The relation between privatespeech and cognitive performance

MARK N. GASKILLUniversity of New Mexico

RAFAEL M. DÍAZStanford University

AbstractStudies of children's private speech have generally failed to find positive relations between

children's spontaneous use of private speech ad their task success or performance. This patternof negative findings has challenged and questioned the validity of Vygotsky's theoiy regardingthe self-regulatory functions of language. In the present study we hypothesize that positive pri-vate speech-performance correlations can be found when such relations are: a) stated in termsof specific speech categories and speafic performance variables, b) analyzed controlling for taskdifficulty, and c) assessed on task improvement over time. Thirty preschoolers, ages 3-5, werevideotaped while peiforming classification, block design, and puzzles with distractor tasks at

itwo points n time. Their spontaneous verbalizations claring the tasks were transcribed and co-ded into different categories of self-regulatory private speech. The results show that, when theaboye methodological constraints are applied, the spontaneous verbalization of labels and des-criptions strongly predicts improvement in a classification task over a tzvo-day period. The useof private speech, however, failed to predict other aspects of performance, such as time on task

or speed of transition between items.Keywords: Private Speech.

Relaciones entre habla privada y ejecucióncognitivaResumen

En la mayoría de estudios que se han realizado hasta el presente, no se ha podido compro-bar con certeza la propuesta Vygotskiana sobre el 'efecto que tiene el habla privada en la rea-lización eficaz de diferentes tareas. En este estudio proponemos que tres problemas metodoló-gicos han impedido el estudiar los efectos del habla privada en la actividad cognoscitiva de losniños. Primero, la relación entre habla y tarea se ha buscado en una forma demasiado ampliay global, sin tomar en cuenta funciones especificas del habla privada con respecto a diferentesaspectos más específicos de las tareas. En segundo lugar, los estudios anteriores no han tomadoen cuenta el efecto que tiene la dificultad de la tarea; tareas de mayor dificultad aumentan eluso del habla privada pero al mismo tiempo disminuyen la posibilidad de una ejecución eficaz.Finalmente, la mayoría de estudios no han examinado el efecto que tiene el habla privada enuna ejecución futura o posterior de la tarea; posiblemente, el mayor efecto del habla privadase puede descubrir en una mejora gradual en la realización de diferentes tareas. El presenteestudio demuestra que, cuando se toman en cuenta estos tres problemas metodológicos, se pue-de observar una relación positiva entre el uso del habla privada y la mejora en tareas de cla-

sificación realizada por niños entre tres y cinco años de edad.

Palabras clave: Habla privada.

Dirección del autor: School of Education, Stanford University, Standford, CA 94305.

O 1991 by Aprendizaje, ISSN: 0210-3702 Infancia y Aprendizaje, 1991, 53, 45-58

46INTRODUCTION

According to Soviet developmentaltheory (Luna, 1961, 1982; Vygotsky,1962, 1978) young children use languagenot only to communicate, but also to gui-de, plan and monitor their own activity.The use of language as a tool of thought,the theory states, frees children from thesalience of the inmediate context or con-crete visual field, allowing a more purpo-seful, planned and reflective course of ac-tion. Furthermore, Vygotsky (1978) pro-posed that the use of language from self-regulatory purposes transforms the cour-se of a child's cognitive development, gi-ving «birth to the purely human forms ofpractical and abstract intelligence» (p. 24).

In recent years, Western investigatorshave tested different aspects of Soviettheory in both laboratory and classroomsettings, confirming the fact that childrendo indeed use self-regulatory privatespeech in the context of goal-directed andproblem-solving activity (see e. g., Berk &Garvin, 1984; Frauenglass & Díaz, 1985).Studies have also found that private speechis consistently related to the child's cog-nitive activity in both its content and ti-ming, that is, the majority of children'sprivate speech is about the task and ismore likely to occur in moments of tran-sition and special difficulty- (Goodman,1981).

In a large number of studies, however,the spontaneous use of self-regulatory lan-guage is not clearly related to success inthe cognitive tasks and, in some cases, thefrequent use of private speech is negati-vely related to the outcome of a child'sperformance (Frauenglass & Díaz, 1985;Yaeger, 1968; Zivin, 1972). Needless tosay, these negative findings have temperedthe enthusiasm for Vygotsky's insights re-garding the significance and function ofprivate speech in a child's cognitive deve-lopment. Studies that have shown positi-ve effects of private speech on behaviorhave involved, for the most part, experi-menter-given and induced speech in theform of self-instructional strategies, ratherthan children's spontaneous production ofprivate speech (see e. g., Balamore &Wozniak, 1984). We believe that such evi-dence, although important, is not suffi-

cient to support Vygotsky's claims on therelation between children's spontaneousprivate speech and cognitive activity. Thepresent study explores this issue in theempirical literature, proposing that, if cer-tain methodological- problems are takeninto account, it is possible to find functio-nal relations between the spontaneous useof private speech and cognitive perfor-mance.

PR! VATE SPEECH ANDCOGNITIVE PERFORMANCE

Even though not all students of Soviettheory expect a positive relation betweenthe use of private speech and success incognitive tasks (see e. g., Frawley & Lan-tolf, 1986), it is clear that Vygotsky hypot-hesized a functional relation between thechild's speech and her ongoing activity(Díaz, 1986b). By functional relation wemean that, according to Vygotsky, thechild's utterances during a given task arenot merely an accompaniment or epiphe-nomenal to the task; rather, the utteran-ces are seen to affect the course and even-tual outcome of the child's activity. Infact, Vygotsky's most famous anecdotalexample of a child's private speech illus-trates how the child's words, «it's broken»(in response to a broken crayon), led thechild to draw a broken car (1962, p. 17).In addition, Vygotsky clearly statod thatthe use of private speech «can be useful inthe solution of the task» (1978, p. 26). Webelieve that, in order to supportVygotsky's propositions, investigatorsmust show that the spontaneous use ofprivate speech predicts differences in per-formance or different behavioral outco-mes during a given task.

