OBJ–OBL “Mike pushed the bookesultativeN–V–OBJ–Xcomp “Mike kissed the childay constructionN–V–[poss-way]–OBL “Mike made his way a

a Adapted from Goldberg, 1995.

heories of the child’s acquisition of verb mean-

bss,-n-

e

-e

re

s

..,)-

y

ings (Gleitman & Gillette, 1995; Naigles, Gleman, & Gleitman, 1992; Pinker, 1989).

Several theories have been proposed toplain this form–meaning relationship. Pink(1989) suggests that it arises from the existeof rules that link verb semantics to argumstructures. A sentence form will be associawith a particular meaning, on this account,cause a cluster of verbs with similar seman(e.g., “transfer” verbs) will commonly occurthe same sentence form. A second theory (Landau & Gleitman, 1985) suggests that thea relationship between sentence forms andscenes that they describe. The form–mealinkage is thus created on the basis of this rtionship between scenes and sentences. F(1996) proposes a similar idea in whichrelationship between a particular type of evand a particular sentence form is the basis oanalogical transfer process which allows knoedge about the old event to be transferredapplied to current events or sentences. A fihypothesis about the form–meaning linkcomes from Goldberg (1995).

Goldberg (1995) claims that certain syntaforms exist as independent units in the languwith their own meaning (called “argumestructure constructions”; see Table 1). Acon-struction (cf., construction grammar, Fillmor1988; Fillmore, Kay, & O’Connor, 1988; sKay & Fillmore, 1999, for a discussion) is d

1

Meaning Pairsa

Hypothesized meaning of form

“X acts on Y”

“X transfers Y to Z”

f the table” “X causes Y to go to Z”

onscious” “X causes Y to become Z”

ss the room.” “X creates and follows a path to

LE

m–

y”

of

unc

cro

fined as a pairing of a form and meaning such

thureun-ticatmaionnt

cifitheh-in

isthetruch tingonverr.g.ctviderbX”)

w sfeb ’s( ruct ssib s ot -

endothe

in

5)ere5)ons trenw

ez-irsg,toodus-

te,

of--odee.g.,seenceder

e orr ex-the

rans-we

icalare

t ofandthe

en-ym-for

tab-hatro-on-

onsr inthearn-for aset-

exi-ceshat

510 KASCHAK AND GLENBERG

that the meaning cannot be predicted onbasis of the form alone. Argument structconstructions are believed to be lexicallyfilled syntactic forms that specify both syntac(e.g., nouns, verbs, and objects) and them(e.g., agents, patients, and recipients) infortion. For example, the transitive constructspecifies reference to the “agent” and “patiethematic roles and links these roles to spesyntactic units (the “agent” is the subject,“patient” is the direct object). It is this hypotesis about form–meaning linkages that wecorporate into the Indexical Hypothesis.

One strength of the constructional viewthat it allows for a parsimonious account ofrelationship between verbs and argument stures. Consider first an alternative approacthis relationship (e.g., Pinker, 1989). Accordto Pinker, for a verb to appear in more thanargument structure, different senses of thatneed to be created. That is, fortoss to appeawith both a transitive argument structure (e“Mike tossed the ball”) and a double-objeargument structure (e.g., “Mike tossed Dathe ball”) requires having two senses of the vtoss.One sense is a transitive sense (“to

hile the other is a transfer sense (“to trany X”). However, if one accepts Goldberg1995) proposal about the existence of constions that carry particular meanings, it is pole to avoid positing these additional sense

oss.On Goldberg’s view,tosscan have a single, general meaning that takes on differshades (e.g., the transfer component of theble-object construction) that are supplied byconstruction in which it is found.

This latter point helps understand certainnovative uses of verbs, such as:

(1) Art sneezed the foam off his beer.

This sentence (adapted from Goldberg, 199unusual in that is presents an intransitive v(sneeze) in a syntactic form in which it must binterpreted transitively. As Goldberg (199notes, such cases are difficult to handle ifrelies on the creation of different verb senseunderstand how verbs can be used in diffeargument structures. It seems unlikely that
have in our lexicon a meaning for “to sneeze”
e

ic-

”c

-

-o

eb

,

r

--f

tu-

-

isb

eot

e

equivalent to “X causes Y to go to Z by sneing.” But, if we have a construction that pathis syntactic form with this type of meaninthen such a sentence can be readily unders(see work on syntactic bootstrapping for discsion of a similar idea; e.g., Gleitman & Gillet1995).

Motivating Goldberg’s (1995) analysisconstructions and their meanings is thesceneencoding hypothesis.According to this hypothesis, argument structure constructions encthe basic scenes of human experience (transfer, acting on, and causing motion;Table 1). Constructions thus aid in sentecomprehension by providing the comprehenwith a cue to the basic nature of the scenevent being described by the sentence. Foample, the double-object construction cuescomprehender that the sentence is about a tfer situation. For the remainder of the article,refer to the sentence forms asconstructionsandthe meanings paired with these forms ascon-structional meanings.

THE INDEXICAL HYPOTHESIS

Glenberg and Robertson’s (1999) IndexHypothesis proposes that three processesused in understanding language. The firsthese is indexing (e.g., mapping) the wordsphrases in the sentence to (a) referents inenvironment or to (b) analog mental represtations (e.g., Barsalou’s, 1999, perceptual sbols; see Barsalou, Solomon, & Wu, 1999,supporting evidence). Thus, indexing eslishes the content of the language: who or wis being talked about. As evidence for this pcess, Glenberg and Robertson (1999) demstrated that the opportunity to index instructito the appropriate actions was a critical factodetermining how well participants acquiredknowledge needed to perform a task (e.g., leing to use a compass; see also Roth, 1999,discussion of this idea in an educationalting).

The second process proposed by the Indcal Hypothesis is the derivation of affordan(after Gibson, 1979) from the referents twere indexed. The termaffordancesrefers to
the ways in which individuals can interact with

eryshee

offor-Tharthethelassthlyusar

.icahe

intstheine

ionctus,cab

h ath

forvicedh.iash

on-if-

plethin&

onthen-il-

othe

or-en-ub-forand

achtaxthethe

.ally,theref-areodelg-Asple,encewhatlou,Theof

cessideula-teps,f theingeserse,”text

t is,so

partthatTheso

vedseen-thed byad),Ex-

xi-


things in their environment. For instance, a pson can interact with a crutch in particular waa crutch can be used to aid one’s walking winjured; to strike something; or to push somthing through a long, narrow crevice. Allthese possibilities for interaction are the afdances that a crutch has for an adult human.affordances that an individual derives from pticular objects reflect both the ways in whichindividual can interact with the object andgoals that the individual has at that particumoment (see Glenberg & Robertson, in preIf the person has an injured leg, for example,“aid in walking” affordance will be more easiderived, whereas if the person wants to psomething through a crevice, the “long and nrow” affordance will be more easily derived

The third process proposed by the IndexHypothesis is that the affordances are mesunder the guidance of (a) intrinsic constraand (b) constraints provided by the syntax ofsentence. Meshing is a process that combaffordances into coherent patterns of act(Glenberg, 1997); that is, actions that can aally be completed to accomplish a goal. Thucrutch affords poking or pushing because itbe held and jabbed. This sort of poking cancombined (meshed) with other actions, sucplacing an apple into a crevice and pushingapple through the crevice. Meshing the afdances of a crutch, an apple, and a creallows for the understanding of, “Lyn pushthe apple through the crevice using a crutcThe meshing process is different from associng, propositionalizing, or parsing in that merespects intrinsic biological and physical cstraints on combination. For example, it is dficult to understand, “Lyn pushed the apthrough the crevice using a thread” becauseusual affordances of threads do not comb(mesh) with the action of pushing (GlenbergRobertson, in press).

The syntax of the sentence provides cstraints on meshing in two ways. First, afterwork on the syntactic form–meaning relatioship (Goldberg, 1995; see also Gleitman & Glette, 1995), the form of the sentence is hypesized to provide cues to the general scen
event that is being described (e.g., a transfe
-:n-

e-

r).e

h-

ld

ss-

anese-e

”t-

ee

-

-or

scene). This “scene” highlights certain affdances and, importantly, the syntax of the stence (in this case, the identification of the sject, direct object, etc.) provides instructionsthe meshing process such that all the objectspeople are placed in the right relations to eother. From the previous example, the synconstrains meshing such that Lyn is holdingcrutch and pushing the apple rather thancrutch is manipulating Lyn in some manner

These three processes interact dynamicnot serially. Upon reading that “Lyn pushedapple through the crevice using a crutch,”erents for Lyn, crutch, apple, and so on,indexed and used to establish a mental m(e.g., Kintsch, 1988; Glenberg, Kruley, & Lanston, 1994; Zwaan & Radvansky, 1998).affordances are derived from Lyn and the apthe meshing process begins. As the sentcontinues, the affordances are meshed intoamounts to a mental simulation (e.g., Barsa1999) of the event depicted in the sentence.syntax of the sentence will be used at allthese stages to both aid in the indexing pro(e.g., forming noun phrases) and to provgeneral constraints on how the mental simtion is to operate. Changes in any of these ssuch as a change in the syntactic analysis osentence, will result in the simulation besystematically altered to accommodate thchanges. Note, however, that in real discoumuch of this work (with regard to the “Lynsentence) would have been done by the conin which the sentence is encountered. Thathe representations for Lyn, the crutch, andon, would have been indexed in an earlierof the discourse, removing the need to dofor the example sentence presented here.discourse would also establish Lyn’s goalsthat the proper affordances could be derifrom the crutch. It is in this way that discouror context will operate to make the comprehsion of sentences felicitous. However, ifaffordances cannot be combined as directethe syntax (e.g., pushing an apple with a threcomprehension suffers (as demonstrated inperiment 3).

The experiments that follow test the Inde
rcal Hypothesis by examining how readers un-

.g.ed-oresingtextotherbtoso, wl to

ea jeca ictei cab

test nd2 eani mst art rov ani odT m-o lang a& ntr ltsE erim ini reh vi-d thai redt

bo gv it-m ld-b si-t las er(

1 ela-t them rmsi ms( en-t e de-g cticf reta-t

artic-i ce-c ncess

.

