Improving theProductivity ofAmerica's SchoolsSyntheses of thousands of research studiesshow the power of nine factorsinfluencing learning.

Education may be our largest en-terprise in terms of the numbersof people involved. the value of

human time required, and the capitaland operating expenditures budgetedThe value of education invested in theAmerican labor force, for example. isnow $815 billion compared to S65 bil-lion in 1900 (Walberg, 1983).

In the last few decades, morcover.spending on schools and colleges accel-erated: it rose from $11 billion to $200billion per year; from 3.4 to 6.8 percentof the gross national product. Duringthe past half century, the inflation-adjusted annual cost of public-schooleducation rose about five-fold from$490 to $2,500 per student (Walberg.1984).

Education: A Declining Industry?Even though costs have risen, the Na-tional Commission on Excellence inEducation (1983) and other groups re-port that students appear to be learningless. For example, comparisons made adecade or two ago showed that Ameri-can students did relatively poorly. Al-though comparing achievements ofU.S. students with those from countrieswith more homogeneous populations,national ministries of education, and

centralized control can be misleading.the differences are striking enough tocompel attention to our assumptionsand practices.

Recent studies provide a grim pictureof U.S. achievement even in the ele-mentanr grades. Stevenson (1983) foundthat in mathematics, U.S. students fellfarther behind the Japanese and Tai-wanese at each grade level; and. byv thgrade. the worst Asian classes in hislarge sample exceeded the best Anmei-can class. Mv research and obsenrvationsin elementanr science classes in Japancorroborate his findings. Recentachievement comparisons in highschool mathematics also showed thatAmerican high school students score onaverage at the first or second percentileof Japanese norms (Walberg, 1983:Walberg, Hamisch. and Tsai, 1984).

Herbert 1. Walberg is Research Professorof Education, College of Education.University of Illinois, Chicago.

Author's note: I thank Benjamin S.Bloom and Ralph W. Tyler for comn-ments on a previous version of this paper;the opinions and remaining shortcom-ings, however, are attributable to me.

19

_ ~~~~~~IAl

IIiIJa

E

I1E

I

Ja

E

I

Ja

E

I

Ja

E

I., A

MAY 1984

Thus, by measurable standards, U S.educational productivity has not kept upeven with that of U.S. smokestack in-dustries such as steel, automobiles, andconsumer electronics-which them-selves are declining as world-class com-petitors in quality and costs. Of course,neither the costs of educational inputs,including human effort, nor the value ofoutputs relevant to immediate and long-term goals are well measured. For thatreason it is difficult to arrive at definitiveconclusions about the causal relations ofeducational investments, services, andvalues beyond the narrow areas indicat-ed by objective achievement tests andreports of attitudes and behavior. Never-theless, since 1975 my colleagues and Ihave tried to develop a comprehensiveframework for the analysis of productivi-ty and test it out in a variety of classroomstudies in the U.S. and other countries.

Research ApproachFollowing the lead of early agriculturalexperimentation, much educational re-search focuses on the relation of singlecauses and effects. Education, however,obviously involves many means andends, each with an explicit or implicitcost or value. The promotion of effi-ciency requires the specification andmeasurement of the' chief causes,means, or "factors" of production.

Experiments and statistical studies ofproductivity data together with cost andvalue estimates have enabled a wide

variety of industries to increase the valueof their output while simultaneouslyreducing costs, thereby raising humanwelfare. Although such thinking mayseem alien to some educators, the pub-lic ranks research on educational pro-ductivity higher in priority than scien-tific investigation in most other naturaland social sciences (Gallup, 1983; Wal-berg, 1983); and we educators may dowell to think more explicitly and unsen-timentally about our business and to tryto found it on the emerging consensusof scientific evidence.

It should also be said, however, thatwe are far from being able to estimateexplicit costs and values. The priorproblem, now being solved, is estimat-ing the magnitude of effects of educa-tional inputs on outputs, which primari-ly involves causal rather than valuequestions.