In the majority of cases, as pointed outin Fuson's (1979) comprehensive review,researchers have not shown the expectedpositive relation between the use of priva-te speech and cognitive performance va-riables. Five studies reviewed by Fuson,where both spontaneous private speechand performance variables were measured,presented inconsistent, and at times puzz-ling, findings. Pechman (1978) found lar-ge (.62-.87) significant correlations bet-ween private speech emited in classroom

47settings and achievement test scores, whi-le Dickie (1973) and Beaudichon (1973)reported small and mostly nonsignificantpositive correlations between privatespeech utterances and task performance.Zivin (1972) and Yaeger (1968), on the ot-her hand, reported negative correlations.In Zivin's study, for example, silent chil-dren performed faster and more accuratein finger mazes than children who emit-ted any kind of private speech during thetasks. A most puzzling finding was repor-ted by Beaudichon (1973): Children ins-tructed to «think aloud» talked less andperformed better than noninstructed chil-dren!

Unfortunately, more recent studieshave continued to present an inconsistentset of findings regarding the effects of pri-vate speech on cognitive performance. Forexample, in Goodman's (1981) study, chil-dren's self-regulatory speech was positi-vely related to faster puzzle solution ti-mes. In contrast, Frauenglass & Díaz(1985) reported higher performance aco-res on both semantic and perceptual tasksfor children who talked less during thetasks. In short, it is clear that the presentpattern of findings does not allow us todistinguish between three possible alterna-tive hypotheses: a) the use of self-regula-tory language, as predicted by Vygotsky,facilitates performance, b) self-regulatoryspeech has an interfering effect on thechild's problem-solving activity, andc) self-verbalizations are epiphenomenalto the child's activity and do not affect theoutcome of a child's performance. A firststep in support of Vygotsky's theorywould be to show, with some consistency,a set of positive correlations between thespontaneous use of private speech and be-havior or performance variables. In thepresent stwiy, we hypothesize that suchpattern of findings in support ofVygotsky's theory can be obtamed if, asdiscussed below, certain methodologicalconstraints are put into effect.

METHODOLOGICAL ISSUES

A careful review of the private speech li-terature shows that the search for speech-performance correlations has been limitedon three accounts: a) the majority of stu-

dies have sought speech-performance re-lations at a very gross and global leve!,b) most studies have not controlled for theeffects of task difficulty in the analyses,and c) researchers have looked for inme-diate effects of private speech on task per-formance at one point in time, rather thanon task improvement over time. Let usdiscuss these problems with some detall.

Global versus specific relations. Syste-matic observations of the content of chil-dren's private speech suggest that suchlanguage could be used for multiple func-tions during task performance (see e.g.,Díaz, 1986a; Díaz & Padilla, 1985). In the-se studies, content refers to what the ut-terance is about; function refers to the ut-terance's possible effect on the child's on-going behavior or performance. Analysesof private speech content reveal that, du-ring a given task, children label and des-cribe materials, verbalize plans, goals andinstructions, emit transitional statementsat the end of discrete items or problem-solving steps, praise themselves for givenaccomplishments, and so forth. These dif-ferent kinds of statements suggest diffe-rent possible functions or effects on theongoing activity. For example, labelingcould facilitate selective attention to diffe-rent features or dimensions of a task, tran-sitional statements could lead to a more ef-ficient pace and faster transitions fromitem to item, questions and answers abouta task could facilitate sustained attentionand engagement on the task, to name afew. In this context, postulating global re-lations between total amount of privatespeech and overall performance might beseriously misleading. In the present study,therefore, we hypothesized that speech-performance relations can be found.whenthe relations are specified in terms of spe-cific types of utterances and specific beha-vioral outcomes, where the content of anutterance and its hypothesized functionare conceptually and meaningfully related.

The effects of task clifficulty. A consis-tent finding in the private speech literatu-re is that the frequency of private speechincreases as a function of task difficulty(see reviews by Berk, 1985 and Fuson,1979). In addition, and not surprisingly,task difficulty increases the likelihood oftask failure or decreased performance sco-

48res. As stated in Frauenglass and Díaz(1985), when these two facts are taken intoaccount, the expectation is that privatespeech should more often co-occur withfailure than with task sucess. In otherwords, speech-performance correlationsare affected by task difficulty as a third-variable confound. By increasing both theamount of private speech and the likeli-hood of failure, task difficulty might maskthe positive effects of such speech on taskperformance. In the present study, wehypothesized that speech-performancecorrelations could be found more reliablywhen a measure of task difficulty is par-tialed out from the analysis.

Improvement versus Performance. Inune with Vygotsky's theoretical state-ments, we should expect that the transfor-mation of preverbal problem-solving skillsinto verbal thinkinl with the use of priva-te speech is a relatively slow and gradualprocess. The effects of using privatespeech during a cognitive task, therefore,might not be evident at the time that suchspeech is being used. Examining the rea-non between private speech and perfor-mance within the time period of a giventask, may be analogous to measuring theperformance of a child learning to walk.His playmate, who still enjoys crawling,will beat him to the cookie jar every time.But we do not conclude that crawling isbetter than walking, since we know thatattempting to walk is positively related tofuture sucess. In the present study, there-fore, we hypothesized that the effects ofprivate speech can be found most reliablyon a child's task improvement over timerather than on individual differences inperformance at a given point in time.