.

i-

e ofe.g.,heheon-en-di-ost

enton-

he[asareb-

Thetheen-irescip-

twodedal-ob-r of

hatlelyen-


derstand innovative denominal verbs (eClark & Clark, 1979); that is, verbs creatfrom nouns (e.g., tocrutch) that have no standard meaning. Denominal verbs of this sprovide an interesting arena within which to tthe Indexical Hypothesis, as their meanseems to rely almost exclusively on the conin which they are found. What it meanscrutch something, for instance, depends onmanner in which crutch is being used as a vto crutch can mean to strike with a crutch,walk with a crutch, to push with a crutch, andon. Understanding what such a verb meansargue, will depend on the perceptual symbowhich the noun (i.e.,crutch) is indexed, th

ffordances that can be derived from that obnd the constraints that the scene being dep

n the sentence provides on how the objecte used.This article presents four experiments to

he Indexical Hypothesis. In Experiments 1 a, we show that adults are sensitive to the m

ngs associated with particular sentence forhat this sensitivity cannot be attributed to picular lexical items, and these meanings pide constraints on the way that sentencesnnovative denominal verbs are understohis finding is of particular interest, as it denstrates that a mechanism that is used inuage acquisition (e.g., Fisher, 1996; GleitmGillette, 1995) continues to play an importa

ole in the language comprehension of aduxperiment 3 expands on the first two expents to demonstrate how syntactic constra

nteract with affordances in language compension. Finally, Experiment 4 provides eence for the idea that specific affordances (

s, actions within specific contexts) are requio understand sentences.

EXPERIMENT 1

Experiment 1 is designed to assess a numf questions. First, it has been shown usinariety of tasks (e.g., Naigles, Gleitman, & Glean, 1992; Gropen, Pinker, Hollander, Goerg, & Wilson, 1989) that children are sen

ive to the meaning associated with particuyntactic forms. We aim to demonstrate h
along with Fisher, 1994; Naigles and Terrazas
,

tt

t

e:

e

t,d

n

t

-,

--d.

-n

.-ts-

t

era

re

998) that adults are also sensitive to this rionship. Second, we demonstrate thateaning conveyed by particular syntactic fo

s not due to previously learned lexical iteespecially verbs) that may occur in these sences. Finally, this experiment assesses three to which the meaning carried by synta

rames can be used to constrain the interpion of innovative denominal verbs.

To assess these claims, we presented ppants with one of two tasks. In the sentenhoice task, participants saw pairs of senteuch as (1) and (2) or (3) and (4):

(1) Lyn crutched Tom her apple so he wouldn’tstarve. (double-object form)(2) Lyn crutched her apple so Tom wouldn’t starve(transitive form)(3) Lindsay bought Sam a sweater to please him(double-object form)(4) Lindsay bought a sweater to please Sam. (transtive form)

These sentence pairs were followed by ontwo inference statements. One statement (“Tom got the apple”) was consistent with tmeaning of the double-object form, while tother (e.g., “Lyn acted on the apple”) was csistent with the meaning of the transitive stence form. Participants were instructed to incate which member of the sentence pair mstrongly implied that the inference statemwas true. Half of the sentence pairs used cventional verbs [as in (3) and (4)], while tother half used innovative denominal verbsin (1) and (2)]. Note that the sentencesslightly more complex than simple double-oject or transitive constructions need to be.additional phrase was included to control fornumber of participants mentioned in each stence. The double-object sentence form requthree participants (an agent, patient, and reient), whereas the transitive only requires(agent and patient). The additional phrase ada motivation for the action described andlowed us to mention the same people andjects in both sentences. Equating the numbeparticipants helps to rule out the possibility tsubjects’ choices in this task will be made soon the basis of the number of participants m
,tioned in each sentence.

ingheithencingjechea

oseno-istnyris

tascons oal

thenci

oerin

de-ugnndIn-est ist ob

r sruc

is

iciromw

ntsngs“tohedth

eanticnsct

tive

eri-nce-ta-

onsay

rbshedter-ard-gIne de-seinal

u-gydi-eachfor

esivedou-nalher’su-irsay

ion,1),

erer a

nt.ncetencethe

andJ2”on-on-“N

the


If participants are sensitive to the meanassociated with these syntactic forms, tchoices should be strongly influenced byinference statement. That is, when the inferestatement is consistent with a transfer meanparticipants should choose the double-obmember of the sentence pair; in contrast, wthe inference statement is consistent with“act on” meaning, participants should chothe transitive member of the pair. As the invative denominal verbs do not have a preexing “transfer component” to their meaning, asuch meaning found in the sentence must afrom the syntactic form.

The second task is a meaning-choicewhich addresses the question of whetherstructional meanings can provide constraintthe interpretation of innovative denominverbs. Clark and Clark (1979) note thatmeaning of these denominal verbs is unpripled (i.e., that it changes greatly dependingthe context in which it is used). As Pink(1989) notes, however, this lack of principlemeaning is detrimental to communication; ially, a language system should provide enoconstraints on meaning that comprehensioless a function of individual interpretation amore a function of general principles. Thedexical Hypothesis describes some of thprinciples: the affordances of the noun thanamed in the denominal verb provide one seconstraints on verb meaning (an issue toaddressed in Experiments 3 and 4). Anotheof constraints may be provided by the consttion in which the verb appears (a point whichalso made in Clark & Clark, 1979).

To explore this idea, we presented partpants with the denominal verb sentences fthe sentence-choice task. Each sentencepresented individually, and the participawere asked to indicate which of two meaniprovided (e.g., “to act on using a crutch” ortransfer using a crutch”) more closely matcthe meaning of the verb in the sentence. Ifargument structure helps to constrain the ming of these innovative denominal verbs, paripants should be more likely to select the trafer definition for verbs in the double-obje
construction and more likely to select the “act
r

e,tnn

-

e

k-

n

-n

his

e

feet-

-

as

e-

--

on” meaning for the same verbs in the transiconstruction.

Note that the two tasks used in this expment address different issues. In the sentechoice task, we probe participants’ interpretion of sentences. These interpretati(especially in the case of denominal verbs) mbe orthogonal to the interpretation of the vein these sentences. For example, “Lyn crutcTom her apple to prove her point” can be inpreted as conveying a transfer meaning regless of whetherto crutchmeans “to act on usina crutch” or “to transfer using a crutch.”contrast, the verb-choice task addresses thgree to which the constructions will imposome structure on the meaning of the denomverbs.

Method

Participants. The 34 participants were stdents enrolled by introductory psycholoclasses at the University of Wisconsin—Mason. Seventeen participants were used intask. They received extra credit in exchangetheir participation.

Materials.Twenty pairs of critical sentenc(double-object and transitive) using innovatdenominal verbs and 20 pairs of sentences (ble-object and transitive) using conventioverbs were written for this experiment. Tconventional verbs were taken from Pinke(1989) list of verbs that participate in the doble-object construction. Twenty additional paof sentences in different constructions (the wconstruction, the caused-motion constructand the resultative construction; see Tablesome of which contained denominal verbs, wgenerated as filler items. See Appendix 1 folist of all critical sentences for this experime

For the sentence-choice task, two inferestatements were generated for each senpair. One statement was consistent withmeaning of the double-object constructionwas presented in the form “OBJ1 got the OB(e.g., “Tom got the apple”). The other was csistent with the meaning of the transitive cstruction and was presented in the formacted on the OBJ1” (e.g., “Lyn acted on
apple”). Four forms were created to counterbal-

douofenoval

hisin

on

efivaceferct)X”anitivfrooicd brepain

andni

chntsthao

erberliedvethace

whon

reto

sfounalastse

uc-va-hedthef a

nstsreen-

ecta

riptndom2.

osefer”tate-

s

ank( erei tiono f 10).R ct on

)7)

d alljects.


ance the order of the sentences in the pair (ble-object first or transitive first) and the typeinference statement (transfer or act-on). A stence pair appeared once on each form andthe course of the experiment appeared equoften in all combinations. Participants in ttask saw 20 pairs of sentences using denomverbs, 20 pairs of sentences using conventiverbs, and 20 filler items.

For the meaning-choice task, two verb dnitions were generated for each of the innotive denominal verbs used in the sentenchoice task. One definition read “to transusing an X” (consistent with the double objewhile the other read “to act on using an(consistent with the transitive). Each participsaw 10 double-object sentences, 10 transsentences, and 10 filler sentences selectedthe fillers generated for the sentence-chtask. Each of the 30 sentences was followetwo definitions of the verb. Four forms wegenerated to counterbalance (a) whether aticular innovative denominal verb appearedthe double-object or transitive construction(b) the order of the transfer and act-on defitions.

Procedure.After signing consent forms, eaparticipant was given one form. Participagiven the sentence-choice task were toldthey were going to read a series of pairssentences followed by an inference. They wto choose (with a pencil mark) which memof the pair of sentences most strongly impthat the inference was true. Participants githe meaning-choice task were informedthey were going to read a series of sentenFor each sentence, they were to determinethe verb in the sentence meant by choosingof the two definitions.

Results

Sentence-choice task.The data of interest ain Table 2. When participants were askedchoose a sentence consistent with the tranmeaning, they overwhelmingly chose the dble-object construction for both conventioand innovative denominal verbs. In contrwhen participants were asked to choose a
tence consistent with the act-on inference, thet
-

-erly

alal

---

temey

r-

-

tfe

nts.ate

er-

,n-

overwhelmingly chose the transitive constrtion for both verb types. Given that the innotive denominal verbs have no preestablismeaning, these results demonstrate howconstruction can determine the meaning osentence.

