Articles

Evidence for a Language Learning Strategy: Onthe Relative Ease of Acquisition of Prefixesand Suffixes

Stan A. Kuczaj IISouthern Methodist University

KUCZAJ, STAN A., II. Evidence for a Language Learning Strategy: On the Relative Ease of Ac-quisition of Prefixes and Suffixes. CHILD DEVELOPMENT, 1979, 50, 1-13. 4 experiments testedSlobin's hypothesis that children are predisposed to learn suffixes rather than prefixes. In general,the results support the hypothesis and demonstrate that this learning bias for suffixes may affectthe initial processing of information about affixes and/or the postinitial processing of such infor-mation. However, the results of experiment 1 demonstrate that previous experience may alterlearning biases to some extent. The findings obtained from the 4 experiments are discussed interms both of the importance of language learning strategies in the language acquisition processand of the influence of experience on the child's utilization of such strategies.

It is well recognized that the child leam-ing his mother tongue faces a formidable task.The language acquisition task hecomes a bitless awesome if one assumes that the childcomes to the task with some predispositionsabout language. Although the precise natureof these predispositions is far from clear (seeChomsky 1972; Derwing 1973), numerous the-orists have suggested that the language-learn-ing child is equipped with a set of strategieswhich somewhat ease the child's task (Bever1970; Brown 1973; Chomsky 1969; Clark1977; Slobin 1973), although a strategy mightoccasionally mislead a child, as in the child'sinterpretation of passive sentences as active(Bever 1970, Maratsos 1974a). Slobin (1973)has provided the most comprehensive list ofpossible strategies (which he terms operatingprinciples) the language-learning child mightemploy, as well as rationale for the same (seeMaratsos, in press, for a critical discussion ofSlobin's posited operating principles and uni-versals of language development). Of particu-lar concern here is the following suggesteduniversal of grammatical acquisition and itscorresponding operating principles. UniversalAj: For any given semantic notion, grammati-

cal realizations in the form of suffixes or post-positions will be acquired earlier than realiza-tions in the form of prefixes or prepositions.Operating principle A: pay attention to theends of words. Operating principle B: the pho-nological forms of words can be systematicallymodified.

In essence, this hypothesized universaland two hypothesized operating principles pre-dict that suffixes marking a particular seman-tic notion will be earlier acquisitions than pre-fixes marking the same semantic notion, thereasons being the child's hypothesized dispo-sition to pay attention to word endings andhis (tacit) knowledge that the phonologicalform of a word can be systematically modifiedin order to alter the meaning of the word (seeSlobin 1966). In support of the purported uni-versal and operating principles, Slobin (1973)reports that children simultaneously learningHungarian and Serbo-Croatian acquire theHungarian locative prior to the Serbo-Croatianlocative. Since the Hungarian locative is ex-pressed by noun suffixes and the Serbo-Cro-atian locative is not, this finding suggests thatthe child may be attending to word endingsand thus acquire forms to express a meaning

I am grateful to Becky Engelhorn, Cynthia Morris, and Mary Nunley for their assistance inpilot testing and collecting the data of the experiments reported in this paper. A version of thispaper was presented at the biennial meeting of the Southwestern Society for Research in HumanDevelopment held in Dallas, Texas, March 1978. Author's address: Department of Psychology,Southern Methodist University, Dallas, Texas 75275. J bJ

[Child Dmlopmenl, 1979, 50. 1-13. © 1979 by the Society for Research in Child Development Ine0009-3920/79/5001-0001$01.09)

2 Child Development

earlier if the forms are suffixes. However, theHungarian locative is linguistically less com-plex than the Serbo-Croatian locative, and soit is not clear what factor determines theearlier acquisition of the Hungarian locative—the fact that it is expressed via suffixes, thefact that it is not as complex a linguistic sys-tem as the Serbo-Croatian locative, or both?Nonetheless, as Slobin notes, the fact that loc-ative inflections appear before locative prepo-sitions in monolingual child speech in severallanguages (including Serbo-Croatian) suggeststhat word endings are perceptually salient, asdoes the finding that the accusative and dativeare earlier acquisitions in languages in whichthey are expressed as suffixes than, for exam-ple, in Cerman where they are expressed asprenominal articles.

Slobin also notes that children's spon-taneous imitations of words are consonant withthe notion that word endings are perceptuallysalient in that children often imitate only thelast part of a word (e.g., raff for giraffe inEnglish). However, the spontaneous imitationscomprise an unsystematic set of data, and itis not immediately clear whether children ex-hibit a strong tendency to imitate the last partof a word if any part of the word is imitated.

The experiments reported in the presentpaper were intended as empirical tests of twoof the main ideas underlying Slobin's positeduniversal and operating principles. The twohypotheses to be tested are: (1) suffixes havegreater perceptual saliency for young childrenthan do prefixes; (2) suffixes which expressa given meaning will be easier acquisitionsthan prefixes which express the same meaning.

Experiment 1

MethodSubjects,—Suhjects were 48 preschool chil-

dren ranging in age from 4-3 (years, months)to 6-9, grouped according to age, condition,and Peabody Picture Vocabulary Test score,as will be explained below.

Design.—A matched-subjects experimentaldesign was employed. Although a within-sub-jects design would have been preferable sinceit would have allowed a direct comparison ofthe relative ease of learning suffixes versusprefixes for individual children, pilot testing re-vealed that children found it very difficult tolearn novel suffix and prefix forms if they wereexposed to novel sufiBxes and novel prefixes.

(Pilot work also revealed that 3-year-old chil-dren found the task most difficult.) Thus, adecision was made to expose individual chil-dren to either a novel suffix or a novel prefixexpressing either past reference or future refer-ence. This yielded four conditions for each agegroup: prefix/past, prefix/future, suffix/past,and suffix/future. Since there were three agegroups (4-year-olds, 5-year-olds, and 6-year-olds), there was a total of 12 groups. Withineach age group, the groups were matched forsex (two males and two females in each group)and for scores on the Peabody Picture Vo-cabulary Test which had been given approxi-mately 1 week earlier. (The matching on thePeabody test was not perfect, since identicalscores were rare. However, an attempt wasmade to match the groups as closely as pos-sible on Peabody scores.)