THE PRESENT STUDY

In the present study we propose a sim-ple regression equation to assess the ef-fects of private speech on performance va-riables that takes into account the threemethodological issues discussed aboye.First, the equation examines relations bet-ween utterances of a specific content andspecific behavioral outcomes. Second,using hierarchical regression techniques,the proposed analyses partial out and con-trol for the effects of task difficulty. Fi-

nally, by measuring both private speechand performance variables at two pointsin time within a two-day period, the equa-tion assesses their relation across time,allowing for an analysis of prívate speechon task improvement over time.

More specifically, the following regres-sion equation was used to analyze the ef-fects of private speech on task improve-ment, controlling for task difficulty:

Performance Score = Performance Sco-re + Private Speech

(at Time 2) (at Time 1) (at Time 1)where «Performance Score at Time 1»

is the measure of task difficulty and is en-tered first in the hierarchical equation.Performance scores at Time 1 were cho-sen as the best available measure of taskdifficulty for two major reasons: a) thedifficulty of a task cannot be adequatelydefined or measured as an objective pro-perty, without taking into account thechild's ability to perform the task, and b)the way the children performed the tasksthe first time around was our best estima-te of how difficult the task was for eachchild, relative to the other children in thesample. An important feature of the abo-ye equation is that, once the variance dueto performance scores at Time 1 is takenout of the analyses, the leftover systema-tic variance in performance scores al Time2 can be attributed to children's improve-ment on the task over the two day period.In other words, we claim that by exami-ning the effects of private speech at Time1 on task performance at Time 2, whilepartialing out performance scores at Time1, we take care of two methodologicalconstraints at once: 1) We control for theeffects of task difficulty in the speech-per-formance relation, and 2) We examine theeffect of private speech on task improve-ment across time.

Using this paradigm, three specifichypotheses were tested:

1. Children's spontaneous labels andverbal descriptions of the task materialswill predict improvement in tasks requi-ring semantic classification of familiar ob-.j ects. We postulated that, through labelingand descriptions, children will focus onthe distinctive features and relations bet-ween the different objects to be classified.Labels and descriptions, therefore, should

49predict a measurable improvement in clas-sification skills.

2. Children's spontaneous use ofquestions and answers about the taskshould promote the development of sus-tained attention by helping the child focuson the task and inhibit possible sources ofdistraction. Our observations of privatespeech in the past (see e.g., Díaz, 1986)have shown that children use a question-answer dialogue form in their spontaneousverbalizations about the task: «Wheredoes the cow go?... Oh, right here!» Webelieve that a possible function of this dia-logue with the self is to keep the child ac-tively engaged and focused on the task.We hypothesized, therefore, that the useof questions about the task in privatespeech should predict improvement in on-task-behavior (or time-on-task, sustainedattention), especially in the presence ofdistractors.

3. Finally, we predicted that children'suse of transitional statements (such as«o.k., I'm done, let's see what's next»)would help children develop the ability togo from item to item efficiently and ra-pidly without adult intervention. Wehypothesized, therefore, that the use oftransitional statements predicts an impro-vement in inter-item latency time, that is,a decrease in the amount of time spent bet-ween the individual items of a given task.

In addition to the three hypotheses sta-ted aboye, the study observed changes inprivate speech over the two-day period.As suggested by Vygotsky's theory, it wasexpected that the amount of private speechwould decline with time, while the num-ber of whispers would increase, indicatingthat private speech becomes internalizedwith increasing competence on a giventask. Furthermore, since the theory statesthat private speech is used to meet the dif-ficulty of a given task, it was hypothesi-zed that children who improved the moston the tasks would show the greatest de-cline of private speech over the two-dayperiod.

METHOD

Subjects

Thirty preschoolers, ages 3-5 and amean mental' age of 4.5 (range 2.5-8.1)

years as measured by the PPVT, were se-lected from several preschool daycare fa-cilities in the Albuquerque area. The eth-nic composition of the sample reflectedNew Mexico's population: 22 Anglo, 1Black, 5 Hispanic and 2 Native Americanchildren; ahl children in the sample, howe-ver, were native speakers of English. Thesample included 17 girls and 13 boys.

Procedure

Children in the sample were administe-red-ihere different tasks designed to elicitprivate speech: A classification task, ablock design task and a puzzle task witha distractor. Each child was randomly as-signed to the six possible orders of taskpresentation,- so that each possible sequen-ce of task presentation was equally repre-sented. After brief instructions (descnbedbelow in the Measures section) and thecompletion of two sample items, childrenwere instructed to work on their own anddo as many items as they could for a pe-riod of five minutes, and also to stop wor-king at the sound of a timer bell. Follo-wing Frauenglass & Díaz's (1985) recom-mendation, children were told that theycould talk all they wanted during thetasks. These procedures were followed si-milarly for each of the tasks. Ah l childrenwere videotaped while working on th.-tasks, while the experimenter was sittinga few feet away with his back turnedagainst the child. Children worked indivi-dually in a space created by a partitionwithin the larger preschool room, wherebackground noise of children at work andat play could be heard. Two days after thefirst administration of the tasks (Time 1),children were administered the tasks a se-cond time (Time 2), following the sameassigned order and procedu res as inTime 1. A few days later, a separate ses-sion was arranged to administer the Pea-body Picture Vocabulary Test (PPVT) toeach child individually.