A Theory of Educational ProductivityNine factors require optimization to in-crease affective, behavioral, and cogni-tive learning (see Figure 1). Potent,consistent, and widely generalizable,these nine factors fall into three groups:

Student aptitude includes:1) Ability or prior achievement, as

measured by the usual standardizedtests,

2) Development, as indexed by chro-nological age or stage of maturation,and

"Since 1975 my colleaguesand I have tried to develop acomprehensive framework forthe analysis of productivityand test it out in a variety ofclassroom studies."20

3) Motivation, or self-concept, as in-dicated by personality tests or the stu-dent's willingness to persevere intensive-ly on learning tasks.

Instruction includes:4) The amount of time students en-

gage in learning and5) The quality of the instructional

experience, including psychological andcurricular aspects.

Four environmental factors also con-sistently affect learning: the educational-ly stimulating, psychological climates of

6) The home,7) The classroom social group,8) The peer group outside the school,

and9) Use of out-of-school time (specif-

cally, the amount of leisure-time televi-sion viewing).

The first five aspects of student apti-tude and instruction are prominent inthe educational models of BenjaminBloom, Jerome Bruner, John Carrol,Robert Glaser, and others (see Walberg,1984, for a comparative analysis); eachappears necessary for learning in school;without at least a small amount of each,the student can learn little. Largeamounts of instruction and high degreesof ability, for example, may count forlittle if students are unmotivated or in-struction is unsuitable.

These five essential factors, however,are only partly alterable by educatorssince, for example, the curriculum interms of lengths of time devoted tovarious subjects and activities is partlydetermined by diverse economic, politi-cal, and social forces. Ability and moti-vation, moreover, are influenced byparents, by prior learning, and by thestudents themselves. Thus educators areunlikely to raise achievement substan-tially by their own efforts alone.

Three of the remaining factors-thepsychological climate of the classroomgroup; enduring affection and academicstimulation from adults at home; and anout-of-school peer group with learninginterests, goals, and activities--influ-ence learning in two ways: studentslearn from them directly, and thesefactors indirectly benefit learning byraising student ability, motivation, andresponsiveness to instruction. In addi-tion, about ten (not the more typical 30)

EDUCATIONAL LEADERSHIP

weekly hours of television time seemsoptimal for learning; more hours thanthis displace homework and other edu-cationally and developmentally con-structive activities outside school.

As Figure I shows, the major causalinfluences flow from aptitudes, instruc-tion, and the psychological environ-ment to learning. In addition, however,these factors also influence one another,and are influenced in turn by howmuch students learn, since those whobegin well learn faster (Walberg,1984a).

Other factors influence learning inschool but are less directly linked toacademic learning. For example, classsize, financial expenditures per student,and private governance (independent orsectarian in contrast to public control ofschools) correlate only weakly withlearning, especially if the initial abilitiesof students are considered. Thus, im-provements in the more direct and morealterable factors hold the best hope forincreasing educational productivity(Walberg and Shanahan, 1983).

Applied ResearchUnlike other national studies of educa-tion that have relied on hearings andtestimony, our investigations of educa-tional productivity followed applied re-search in the natural sciences in severalrespects (Walberg, 1983a). The theoryof educational productivity (discussedabove) which guided the inquiry (Wal-berg, 1981) is sufficiently explicit to test;and, using large bodies of national andinternational data, a wide variety ofempirical studies of it were conducted.

We published about two dozen ofthese empirical studies in research jour-nals of the American Educational Re-search Association and the AmericanPsychological Association that requirereview by referees as in other scientificdisciplines. Only after extensive obser-vation and some modifications of thetheory (notably the addition of televisionand peer group to the list of majorfactors) were the implications drawn inprofessional and policy journals such asEducational Leadership and Daedalus.Like other explicit scientific theories,however, the theory of educational pro-ductivity should be considered open to

fre 1. Cmdl _mo m Sblmlum L-b__ik

,Aptid. instcuion, and the pywchdoll anvlonamn ae mqor di e , dofmin(shown im doiUM OX, Y. and Z. Theq hduen~ on e W _ashnat ta, b, ande , and Mb in irn dnluenad by s on Me aod n

that 1tads plce (shown brolen auws).

disproof in part or whole by empiricalcontradiction.