Exposure and test sentences.—The non-sense syllable ip was used as the novel suffixor prefix for each group. Each group receivedthe following eight exposure sentences with ipserving as a verb suffix or prefix dependingupon the condition (for each instance of anaffixed word, the root was stressed). (1) Theboy jump on the bed. (2) The lady drive thecar. (3) The mother swing the baby. (4)The boy kick the car. (5) The dog fall down.(6) The man push the car. (7) The babykiss the dog. (8) The boy hit the girl.

Children in the past tense conditions re-ceived the following eight test sentences: (1)The horse ran fast. (2) The monkey climbedthe tree. (3) The girl brushed her hair. (4)The bird flew high. (5) The lion roared veryloud. (6) The plant grew tall. (7) The babycried three times. (8) The fish swam upstream.

Children in the future reference conditionsreceived the following test sentences. (1) Thehorse will run fast. (2) The monkey will climbthe tree. (3) The girl will brush her hair. (4)The bird will fly high. (5) The lion will roarvery loud. (6) The plant will grow tall. (7)The baby will cry three times. (8) The fishwill swim upstream.

ProcedureEach child was trained and tested indi-

vidually with the entire procedure being re-corded on a Sony portable tape recorder. Theexperimenter sat at a table with the child andallowed the child to choose one of two avail-able puppets to use in the experiment. Theexperimenter then took the remaining puppet

and explained to the child that the experi-menter's puppet would speak a little bit dif-ferently than everybody else and that the childshould listen very carefully to what the experi-menter's puppet said so that the child couldteach his or her puppet to talk like the experi-menter's puppet.

For the two past reference groups, theexperimenter acted out the action expressedin an exposure sentence (with toys), then hadhis puppet say the sentence with ip in eithera suffix or prefix position depending on thecondition, and then asked the child to havehis puppet repeat the exposure sentence. Ifnecessary, the exposure sentence was repeatedin order to allow the child to attempt to re-produce it. Then this procedure was repeatedwith the next exposure sentence, and so on,until all eight exposure sentences had beenpresented to the child. Immediately followingthis, the experimenter presented the test sen-tences to the child. For the past referencecondition, the experimenter told the child thathe was going to say something and that thechild should make his puppet say it the waythe experimenter's puppet would say it, inother words, to talk like the experimenter'spuppet did. The experimenter then said anEnglish past tense sentence (see above), andasked the child to have his puppet say thesentence the way the experimenter's puppetwould say it. This procedure was repeateduntil the child had responded to each of theeight test sentences.

The procedure for the children in thefuture reference condition differed only in thatduring the presentation of the exposure sen-tences, the experimenter's puppet said one ofthe sentences, the child attempted to have hispuppet repeat the sentence, and then the ex-perimenter acted out the action described inthe sentence. During testing, these childrenwere given future tense sentences (see above)to transform to sentences that the experimen-ter's puppet would understand.

ScoringThe children's imitations of the exposure

sentences were scored as correct if the childretained ip in the position (prefix/suffix) it hadoccupied in the exposure sentence. There werethree categories of incorrect imitations: (1)those which omitted ip (e.g., the lady ipdrivethe car imitated as the lady drive the car), (2)those which moved ip to a different positionthan that it had occupied in the exposure sen-

Stan A. Kuczaj n 3

tence (e.g., the lady ipdrive the car imitatedas the lady drive-ip the car), and (3) thosewhich kept ip in the prefix or suffix positionbut added a modal auxiliary or past tensemarker (e.g., the lady ipdrive the car imitatedas the lady will ipdrive the car or the ladyipdrived the car).

Responses to the test sentences werescored according to whether ip was used andaccording to the position (prefix or suffix) inwhich it appeared. A correct response was onein which the child used ip in the appropriateposition.

ResultsImitations.—The elicited imitation task was

designed as a means to insure that the childrenhad in fact attended to the prefix/suffix ip. Theexposure sentences were relatively short inorder to reduce the possibility that the lengthof the model sentences would prohibit thechildren from processing the novel prefix/suf-fix. Nonetheless, children were unable to imi-tate each model sentence correctly. Table 1shows that older children were more likely toimitate the model sentences correctly than weretheir younger counterparts, F(2,36) = 6.59,p < .01. There was virtually no difference be-tween children's ability to correctly imitatemodel sentences containing a prefix and thosecontaining a suffix. However, children weremore likely to imitate past tense model sen-tences correctly than they were to imitate futuretense model sentences correctly, F(l,36) = 4.65,p < .05. In spite of these differences, the re-sults suggest that children in all conditionswere able to attend to the novel prefix/suffix.That there was no difference between children'simitations of prefix sentences and suffix sen-tences does not reflect on Slobin's hypothesizeduniversal and operating principles since themodel sentences were short enough to allowverbatim reproduction in many cases.

Although children were much more likelyto imitate the model sentences correctly thanto imitate the model sentences incorrectly,F(l,36) = 21.55, p < .001, the few incorrectimitations do provide some information con-cerning the manner in which the childrenprocessed the information in the exposure sen-tences since the incorrect imitations did notoccur randomly. Omissions of ip from imita-tions were more likely for future tense sen-tences than for past tense sentences, F(l,36)= 9.37, p < .01. Moreover, although errors in-volving the shifting of ip from its position

4 Child Development

TABLE 1

SUMMARY OF RESPONSES DURING EXPERIMENT 1 GROUPEDACCORDING TO CONDITION AND CHRONOLOGICAL A G E

CONDITION

RESPONSE MEASURES AND AGE Prefix/Past

Mean number of correct responses in imitation task:4 6.755 8.06 8.0

Mean number of correct responses in test task:4 1.755 2.256 4.5

Mean number of omissions of novel prefix/suffix intest task:

4 5.05 4.06 2.25

Mean number of times children used prefix as suffix(or vice versa) in test task:

4 1.255 1.756 1.25

Suffix/Past Prefix/Future Suffix/Future

7.07.758.0

4.05.05.75

4.03.02.25

000

5.256.757.75

2.252.755.25

5.755.252.75

000

5.257.257.75

1.52.253.75

6.04.01.5

.51.752.75

in the exposure sentence to a new positionwere quite rare, no such errors occurred inthe suffix/past condition or the prefix/futurecondition. Errors of this type consisted ofmoving ip from a prefix position to a suffix po-sition in the prefix/past condition or movingip from a suffix position to a prefix positionin the suffix/future condition. The relation ofthese errors, the deletion errors, and the cor-rect imitations to the children's responses inthe test situation will be discussed later.