Measures

Classifications Skills. In order to mea-sure children's classification ability, chil-dren were asked to classify cards into mea-ningfully related pairs. The task included

508 envelopes, each containing a set of fivecards. Only four cards of each set couldbe classified into two meaninfully relatedpairs, the other card had to be set aside asnonrelevant. Ah the classifiable pairs inthe task could be classified according to arelationship of the general form: agent Xuses object Y. For example, a typical en-velope contained five cards portraying:1) a baseball player, 2) a mechanic, 3) apoliceman, 4) a wrench, and 5) a baseballbat. For each correctly classified pair, chil-dren received 1 point for a possible maxi-mum total score of 16.

Inter-Item Transition Time. In order tomeasure time spent between individualitems, children were given a block designtask containing 12 cards with pictured de-signs and a set of colored blocks to matchthe designs. Children had to work on theirown, copying with colored blocks the pic-tured designs, moving on to the next cardonce a given design was finished. Theamount of time spent between the itemswas measured with a stopwatch from thevideotapes. Specifically, inter-item transi-tion time was defined as the time betweenleaving a finished design aside and takingthe card with the next design.

Sustained Attention. Sustained attentionwas defined as the total amount of timegazing at a puzzles task, within the five-minute allotted period. Children were gi-ven a set of puzzles of age-appropiate le-vel of difficulty. Children were asked todo as many puzzles as they could andavoid looking at a distractor set of slidessituated on the near left of the child at a90 degree angle. They were told that theycould see all the sudes after they finishedthe puzzles. Such instructions were givenin order to make clear that the task wasto complete the puzzles, while the slideswere indeed distractors that they shouldtry to avoid attending to. Distractor sudesportrayed different attractive animal andzoo scenes that were projected conti-nuously with an audioviewer. With astopwatch, the amount of time looking atdistractor slides or away from the puzzleswas measured from the videotapes. Sustai-ned attention was computed as five minu-tes minus time spent looking away fromthe task.

Private Speech Categories. In order to

test the three specific hypotheses outlinedaboye, children's speech emitted duringthe task was transcribed from the video-tapes and categorized first as social or pri-vate speech. Social speech was defined asspeech inmediately preceded, accompai-ned or inmediately followed with a gazetowards the experimenter (see Frauenglass& Díaz, 1985, for further details on defi-ning and categorizing private speechunits). Al! other speech not obviously di-rected at the experimenter was classifiedas private. From the private speech proto-coís the number of items under the follo-wing categories were counted for eachchild: 1) Labels and descriptions of mate-rials (e.g., «The man is going to work»,«Mailman, letter»); 2) Transitional state-ments, (e.g., Another one», «O.K., I'mdone... what's next?»), and 3) Questionsabout the task, (e.g., «Where does this onego?» «Is this the nght one?»). In addition,for the sake of more global analyses, pri-vate speech was also classified as task-re-levant, task-irrelevant and whispers.

RESULTS

Speech and Performance Changes overTime

Table I shows means and standard de-viations for all speech and performancemeasures in the study at both points intime. As expected, children in the sampleshowed some improvement in all perfor-mance measures over the two day period.However, only the improvement in clas-sification scores was statiscically reliable[t (28) = 2.44, p<.02]. There was a decli-ne over time in the amount of privatespeech emitted, except for whispers whichshowed a slight increase. Only questionsabout the task, however, showed a statis-tically significant decline over the two dayperiod [t (28) = 2.58, p<.01].

Even though the overall decline in pri-vate speech and the increase in whisperswere slight and nonsignificant, possiblydue to the short period of time betweentesting sessions, the obtained pattern ofchanges over time suggests that privatespeech tends to diminish with increasingcompetence and mastery of a given task.

51TABLE I

Means (and standard deviations) for speech and performance variables at bothtimes of testing

Speech variables Time I Time 2

Labels and description 3.96 2.86 1.04 n.s.(6.39) (5.01)

Transitional statements .90 .45 1.56 n.s.(1.72) (.83)

Questions .38 .03 2.58 <.01(.78) (.18)

Task relevant speech 6.21 4.11 1.45 n.s.(8.81) (6.05)

Whispers 6.83 7.96 .91 n.s.(6.80) (7.27)

Task-irrelevant speech .86 1.76 1.43(2.69) (3.51) n.s.

Performance variables Time I Time 2

Classification 6.43 7.70 2.44 <.02(4.38) (4.99)

Inter-item transition 21.78 20.15 .94 n.s.Time (in seconds) (22.78) (19.93)Time on task 275.10 280.93 .76 n.s.(in seconds) (21.54) (39.18)

In order to test this possibility, the sam-ple was divided into high and low impro-vement groups with a median split, accor-ding to the amount of improvement ob-served in the different tasks. Improvementwas defined as the difference betweenTime 2 and Time 1 scores in the tasks.