Statistical support for these conclusiocomes from a two-factor within-subjecANOVA in which the independent variables aconstruction type and verb type and the depdent variable is the proportion of double-objsentences chosen.1 Analyses with subjects asrandom factor will be denoted with the subsc1; analyses using texts or sentences as a rafactor will be denoted with the subscriptParticipants were much more likely to chothe double-object sentence when the “transstatement was used than when the act-on sment was used [F1(1,16)5 250.66,p , .001;F2(1,19) 5 504.85. p , .001]. Participant

1 Two participants left one of the items on the form bla different item for each participant). Their data wncluded in this analysis after we calculated their proporf double-object sentences chosen of 9 (rather than oemoving these subjects from the data pool had no effe

TABLE 2

Results of Experiment 1

Verb type

Inference type

Transferinference

Act-oninference

Sentence-choice task: Proportion of double-objectsentences chosena

Conventional .92 (.17) .06 (.06Innovative denominal .80 (.19) .04 (.0

Sentence type Proportion

Meaning-choice task: Proportion oftransfer

definitions chosena

Double object .61 (.23)Transitive .42 (.18)

a Standard deviations in parentheses. For this antables, standard deviations were computed across sub

yhe pattern of results observed.

blened

F sce

rmof

jecfo

m,esi-

s2

parthetivet-ontsionnc

p

D

thh area thi icip bleo ncT de-n t thm tot tived nom ont dino den e

ici-theontha

no-

urantwithoicet iseri-le-andt torbs,

insfer

lowni-e.g.,es).ceeon-. Weoftheont ashisofer

ntsble-an-lsothe

inaltruc-thehean-

ncesp afterverb


were also more likely to choose the douobject sentence when the sentences contaiconventional verb [F1(1,16)5 12.96,p , .01;

2(1,19) 5 7.51, p , .025]. This effect iqualified by a significant verb type by inferenstatement interaction [F1(1,16) 5 20.44, p ,.001; F2(1,19) 5 4.61, p , .05]. When theinference statement was of the act-on foparticipants did not differ in the percentagetrials on which they chose the double-obsentence (6% for the conventional verbs, 4%the denominal verbs;F1 and F2 both ,1).When the statement was in the “got” forthough, participants did differ in their choic[92% for conventional verbs, 80% for denomnal verbs;F1(1,16)5 20.91;F2(1,19)5 6.92].

Meaning-choice task.The results from thitask are presented in the bottom of TableAfter reading the double-object sentence,ticipants preferred the transfer definition forverb. In contrast, after reading the transisentence, participants preferred the acmeaning of the verb. Statistically, participawere more likely to select the transfer definitafter having read the double-object sente[F1(1,16) 5 8.78, p , .01; F2(1,19) 5 7.37,

, .025].

iscussion

The data from these two tasks supportypothesis that particular syntactic formsssociated with particular meanings. When

nference statement implied transfer, partants were far more likely to choose the doubject sentence than the transitive sentehis was true for both conventional andominal verbs. These data demonstrate thaeaning of the construction is not purely tied

he semantics of the verb. First, the innovaenominal verbs (e.g., “to crutch”) haveeaning outside of the sentence frame. Sec

he meaning of these verbs changes depenn the sentence frame. When an innovativeominal verb (e.g.,crutch) is presented in th

double-object construction, 80% of the partpants agree that it implies transfer; whensame verb is presented in the transitive cstruction, however, 95% of the people agree
it implies “acted on.” Thus, syntactic forms can
a

,

tr

.-

n

e

e

e--e.

e

d,g-

-t

be used to constrain the interpretation of invative denominal verbs.

There are two minor qualifications to oconclusions. The first arises from the significdifference between the verb types observedthe “transfer” statements in the sentence-chtask. One possible explanation for this effecthat the conventional verbs used in this expment occurred fairly frequently with the doubobject construction (e.g., give, send, bring),many of them have a transfer componentheir meaning (e.g., give). The denominal vehowever, lacked both the frequency of usethe double-object construction and the trancomponent of meaning.

The second qualification arises from theproportion of construction-consistent defitions selected in the meaning-choice task (61% for verbs in the double-object sentencWhile this proportion is different from chan[t(16) 5 2.14, p , .05], it is lower than onmight expect given the hypothesized relatiship between sentence forms and meaningbelieve that this low proportion is a functionthe fact that neither definition presented tosubjects was “wrong” in any sense. Actingsomething is one component of transfer, justransfer is one way of acting on something. Tambiguity may have led to the proportionconstruction-consistent definitions being lowthan expected.

EXPERIMENT 2

Experiment 1 demonstrated that participaare sensitive to the meaning of the douobject construction and that this meaning cnot be solely a product of verb semantics. It ademonstrated that syntactic forms constrainmeaning ascribed to innovative denomverbs. Nonetheless, for both tasks the constional meaning was presented explicitly toparticipants. In Experiment 2, we ask if tparticipants would come up with these meings on their own.

We presented the denominal verb sentefrom Experiment 1 with a context that sets upotential transfer scene (see Table 3). Areading these contexts and the denominal
sentences, the participants were asked to per-

thde

o-er-c-anar

ederbdixpoxt

nt 4ntiaion

m-firstsingas-

hing

erb-d to, (b)en-as-withsen-ansi-

eyhey

en-that

idding

a

test civilbeatenTom andwasove agot an


form one of two tasks, either to paraphrasecritical sentence or to define the innovativenominal verb.

Method

Participants.The participants were 64 intrductory psychology students from the Univsity of Wisconsin—Madison. Thirty-two partiipants were used in each task. The participreceived extra credit in exchange for their pticipation.

Materials. A pair of contexts was generatto introduce each of the 20 denominal vsentences used in Experiment 1 (see Appenfor critical sentences; passages available urequest from the authors). Pairs of contewere generated to fit the needs of ExperimeEach member of the pair presented a potetransfer situation, and each of these situat

TAB

Example Pair of Pa

(1) Tom and Lyn competed on different baseball teaher about striking out three times. Lyn said, “It was ananything to any field using anything!” To prove it, shetwisted ankle, and she grabbed his crutch.

Critical sentencesLyn crutched Tom her apple to prove her point. (doLyn crutched her apple to prove her point to Tom (t

Probes (used in Experiment 4)The crutch is sturdy. (Most Important affordance)The crutch is long. (Not Important affordance)The crutch an help with injuries. (Most Frequent As

(2) Tom and Lyn had made a bad miscalculation. Brights abuses in the dictatorship. But now they were bso badly that she needed a crutch to help her to walkLyn were able to create a long, narrow crevice in thebeing deprived of food in an effort to get him to reveapiece of apple through the crevice, but the wall was toidea.

Critical sentencesLyn crutched Tom the apple so he wouldn’t starve.Lyn crutched the apple so Tom wouldn’t starve. (tra

Probes (used in Experiment 4)The crutch is sturdy. (Not Important affordance)The crutch is long. (Most Important affordance)The crutch can help with injuries. (Most Frequent A

proposed using the object that was named by th

e-

ts-

1n

s.ls

denominal verb in a different way. As an exaple, consider the passages in Table 3. In thepassage, the transfer is accomplished by uthe crutch to hit the apple; in the second psage, the transfer is accomplished by pusthe apple through a crack with the crutch.

For both the sentence-paraphrase and vdefinition tasks, eight forms were constructecounterbalance (a) version of the passagedouble-object and transitive concluding stence, and (c) two random orders of the psages. Each form presented the participant10 passages that ended with double-objecttences and 10 passages that ended with trtive sentences.

Procedure.Participants were told that thwere going to read a series of passages. Twere informed that some of the concluding stences would contain unusual verbs and

3

ges from Experiment 2

. After the game, Tom, who had been pitching, was kerration! I was distracted by your ugly face. I can hitk her apple over to manager who was recovering from

object)sitive)

iate)

use they are U.S. citizens they thought they could prog held incommunicado in a prison dungeon. Lyn wascause the mortar between the bricks was crumbling,e-foot wall separating the cells. Lyn learned that Tomer members of their human rights group. Lyn tried shide, and her arm couldn’t reach through it. Then she

uble object)ive)

ciate)

LE

ssa

msab

too

ubleran

soc

ecaein. Bethrel otho w

(donsit

sso

ethey should try their best to understand what the

ipan iththef-hae.ice

u 1)i yet ctt der het d a0 omg

ot cow 89l on-s rap it tog ife

twoc ni-t ndt oun tchm gc l tot tasi didt 89l

R

ew rap ateT

fep rans seh cec aly aie ,p

p ra-p thed if-i ysis[

asnsTheThe

sfera

nalble-

oore

theon-s inthisthe

ectthethe

uc-n-nces

ore

)

)


passages and sentences mean. The particin the sentence-paraphrase task were thestructed to write a paraphrase for each ofconcluding sentences. The participants inverb-definition task were instructed to write dinitions that best matched their sense of wthe verb meant in the context of the passag

Scoring.The paraphrases were scored twsing two criteria. The transfer score (0 or

ndicated whether the paraphrase converansfer by explicitly indicating that the objeo be transferred was received by the intenecipient. Thus, the paraphrase, “Lyn pushe apple through the crack” would be score

because it does not explicitly state that Tot the apple.The verb score (0 or 1) was based solely

he verb used in the paraphrase. The verb sas 1 if the verb appeared on Pinker’s (19

ist of verbs that take the double-object ctruction. If two verbs were used in the pahrase (e.g., “Lyn found a crutch and usedive Tom the apple”), the verb score was 1ither verb was on Pinker’s list.The definitions were also scored using

riteria. The transfer score was 1 if the defiion included both a transfer of an object ahat the means of the transfer was the named in the denominal verb (e.g., “to crueans to hit something to someone usin

rutch”). The second scoring was identicahe verb score for the sentence-paraphrasen defining the innovative denominal verbs,he participant use a verb from Pinker’s (19ist?

esults

Sentence-paraphrase task.The transfer scoras assigned by M.P.K. One-third of the pahrases were scored by an independent rhe agreement between the two was 95%.Table 4 presents the proportion of trans

araphrases (of 10) for double-object and titive sentences for each participant. Becauad anF ratio of greater than 1, counterbalanondition was included as a factor in this ansis. For the transfer score, there was a mffect of construction type [F1(1,24)5 16.43
, .001;F2(1,19)5 26.2,p , .001]. Partici-
ntsn-ee

t

d

dds

nre)

-

n

a

k:

)

-r.

r-it

-n

ants were more likely to give a transfer pahrase for the sentence when it was inouble-object form. This effect was also sign

cant for the verb score in the subject analF1(1,24) 5 4.40, p , .05], but not in the

analysis by items [F2(1,19)5 3.39,p 5 .08].Verb-definition task.The transfer score w

assigned by M.P.K. One-third of the definitiowere also scored by an independent rater.agreement between the raters was 94%.data are shown in Table 4. For the transcore, participants were more likely to givetransfer definition for the innovative denomiverbs when they were presented in the douobject construction [F1(1,24) 5 18.36, p ,.001;F2(1,19)5 9.78,p , .01]. This result alsshowed up in the analysis of the verb sc[F1(1,24)5 13.38,p , .01; F2(1,19)5 6.18,p , .025].