In our investigations, we tried to fol-low three scientific canons-parsimony,replication, and generalizability. Parsi-mony means that the theory convergeson the least number of factors thatpowerfully and consistently predict orexplain cognitive, affective, and beha-viorial learning.

In this regard, the theory is reduction-ist and psychological: it fundamentallyassumes that academic learning is anindividual affective, behavioral, andcognitive activity that mainly takes placein the social context of the classroomgroup as well as in the home and peergroups. This is not to deny the influenceof Washington, the statehouse, thecommunity, superintendent, and prin-cipal but to encourage examination oftheir effects on the nine factors directlyimpinging on individual students.

Thus, from our view, school anddistrict economic, political, and socio-logical characteristics are less relevant tolearning because their influences areless alterable, direct, and observable.They are not substitutes for the ninefactors, but more distant forces that cansupport or interfere with them.

More and less productive classes,moreover, may be expected in the sameschool; and it is somewhat misleading tocharacterize a whole school or district aseffective-just as it is less accurate tospeak of the optimal condition for plantgrowth as being the average annualrainfall in a state rather than the amountof moisture reaching the roots of a singleplant.

The educational productivity theoryitself is admittedly simplified becauselearning is clearly affected by school anddistrict characteristics as well as by manyeconomic, sociological, and political

MAY 1984 21

forces at the school, communit., state.and national levels. Yet these character-istics and forces-such as the sex, eth-nicity, and socioeconomic status of thestudent, the size and expenditure levelsof schools and districts, and their politi-cal and sociological organization-areless alterable in a democratic, pluralisticsociety; are less consistently and power-fully linked to learning; and appear tooperate mainly through the nine factorsin the determination of achievement.Thus, we offer our theory not as a threatto those concerned about these otherfactors but as a friendly, collegial in ita-tion to demonstrate their effects on thenine factors or directly on the outcomesof schooling.

The canon of replication. means thatthe findings in similarly designed studiesshould reproduce one another fairlyclosely. For example, reinforcement orreward of learning has been implement-ed in various forms such as candy,tokens, symbols, and social recognition;it can be and usually is operationallydefined in various studies. The questionis whether these forms are the same ordifferent in their effects

To answer this question, the variousimplementations or strategies groupedunder the same category may be morefinely categorized and empirically com-pared in their effects on learning to see iftheir magnitudes are the same or differ-ent. Simple rather than complicated,detailed classifications usually serve tosummarize the findings; and these rela-tively simple findings suggest education-al implications that are convenient andpractical to implement.

Ceneralizability means that studiesshould yield similar results in nationaland international samples of students ofdifferent characteristics such as sex andage, in different subjects such as civicsand science, and using different re-search methods such as surveys, casestudies, and experiments (Walberg,1983a). For example, the effects of mas-tery learning on different types of stu-dents and in different school subjectsand grade levels may be estimated todetermine the extent of their generality.

What has been empirically found inthousands of studies is that generally theresults are surprisingly robust. Echoing

the folk adage, *what's good for the gooseis good for the gander.

But there are exceptions to the resultsreported below. The more powerful fac-tors appear to benefit all students in allconditions; but some students appear tobenefit somewhat more than others un-der some conditions. In addition, somestudies report larger effects than theaverages given belo.w; others, of course,report smaller effccts than the average.The cited research should be consultedfor details.

Methods of ResearchSince our concern was productivity, Nwchoped that our own research wouldefficiently capitalize on previous inqui-ry; and, under the support of the Na-tional Institute of Education and theNational Science Foundation, our teamof investigators started by compiling re-views of the 1970s on the productivefactors in learning (Walberg, Schiller,and Haertel, 1979; Waxman and Wal-berg, 1982). Next, quantitative svnthc-ses of all available studies of productivefactors were conducted; syntheses ofnearly 3,000 investigations-summa-rized below-were compiled (see Wal-berg, 1984c, for a more detailed ac-count). Case studies of Japanese andAmerican classes were carried out tocompare educational productivity in thetwo countries (Schiller and Walberg,1982; Walberg, 1983).