First, the effects of classification impro-vement on the decline of private speechover time was tested in a 2 x 2 ANO VAwith factors IMPROVEMENT (high vs.low) and TIME (Ti vs. T2), where TIMEwas a repeated measures factor and theamount of private speech emitted the de-pendent variable. A significant effect ofclassification improvement was found[F (1,27) = 6.95, p<.01], indicatinl thatchildren who improved the most in theclassification task emitted more privatespeech overall. As hypothesized, therewas an improvement by time interaction[F (1,27) = 4.02, p<.05] indicating that,for children whose classification scoresimproved the most, their private speechdeclined more sharply over the two day

period. The same pattern of decline wasnot found, however, when improvementscore for time on task or transition timewere used as independent variables. Con-sidering that the hypothesized effect wasfound only for the classification task, andthat the interval between times of testingwas relatively brief, the decline of speechover time as a function of task improve-ment must be considered with caution andonly as suggestive for further research onthe changes of private speech over time.

Private Speech as Predictor of TaskPerformance and Task Improvement

Through the use of hierarchical multi-ple relression equations it was possible toinvestigate the effect of private speech onboth performance at one point in time andon task improvement over the two day pe-riod. As stited earlier, the following equa-tion was used to test the effects of privatespeech on improvement:

52TABLE II

Effects of private speech on task performance (within time) and task improvement(across time)

Label and descriptions on classificationA. (across time) á R2 = . 10 F = 11.36, p<.001B. (within time) A R 2 = 0,07 F = 6.88, p<.01

2. Transitional statements on inter-item transition time

1.

A. (across time) A R2 = .003B. (within time) á R2 = .001

3. Questions on time on taskA. (across time) A R2 = .006B. (within time) A R2 = .002

F<1, n.s.F<1, n.s.

F<1, n.s.F<1, n.s.

A) Performance = Performan-ce + Private Speech

(at Time 2) (at Time 1) (at Time 1)

In addition, by using Private Speech atTime 2 rather than Private Speech atTime 1 as the predictor, as in the folio-wing equation:

B) Performance = Performan-ce + Private Speech

(at Time 2) (at Time 1) (at Time 2)

the effects of speech on performance atone point in time (i. e., at Time 2) couldbe afso assessed. Note that in both equa-tions A and B Performance at Time 1 isentered first in the equation as a measureof task difficulty. The effects of privatespeech can then be determined by exami-ning the changes in R square produced byentering private speech as a predictor.

Table II shows the changes in R squarefor private speech as a predictor of taskimprovement (across time) and task per-formance (within time), testing the threehypothesized specific relations betweenprivate speech and performance. Neithertransitional statements nor questionsabout the task predicted any significantportions of the corresponding performan-ce variance. The failure to show significanteffects could be due in part to the fact thatneither transition time nor time on taskshowed significant improvement duringthe short duration of this study, and also

to the fact that the actual amount of pri-vate speech emitted under these categoriesof speech was considerably lower thanspeech emitted under the labels and des-criptions category.

On the other hand, labels and descrip-tions of the task predicted significant por-tions of both the performance variance onclassification within time (delta R squa-re = .07, F = 6.88, p<.01) and the impro-vement variance across time (delta R squa-re = .10, F = 11.36, p<.001). Note that,as hypothesized, the predictability of pri-vate speech on classification performanceincreases substantially when improve-ment, rather than performance at onepoint in time, is used as the criterion.

In order to further investigate the ef-fects of private speech on performance andtest the three main hypotheses of thestudy, children were divided into high andlow private speech groups with a mediansplit, according to the number of utteran-ces emitted at Time 1 under the three dif-ferent private speech categories. The ef-fects of private speech on performance va-riables was tested in a 2 x 2 ANO VA de-sin with factor PRIVATE SPEECH(high vs low) and Time (T1 vs T2), whereTIME was a within-subject factor andperformance measures the dependent va-riables. Three ANOVAS were done to as-ses separately the effects of private speechon classification scores, time on task, andtransition time. Table III shows the meansand standard deviations for the variablestested in these analyses.

High

Low

on task in seconds

High

Low

15.6 13.80(6.85) (6.80)25.73 23.17

(26.85) (18.90)

271.86 286.00(22.51) (10.86)275.54 278.95(22.01) (45.51)

Transitional statements

3. Dependent variable: time

Questions

53TABLE III

Means (and standard deviations) of pvformance variables and both points in timefor high and low speech groups

Time I Time 2

1. Dependent variable: classification scores

HighLabels and descriptions

Low

7.71(4.27)

5.67(4.18)

10.36(4.50)

5.73(4.18)

2. Dependent variable: inter-item transition time in seconds

As in the multiple regression analyses,no significant effects were found for thetransition time and time on task measures.However, a close inspection of the meansreported in Table III indicates severaltrends in the predicted direction. Forexample, on the average, inter-item tran-sition times were approximately 10 se-conds faster for children who emitted ahigher number of transitional statements.In addition, time on task increased ap-proximately 15 seconds for children whoemitted a higher number of questionsabout the task, while only about 3 secondsfor children. in the low questions group.

On the other hand, a very strong statis-tical relation between private speech andclassification skills was found. A signifi-cant TIME main effect was found[F (1,27) = 7.30, p<.01], indicating a sig-nificant increase in classification scoresover time. Also, a significant PrivateSpeech main effect was found[F (1,27) = 4.84, p<.05], indicating thatchildren who emitted more labels and des-criptions obtained overall higher classifi-cations scores. And, as hypothesized, asignificant and strong PRIVATESPEECH by TIME interaction was found

[F (1,27) = 7.05, p<.01 ], indicating thatchildren who emitted more labels and des-criptions during the classification task atTime 1 improved the most on the task.