Discussion

The data from both tasks complementresults of Experiment 1. Participants demstrated sensitivity to constructional meaningtheir paraphrases, and they showed thatmeaning shapes the definition imposed oninnovative denominal verbs. One might objto our interpretation on the grounds thatpassage, not the construction, was providing“transfer” meaning. On this view, the constrtion would not in fact be providing any costraints on the interpretation of these sente

TABLE 4

Results from Experiment 2

Construction Transfer score Verb sc

Sentence-paraphrase taska

Double object .65 (.14) .49 (.15Transitive .52 (.17) .43 (.12)

Verb-definition taska

Double object .50 (.24) .49 (.23Transitive .35 (.21) .32 (.21)

a Standard deviations in parentheses.

and verbs. This hypothesis is supported by the

arath

thear

ha, ath

onterm

web

onn-

tratheenon(i.eitmseateupion

seomgnetexthe

ord

bybe

e 5,rdsair.ingsilyhelThe

thenalic-

ical

is,cesn-thef a

venwith

be(asiontheom-notuldls

n (incesro-

anthe

mail tog the


presence of the large number of transfer pphrases for the transitive sentences. Noneless, this objection cannot explain whydouble-object sentences and verbs were pphrased as meaning “transfer” more often tthe transitive sentences and verbs. Thusexplanation of these data on the basis ofpassage alone is not justified.

EXPERIMENT 3

Experiments 1 and 2 demonstrated how cstructions can provide constraints on the inpretation of sentences and innovative denonal verbs. In Experiments 3 and 4,demonstrate how the constraints providedconstructions (again, the double-object cstruction) interact with the embodied costraints on meshing. We do this by demonsing that the constraints provided byconstructional scene (i.e., the general scspecified by the meaning of the constructiare not enough to ensure comprehensionsuccessful meshing) of sentences. Rather,these constraints in combination with the ebodied constraints that determine whether atence will be understood. As we demonstrwhen the affordances of a situation do not sport the scene provided by the constructcomprehension suffers.

Table 5 presents an example of the texts uin this experiment. As in the passages frprevious experiments, these texts are desito set up a potential transfer situation. Eachhas two versions: one that easily affordstransfer by means of some object (theaffordedversion) and one that does not easily aff

TAB

Example Passa

Rachel worked for a scientist in a research firm. As pahis office so he could open it after lunch. On this pamail addressed to the scientist. The boxes were wa

Affordance manipulating sentence:In the corner of the room, though, Rachel noticed a

Critical sentenceRachelbrought/chairedthe scientist his mail.

transfer by means of that same object, but doe

-e-

a-nne

--i-

y-

t-

e).,is-n-,-,

d

dt

not rule out transfer altogether (thenonaffordedversion). This manipulation was achievedchanging a property of the object that is toused in the transfer. In the example in Tablwhen the chair has four good wheels, it affotransferring the mail by means of the chHowever, when the chair has four misswheels, transfer via the chair is no longer eaafforded, although it is still possible that Raccould somehow get the mail to the scientist.passage ended with a critical sentence indouble-object form containing a conventioverb or an innovative denominal verb. Predtions focus on the reading time for these critsentences.

According to the Indexical Hypothesmeaning arises from the mesh of affordanguided by intrinsic biological and physical costraints and the scene or goal specified byconstruction. Thus, when the affordances ochair with wheels (it can easily be pushed ewhen stacked with items) can be meshedthe affordances of large boxes (they canstacked on a chair) to accomplish transferspecified by the construction), comprehensshould be easy and fast. However, whenaffordances cannot be easily meshed to accplish transfer (a chair without wheels doeseasily afford pushing), comprehension shobe difficult and slow. This prediction parallethe data reported in Glenberg and Robertsopress) that shows that lack of proper affordanfor the action described by a sentence will pduce comprehension difficulties.

The Indexical Hypothesis also predictsaffordance by verb-type interaction. When

5

from Experiment 3

f her duties, she was required to bring the scientist’sular day, Rachel encountered three large boxes amono big for her to carry.

ffice chair with fourgood/missingwheels

LE

ge

rt orticy to

n o

scritical sentence contains a conventional verb,

deelyonna

themag)inabeingsileriing

u-d-ex

tettegns

saonnsthont dfe(onde-fornca

sagentive

s-nduldony tedar

12tionsaf-of

mi-domf 4be

antsrom

via-vedtheddi-/noncerage%),

ues-ur

thed by

areent

e-ncess. In

asen-es.d


we expect the difference between the afforand nonafforded conditions to be relativsmall. This is because the meaning of the cstruction and the semantics of the conventioverb should allow the reader to infer thattransfer took place even though he or shenot be sure how it occurred (e.g., by bringinWhen the critical sentence contains a denomverb, however, we expect this difference tomuch larger. In this case, the noun underlythe denominal verb either can or cannot eaeffect the transfer, and the reader will expence greater difficulty in determining a meanfor the sentence.

Method

Participants. The participants were 36 stdents from the University of Wisconsin—Maison. The students received extra credit inchange for their participation.

Materials. Twelve passages were generafor this experiment. The passages were writo suggest a potential transfer scene usinparticular object as a potential means of trafer. The penultimate sentence of each paswas manipulated to produce afforded and nafforded versions of the text. Afforded versiodescribed the object as having affordancessupported use of the object in the transfer; nafforded versions presented affordances thanot support the use of the object in the transFor each passage, two critical sentenceswith a conventional verb and one with anominal verb) were written (see Appendix 2a list of critical sentences). Four counterbalaconditions were created to ensure that theforded and nonafforded version of each paswas presented equally often with critical stences containing conventional and innovadenominal verbs.

Procedure.Participants were randomly asigned to one of the four counterbalance cotions. The participants were told that they woread each paragraph sentence by sentencescreen of a computer and that they should trunderstand each sentence fully. To advancthe next sentence, the participant needepress the space bar on the computer keybo

After reading the instructions and going

d

-l

y.l

y-

-

dna-

ge-

at-idr.e

ef-e

-

i-

theototod.

through a practice text, each participant readpassages [3 passages in each of 4 condiformed by crossing the two types of texts (forded and nonafforded) with the two typesverbs (conventional and innovative denonal)]. The passages were presented in ranorder with the constraint that in each block otrials, 1 passage from each condition wouldpresented. Following each passage, participanswered a “yes/no” question about details fvarious points in the text.

Results

Reading times greater than 3 standard detions from their respective mean were remofrom the dataset as outliers. This involvedloss of less than 1% of the responses. In ation, we checked for accuracy on the yesquestions that followed each text. Performawas quite good on these questions (aveperformance across participants was 96.8and no participants missed more than two qtions, indicating that they were following oinstructions.

The dependent variable of interest wastime to read the critical sentence as measurethe time between key presses. These datapresented in Table 6. The data are in agreemwith the predictions of the Indexical Hypothsis. Participants read the afforded sentemore quickly than the nonafforded sentenceaddition, this difference was more than twicelarge for the innovative denominal verb stences as for the conventional verb sentenc

A within-subjects ANOVA demonstrate

TABLE 6

Reading Time (in Seconds) for CriticalSentences in Experiment 3a

Verb type

Passage type (Nonafforded2 afforded)differenceAfforded Nonafforded

Conventional 2.37 (.90) 2.83 (.85) .46Denominal 3.21 (1.09) 4.31 (1.33) 1.10

a Standard deviations in parentheses.

main effects of both verb type and affordances.

enm

othethag

,of

or-nsfutn-

y twn

pretannal

orce

ac-owrbiaarenteingnc

ceenthaoueneages

boththeone

rayed. Weafter

on.to

achd themi-m-am-thatr-

alsoounn ofe inot

oware

for-af-irdAs-y a

ar-antcence,ene isardto

or-to

thaniate

o-er-x-

datp tes ncw.


Participants read critical sentences with convtional verbs faster than sentences with denonal verbs [F1(1,35) 5 120.04, p , .001;F2(1,11)5 28.30,p , .001]. Participants alsread the critical sentences faster followingafforded version of the passages than whensentences followed the nonafforded pass[F1(1,35)5 24.74,p , .001;F2(1,11)5 20.61p , .001]. Finally, there was an interactionaffordance and verb type [F1(1,35)5 4.63,p ,.05; F2(1,35)5 3.80,p 5 .077].2

Discussion

When the context does not provide affdances needed to understand how the traimplied by the critical sentence was carried oparticipants had a much more difficult time uderstanding the sentences (as evidenced blonger reading times). Thus, we have shothat affordances are important to the comhension of sentences and particularly importo the manner in which innovative denomiverbs are understood.

The effect of verb type is only of mininterest. The slower reading times of sentenwith denominal verbs may be attributed to ftors such as their novelty, the relatively lfrequency with which they are used as ve(i.e., zero), or other similar factors. The crucfinding in these data is that affordancesconsidered in understanding the critical stences. When affordances are proportionaimportant for understanding (i.e., when readthe innovative denominal verbs), the affordamanipulation is proportionately large.

EXPERIMENT 4

Experiment 3 demonstrated that affordanare important to the comprehension of stences. In this experiment, we demonstrateparticular affordances are derived to carrythe action specified by the constructional sc

Participants were presented with the passand critical sentences from Experiment 2. Th

2 The data were also analyzed including outliers. Theroduced the same pattern of results as reported in theave for the fact that the interaction of verb and affordaas significant in the analysis by items in this case (p 5

05).