Test ResponsesThe children's responses to the test sen-

tences are summarized in table 1. Older chil-dren were more likely to respond correctlythan were younger children, F(2,36) = 14.69,p < .001. Children were also more likely torespond correctly to past tense test sentencesthan to future tense test sentences, F(l,36) =5.87, p < .01. Children were also better at re-sponding to suffix test sentences than to prefixtest sentences in accordance with Slobin's hy-pothesis, although this difference failed toachieve statistical significance, F(l,36) = 2.38,p > .1. Significantly, there was a strong inter-action between the prefix/suffix dimension andthe past/future dimension, F(l,36) = 15.96,p < .001. The implications of this interactionare important, as will be noted in the discus-sion.

Older children were less likely to omitip in their responses to the test sentences than

were the younger children, F(2,36) = 15.58,p < .001. Children were also more likely toomit ip if they had been exposed to it as aprefix than if they had been exposed to it asa suffix, supporting Slobin's hypothesis aboutthe relative perceptual salience of suffixes. Thisdifference failed to achieve statistical signif-icance, however, F(l,36) =2.39, p > 1. In-terestingly, age was not a significant factorinsofar as the use of ip in the wrong position(in a nonprefix position after exposure to ipas a prefix and in a nonsuffix position for ipas a suffix) was concerned, F(2,36) =1.88,p > .25. There was, however, a strong inter-action between the prefix/suffix factor and thepast/future factor, F(l,36) = 38.54, p < .001.

Relation between Imitation and Test ResponseThere was a significant relation between

correct responding on the imitation task andcorrect responding on the test task, r(47) =.43, p < .01, and also between deletion of ipin imitations and omission of ip in the test task,r(47) = .60, p < .01. This is not surprising,since a child who could not imitate ip in itsproper position in the elicited imitation taskcould hardly be expected to use ip correctlyin the test task condition. In fact, given thisprobable dependence relation, one might haveexpected the correlations to have been higherthan they were. Recall, however, that theelicited imitation task was constructed in orderto facilitate correct responding. Thus, children

were much better at the imitation task thanat the test task. The difference between correctresponding in the imitation task and correctresponding in the test task was substantial (X =7.13 for imitation task, X = 3.42 for test task).It is also interesting that no child respondedcorrectly in the test task as often as he didin the imitation task (p < .001, by sign test).It would appear, then, that the imitation taskserved its function as an indicator of whetheror not the children had even superficially ana-lyzed the novel nonsense suffix/prefix, whilethe test task assessed the degree to and themanner in which the children used this initialanalysis.

DiscussionThe results from experiment 1 provide

some support for Slobin's hypothesis that suf-fixes will be easier acquisitions than prefixes,since in general children were more likely torespond correctly in suffix test conditions thanin prefix test conditions. However, this wasan overall difference, and there was a signif-icant interaction between the prefix/suffix di-mension and the past/future dimension. Chil-dren were more likely to respond correctly inthe suffix/past test condition than in the pre-fix/past condition, t{22) = 10.61, p < .001,and were more likely to respond correctly inthe prefix/future condition than in the suffix/future condition, t(22) = 1.47, p > .1. Al-though the latter difference is not statisticallysignificant, it is noteworthy because it helps toillustrate the interaction observed between pre-fix/suffix and past/future conditions. This in-teraction is also borne out in the movement ofip to a different position than it had occupiedin the exposure sentences. Children in the suf-fix/past condition did not move ip to a prefixposition in the test situation, while childrenin the prefix/past condition did occasionallymove ip to a suffix position in the test situa-tion, this being a significant group difference,t{22) = 3.94, p < .001. Children in the pre-fix/future condition did not move ip to a suf-fix position in the test situation, whereas chil-dren in the suffix/future condition did some-times move ip to a prefix position in the testsituation, t{22) = 4.28, p < .001.

These results do not provide straightfor-ward support for Slobin's claim that suffixeswill be easier acquisitions than prefixes for a

Stan A. Kuczaj II 5

given semantic notion. The prefix ip was aneasier acquisition than the suffix ip in the futurecondition, while the suffix ip was an easieracquisition than the prefix ip in the past con-dition. However, these findings do not neces-sarily repudiate Slobin's hypothesis. Instead,they demonstrate the influence of past knowl-edge on the processing of new information.Children acquiring English as their first lan-guage have gained considerable knowledgeabout past tense and future tense forms priorto the age of 4-3 (Brown 1973; Clark 1970,1973; Cromer, Note 1) this being the age ofthe youngest child in experiment 1. Thus, thechildren had (tacit) knowledge that Englishexpresses the simple past tense via a suffix (theregular -ed form) and numerous irregularforms and the future tense via a prefix (usuallywill). The results refiect this tacit knowledgeand demonstrate that the processing of newlinguistic forms is infiuenced by past experi-ence in the sense that children appear to de-velop expectancies as to where certain semanticinformation is most likely to be syntacticallyencoded in their language. If a child haslearned to express a semantic notion with asuffix (the English past tense), he finds iteasier to learn a novel suffix than a novel pre-fix with which to express the semantic notion.If, however, he has learned to express a se-mantic notion with a prefix (the English futuretense),! he finds it easier to learn a novel pre-fix than a novel suffix with which to expressthe semantic notion. Moreover, if the child haslearned to express a semantic notion with aprefix (suffix), he will sometimes move a novelsuffix (prefix) to the previously learned prefix(suffix) position. Thus, children seem to learnsyntactic positions as well as linguistic formswith which to express a given semantic notion.This knowledge appears to infiuence both theinitial and postinitial processing of linguisticinformation (at least that concerning affixes).