Increasing the Predictability of PrivateSpeech

In the present study we hyothesizedthat by applying several methodologicalconstraints, it was possible to show signi-ficant effects of private speech on chil-dren's task improvement over time. Theconstraints included: a) hypothesizingspecific instead of global speech-perfor-mance relations, b) controlling of task dif-ficulty in the analyses, and c) assessing theeffects of private speech on inprovementover time. The results reported abo ye, ho-wever, do not clearly specify the indepen-dent effects of each methodological cons-traint on the speech-performance relation.The question remains as to which para-digm or set of constraints maximizes thepredictability of private speech on cogni-tive performance.

In order to answer this question, the ef-fects of applying one constraint at a timewas examined on the speech-performance

Pearson rwithin time

Pearson racross time

Control fordifficulty

within time'

Control fordifficulty

across timeb

GLOBAL Time 1: r = —.07, n.s. A R2 = .06(Total private r = .18, n.s. A R2 = .003

F = 5.13, p.<.05

speech as pre- Time 2: r = —.01, n.s. F<1dictor)

SPECIFIC Time 1: r = —.08, n.s. A R2 = .07 á R2 = .10(Labels & des- r = .26, n.s. F = 11.34, p<.001criptions as Time 2: r = —.08, n.s. F = 6.88, p<.01predictor)

54 TABLE IV

Increasing predictabdity of private speech on classification skills as a function onmethodological constraintsHypothesized Relation

' equation used: performance T2 = Performance TI + Private speech T2.b equation used: Performance T2 = Performance Ti + Private speech TI.

correlation, using the classification task asan example. Table IV reports the resultsof these analyses. As can be seen in the ta-ble, the Pearson correlations between pri-vate speech and cognitive performance atboth points in time are nonsignificant andin the negative direction, a finding that iscommon in the private speech literature.Interestingly, the correlation coefficientdid not change dramatically or achievedsignificance when the specific and mea-ningfully related category of privatespeech was used. Relating private speechand performance across time augmentedand changed the correlations to the posi-tive direction. However, both specific andglobal relations across time failed to reachstatistical significance with this manipula-tion.

When the measure of task difficulty iscontrolled for (using equation B, aboye)the relationship between speech and per-formance increases and achieves signifi-cance for the specific type of speech. Thisfinding suggests that the positive effects ofprivate speech on performance can be seenwithin time for specific relations, as longas the difficulty of the task is controlledfor in the analyses. However, as eviden-ced by the changes in R square and thecorresponding F values, the paradigm thatmaximizes the predictability of privatespeech on performance variable is one that

simultaneously applies the three cons-traints mentioned aboye. A change in Rsquare of .10 as a function of privatespeech is quite impressive, consideringthat the control variable (i.e., classificationat Time 1) predicted on its own 65 % ofthe variance in the equation. Since only35 % of the remaining variance in classi-fication scores at Time 2 could be attribu-ted to improvement, the aboye finding in-dicates that private speech accounted foralmost one-third of the improvement va-riance in the classification task. Such im-pressive contribution of private speech tochildren's classification improvement, ob-tained with the proposed regression equa-tion, is certainly not evident in the non-significant negative correlations that areusually reported in the current literature.

DISCUSSION

The present study investigated the rela-tion between spontaneous private speechand cognitive performance, proposing andapplying three important methodologicalconstraints. It was hypothesized that, inorder to assess private speech-performan-ce correlations, investigators must: a) sta-te relations between specific privatespeech categories and meaningfully-rela-ted specific behavioral outcomes, b) con-

55trol for task difficulty in the analyses, andc) assess the effects of private speech ontask improvement over time. When thesemethodological issues were taken into ac-count, the present investigation revealed astrong relauon between children's sponta-neous labeling and descriptions and theirsemantic classification slulls. Specifically,private speech was positively related toperformance at one point in time when therelation was stated in terms of specific ca-tegories of private speech and performan-ce, and when the task difficulty was con-trolled for in the analysis. However, therelation became even stronger when im-provement over time, rather than perfor-mance at one point in time, was the pre-dicted variable. Children's spontaneousverbalizations predicted close to one-thirdof the improvement variance in the classi-fication task when the three methodologi-cal constraints were applied simulta-neously.

Even though causality cannot be appro-priately inferred from correlational data,these findings provide some support toVygotsky's (1962, 1978) notion that pri-vate speech is not merely an accompani-ment to the child's activity, but rather af-fects and transforms the outcome of suchactivity. Further support for Vygotsky'stheory was found in the patterns of chan-ge observed over time: a) private speechdeclined, b) whispers increased, and c) thedecline in private speech over time wasmore pronounced for children who im-proved the most on the task. Even thoughfurther research is needed to reliably ve-rify these changes in private speech overtime, the finding, s suggest that overt pri-vate speech might serve important func-tions in the early development of strate-gies and skills, and then becomes subvo-cal or disappears with increasing mastery.

Not all the hypothesized relations bet-ween private speech and behavioral out-comes were supported by the data. Speci-fically, even though the observed trendswere in the right direction, no significanteffects of private speech were found on in-ter-item transition times nor on the mea-sure of sustained attention. Two alterna-tives should be considered when explai-ning these negative findings. First, the ne-gative findings might be due to the follo-

wing two facts: a) children actually em-mitted very few questions and transitionalstatements, and b) these two measures fai-led to show reliable improvement over thetwo-day period. A second possible expla-nation, however, is that private speechplays a major role only in tasks involvinghigher levels of semantic organization(Frauenglass & Díaz, 1985), while onlyminimally affects nonsemantic skills suchas speed and distractability. This secondpossibility merits further investigationconsidering the fact that training self-ver-balizations is frequently the treatment ofchoice for impulsive children (Kendall &Braswell, 1985).