-i-

ee

er,

he

-t

s

sle-ly

e

s-t

t.

ese

passages were generated in pairs such thatmembers of the pair involved the use ofsame object to transfer something to someelse (see Table 3), and each passage portthe object as being used in a unique mannerasked the participants to read the passages,which we presented a probe for verificatiThe participants answered “yes” or “no”these probes (see Table 3).

The probes were of three types. For epassage, there was a probe that describeaffordance of the noun underlying the denonal verb that was Most Important to the coprehension of the critical sentence. For exple, in the first passage in Table 3, the fact“the crutch is sturdy” is important to undestanding how Tom got the apple. There wasa probe that named an affordance of the nthat was Not Important to the comprehensiothe critical sentence. For the first passagTable 3, the fact that the “crutch is long” is nparticularly important to understanding hTom got the apple. Note that the materialsconstructed so that the Most Important afdance from version 1 was a Not Importantfordance in version 2 and vice versa. The thprobe type was named the Most Frequentsociate to the noun. It was determined bnorming study.

According to the Indexical Hypothesis, pticipants should be faster to verify the ImportAffordance than the Not Important affordanbecause, in understanding the critical sentethe Most Important affordance will have bederived. The Most Frequent Associate probintended as a control to provide a standagainst which to compare the reaction timethe other probes. If the Most Important affdance is truly being highlighted, responsesthese probes should be as fast or fasterresponses to the Most Frequent Assocprobes.

Method

Participants.The participants were 50 intrductory psychology students from the Univsity of Wisconsin—Madison. They received e

axt,e

tra credit in exchange for their participation.

omFoesorafof

orm3).bethedr-

ox-r o5 tthaons n.o

steas

itioostalsewitore”e atheth

e”-moffecFretionm-alboth

or-e”s tbe.ee

thi

ext)wnfter

ostMostthe

thes.ntedbothr-ratech

n ofem-

tingdedgeshyri-

or-pre-hatdueon-

for-in-

tive.d tolysisu-m,terted-s ofre-curm,be

ededrb,heses ofvec-. If

woexts


Materials. Eighteen pairs of passages frExperiment 2 were used in this experiment.each passage, there were three probe typMost Frequent Associate probe, a Most Imptant affordance probe, and a Not Importantfordance probe (see Appendix 3 for a listprobe items). Probes were generally of the f“the [noun] is/can [affordance]” (see TableOf the 54 probes used in this study (3 proeach for 18 denominal verbs), 33 were of“can” form, 16 of the “is” form, and 5 useanother form (e.g., “the [noun] has [affodance]”). All three probe types were of apprimately the same length; the mean numbecharacters in each probe type ranged from 226.4. In addition, 20 distractor passageswere similar to the original passages were cstructed. For each passage, a probe that watrue was generated (e.g., “crayons can fly”)

To increase the number of observationseach probe type, the type of probe contrawith the Most Important affordance probe wmanipulated between subjects. In one cond(i.e., the Most Important affordance vs MFrequent Associate condition), four counterbanced sets of items were generated to preboth versions of each passage equally oftenthe Most Frequent Associate and Most Imptant affordance probes. Here, the “affordancbased probes (i.e., the probes based on thfordances of the object used to effecttransfer) were presented equally often as“Most Important” probes, and the “associatbased probes (i.e., the probes based on thefrequent associate to the object used to ethe transfer) were presented as the “Mostquent Associate” probes. In the other condi(i.e., the Most Important affordance vs Not Iportant affordance conditions), four counterbanced sets of items were created to presentversions of each passage equally often withMost Important affordance and the Not Imptant affordance probes. Each “affordancbased probe was presented equally often aMost Important and the Not Important proAny one participant saw only two of the thrprobe types in the experiment.

Three phases of norming were used in
experiment. First, the nouns named by the de
r: a--

s

fot-ot

nd

n

-nth--f-

e

stt-

-the

-he

s

nominal verbs were presented (out of contto 46 participants who were asked to write dothe first three words that came to mind areading each noun. For each noun, the mfrequent response was used to generate theFrequent Associate probe. In no cases didMost Frequent Associate overlap with eitherMost Important or Not Important affordance

In the second norming phase, we presethe passages to 40 participants, along withthe Most Important and Not Important affodance probes. We asked the participants to(on a scale from 1 to 7) how important eaaffordance was to understanding the actiothe critical sentence. Passages that did not donstrate a mean difference of at least 2 rapoints between each probe type were exclufrom the experiment. Two pairs of passawere excluded by this criterion, which is wwe used only 18 of the 20 pairs from Expement 2.

Note that in this experiment, the Most Imptant affordance probe is determined by theceding context. Thus, there is a possibility tdifferences in reaction times to the probe isto direct associations between words in the ctext and words in the probe rather than afdances. The final phase of norming wastended to counter this association alternaThe passages and probes were submittethree analyses using Latent Semantic Ana(LSA; see Landauer, 1999, Landauer & Dmais, 1997, and Landauer, Foltz, & Laha1998, for a discussion). LSA is a compuprogram that computes an index of the relaness between sets of words on the basioccurrences in similar contexts. Words arelated, on this view, to the degree that they ocin similar texts (see Landauer, Foltz, & Laha1998). We used LSA to compare the proconcepts to (a) the whole passage that precit, (b) the noun named by the denominal veand (c) the last sentence of the passage. Tcomparisons yield a metric of the relatednesthe two concepts: the cosine between thetors representing the stimuli being comparedthe cosine is low (i.e., close to zero), the tconcepts appear in nearly orthogonal cont
-and hence are not likely to be associatively

two

ineobageafanch

eiat

ately

t tha

the

oneddeineorinath

beor-rel. Thonyinnnfo

ffer.02

dife f

eycinth

thre1

be

d towasreas

o aobentheor-the

iateantr-

for-and

iateof

obe

evia-ereltedses.tererethe

antdi-

asre-nted

m-erean

entlsoce

dif-ach

SAT useb wsp


related. A cosine of 1.0 indicates that thestimuli appear in identical contexts.3

In the first analysis, we calculated the cosbetween the main concept from each pr(e.g., “sturdy”) and the accompanying passThe average cosine for the Most Importantfordance probes (.06) and the Not Importaffordance probes (.05) did not differ from eaother (F , 1), but they did differ from thaverage cosine of the Most Frequent Assoc(.14) [F(1,53)5 12.33,p , .01; andF(1,53)517.08,p , .01, respectively]. This indicates ththe Most Frequent Associate is more closrelated to the passage (in the sense thaassociate and words in the passage tend topear in similar texts) than were either ofaffordance probes.

In the second LSA analysis, the probe ccepts (e.g., “sturdy” and “long”) were comparto the noun named by their accompanyingnominal verb (e.g., “crutch”). The mean cosfor the Most Frequent Associate (.36) was mrelated to the noun named by the denomverb than were the mean of the cosines fortwo affordances (.16), [F(1,33) 5 38.70,p ,.001]. These analyses do not differentiatetween Most Important and Not Important affdances because importance is determinedtive to the context presented by the passagethird LSA analysis compared the probe ccepts to the last sentence of their accompanpassage (i.e., the sentence containing the ivative denominal verb). The mean cosinethe Most Frequent Associate (.01) did not difrom the mean cosine for the affordances ([F(1,35)5 1.57,p 5 .22].

The LSA analyses demonstrate either noference between probe types or an advantagthe Most Frequent Associate probes.

Procedure.Participants were told that thwere to read passages line by line, advanfrom one sentence to the next by pressingspace bar. At the end of each passage,asterisks were presented on the screen forand this was followed by one of the three pro

3 The analyses were conducted using the LASA-WK space. In this space, the corpus of languagey LSA in computing relatedness come from novels, ne
apers, and other types of texts.
e.

-t

e

ep-

-

-

ele

-

a-e

-go-r

)

-or

gees,s

for the passage. The participants were tolrespond as to whether the probe statementtrue by pressing a “yes” or “no” key. They wetold to answer as quickly and accuratelypossible.

Participants were randomly assigned tcondition that presented two of the three prtypes. All participants received Most Importaaffordance probes. Approximately half of tparticipants also saw the Not Important affdance probes (24 participants), whereasother half saw the Most Frequent Assocprobes (26 participants). In all, each participsaw 18 critical probes (9 Most Important Affodance probes and either 9 Not Important Afdance or 9 Most Frequent Associate probes)18 distractor probes to which the appropranswer was “no.” The dependent variableinterest was the response time to the pritems.

Results

Analysis of response times for all texts.Theresponse times greater than 3 standard dtions from the mean in each condition wremoved from the data as outliers. This resuin a loss of less than 2% of the responAdditionally, participants who made greathan 22% errors across all of the trials wexcluded from the analysis. This resulted inloss of one participant from the Most Importaffordance vs Not Important affordance contions. Finally, all trials on which an error wmade was excluded from the analysis ofsponse times. The data of interest are presein Table 7.

In the Most Important affordance vs Not Iportant affordance conditions, participants wfaster to verify the Important Affordance ththe Not Important affordance [F1(1,23)5 6.25,p , .025; F2(1,35) 5 8.69, p , .01]. In theMost Important affordance vs Most FrequAssociate conditions, participants were afaster to verify the Most Important affordan[F1(1,25)5 14.03,p , .001;F2(1,35)5 7.14p , .025].

Further analysis demonstrated significantferences in the number of errors made in ed-
condition. Participants made more errors on the

re-osreino

ost

tesauwit

teses

iateinaeathema

se,eenthis

etryTherob-

sureoddwithforex-

ges5%.ges

animi-tici-

otar-forndtli-

m-

ent 4

ociate

affo


Not Important affordance (31%) and Most Fquent Associate (28%) probes than on the MImportant affordance probes (8 and 14%,spectively). This difference was significantboth the Most Importance affordance vs NImportant affordance condition [F1(1,25) 531.50,p , .001; F2(1,35)5 18.78,p , .001]and the Most Important affordance vs MFrequent Associate condition [F1(1,25) 517.41,p , .001; F2(1,35)5 5.55, p , .025].The direction of the difference in error raprecludes a speed accuracy trade-off becthe higher error rates were associatedlonger response times.