At any rate, the results demonstrate thatSlobin's hypothesis concerning the relative dif-ficulty of learning prefixes and suffixes for agiven semantic notion does not hold for seman-tic notions for which a linguistic form (orforms) has been acquired. In such cases, theconsolidation of previously acquired informa-tion proves to be as important as whether thenovel form appears in either a suffix or prefix

1 It is not clear that English actually utilizes a future tense in the same sense that it utilizesa present tense and a past tense (see Jespersen 1969; Lyons 1969). Speakers of English do engagein future reference, and this behavior is what is referred to by the phrase "future tense' here.

6 Child Development

condition. However, Slobin's hypothesis neednot account for the acquisition of novel formswith which to express a semantic notion whichis already expressed by either a prefix or suffixform since the hypothesis was not supposedto refiect an unyielding bias but a relative ten-dency. Thus, a better test of the hypothesiswould involve the utilization of semantic no-tions which children have not yet learned toexpress primarily with either a prefix or a suf-fix form. This is what was attempted in thefollowing experiment.

Experiment 2

MethodSubjects,—Subjects were 48 preschool chil-

dren ranging in age from 3-2 to 5-10. As inexperiment 1, children were grouped accord-ing to age, condition, and Peabody Picture Vo-cabulary Test score. Unlike experiment 1, 3-year-old children were tested in experiment 2in the hope that at least some informationcould be gained about the younger child'sbiases in regard to prefix/suffix learning. Thus,the three age groups in experiment 2 were3-year-olds, 4-year-olds, and 5-year-olds. Thefour groups within each age group werematched for sex and for Peabody Picture Vo-cabulary Test scores (again, only approximatematching was possible for the Peabody scores.)

Conditions,-The domains of applicationin experiment 2 were size and color, more spe-cifically, big and red. Young preschool childrenappear to have at least some knowledge ofthese concepts (Heider 1971; Maratsos 1973,1974b), and so the concepts themselves shouldnot present undue difficulty to the children.More importantly, English does not consistent-ly mark size or color via prefixes or suffixes.Instead, linguistic forms marking size and/orcolor can be used in a prenominal position (thebig boy hit the girl; the red ball rolled downthe hill) or a variety of postnominal positions(the boy that is big hit the girl; the boy is big;the ball that rolled down the hill is red).Admittedly, when an adjective denoting sizeor color is used in a noun phrase, the adjec-tive precedes the noun. However, the childis unlikely to learn a bias to place color andsize adjectives before nouns since such adjec-tives can also appear in postnominal positions.Thus the size and color domains appear toavoid the possibility of previous experience es-tablishing a disposition in regard to markingthese concepts in sentences in either a prefix

or a suffix position. Nonetheless, both size andcolor were used in order to determine if chil-dren had learned to consistently use big or redin a prefix or suffix position, if so, this shouldexhibit itself in different developmental pat-terns (unless a specific prenominal or post-nominal position had been learned for bothterms). "The four conditions in experiment 2were big/prefix, big/suffix, red/prefix, andred/suffix. (Children in the red conditionswere tested prior to the start of the rest of theprocedure to insure that they could discrimi-nate red from other colors.)

Exposure Sentences and Test ConditionsThe nonsense syllable ip was used as the

novel suffix or prefix for each group. Childrenin the big conditions received eight exposuresentences, with ip immediately preceding orsucceeding the noun in either the first or sec-ond noun phrase, depending on the condition.For each condition, the children were exposedto ip in connection with the first noun phrasein four sentences and in connection with thesecond noun phrase in four sentences. Theexposure sentences for the big conditions wereas follows: (1) The boy drove the car. (2)The lady pushed the swing. (3) The dogjumped on the bed. (4) The man drove thecar. (5) The rock fell off the truck. (6) Thegiraffe kicked the hippo. (7) The elephantpushed the car. (8) The girl kissed the man.In the first four sentences, ip was placed im-mediately before or after the noun in the firstnoun phrase depending on the condition (suf-fix vs. prefix). In the last four sentences, ipwas placed immediately before or after thenoun in the second noun phrase depending onthe condition. The order of presentation of theeight exposure sentences was random.

The eight exposure sentences in the redcondition were as follows: (1) The boy drovethe car. (2) The lady pushed the swing. (3)The pig jumped on the bed. (4) The mandrove the car. (5) The rock fell off the truck.(6) The giraffe kicked the tree. (7) The ele-phant pushed the block. (8) The girl kissedthe man. As was the case for the big condition,the first four sentences contained ip in the firstnoun phrase while the last four sentences con-tained ip in the second noun phrase. The eightsentences were presented in a random order.

As in experiment 1, stress was placed onthe root of the affixed words.

Children in the big condition saw theexperimenter perform the following eight ac-

tions with toys for the test aspect of the pro-cedure: (1) a big lion kissing a frog; (2)a big man standing on his head; (3) a bigwheel rolling off a truck; (4) a big turtle get-ting in a truck; (5) a duck pushing a bigblock; (6) a boy moving a big rock; (7) a catchasing a big dog; and (8) a man kissinga big lady.

Children in the red condition saw theexperimenter perform the following actionswith toys for the test aspect of the procedure:(1) a red goat jumping a fence; (2) a redcar pushing a cow; (3) a red turtle gettingin a truck; (4) a red tree falling down; (5)a boy pushing a red car; (6) a man drivinga red motorcycle; (7) a dog chasing a red cat;and (8) a dinosaur jumping on a red balloon(uninflated).

In both the big and red conditions, theeight test actions were presented in a randomorder.

ProcedureEach child was trained and tested indi-

vidually with the entire procedure being re-corded on a Sony portable tape recorder. Asin experiment 1, the child was asked to choosea puppet to play with and then told that theexperimenter's puppet spoke a little differentlythan everyone else and that the child shouldlisten very carefully so that he could teachhis puppet to speak like the experimenter'spuppet.