It should be clear that to show a posi-tive relation between private speech andcognitive performance represents only afirst step in the understanding of the is-sues at hand. The question remains, forexample, as to how or through which me-chanisms the verbalizations during a giventask predict sucess and improvement onthe task. It is possible, as Vygotsky sug-gested, that pnvate speech signals the be-ginning of a new and more advanced typeof thinking, verbal thought, that facilita-tes accurate performance. On the otherhand, it is possible that verbalizations du-ring a given task represent children's ac-tual effort and mouvation to solve thetask. By the same token, improvement inclassification scores could index improve-ment on a number of performance varia-bles. For example, did private speech fa-cilitate recogniuon of the relations presen-ted in the classification items? Or, did pri-vate speech simply speed-up the sorting ofthe items according to relations already re-cognized without the aid of speech? Ob-viously, further research is needed to fullyvalidate the Vygotskian hypotheses onthought and language.

Most likely, we have oversimplified thetrue relation between language and cogni-tive performance. Our categories of priva-te speech were devised primarily for expe-rimental precision and not to describe ex-haustively be multifunctional aspects ofself-directed language. Moreover, a child'sprivate speech utterances are usually nes-ted within the social/parental context rat-her than being exclusively oriented towardindividual achievement or failure. In the

56present study we asked children to workon their own in order to reduce their at-tempts to engage the experimenter forhelp, and elicit the use of self-regulatorylanguage in the relative absence of exter-nal social structure. Even though theseprocedures probably do not reflect the na-tural context of children's private speechnor maximize the absolute frequency ofsuch speech (Goudena, 1983), the experi-mental manipulations did indeed elicitself-regulatory language and also facilita-ted a more precise classification of utte-rances into social and private speech cate-gories.

The overwhelming majority of privatespeech used during the tasks was catego-nzed under labels and descriptions. Thisfinding is not only consistent with theexisting private speech literature, but alsoprovides some support to Vygotsky'sclaims regarding the social origins of pri-vate speech. Maternal teaching strategiesare characterized by a high frequency ofquestions about the task (Laosa, 1980)and, in adult-child teaching situations,children's verbal participation consistsmostly of brief responses about the task,usually names, labels, and other descrip-

tions of the materials at hand. (Mother:«what goes here?» Child: «the cow...»Mother: «And, what color is this one?»Child: «a red one...»). It is interesting tonote that, when the child is asked to workalone, the child brings to the task not theexact question-and-answer teaching for-mat but, as Vygotsky would predict, amore simplified or predicated form of so-cial interaction. Also, the use of labels andthe verbal description of materials reflectthe child's connection to the conceptualworld of the adult culture. Labels and des-criptions are precisely the conceptual toolsand symbols that the child has receivedfrom his social-cultural context while dea-ling with and adapting to a changing en-vironment (Vygotsky, 1978; Werstch,1985). Labelin, therefore, is probably amore socially integrative form of speechthan other types of speech, such as tran-sitional statements or questions and ans-wers about the task. As such, the presentstudy offers some support to Vygotsky'sclaim that higher levels of conscious, goal-directed activity originate in social interac-tions and come to be mediated by the se-miotic functions of language.

ReferencesBALAMORE, U. & WOZNIAK, R. H. (1984). Speech-action coordination in young children. Developmental

Psychology, 20, 850-858.BEAUDICHON, J. (1973). Nature and instrumental function of private speech in problem solving situations.

Merrill-Palmer Quarterly, 19, 117-135.BERK, L. E. (1985). Why children talk to themselves. Young Children, July, 46-52.BERK, L. E. & GARVIN, R. A. (1984). Development of private speech among low-income Appalachian chil-

dren. Developmental Psycholop, 20, 271-286.DÍAZ, R. M. (1986a). The union o thought and language in children's private speech. The Quarterly News-

letter of the Laboratory of Comparative Human Cognition, 8, 90-97.DÍAZ, R. M. (19866). Issues in the empirical study of private speech: A response to Frawley & Lantolf's

commentary. Developmental Psychology, 22, 709-711.DÍAZ, R. M. & PADILLA, K. A. (1986). Teoría e investigaciones empíricas sobre el lenguaje privado. Anuario

de Psicología, 33-1985(2).DICKIE, J. R. (1973). Private speech: The effect of presence of others, task and intrapersonal variables. Dis-

sertation Abstracts International, 34 (3-B), 1292.FRAUENGLASS, M. H. & DÍAZ, R. M. (1985). Self-regulatory functions of children's private speech: A critical

analysis of recent challenges to Vygotsky's theory. Developmental Psychology, 21, 357-364.FRAWLEY, W. & LANTOLF, J. P. (1986). Private speech and self-regulation. A commentary on Frauenglass

and Díaz. Developmental Psychology, 5, 706-708.FUSON, K. C. (1979). The development of self-regulation aspects of speech: A review. In G. Zivin (ed.) The

development of self-regulation through private speech. New York: John Wiley & Sons, pp. 135-217.GOODMAN, S. H. (1981). The integration of verbal and motor behavior in preschool children. Child Deve-

lopment, 52, 280-289.GOUDENA, P. P. (1983). Private speech: An analysis of its social and self-regulatory functions. Unpublished

Doctoral Dissertation, Sociale Wetenschappen aan de Rijksunivessiteit te Utrecht, Holland.