Analysis of Response Times for Low-ErrorRate Texts

It appeared as if the difference in error rawas due to difficulties with specific passagFor example, the Most Frequent Assocprobe for a passage involving the denomverb “to mouse” was “the mouse cancheese.” This was true for one version ofpassage, in which the mouse was a live ani

TAB

Response Times in Seconds and Error Proportion

Probe Type Most Important affordance

All

ConditionMIA vs NIA a

Response time 2.37 (.53)Error rate .08 (.10)

MIA vs MFA a


Exclu

ConditionMIA vs NIA


MIA vs MFAResponse time 2.35 (.63)Error rate .06 (.10)

a IA, Most Important affordance; NIA, Not Important

however, it was false of the other version, in

t-

t

seh

.

lt

l;

which the mouse was a toy. In the latter cathe correct response (i.e, “no”) would have bscored as an error. Of the 36 texts used inexperiment, 18 showed this type of asymmand displayed error rates in excess of 25%.passages that did not show these types of plems had much smaller error rates. To enthat our conclusions are not affected bypassages, we performed analyses on textslow error rates. We examined the error rateseach passage across all participants andcluded from analysis those pairs of passathat produced an error rate of greater than 2This resulted in the loss of 9 pairs of passa(of 18). After this step, all participants witherror rate of greater than 22% were also elnated. This resulted in the loss of one parpant from the Most Important affordance vs NImportant affordance condition (the same pticipant who was excluded from the analysisall texts). Finally, all incorrect responses aoutliers were excluded from the analysis (ouers defined as above).

In the Most Important affordance vs Not I

7

with Standard Deviations in Parentheses) for Experim

Not Important affordance Most Frequent Ass

ts

2.74 (.92) —.31 (.19) —

— 2.96 (1.05)— .28 (.19)

texts

2.85 (1.26) —.25 (.22) —

— 2.76 (1.15)— .13 (.18)

rdance; MFA, Most Frequent Associate.

LE

s (

tex

ding

portant affordance conditions see Table 7), par-

ntces

p vsM ntw r-d argF

D

nsd ici-p im-p ent nta hir pef nay inp nyo

oft e oa thg rtyv onw edF osI end det e ot anb gen saa , as eci n-d f tht oca dinc ncet re-fl wes est din thes ers

theteding

om-nd 2itiveyn-due

rbs.trate

d toinalhaten-andy ah of

theas,ningandys.nalpar-dif-hen

aveces.nal

eta-

ndold-

earsuisi-ibly,and

ificlop-

onan,ideduc-of


ticipants were faster to verify the ImportaAffordance than the Not Important affordan[F1(1,23) 5 8.20, p , .01; F2(1,17) 5 7.86,

, .025]. In the Most Important affordanceost Frequent Associate condition, participaere faster to verify the Most Important affoances, although the difference was only minally significant [F1(1,25)5 3.81, p 5 .06;2(1,17)5 3.17,p 5 .09].

iscussion

The data are in accord with the predictioerived from the Indexical Hypothesis. Partants were faster to verify the affordanceortant for the understanding of the critical s

ence than to verify either the Not Importaffordance or the Most Frequent Associate. Tesult held across both analyses that weormed (the weaker effects in the second asis were likely due to the dramatic lossower that resulted from excluding so mabservations).One might object to our characterization

hese data as a demonstration of the rolffordances in sentence comprehension onrounds that our task was simply a propeerification task. There are a number of reashy we believe this objection to be misguidirst, the faster response times for the M

mportant affordance probes is context depent. That is, the affordance that was respon

o more quickly was determined by the naturhe transfer action that needed to be takeny the goals of the characters in the passaot by a simple association between the pasnd the property named in the probe. Alsoimple associative account of this context spficity is ruled out by the LSA analyses coucted on the probes and passages. In two o

hree LSA analyses, the Most Frequent Asste was more strongly related to the preceontext than were either of the two affordaypes. If the reaction time data were onlyecting associative priming by the context,hould have seen the fastest response timhe Most Frequent Associate probes, but weot. Further, if associative priming wasource of this observed pattern of data, th
hould have been no difference between th
s

-

-

sr-l-

fe

-s

.t-dfds,ge

-

ei-g

tod

e

response times to the Most Important andNot Important affordances, as both were rela(in the LSA analyses) equally to the precedcontext.

GENERAL DISCUSSION

The experiments reported here have accplished several goals. First, Experiments 1 ademonstrate that adult participants are sensto the meanings associated with particular stactic forms and that these meanings are notto previously established meanings of veSecond, these same experiments demonsthat constructional meanings can be useconstrain the meaning of innovative denomverbs. Finally, Experiments 3 and 4 show tconstructional meanings are not enough tosure felicitous understanding of sentencesdenominal verbs. Rather, affordances plarole in sentence comprehension as well. Eacthese is discussed in turn.

In Experiments 1 and 2, we replicatedfinding (Fisher, 1994; Naigles & Terraz1998) that adults are sensitive to the meaassociated with particular sentence formsexpanded on the existing literature in two waFirst, we demonstrated that constructiomeanings are not tied to the semantics ofticular verbs. Participants were sensitive toferences in constructional meanings even wdifferential cues to this meaning could not hcome from the verbs and nouns in the sentenSecond, we have shown that constructiomeanings provide constraints on the interprtion of innovative denominal verbs.

The relationship between syntactic form ameaning discussed here and elsewhere (Gberg, 1995; Pinker, 1989; Fisher, 1996) appto be an important aspect of language acqtion, language comprehension, and, ostenslanguage production (see Goldberg, 1999,Tomasello & Brooks, 1999, for more specconstructional approaches to language devement). A large body of work on verb acquisiti(e.g., Pinker, 1989, 1984; Landau & Gleitm1985) has demonstrated that the cues provby such form–meaning links (i.e., constrtions) may be vital to learning the meaning
enew verbs. As the work of Fisher (1994; see

ermaverticismon

ageimth

oran

es’seeigeoniveveanit-ns

avenaratrticdeeasethiWeexnecaa

lesat iceof

ify.ingale aonci-

rsaer,

old-ts in.howac-de-chn toingas

spe-fectts toSuchsioneirthe

rk,eal

ountinaloun79)In

on-d by

t topicalfor-y to

ofp usow

m isthe

ym-sticbe

g ofac-

n offor-(or,thatideuc-


also Naigles & Terrazas, 1998) and the expments reported here demonstrate, this linkbe important for adults’ understanding of noutterances as well. These findings are of paular interest, for they suggest that a mechanthat is used in the acquisition of language ctinues to play an important role in the languprocessing of adults. This idea has strongplications for the study of language in boadults and children (see MacDonald, 1999, fdiscussion of some ways to relate adultchild language research).

While we consider our experiments a succin providing basic support for Goldberg(1995) approach to the relationship betwsentence forms and meaning, this success mbe qualified by two limitations. First, thpresent experiments examined only two cstructions: the double object and the transitSecond, our data do not, in and of themselrule out alternative approaches to form–meing links (e.g., Fisher, 1996; Landau & Gleman, 1985; Pinker, 1989). These limitatiolead to the criticism that our experiments hnot demonstrated sensitivity to constructiomeanings, but, rather, that they demonstsensitivity to the presence or absence of paular thematic roles. Participants may havecided that the double-object sentences mtransfer when the denominal verbs were prebecause they detected the “recipient” role insentence form and not in the transitive form.would argue, however, that this alternativeplanation is not much different from the ooffered by the constructional approach. Rethat constructions carry both syntactic informtion and information about the thematic rolinked to each syntactic category. Part of whis to recognize these constructions and actheir meaning may well be the recognitionthe pattern of thematic roles that they spec

This criticism does raise an issue regardthe linguistic knowledge to which an individuhas access. The type of linguistic knowledgwork in these experiments is knowledge of cstructions, or knowledge of some other prinple, such as that specified by the UniveTheta Assignment Hypothesis (UTAH; Bak
1988). Unfortunately, our data do not distin-
i-yl-

-

-

ad

s

nht

-.s,-

le--ntnts

-

ll-

tss

t-

l

guish among these alternatives, although Gberg (1995) presents a series of argumenfavor of adopting a constructional approach

The experiments reported here also show the Indexical Hypothesis provides ancount of the comprehension of innovativenominal verbs. The syntactic form in whithese verbs were encountered was showplay a large role in determining the meanthat was ascribed to them. In addition, it wshown that particular affordances (e.g., thecific actions required to use the object to efa transfer) were necessary for the participanunderstand sentences with these verbs.constraints are important to the comprehenof denominal verbs, as, on the whole, thmeaning cannot be well predicted outside ofcontext in which they appear (Clark & Cla1979; but see Kelly, 1998). Note that our appto the concept of affordances does not amto saying that in using an innovative denomverb, we employ the object named by the nin a stereotypic manner. Clark and Clark (19argue that this characterization is wrong.addition, the results from Experiment 4 demstrate that unusual uses of the objects namethe denominal verbs were more importanunderstanding sentences than were stereotyuses. We believe that the computation of afdances for the noun is perhaps the only wadetermine the use of an object in context.

Experiments 3 and 4 build on the workGlenberg and Robertson (in press), and helto answer the question that we started with: hare sentences understood? Our basic claithat meaning is not achieved solely throughcombination of words and other abstract sbols (e.g., nodes) by syntactic and probabilimanipulations. Instead, the symbols mustgrounded, and we propose that the meaninsituations and of sentences is grounded intion.