For the two big conditions and the twored conditions, the experimenter acted out anaction corresponding to an exposure sentence,had his puppet say the exposure sentence withip in the appropriate position, and then askedthe child to have his puppet repeat the expo-sure sentence. This was repeated until the child

Stan A. Kuczaj II 7

had been exposed to each of the eight trainingsentences.

In order to facilitate the child's leamingthat ip corresponded to red or big, the toysused by the experimenter to act out the ex-posure sentences contrasted big with small andred with nonred. For example, in the big con-ditions, the experimenter showed the child abig {tall) lady and a small (short) lady, spe-cifically informed the child that one was bigand one was small, and then had the big ladypush the swing, followed by the experimenter'spuppet saying the ip-lady pushed the swing(or the lady-ip pushed ). The sole dif-ference between the two lady dolls was thatone was tall and one was short. For each ofthe other sentences acted out by the experi-menter, the two objects corresponding to thecrucial noun differed only in that one wasbigger than the other (for the big conditions)or that one was red and the other was not(for the red conditions).

For the test conditions, the experimenterperformed one of the test actions with toys andthen asked the child to have his puppet tellthe experimenter's puppet what happenedusing the experimenter's puppet's different lan-guage. As was the case for the exposure sen-tences, there were two objects (big/small, red/nonred) corresponding to one of the nounsused to describe the action the experimenterhad performed.

ResultsImitations.—As shown in table 2, older

children were more likely to correctly imitatethe model sentences than were their youngercounterparts, F(2,36) = 6.06, p < .01. Chil-dren were as likely to correctly imitate sen-tences containing ip as a prefix as they were

TABLE 2

SUMMARY OF RESPONSES DURING EXPERIMENT 2 GROUPEDACCORDING TO CONDITION AND CHRONOLOGICAL A G E

RESPONSE MEASURES AND AGE

Mean correct responses in imitation task:.J

45

Mean correct responses in test task:J45

Prefix/Big

6.257.758.0

1.52.253.25

CONDITION

Suffix/Big

6.757.257.75

2.253.255.0

Prefix/Red

6.06.757.5

2.02.754.0

Suffix/Red

6.56.757.5

3.03.755.5

8 Child Development

to correctly imitate sentences containing ip asa suffix. There was no statistically significantdifference in the frequency of correct imitationsin the big and red conditions.

Children apparently were able to processip and its sentential position (prefix vs. suffix),regardless of whether it referred to big or tored, at least insofar as the ahility to imitatean immediately preceding model is concerned.(Supplementary analysis revealed that whetherip occurred in the first or second noun phrasehad no effect on children's ahility to correctlyimitate ip in the model sentence.) Again, themodel sentences were deliberately constructedto facilitate correct imitation (i.e., they werekept as short as possible), and so these resultsdo not directly bear on the hypothesis of con-cern here, namely whether suffixes are easieracquisitions than prefixes.

TestOlder children were more likely to re-

spond correctly (i.e., put ip in the positionin which it had appeared in the exposure sen-tences) than were younger children, F(2,36)= 16.11, p < .001. Suffix ip proved to beeasier than prefix ip, F(l,36) =12.67, p<.001. (Whether ip appeared in the first orsecond noun phrase was not a significant fac-tor. ) The difference between the red conditionsand the big conditions was not significant,F(l,36) =3.17, p < . l . These data are sum-marized in table 2.

Younger children were more likely not touse ip in their responses to the test situationsthan were older children, F(2,36) = 14.28,p < .001. Children were more likely to omitip in the big conditions than in the red con-ditions, F(l,36) =5.11, p < . 0 5 . These dataare summarized in table 2.

One last point regarding the children'sresponses in the test situation. In experiment 1,children occasionally responded with ip in asuffix position in the past conditions regard-less of whether they had been exposed to itas a prefix or a suffix. Children occasionallyresponded with ip in a prefix position in thefuture conditions regardless of whether theyhad heard ip in a prefix or suffix position. Sucheffects were not observed in experiment 2.Children exposed to ip as a prefix did not useit as a suffix in the test situations, nor didchildren exposed to ip as a suffix use it as aprefix in the test situations. Although childrendid occasionally use ip in an inappropriate

sentential position, these inappropriate posi-tions were rarely readily identified as inappro-priate prefix positions for what should havebeen a suffix position, or vice versa. Instead,they appeared to be instances where the childrealized that he should use ip but did not knowexactly where to insert it in his sentence planand so just put it in anywhere. In other words,the consistent misplacement patterns observedin experiment 1 did not have any parallelsin experiment 2. Thus it would appear thatthe domains tested in experiment 2 (big, red)were not subject to previously learned place-ment biases as were the domains in experi-ment 1 (past, future), and so experiment 2provides a better (purer) test of Slobin's hy-pothesis than experiment 1.

Relations between Imitationand Test Responses

There were no dramatic relations betweenchildren's responses to the imitation task andtheir responses to the test task. The highestcorrelation was that between correct responseson the imitation task and correct responses onthe test task, r(47) = .38, p < .01. This lackof strong correlations suggests that the chil-dren's imitations refiect at best superficial anal-yses of ip, while the test responses reflect theextent to which these superficial analyses havebeen transformed, organized, and reorganizedin a more long-term memory store. Again, thisdiscrepancy between imitation responses andtest responses is not surprising. Model sen-tences used in the imitation task were delib-erately constructed to facilitate immediate cor-rect superficial analyses since such analysesare the minimum requirement for acquiringnovel linguistic forms. Thus, if the initial acqui-sition task had been too difficult, the childrenwould likely have failed to acquire any knowl-edge of ip at all (as was the case for childrenin various pilot studies where the acquisitiontasks proved to be too difficult).