The productive factors were furtherprobed for their significance in promot-ing learning in three large sets of statisti-cal data on elementary and high schoolstudents-the National Assessment ofEducational Progress, High School andBeyond, and the International Study ofEducational Achievement (Walberg,1984c; Walberg and Shanahan, 1983;and Walberg, Harnisch, and Tsai,1984). Finally, large-scale studies weremade of the most effective ways of assist-ing educators to bring about construc-tive changes in schools (Walberg andGenova, 1982).

ResultsCollectively the various studies suggestthat the three groups of previously-de-fined nine factors are powerful and con-sistent in influencing learning. Synthe-ses of about 3,000 studies suggest thatthese generalizable factors are the chief

influences on cognitive. affecti-c. andbehavioral learning (see Figures 2through 4). Many aspects of these fac-tors can be altered or influenced bseducators.

The first five essential factors appearto substitute, compensate, or trade-offfor one another at diminishing rates ofreturn. Immense quantities of time, forexample, mav be required for a moder-ate amount of learning if motivation.ability, or instructional quality is mini-mal. Thus, no single essential factoroverwhelms the others, all appear im-portant.

Although the other factors are con-sistent correlates of academic learning,they may directly supplement as well asindirectly influence the essential class-room factors. In either case, the power-ful influences of out-of-school factors.especially the home env ironment, mustbe considered.

For example, the 12 years of 180 six-hour davs in elemcntarD and secondaryschool add up to only about 13 percentof the waking, potentially-educativetime during the first 18 years of life. If alarge fraction of the student's wakingtime nominally under the control ofparents that is spent outside school wereto be spent in academically-stimulatingconditions in the home and peer group,then the total amount of the student'stotal learning time would be dramatical-ly raised beyond the 13 percent of thetime in conventional American schools.

For instance, the mere four or fivehours per week high school studentstypically devote to homework might besupplemented by some of the 28 hoursper week they spend viewing television(Walberg and Shanahan, 1983). Euro-peans and Japanese believe homeworkhelps learning; empirical results ofAmerican research summarized belowsupport their belief.

Specific EffectsFigures 2 through 4 show the numericalresults of syntheses of several thousandstudies of academic learning conductedduring the past half century Interestedreaders and those who wish technicaldetails may examine the findings andmethods reported in the compilations ofthese syntheses (cited in the references),which in turn, contain references to the

EDUCATIONAL LEADERSHIP22

original studies. (In several instances,separate estimates of correlations andeffects are available for science andmathematics because the National Sci-ence Foundation awarded grants forspecial synthesis projects on these twosubjects. The tables contain both effectsand correlations, and the correlationsassume a one-standard deviation rise inthe independent variable).

Student AptitudeFigure 2 shows that IQ is a strongcorrelate of general academic learningbut only a moderately strong correlate ofscience learning. A student's Piagetianstage of development correlates moder-ately with both general and sciencelearning. By comparison, motivationand self concept are weaker correlates.

Student aptitudes as a set may be lessalterable than instruction. Yet positivehome environments and good instruc-tion affect them (Figure 1); and, sincethey are powerful correlates of learning,they deserve inclusion in theories ofeducational productivity.

The Largest Instructional EffectsFigure 3 shows the effects of variousaspects and methods of instruction. Ofall the factors in the table, the psycho-logical components of mastery learningrank first and fourth in their effects oneducational outcomes: Skinnerian rein-forcement or reward for correct per-formance has the largest overall averageeffect-1. .17 standard deviations; in-structional cues, engagement, and cor-rective feedback have effects equal toapproximately one standard deviation.Separate syntheses of mastery programsin science show an average effect of .8.

Acceleration programs, ranked sec-ond in effect, provide advanced activi-ties to elementary and high school stu-dents with outstanding test scores.Students in these programs gain muchmore than comparable control groups.