But just as language is not a concatenatiowords, meaning is not a concatenation of afdances. The affordances must be combinedmeshed) into a coherent pattern, that is, onesupports action. Experiments 3 and 4 provthe first evidence for how this is done: constr
tional meanings provide a framework for the

ofsi

ainpeforn

n-ofwit.g.is

ifytheificc-arlbois

dehre7)

phamdinmina

t caretsisabucr-toplt ousis

ssee-turededtheivege

racttedab-our

riseon-oryellces

ex-tionsayul-

ary.tract

onserv-ted

iser-m-

g ofirs,lanstractreteate

up-larar-y-nderthe

an-lan-tionhevan-onand9).

ningxi-esson

riz-


selection (Experiment 4) and combination(Experiment 3) affordances. That is, the bascene specified by the construction constrthe mesh of affordances so that the goals sified by the scene are achieved. If the afdances cannot be meshed, comprehensioslow and unsuccessful (Experiment 3).

Work on the combination of lexical and cotextual information in the comprehensionsentences (e.g., Swinney, 1979) has dealtsimilar issues, as has work in linguistics (eFauconnier’s, 1999, “mental spaces” analysThe novel contribution of this work is to specthat a particular type of knowledge aboutworld (i.e., affordances) interacts with a spectype of linguistic knowledge (i.e., construtions) to produce meaning. There are clemany questions that need to be answered aour approach. One important questionwhether the idea of affordances can be extento abstract concepts. There are at least tmechanisms for doing so. First, Lakoff (198discusses how image schemas can be metaically extended to abstract domains. For exple, Lakoff suggests that human understanof a container is a structured image schebased on direct experience. The schemacludes the information that a container hasinside and an outside and that another objecbe inside or outside, but not both. This concunderstanding of containers forms the baLakoff asserts, for our understanding of anstract logical principle that has a related strture, namely “p or not-p, but not both.” Basalou (1999) offers another approachunderstanding abstract ideas. As an examconsider Barsalou’s analysis of the conceptruth. He suggests that truth arises from a mtistep comparison process. First, one formsimulation of a situation. This simulationoften prompted by language, such as the ation, “The airplane is in the sky.” Second, bcause the simulation makes use of percep(i.e., analogical) symbols, it can be compawith an actual situation that is being perceivThird, the language is judged as “true” whensimulation substantially matches the percesituation. Fourth, the concept of truth emer
as a conscious application of this procedure.
csc--is

h,).

yut

de

or--ga-

nn

e,--

e,fl-a

r-

al

.

ds

The final approach to understanding abstideas within an embodied system is illustrain the previous paragraphs: we understandstract ideas from concrete examples. Thus,understanding of containment may well afrom experience with specific examples of ctainers, just as our understanding of a the(e.g., about how language works) may warise from specific examples (e.g., sentenabout crutches). Importantly, the individualamples use language about concrete situafrom which affordances can be derived. It mbe that abstract symbols emerge from the mtiple examples, but it is by no means necessFor instance, an understanding of the absconcept of democracy may well be basedspecific experiences and actions such as obing a parent casting a ballot. A sophisticaunderstanding of the notion of democracymore than this, but that sophisticated undstanding may well be based on multiple exeplars, just as a sophisticated understandinthe notion of a chair must include dining chastuffed chairs, beanbag chairs, and bachairs. This approach to understanding absideas is consonant with the use of concexamples at all levels of pedagogy to explicabstractions.

In this article we have demonstrated sport both for the hypothesis that particusyntactic forms (i.e., constructions) carry pticular meanings and for the Indexical Hpothesis that affordances are meshed uthe guidance of constructions. Whereasfocus of this work has been on deriving meing from sentences, the broad scope ofguage use indicates that the communicaof meaning is opportunistic. In addition to tdevices discussed here, language takes adtage of gesture (MacNeill, 1998), commground (Clark, 1996), and the real peopleobjects in the environment (see Roth, 199We believe that these components of meamaking can be interpreted within the Indecal framework. At the very least, the succof this project demonstrates how a focusmeaning can profitably constrain our theo
ing about language processing.

2

ndmaageublesitiv

keikebal-

t heary

almbo

himhimp hsist

ul

uitauita

forfor

eeeer

er. J

coufin

any

he

him

ouldent

rvena-yn

himmilehe

rik

ArtArte to

1:nce

aked

axed

nt a

herrm.rk

isit.

akeister

himppy.im.

ax

ed

him.

Paul

r life.

rtha

aids.

gry.

her.

rom.


APPENDIX 1

Critical Sentences from Experiments 1 and

Denominal verb sentences used in Experiments 1 aAlternative versions of the sentences refer to changesin the sentences so they would fit better with the passused in Experiment 2. The sentence on top is the doobject sentence; the sentence on bottom is the transentence.

(1) Mike ballooned David the toy to help him out. Miballooned the toy to help David out. (Alternative: Mballooned David the toy to continue the game. Mikelooned the toy to continue the game with David.)

(2) Pauline spatulaed Mary the cookie dough to meerequest. Pauline spatulaed the cookie dough to meet Mrequest.

(3) The old man cupped the boy some popcorn to chim down. The old man cupped the popcorn to calm thedown.

(4) Lois blanketed the neighbors her baby to saveLois blanketed her baby so the neighbors could save

(5) Maureen postcarded her sister the news to keeinformed. Maureen postcarded the news to keep herinformed.

(6) Paul rocked Bill the lure to give him luck. Parocked the lure to give Bill luck.

(7) Adam booked Jareb his request to stop the gplaying. Adam booked his request that Jareb stop the gplaying.

(8) George bindered Lydia the note so he could askdate. George bindered the note so he could ask Lydiadate.

(9) Blind Willie saxophoned John a happy story to chhim up. Blind Willie saxophoned a happy story to chJohn up.

(10) Jeb tractored Gwenda a message to reassure htractored a message to reassure Gwenda.

(11) Doug bottled the rescuers a message so theyfind him. Doug bottled a message to the rescuers couldhim.

(12) Sally pursed Pete the fish to keep him compSally pursed the fish to keep Pete company.

(13) Duncan Tonkaed mom his laundry to appeaseDuncan Tonkaed his laundry to appease his mom.

(14) Todd branched Rick a scratch to get back atTodd branched a scratch to get back at Rick.

(15) Rachel chaired the scientist the mail so he copen his packages. Rachel chaired the mail so the scicould open his packages.

(16) Lyn crutched Tom her apple so he wouldn’t staLyn crutched the apple so Tom wouldn’t starve. (Altertive: Lyn crutched Tom her apple to prove her point. Lcrutched her apple to prove her point to Tom.)

(17) Mina enveloped Larry some chocolate to makesmile. Mina enveloped some chocolate to make Larry s

(18) The operator craned the mason the cement to
him. The operator craned the cement to help the mason.
2.des-e

r’s

y

.

.erer

rr

aa

r

eb

ldd

.

r.

.

ist

.

.lp

(19) Erik fanned his cat the ball to satisfy him. Efanned the ball to satisfy his cat.

(20) Art moused Jim a greeting to make him laugh.moused a greeting to make Jim laugh. (Alternative:moused Jim a joke to make him laugh. Art moused a jokmake Jim laugh.)

Conventional Verb sentence pairs used in Experiment(21) Vince cooked Frank a meal for his retirement. Vi

cooked a meal for Frank’s retirement.(22) Peter baked John a cake to surprise him. Peter b

a cake to surprise John.(23) Josh faxed his boss a letter to satisfy him. Josh f

a letter to satisfy his boss.(24) Wendy sent Kyle a card to apologize. Wendy se

card to apologize to Kyle.(25) Thomas flung his girlfriend his coat to keep

warm. Thomas flung his coat to keep his girlfriend wa(26) Mark painted Sally a portrait for her birthday. Ma

painted a portrait for Sally’s birthday.(27) Bruce bought Seline a present during his v

Bruce bought a present during his visit to Seline.(28) Jack won his little sister a stuffed animal to m

her smile. Jack won a stuffed animal to make his little ssmile.

(29) Lindsay purchased Sam a sweater to makehappy. Lindsay purchased a sweater to make Sam ha

(30) Kevin blasted Simon the rocket ship to impress hKevin blasted the rocket ship to impress Simon.

(31) Max poured Luke a drink to please him. Mpoured a drink to please Luke.

(32) Julie fixed Ivor a dinner to surprise him. Julie fixa dinner to surprise Ivor.

(33) Fred tossed Bob a rotten egg to get back atFred tossed a rotten egg to get back at Bob.

(34) Paul recorded Marge a song to show his love.recorded a song to show Marge his love.

(35) The scout radioed Mandy a message to save heThe scout radioed a message to save Mandy’s life.

(36) Bertha mailed Betty a letter to make amends. Bemailed a letter to make amends to Betty.

(37) The woman gave the fire victims ten dollars tothem. The woman gave ten dollars to aid the fire victim

(38) Liz tossed Brain a salad so he wouldn’t be hunLiz tossed a salad so Brian wouldn’t be hungry.

(39) Tony shipped his wife a souvenir to placateTony shipped a souvenir to placate his wife.

(40) The mother made her daughter a dress for the pThe mother made a dress for her daughter’s prom.

APPENDIX 2

Critical Sentences from Experiment 3

(1) David gave/flasked John some juice.(2) Art dropped/ballooned Mina the paper.(3) Joan sent/belled the hikers a warning.(4) Jeb brought/minivanned Gwenda the sculpture.
(5) Patty played/guitarred the judges her song.

ers.ds

saine.

fforuen

n. T

oft

The

n b

h is

The

use

s in

Th

ce

Th

an

turn

The

he

ngs

rive

fan

e eo-l-

t-o

-

s).K.

tivee ofion

andon-

as

lities,

.tionics

g-ns:

Then-

F lexi-ure.

F ,

G al

G ins.),:

G l

G 4).n of

ic

G talm-

G cals,

G bol


(6) Rachel brought/chaired the scientist his mail.(7) Kate got/dumb waitered June some water.(8) Keith brought/glassed his mother-in-law the flow(9) Lindsay and Sam sent/flashlighted the farmhan

message.(10) Fred sent/moused Bob a message.(11) Brian presented/Appled Jen his marriage propo(12) Tom gave/oranged his grandmother her medic

APPENDIX 3

Probe Statements from Experiment 4

The top two probes in each set of three are the “adance-based” probes. The third probe is the “Most FreqAssociate” probe.