DiscussionThe results of experiment 2 provide more

straightforward support for Slobin's hypoth-esis than did the results of experiment 1. Thetest results indicate that children do find iteasier to learn suffixes than prefixes insofar asa given semantic notion is concerned. In boththe big condition and the red condition, chil-dren were more likely to respond correctly inthe test situation when they had been exposedto ip as a suffix than when they had beenexposed to ip as a prefix. This is strong sup-

port for Slobin's hypothesis because althoughEnglish allows terms making reference to sizeand color to occur before or after the appro-priate noun, when the modifying term is to beused in the same noun phrase as the modifiednoun, the modifier occurs before the noun(e.g., the red barn, the big dog). In spite ofthis, the children in experiment 2 found iteasier to learn (i.e., respond correctly in thetest situation) ip as a suffix (postnominal po-sition) than as a prefix (prenominal position).Thus, the effect of previous experience at ex-pressing the target concepts was not evidentin experiment 2 although such an effect waspresent in the results of experiment 1.̂

The discrepancies between imitation re-sponses and test responses demonstrate thatthe relatively easier acquisition of suffixes thanprefixes is not restricted to initial processing.The children in experiments 1 and 2 were ableto superficially analyze the sentential positionof ip and correctly imitate the model sentencesregardless of whether the sentential positionwas that of a prefix or that of a suffix. In spiteof this initial equal success with ip as a prefixor suffix, the children's test responses in ex-periment 2 showed that sufiix ip was an easieracquisition than prefix ip. Thus, even wheninitial (superficial) processing does not pro-vide a distinction between prefix and suffixforms, the later processing of this informationseems to favor suffix forms and result in betterperformance in the test situation if the childwas exposed to ip as a suffix than as a prefix.Thus, the developmental implications of thesuffix/prefix distinction seem to be at twolevels—the initial processing of information(perceptual salience of suffixes as opposed toprefixes) and the later processing of the infor-mation processed initially. Suffixes may be at-tended to more than prefixes during initial pro-cessing, later processing, or both.

Experiments 1 and 2 both dealt with thechild's ability to gain productive knowledge

Stan A. Kuczaj II 9

of a new prefix or suffix form. The two experi-ments do not provide much information con-cerning the initial processing of such forms(recall that there was little difference insofaras the correct imitation of prefix ip and suffixip was concerned), but instead provide con-siderable information about differences in thelearning of suffixes and prefixes (for a givensemantic domain), implicating differential post-initial processing for prefixes and suffixes. Ex-periments 3 and 4 were meant to assess anyimmediate processing differences which mightexist between suffixes and prefixes.

Experiment 3

MethodSubjects,—Thirty children participated as

subjects in experiment 3. This sample was di-vided into three groups on the basis of chrono-logical age. Children in group 1 ranged in agefrom 3-2 to 3-7; children in group 2 rangedin age from 4-0 to 4-6; and children in group 3ranged in age from 5-0 to 5-7.

Task,—An elicited imitation task was used.In such a task, the child is asked to repeatwhatever the experimenter has just said. Pre-vious research using this method has demon-strated that if the model sentences are suffi-ciently long and/or complex to prohibit roteimitation, then the child's imitations may re-veal the extent to which he processed the infor-mation in the model sentence through his syn-tactic and semantic systems (Kuczaj 1975;Kuczaj & Maratsos 1975; Slobin & Welsh 1973;Smith 1970; Thieman 1975). Thus, if modelsentences containing a novel prefix or novelsuffix are sufficiently long, the child's imita-tions of these model sentences should revealthe relative ease with which he is able to ini-tially process prefixes and suffixes.

The present task consisted of 36 modelsentences, 18 of which contained a novel pre-fix and 18 a novel suffix. Five of the children

2 It is possible that the children in experiment 2 failed to apprehend the intended meaningof ip (big or red), and only learned to respond correctly by using ip in the correct sentential posi-tion without knowing wbat meaning ip was supposed to bave. Tbis is unlikely, given tbe facttbat tbere was no effect of noun phrase (first vs. second). If cbildren bad learned to respond onthe hasis of sentential position independent of meaning, they would bave heen extremely unlikelyto respond correctly since a correct response consisted of placing ip in tbe appropriate prefix orsufBx position in tbe appropriate noun pbrase. Altbougb some correct responses could occur hychance if tbe child bad learned to correctly use ip solely on the hasis of syntactic position (sincehe would have a 50% chance of choosing tbe rigbt noun pbrase) tbere should have been manyerrors in wbicb tbe child used ip in tbe wrong noun pbrase hut as an appropriate syntactic suffix(or prefix). Tbese errors were exceedingly rare, and so it would seem tbat the children's correctresponses were based on syntactic and semantic information.

10 Child Development

in each age group heard ip as a prefix and ufas a sufiSx while the other five children in eachage group heard ip as a suffix and uf as aprefix. Half of the model sentences containinga novel prefix had the prefix attached to thenoun in the first noun phrase (e.g., the ip-girlran up the steep stairs), while the other modelsentences containing a novel prefix had theprefix attached to the noun in the second nounphrase (e.g., the little frog jumped over theip-log). This was also true of the model sen-tences containing a novel suflBx. As in the pre-vious experiments, the root of the afiixed wordwas stressed rather than the affix. The orderof presentation of the model sentences wasrandom for each child.

ProcedureEach child was tested individually. Prior

to heginning the elicited imitation task, thechild was asked to choose a puppet to playa game with. He was then asked to have hispuppet say what the experimenter's puppetsaid and then given several short model sen-tences to repeat in order to insure that he hadcomprehended the task. Following this, thechild was given the elicited imitation task andattempted to imitate each model sentence im-mediately after it was presented.

ResultsThe results are summarized in tahle 3.

Older children were more likely to respondcorrectly (i.e., imitate the novel suflBx or pre-fix in the appropriate position) than wereyounger children, F(2,27) = 56.19, p < .001.Sentences containing a novel suffix were morelikely to he imitated correctly than were sen-tences containing a novel prefix, F(l,27) =5.19, p < .05, although this tendency heldonly for 3-year-olds and 4-year-olds.

Younger children were more likely to omitthe novel prefix or suffix from their imitations

TABLE 3MEAN NUMBER OF CORRECT AND INCORRECT RESPONSES

IN THE IMITATION TASK OF EXPERIMENT 3 GROUPED

ACCORDING TO CHRONOLOGICAL A G E , RESPONSE

TYPE, AND PREFDC VS. SUFFIX

Age

34 . . . .5

CorrectImitationof Prefix

2.27.1

14.0

CorrectImitationof Suffix

5.48.3

13.9

Omissionof Prefix

13.89.52.5

Omissionof Suffix

11.06.92.4

than were the older children, F(2,27) = 42.33,p < .001. Prefixes were more likely to be omit-ted from the children's imitations than weresuffixes, F(l,27) =11.23, p < .001, althoughthis was only true for 3-year-olds and 4-year-olds.