Reading training, ranked third in in-structional impact, refers to programsthat coach learners in adjusting readingspeed and techniques to purposes suchas skimming, comprehension, and find-ing answers to questions. The usuallearning criterion in evaluating theseprograms is learner adaptability to pur-pose.

Other Large EffectsSeveral other instructional programsand methods have strong effects rangingfrom .3 to .8. These include coopera-tive-team learning in which some au-tonomy over the means and pace oflearning is delegated to students whohelp each other in small groups.

Personalized and adaptive instruc-tion, tutoring, and diagnostic-prescrip-tive methods also have strong effects.Personalized learning, sometimes called"the Keller Plan," is similar to masterylearning in mainly eliminating lecturesand recitations but guiding each studentby entry tests and written lessons plusindividual help. Adaptive instructionuses similar techniques plus work insmall groups and differentiated staffingto increase learning.

Tutoring and lesson prescriptionsbased on diagnosed individual needs aresimilar ways to adapt instruction tolearners rather than batch-processingthem. These related methods may attaintheir success by helping students to con-centrate on the specific goals they indi-vidually need to achieve, or by freeingthem from the pervasive seatwork andrecitation in groups that may suit onlythe middle third of the students.

Moderate and Small Effects ofInstructionAlthough many schools no longer usethe science and mathematics curriculacreated in the decade after Sputnik in

1957, several svntheses of their evalua-tions show that thev had moderate ef-fects on leaming.

High teacher expectations for studentperformance also have a moderate ef-fect. on average, as do advance organiz-ers, which are "cognitive maps" show-ing the relationship of material to belearned in a lesson to concepts learnedin previous lessons.

Some highly touted programs havehad small and even negative efects onaverage (shown in the lower part ofFigure 3). Reduced class size, for exa-pie, has small positive effects but isexpensive and draws money and effortaway from factors with large effects.

(Japanese school classes often runthree times the current U.S. average of17 students per class; yet thev consistent-ly rank highest among nations com-pared in mathematics and scienceachievement. With fixed or decliningbudgets for U.S. education, we mayface the trade-off of sharply increasingteachers' salaries and incentives-wvhichmay help the morale and productivity ofa smaller, elite worlfore--vems keep-ing class sizes far smaller than the aver-ages of the first seven decades of Ameri-can education in this century.)

The effect reported for computer-as-sisted instruction is deceptively small.Most of the research was conductedwith drill-and-practice or "page-turn-ing" programs rather than the morepsychologically sophisticated ones nowbeing developed. Because future pro-grams will be able to adapt to learnerinterests and abilities, they are likely toshow large effects (Walberg , 1983).(However, educators may have to wait adecade or two before such effects aredemonstrated. Accumulating closetfull of unused or usable computers to-day may deter valid and efficient use ofmuch better ones later.)

Quantity and IntructionInstructional time, as shown in the lastline of Figure 3, has an overall correla-tion of about .4 with learning outcomes.It is neither the chief detenninant nor aweak correlate of learning; like the otheressential factors, time appears to be anecessary ingredient but insicient byitself to produce learning.

MAY 1984

flpm5 2. = W=s A1111111itu la di-

Abs -- - d

IQ .71 r (00001IQ mu~ece AS ^xocX

Sd~se"411:110 .16 )DC

mass Vie X vpdmk lowss. ~

scI k-' 0mtU sudm h

a l m du

ra J-l Bala o 2100

Nose: lhetaX _latt cln aad X , _ Ace adze-Wn~lb~at . 0

23

For at least two reasons, time is aparticularly interesting factor: first, sev-eral national reports have called atten-tion to the need for lengthening theschool day and year to the levels of othercountries, particularly Japan and West-em Europe (National Commission,1983; Walberg, 1983).