(1) The mouse can be heard. The mouse can be seemouse can eat cheese.

(2) The blanket is long and sturdy. The blanket is sThe blanket is warm.

(3) The book can be thrown. The book can slide.book can be read.

(4) The saxophone is tube-like. The saxophone caheard. The saxophone can play music.

(5) The crutch is sturdy. The crutch is long. The crutcused for an injury.

(6) The bottle can reflect light. The bottle can float.bottle can hold beer.

(7) The spatula can carry things. The spatula can beas a catapult. The spatula can be used for cooking.

(8) The binder can open. The binder can hold thingplace. The binder holds paper.

(9) The postcard is flat. The postcard can be mailed.postcard depicts vacations.

(10) The purse can carry things. The purse can conthings. The purse can hold money.

(11) The cup can float. The cup can contain things.cup is used for drinking.

(12) The branch has whippy ends. The branch is longfirm. The branch is from a tree.

(13) The tractor can hold things. The tractor can overearth. The tractor is used for farming.

(14) The chair can roll. The chair can hold things.chair is used for sitting.

(15) The balloon is buoyant. The balloon is hollow. Tballoon holds air.

(16) The envelope is firm. The envelope can hold thiThe envelope can hold letters.

(17) The Tonka can carry things. The Tonka can dover things. The Tonka is a truck.

(18) The fan can rotate. The fan can blow things. Thecools things.

REFERENCES

Andrews, S. (1992). Frequency and neighborhood sizfects on lexical access: Lexical similarity or orthgraphic redundancy.Journal of Experimental Psycho
ogy: Learning, Memory, and Cognition,18, 234–254.
a

l.

-t

he

.

e

d

e

al

e

d

.

f-

Baker, M. C. (1988).Incorporation: A theory of grammaical function changing.Chicago: Univ. of ChicagPress.

Barsalou, L. W. (1999). Perceptual symbol systems.Behavioral and Brain Sciences,22, 577–660.

Barsalou, L. W., Solomon, K. O., & Wu, L. L. (in presPerceptual simulation in cognitive tasks. In M.Hiraga, C. Sinha, & W. Wilcox (Eds.),Cultural, typo-logical, and psychological perspectives in cognilinguistics: The Proceedings of the 4th Conferencthe International Cognitive Linguistics Associat(Vol. 3). Amsterdam: John Benjamins.

Clark, H. H. (1996). Communities, commonalities,communication. In J. J. Gumperz & S. C. Levins(Eds.),Rethinking linguistic relativity.Cambridge, England: Cambridge Univ. Press.

Clark, E. V., & Clark, H. H. (1979). When nouns surfaceverbs.Language,55, 767–811.

Fauconnier, G. (1999). Mental spaces, language modaand conceptual integration. In M. Tomasello (Ed.),Thenew psychology of language.Mahwah, NJ: Erlbaum

Fillmore, C. J. (1988). The mechanisms of “construcgrammar.” Proceedings of the Berkeley LinguistSociety,14, 35–55.

Fillmore, C. J., Kay, P., & O’Connor, M. C. (1988). Reularity and idiomaticity in grammatical constructioThe case oflet alone. Language,64, 501–538.

Fisher, C. (1996). Structural limits on verb mapping:role of analogy in children’s interpretation of setences.Cognitive Psychology,31, 41–81.

isher, C. (1994). Structure and meaning in the verbcon: Input for a syntax-aided verb learning procedLanguage and Cognitive Processes,9, 473–518.

razier, L., & Clifton, C. (1996).Construal. CambridgeMA: MIT Press.

ibson, J. J. (1979).The ecological approach to visuperception.New York: Houghton Mifflin.

leitman, L. R., & Gillette, J. (1995). The role of syntaxverb learning. In P. Fletcher & B. MacWhinney (EdThe handbook of child language.Cambridge, MABasil Blackwell.

lenberg, A. M. (1997). What memory is for.Behavioraand Brain Sciences,20, 1–19.

lenberg, A. M., Kruley, P., & Langston, W. E. (199Analogical processes in comprehension: Simulatioa mental model. In M. A. Gernsbacher (Ed.),Hand-book of psycholinguistics.San Diego, CA: AcademPress.

lenberg, A. M., Meyer, M., & Lindem, K. (1987). Menmodels contribute to foregrounding during text coprehension.Journal of Memory and Language,26,69–83.

lenberg, A. M., & Robertson, D. A. (1999). Indexiunderstanding of instructions.Discourse Processe28, 1–26.

lenberg, A. M., & Robertson, D. A. (in press). Sym
grounding and meaning: A comparison of high-dimen-

n:

cs oney-

en

, &of

R.o

en

acon

uc--

edan

K rsedel.

L gs:.

L -,

L ory

L tory owl-

L An

M initiond.),

M . S.lu-

M . In-ni-

M n,nn

cho-,

M ns-

N erb

N 2).syn-d:

N er-lan-.

P e-

P i-

P uc-ive.

R ence

S ncects.

T cticM.

v -

Z els-

((


sional and embodied theories of meaning.Journal ofMemory and Language.

Goldberg, A. E. (1995).Constructions: A constructiogrammar approach to argument structure.ChicagoUniv. of Chicago Press.

Goldberg, A. E. (1999). The emergence of the semantiargument structure constructions. In B. MacWhin(Ed.), The emergence of language.Mahwah, NJ: Erlbaum.

Goldberg, A. E. (1992). The inherent semantics of argumstructure: The case of the English ditransitive.Cogni-tive Linguistics,3, 37–74.

Gropen, J., Pinker, S., Hollander, M., Goldberg, R.Wilson, R. (1989). The learnability and acquisitionthe dative alternation in English.Language,65, 203–257.

Gropen, J., Pinker, S., Hollander, M., & Goldberg,(1991). Affectedness and direct objects: The rolelexical semantics in the acquisition of verb argumstructure.Cognition,41, 153–195.

Johnson-Laird, P. N. (1983).Mental models: Towardscognitive science of language, inference, andsciousness.Cambridge, MA: Harvard Univ. Press.

Kay, P., & Fillmore, C. J. (1999). Grammatical constrtions and linguistic generalizations: TheWhat’s X doing Y?construction.Language,75, 1–33.

Kelly, M. H. (1998). Rule- and idiosyncratically derivdenominal verbs: Effects on language productioncomprehension.Memory & Cognition,26, 369–381.

intsch, W. (1988). The role of knowledge in discoucomprehension: A construction-integration moPsychological Review,95, 163–182.

akoff, G. (1987). Women, fire, and dangerous thinWhat categories reveal about the mind.Chicago: Univof Chicago Press.

andau, B., & Gleitman, L. R. (1985).Language and experience: Evidence from the blind child.CambridgeMA: Harvard Univ. Press.

andauer, T. K. (1999). Latent Semantic Analysis: A theof the psychology of language and mind.DiscourseProcesses.

andauer, T. K., & Dumais, S. T. (1997). A solutionPlato’s Problem: The latent semantic analysis theoacquisition, induction, and representation of knoedge.Psychological Review,104,211–240.

andauer, T. K., Foltz, P. W., & Laham, D. (1998).introduction to latent semantic analysis.DiscourseProcesses,25, 259–284.

acDonald, M. C. (1999). Distributional informationlanguage comprehension, production, and acquisThree puzzles and a moral. In B. MacWhinney (EThe emergence of language.Mahwah, NJ: Erlbaum.

f

t

ft

-

d

f

:

acDonald, M. C., Pearlmutter, N. J., & Seidenberg, M(1994). Lexical nature of syntactic ambiguity resotion. Psychological Review,101,676–703.

acNeill, D. (1998). Speech and gesture integrationJ. M. Iverson and S. Goldin-Meadow (Eds.),The nature and functions of gesture in children’s commucations.San Francisco: Jossey–Bass.

arlsen-Wilson, W. D. (1990). Activation, competitioand frequency in lexical access. In G. T. M. Altma(Ed.),Cognitive models of speech processing: Psylinguistic and computational perspectives.CambridgeMA: MIT Press.

itchell, D. C. (1994). Sentence parsing. In M. A. Gerbacher (Ed.),Handbook of psycholinguistics.San Di-ego: Academic Press.

aigles, L. R. (1990). Children use syntax to learn vmeanings.Journal of Child Language,357–374.

aigles, L. R., Gleitman, L. R., & Gleitman, H. (199Children acquire word meaning components fromtactic evidence. In E. Dromi (Ed.),Language ancognition: A developmental perspective.Norwood, NJAblex.

aigles, L. R., & Terrazas, P. (1998). Motion verb genalization in English and Spanish: Influences ofguage and syntax.Psychological Science,9, 363–369

inker, S. (1984).Language learnability and language dvelopment.Cambridge, MA: MIT Press.

inker, S. (1989).Learnability and cognition: The acquistion of argument structure.Cambridge, MA: MITPress.

inker, S., Lebeaux, D. S., & Frost, L. A. (1987). Prodtivity and constraints in the acquisition of the passCognition,26, 195–267.

oth, W. (1999). Discourse and agency in school scilaboratories.Discourse Processes,28, 27–60.

winney, D. A. (1979). Lexical access during sentecomprehension: (Re)consideration of context effeJournal of Verbal Learning and Verbal Behavior,18,645–659.

omasello, M., & Brooks, P. J. (1999). Early syntadevelopment: A construction grammar approach. InBarrett (Ed.),The development of language.London:UCL Press.

an Dijk, T. A., & Kintsch, W. (1983).Strategies in discourse comprehension.New York: Academic Press.

waan, R. A., & Radvansky, G. A. (1998). Situation modin language comprehension and memory.Psychological Bulletin,123,162–185.

Received July 26, 1999)Revision received January 6, 2000)

Constructing Meaning: The Role of Affordances and ...martin/Csci6402/Papers/kaschak00.pdfthe objects...

Documents

Transcript of Constructing Meaning: The Role of Affordances and ...martin/Csci6402/Papers/kaschak00.pdfthe objects...