Experiment 4

This experiment was conducted in aneffort to ohtain some information ahout theinitial processing of prefixes and suffixes byyoung 2-year-old children. Pilot testing re-vealed that the model sentences used in ex-periment 3 were far too difficult for such chil-dren, and so an elicited imitation task was con-structed which utilized single words (i.e., thechild was asked to imitate isolated words).Half of the words the child was asked torepeat contained a novel prefix with a commonroot (e.g., ip-car). The remaining words hada novel suffix with a common root (e.g., ball-uf). To prevent possible hiases due to the pre-fix and suffix words qua sounds, half of thechildren tested were exposed to ip as a prefixand uf as a suffix, while the other childrenheard ip as a suffix and uf as a prefix.

SubjectsEight children (5 males, 3 females) par-

ticipated as subjects. The children ranged inage from 1-10 to 2-5, with a mean age of 2-2.

ProcedureEach child was tested individually in his

home. The child was given a puppet to playwith and was instructed to make his puppetsay what the experimenter's puppet said. Fol-lowing several practice trials with verb baseforms (e.g., eat), the child was presented withthe test words. There were 20 test words (10prefix and 10 suffix). Interspersed in this listwere six verb base forms (e.g., eat) in order toprovide some relief from the novel prefix andsuffix forms.

The entire procedure was recorded on aSony portable tape recorder.

ResultsThe children were more likely to imitate

suffix forms correctly (X = 7.63) than theywere to imitate prefix forms correctly (X =5.38), t{7) =2.85, p < .05. Moreover, pre-fixes were more likely to be moved to a suffixposition in imitations (X = 1-12) than suffixesto a prefix position (X = .12), t{7) =2.38,p < .05.

DiscussionThe results of experiments 3 and 4 provide

further support for Slobin's hypothesis thatsuffixes are easier acquisitions than prefixes ifeverything else is equal.^ Children in experi-ments 3 and 4 were better able to correctlyimitate suffixes than prefixes indicating thatinitial processing favors word endings ratherthan word beginnings. Thus, as Slobin hassuggested, children do seem prone to payattention to the ends of words. Interestingly,however, the disparity between prefix and suf-fix responses diminishes with increasing ageinsofar as initial processing is concerned. How-ever, this developmental tendency was not ob-served insofar as learning to productively usenovel prefix and suffix forms was concerned.In other words, suffixes remain easier acqui-sitions than prefixes even after the initial pro-cessing differences between prefixes and suf-fixes disappear. This suggests that the laterutilization of the information provided by ini-tial processing is quite important and that pre-vious experience is a major determinant oflater development. Both of these points areexpanded below.

Initial processing of information is obvi-ously quite important. If one does not pick upinformation from a present array then it willnot be available for later processing. RegardingSlobin's hypothesis, the better initial process-ing of suffixes than prefixes evidenced in ex-periments 3 and 4 (for children younger than5 years, at any rate) provides the postinitialprocessing mechanisms with more informationabout suffixes than about prefixes and thusallows for the easier acquisition of suffixes thanprefixes. Moreover, young children exhibited agreater tendency to repeat prefixes as suffixesthan vice versa, indicating that the "processingset" for suffixes may yield some inaccurate ini-tial processing of prefixes as suffixes and thusincrease the information bias (in terms ofamount available for postinitial processing).

The finding in experiment 3 that 5-year-old children do not find it easier to processinitial information about suffixes than to pro-cess initial information about prefixes is curious

Stan A. Kuczaj II 11

in that the 5-year-oId children in experiment 2found it easier to learn to productively usesuffixes than to productively use prefixes (thesechildren also did not evidence any differencein initial processing of prefixes and suffixes).Theoretically, the lack of any significant dif-ference in initial processing of suffixes and pre-fixes should provide the later processing mech-anisms with equivalent information about thesuffixes and prefixes. If the difference in learn-ing is not due to a difference in initial pro-cessing, then it must be due to a transforma-tion of the information which was initiallyprocessed. This suggests that the infiuence ofthe strategy to pay attention to ends of wordsexists even after the strategy has waned onthe initial processing level. Even children whohave come to pay attention to both word be-ginnings and endings at the level of immedi-ate perception and processing may pay moreattention to word endings at the level of post-initial processing. The result is the memorydecay of the information about prefixes due,most likely, to two factors: (1) a lack of re-hearsal of the information provided by initialprocessing, (2) some of the information aboutprefixes being transformed such that it nowbecomes information about suffixes, that is, thechild comes to use information about prefixesas if it was information about suffixes. Thislatter possibility gains some support from thefact that the children in experiment 4 occa-sionally used prefixes as suffixes in a task in-volving immediate processing. If this can oc-cur at the initial processing level, then it mayalso occur at later processing levels. In fact,children of all ages in experiment 1 demon-strated a tendency to transform initially pro-cessed information so that it would accordwith their earlier acquired knowledge aboutforms used to refer to the past and the future.In the same vein, children may also transforminformation about prefixes into informationabout suffixes at the postinitial processing levelbecause of the emphasis on suffixes rather thanprefixes in the early portion of the languageacquisition process. Not only could the strat-

^egy hypothesized by Slobin and supported bythe experiments in this paper lead to this ten-

3 It should be noted that the following discussion is based on a relative rather than an ab-solute difference. Children at all ages in all tasks exhibited an ability to deal with prefixes success-fully, but children were significantly more successful at dealing with suffixes than they were atdealing with prefixes. Thus, the results support Slobin's hypothesis that children learning theirfirst language may use a strategy of attending to word endings rather than word beginnings, butthe results do not demonstrate that children cannot attend to and learn prefixes. They can and do,hut they seem to be biased toward selectively attending to suiBxes.