Second, time is the only factor thatcan be roughly measured on a ratioscale with equal intervals between scalepoints and a true zero point. Perhapsbecause it can be measured on an abso-lute scale resembling capital and laborinputs to production processes in agri-culture and industry, time has showndiminishing returns (Frederick andWalberg, 1980): equal additions oftime, with other factors held fixed, yieldever smaller gains in learning, whichsuggests that neither time alone nor anyother factor by itself can solve the pro-ductivity problem.

It is also reasonable to think that zerotime results in zero learning no matterwhat the level of the other factors, and,to generalize, that each of the otheressential factors, if well measured,would prove necessary but insufficientby itself and would show diminishingreturns-thus the possible danger ofconcentrating on any one factor alone.

Since the other factors are not mea-sured as universally and precisely astime, this remains a matter of specula-tion. But it can be concluded that learn-ing is produced jointly by several factorsrather than any one by itself. A prelimi-nary estimate suggests that optimizingall the factors simultaneously is associat-ed with an effect of about 3.7, which isabout three times the 1.2 effect of themost powerful factor, reinforcement, byitself and nearly 15 times the effect ofsocioeconomic status (Horn and Wal-berg, 1984).

EnvironmentsFigure 4 shows the major results ofsyntheses of the supportive or supple-mentary factors. Ignored in several na-tional reports and in instructional theo-ries, these factors have strong influenceson learning. The psychological moraleor climate of the classroom group, forexample, strongly predicts end-of-course measures of affective, behavioral,

and cognitive learning. Morale refers tothe cohesiveness, satisfaction, goal di-rection, and related social-psychologicalproperties or climate of the classroomgroup perceived by students. By com-parison, the influence of the peer-groupoutside of school is moderate and com-parable to the influence of the student'ssocioeconomic status.(SES).

As also shown in Figure 4, homeworkthat is graded or commented upon hasthree times the effect of SES. By corn-

MeihodReinforcementAccenratonReading TrainingCues and FeedbackSdcience Mastery LearningCoTprative earningReadin _primentsPersonalized InstructionAdaptve InstructionTutoringndivrdualized Science

Higher-Order QuestionsDianotic Prescriptive MethodsIndivdualized InstructionIndividualiled MathematicsNew Science CurriculaTeacher SpectatlonsComputer Assisted InstructionSequenced LessonsAdvance OrganizersNew Mathematics CurriculaInquiry lobeyHomogeneous GroupsClass SizeProgrammed InstructionMainstreaming

Instructional Time

parison, homework that is merely as-signed has an effect comparable to SES.More than about 12 per weekly hours ofleisure-time television viewing, perhapsbecause it displaces more educationally-constructive home activities, has a weaknegative or deleterious influence onschool learning.

In addition to increasing supervisedhomework and reducing televisionviewing, school-parent programs to im-prove academic conditions in the home

Eled1.171.00

.97

.97

.81

.76

.60

.57

.45

.40

.35

.34

.33

.32

.32

.31

.28

.24

.24

.23

.18

.16

.10

.09-.03-. 12

.38

SizeXXVJ00MV=00

)0000000((XX)000)(X)CO(

XXXXXX

XXX0000(

XXXX)O(XXXOO

XXXX

XXXXXXXOXXXXXXXXXXX

-X.

XXX

XXX

XXXX

-X.

XXW

Note: The X symbols represent the sizes of effects in tenths of standard deviations.

EDUCATIONAL LEADERSHIP

Figure 3. Instructnal Quity and Time lects on Leaing

Figure 4. Home, Peer, Ca Moral, and Mefd Elecbt

Elect

Graded Homework .XXXxOxxCls Morale .60 XXXXXX)OOHome Intventions .50 XXXXXHome Evonment .37 X)OXAssid Hom k .28 XXXSocIeconomic Status .25 XXXeer Goup .24 XX

Televsion -.05 X.

Note: The X symbos reprent t sim of eects In tenths of stndKld deviations orcorrMllos.

24

have an outstanding record of success inpromoting achievement. What mightbe called "the alterable curriculum ofthe home" is twice as predictive ofacademic learning as is family SES.