57KENDALL, P. C. 8t BRASWELL, L. (1985). Cognitive-behavior therapy for impulsive children. New York: Guil-

ford Press.LAOSA, L. (1980). Maternal teaching strategies in Chicano and Anglo-American families: The influence of

culture and education on maternal behavior. Child Development, 51, 759-765.LURIA, A. R. (1961). The role of speech in the regulation of normal and abnormal behavior. Oxford: Perga-

mon Press.LURIA, A. R. (1982). Language and cognition. New York: John Wiley & Sons.PECHMAN, E. M. (1978). Spontaneous verbalization and motor accompaniment to children's task orientation

in elementary classrooms. Unpublished doctoral dissertation, University of Michigan.SUGARMAN, S. (1983). Children's early thought: Developments in classifcation. Cambridge: Cambridge Uni-

versity Press.VYGOTSKY, L. S. (1962). Thought and language. Cambridge, MA: MIT Press.VYGOTSKY, L. S. (1978). Mind in society. Cambridge, MA: Harvard University Press.WERTSCH, J. V. (ed.) (1985). Culture, Communication and Cognition: Vygotskian Perspectives. Cambridge:

Cambridge University Press.YAEGER, J. A. (1968). Self-directing speech and puzzle-solving in children. Unpublished doctoral disserta-

tion, University of Chicago.ZIVIN, G. L. (1972). Functions of private speech during problem-solving in preschool children. Dissertation

Abstracts International, 33, (4-B), 1834; Ann Arbor, Michigan.: University Microfilms. N." 72-26,224.

Resumen Extenso

Uno de los temas centrales en la teoría de Vygotski es que, a una temprana edad,los niños comienzan a usar el lenguaje no sólo para comunicarse sino también para pla-nificar, guiar y evaluar la propia actividad. Este uso del lenguaje, llamado «habla pri-vada», representa el origen de las funciones autorreguladoras que tipifican la inteligen-cia humana. En su mayoría, las investigaciones hechas en Europa y América han apo-yado y dado validez a las principales tesis Vygotskianas sobre el habla privada. Sin em-bargo, en el trabajo de investigación empírica sobre el tema existe un punto controver-tido de gran importancia, ya que muchas investigaciones no han podido documentaruna relación positiva entre el uso del habla privada y la ejecución eficaz de las tareas.

En el presente estudio se propone que la ausencia de una relación positiva entre ha-bla privada y ejecución de tareas se debe, principalmente, a varios problemas metodo-lógicos. Específicamente, los autores proponen que una relación positiva entre hablaprivada y realización de tareas se puede encontrar si se toman en cuenta las tres si-guientes sugerencias metodológicas:

Primero, la relación se debe buscar en términos específicos y no globales. Es decir,se deben buscar relaciones entre un tipo determinado de verbalización privada y un as-pecto específico de la actividad del niño. Los estudios deben buscar el documentar re-laciones, por ejemplo, entre el uso de descripciones de materiales y la atención selec-tiva, ya que, al describir diferentes aspectos de la tarea, el niño puede enfocar su aten-ción selectivamente. La observación de relaciones específicas, en vez de relaciones glo-bales, puede manifestar con mayor claridad las relaciones habla-actividad que realmen-te existen en la ejecución de diferentes tareas.

En segundo lugar, las investigaciones deben controlar los efectos de la dificultad dela tarea en el habla privada. Se sabe que la dificultad de una tarea influye en la cantidadde habla privada que un niño emite; tareas que son más difíciles evocan una mayor can-tidad de habla privada. También se sabe que, por supuesto, mientras más difícil es unatarea para un niño más pobre será la realización de la tarea para ese niño. Dado que ladificultad de una tarea simultáneamente aumenta el habla privada y disminuye la efi-cacia de la ejecución, no nos debe sorprender que las relaciones positivas entre hablay ejecución son muy difíciles, si no imposibles, de encontrar. El presente estudio pro-

58pone un método donde se puede controlar estadísticamente la dificultad de una tareaen el análisis de relaciones habla-actividad.

Finalmente, el estudio propone que los efectos del habla privada en la actividad delniño se deben buscar no en un momento dado, sino en la mejoría que se observa gra-dualmente en la ejecución de diferentes tareas. El habla privada emitida durante unatarea quizá no mejora tanto la ejecución de esa tarea específica en el momento presen-te, ya que el habla privada se emite al encontrar obstáculos y dificultades en la tarea.Sin embargo, el habla privada debe conducir a una mejor ejecución de la tarea la próxi-ma vez que se realiza, ya que a través del habla privada el niño desarrolla los recursosy estrategias necesarios para una ejecución más eficaz de esa tarea.

El presente estudio investiga la relación entre habla privada y ejecución de tareastomando en cuenta los tres puntos metodológicos mencionados. La investigación ana-liza tres relaciones específicas: 1) el uso de etiquetas (labels) y descripciones en relacióncon una tarea de clasificación que requiere atención selectiva, 2) el uso de preguntas yrespuestas en relación con el tiempo que el niño pasa actualmente trabajando en la ta-rea, ya que el uso de preguntas y respuestas llevan a una mayor motivación e implica-ción en la tarea, y 3) el uso de transiciones verbales en la eficacia de moverse de unítem a otro, ya que la transición eficaz entre ítems se puede ayudar verbalmente conpalabras como: «bueno ya hice ésta... a ver cuál otra debo hacer».

Los resultados del estudio indican que el habla privada tiene un impacto significa-tivo en la ejecución de tareas de clasificación. Sobre todo, el habla privada predice unamejor ejecución de la tarea, la segunda vez que se intenta.