12 Child Development

dency to use information about prefixes as ifit were information about suffixes; the cbild'searly acquisition of various noun and verbinflections could also contribute to this ten-dency. In the early periods of language acqui-sition, the child learns to use the progressive-ing, the plural -s, the regular past tense -ed,the possessive -s, and the third person singularpresent -s (Brown 1973; Cazden 1968; Kuczaj1977, 1978). All of these forms are suflBxes,and most are regular forms which are over-generalized to irregular forms. (It would beinteresting to compare the frequency and typesof overgeneralization errors which occur fora given semantic notion expressed by a suffixor system of suSixes in one language and bya prefix or system of prefixes in another lan-guage.) At any rate, the child acquiring En-glish as his mother tongue has encounteredmuch success by paying attention to suffixesand so even after he has begun to pay asmuch attention to prefixes at the level of ini-tial processing, he may continue to pay moreattention to suffixes during the later process-ing of the information obtained during initialprocessing and perhaps even treat informationabout prefixes as if it were information aboutsuffixes. Thus, the easier acquisition of suffixesthan prefixes may be the result of the better ini-tial processing of suffixes (occurring early in thechild's acquisition of his first language) or theresult of postinitial processing favoring suffixes(occurring later in the child's acquisition ofhis first language). This latter effect occursbecause of previous experience which biasespostinitial processing toward suffixes, at leastfor children learning English as their first lan-guage. Past experience, then, is doubly im-portant. It provides a basis for establishing an"attention set" in postinitial processing. If pay-ing attention to suffixes during initial process-ing has paid off then the child is likely tocontinue to attend to suffixes rather than pre-fixes in postinitial processing. Thus childrenlearning a first language which rewards thisstrategy will exhibit this postinitial processingbias for suffixes whereas children learning afirst language which does not reward this strat-egy will not. Past experience also providesa basis for deciding what type of form (prefixvs. suffix) is commonly used to express a givenfunction, as shown in experiment 1. The in-teraction between language processing strate-gies used by the child in learning his firstlanguage and the child's success with thesestrategies is in need of further investigation

since such research could provide informationconcerning both the degree to which childrenpersist in using inappropriate strategies (fora particular acquisition or even a particularlanguage) and the universality of particularstrategies.

Reference Note

1. Cromer, R. The development of temporal ref-erence during the acquisition of language. Un-published doctoral dissertation. Harvard Uni-versity, 1968.

References

Bever, T. The cognitive basis for linguistic struc-tures. In J. R. Hayes (Ed.), Cognition and thedevelopment of language. New York: Wiley,1970.

Brown, R. A first language: the early stages. Cam-bridge, Mass.: Harvard University Press, 1973.

Cazden, C. The acquisition of noun and verb inflec-tions. Child Development, 1968, 39, 433-438.

Chomsky, C. The acquisition of syntax in childrenfrom 5 to 10. Cambridge, Mass.: M.I.T. Press,1969.

Chomsky, N. Language and mind (2d ed.). NewYork: Harcourt, Brace & World, 1972.

Clark, E. How young children describe events intime. In C. B. Flores d'Arcais & W. J. M. Levelt(Eds.), Advances in psycholinguistics. Amster-dam: North-Holland, 1970.

Clark, E. How children describe time and order. InC. A. Ferguson & D. I. Slohin (Eds.), Studiesof child language development. New York:Holt, Rinehart & Winston, 1973.

Clark, E. Strategies and the mapping problem infirst language acquisition. In J. MacNamara(Ed.), Language learning and thought. NewYork: Academic Press, 1977.

Derwing, B. Transformational grammar as a theoryof language acquisition. New York: CambridgeUniversity Press, 1973.

Heider, E. "Focal" color areas and the developmentof color names. Developmental Psychology,1971, 4, 447-^55.

Jespersen, O. Essentials of English grammar. Uni-versity, Ala.: University of Alabama Press,1969.

Kuczaj, S. On the acquisition of a semantic system.Journal of Verbal Learning and Verbal Behav-ior, 1975,14, 340-358.

Kuczaj, S. The acquisition of regular and irregularpast tense forms. Journal of Verbal Learningand Verbal Behavior, 1977, 16, 589-600.

Kuczaj, S. Why do children fail to overgeneralize

the progressive inflection? Journal of ChildLanguage, 1978, 5, 167-171.

Kuczaj, S., & Maratsos, M. What children can saybefore they will. Merrill-Palmer Quarterly,1975, 21, 89-111.

Lyons, J. Introduction to theoretical linguistics.New York: Cambridge University Press, 1969.

Maratsos, M. Decrease in the understanding of theword "big" in preschool children. Child Devel-opment, 1973, 44, 747-752.

Maratsos, M. Children who get worse at understand-ing the passive: a replication of Bever. Journalof PsychoUnguistic Research, 1974, 3, 65-74.(a)

Maratsos, M. When is a high thing the big one?Developmental Psychology, 1974, 10, 367-375.(b)

Maratsos, M. How to get from words to sentences.In D. Aaronson & R. Rieber (Eds.), Perspec-tives in Psycholinguistics. Hillsdale, N.J.: Erl-baum, in press.

Stan A. Knczaj II 13

Slohin, D. The acquisition of Russian as a nativelanguage. In F. Smith & C. Miller (Eds.), Thegenesis of language: a psycholinguistic ap-proach. Cambridge, Mass.: M.I.T. Press, 1966.

Slobin, D. Cognitive pre-requisites for the acquisi-tion of grammar. In C. A. Ferguson & D. I. Slo-bin (Eds.), Studies of child language develop-ment. New York: Holt, Rinehart & Winston,1973.

Slohin, D., & Welsh, C. Elicited imitation as a re-search tool in developmental psycholinguistics.In C. A. Ferguson & D. I. Slobin (Eds.),Studies of child language development. NewYork: Holt, Rinehart & Winston, 1973.

Smith, C. An experimental approach to children'slinguistic competence. In J. R. Hayes (Ed.),Cognition and the development of language.New York: Wiley, 1970.

Thieman, T. Imitation and recall of optionally delet-able sentences by young children. Journal ofChild Language, 1975, 2, 261-270.