This curriculum refers to informedparent-child conversations about schooland everyday events; encouragementand discussion of leisure reading; moni-toring and joint critical analysis of tele-vision viewing and peer activities; defer-ral of immediate gratifications toaccomplish long-term human-capitalgoals; expressions of affection and inter-est in the child's academic and otherprogress as a person; and perhaps,among such unremitting efforts, smiles,laughter, caprice, and serendipity.

Cooperative efforts by parents andeducators to modify these alterable aca-demic conditions in the home havestrong, beneficial effects on learning. In29 controlled studies of the past decade,91 percent of the comparisons favoredchildren in such programs over non-participant control groups. Althoughthe average effect was twice that of SES,some programs had effects ten times aslarge; and the programs appear to bene-fit older as well as younger students.

Since few of the programs lasted morethan a semester, the potential for thosesustained over the years of schooling isgreat. On the other hand, it should berecognized that educators cannot carryout these programs by themselves; theyrequire the concerted cooperation ofparents, students, and other agents inthe community.

Autonomous LearningIf education proceeds by fads rather thancumulative research, it will fail to makethe great advances in productivity thathave characterized agriculture and in-dustry in this century. Syntheses of re-search on the effects of open educationillustrate the dangers of basing conclu-sions, policies, and practices on singlestudies no matter how large or widelypublicized. They also illustrate thestrengths of replication and improvedmethods of synthesis, and a shortcom-ing of some of the research discussedabove that employs grades and standard-ized achievement as the only outcomeof educaton.

Open education has been dismissedby many educators, but syntheses ofresearch now illuminate its beneficialeffects. From the start, open educatorstried to encourage educational out-comes that reflect teacher, parent, stu-dent, and school board goals such ascooperation, critical thinking, self reli-ance, constructive attitudes, life-longlearning, and other objectives that tech-nially oriented psychometrists seldommeasure. Raven's (1981) summary ofsurveys in Western countries, includingEngland and the United States, showsthat, when given a choice, educators,parents, and students rank these goalsabove standardized test scores andschool marks.

Moreover, a synthesis of the relationbetween grades and adult success showstheir slight association (Samson andothers, 1984). Thirty-three post-1949studies of the college and professional-school grades of physicians, engineers,civil servants, teachers, and othergroups show an average correlation of1 155 of these educational outcomes with

life-success indicators such as income;self-rated happiness; work performanceand output indexes; and self-, peer-, andsupervisor-ratings of occupational effec-tiveness.

These results should challenge edu-cators and researchers to seek a balancebetween continuing autonomy, motiva-tion, responsibility, and skills to learnnew tasks as an individual or groupmember on one hand and memoriza-tion of teacher-chosen, textbook knowl-edge that may soon be obsolete or for-gotten on the other. Perhaps sinceSocrates, however, these views have re-mained so polarized that educators findit difficult to stand firmly on the highmiddle ground of balanced or coopera-tive teacher-student determination ofthe goals, means, and evaluation oflearning.

Progressive education, the Daltonand Winnetka plans, team teaching,and the ungraded school, and otherinnovations in this century-all heldforth this or a similar ideal but driftedinto authoritarianism, permissiveness,or confusion. They were difficult tosustain as idealized.

"The 'alterablecurriculum of thehome' is twice aspredictive ofacademic learningas is familysocioeconomicstatus."

Although open education, like itsprecursors, faded from view, it was moremassively researched bv dozens of inves-tigators whose work goes little noted.Perhaps the syntheses of this researchmay be useful to educators who want tobase practice on s nthesized knowledgerather than on fads, or to those who willevaluate future descendants of openeducation.

Hedges, Giaconia, and Gage (1981)synthesized 153 studies of open educa-ton, including 90 dissertations. Theaverage effect was near zero for achieve-ment, locus of control, self conceptand anxiety (which suggests no differ-ence between open and control classeson these criteria); about .2 for adjust-ment, attitude toward schools andteachers, curiosity, and general mentalability; and about a moderate .3 forcooperativeness, creativity, and inde-pendence. Thus, students in open class-es do no worse in standardized achieve-ment and slightly to moderately betteron several outcomes that educators, par-ents, and students hold to be of greatvalue.

Unfortunately, the negative conclu-sion of Bennett's (1976) single stud--introduced by a prominent psychologist,published by Hanrvard University Press,

MAY 1984 25

i

publicized by The New York Times andby experts that take the press as theirsource-trumpeted the failure of openeducation, even though the conclusionof the study was later retracted (Aitkin,Bennett, and Hesketh, 1981) because ofobvious statistical flaws in the originalanalysis (Aitkin, Anderson, and Hinde,1981).

Giaconia and Hedges (1982) took an-other recent and constructive step in thesynthesis of open education research.From their prior synthesis, they identi-fied the studies with the largest positiveand negative effects on several outcomesto differentiate more and less effectiveprogram features. They found that pro-grams that are more effective in produc-ing the nonachievement outcomes-attitude, creativity, and self concept-sacrificed academic achievement onstandardized measures.

These programs were characterizedby emphasis on the role of the child inlearning, use of diagnostic rather thannorm-referenced evaluation, individ-ualized instruction, and manipulativematerials but not three other compo-nents sometimes thought essential toopen programs--multi-age grouping,open space, and team teaching. Gia-conia and Hedges speculate that chil-dren in the most extreme open programsmay do somewhat less well on conven-tional achievement tests because theyhave little experience with them. At anyrate, it appears from the two most com-prehensive syntheses of effects that openclasses on average enhance several non-standard outcomes without detractingfrom academic achievement unless theyare radically extreme.

Caveats and ConclusionsResearch workers and educators shouldretain both openmindedness and skepti-cism about educational productivity andsyntheses of research. Yet the presentdoes seem a period of quiet accomplish-ment. In a short time, research synthesisand other comprehensive approacheshelped sort what is known from whatneeds to be known about some impor-tant means and ends of education.

Agriculture, engineering, and medi-cine made great strides in improving

human welfare as doubts arose abouttraditional, natural, and mystical prac-tices, as the widened measurement ofresults intensified, as experimental find-ings were synthesized, and as their theo-retical and practical implications werecoordinated and vigorously implement-ed and evaluated. Education is no lessupen to humanistic and scientific inqui-ry and no lower in priority since half theworkers in modern nations are in knowl-edge industries, and the value of invest-ments in people is now more apparentthan ever (Walberg, 1983). Althoughmore and better research is required,synthesis points the way toward im-provements that seem likely to increaseteaching effectiveness and educationalproductivity.

In addition, we educators can learnmore from our past successes and fail-ures in using scarce resources, especiallyhuman time, to meet competing goals.Recent national reports may rightly callfor more emphasis on academic subjectmatter, and the National Commissionon Excellence (1983) seems right inemphasizing the need for more time inschool. But students should also beemploying more time in academic pur-suits outside the school and using bothin-school and out-of-school time moreefficiently.

Synthesis of educational and psycho-logical research in ordinary schoolsshows that improving the amount andquality of instruction can result in vastlymore effective and efficient academiclearning. Educators can do even moreby also enlisting families as partners andengaging them directly and indirectly intheir efforts.

The present overview of a vastamount of research cannot substitute forselectively reading some of the severaldozen syntheses and thousands of stud-ies conducted in the past half century.Since many details are omitted here,reading the original material might pro-mote a more complete and critical un-derstanding of specific factors and meth-ods. For example, although the factorsthat have large effects are robustly posi-tive, exceptional conditions can reducetheir effectiveness.

Finally, educational costs and goalsbeyond immediate measurement areworth remembering. But great accom-

plishments also result from sustainedhard work, supportive parents, andworld-class standards and instruction.Psychological studies of the lives of emi-nent painters, writers, musicians, phi-losophers, scientists, and religious andpolitical leaders of past centuries as wellas prize-winning adolescents of todayreveal early, intense, and sustained con-centration as well as parents and teach-ers who sacrificed much to help them.

World-class performance may require70 hours of effective instruction andpractice per week for a decade (Walberg,1983). It may take considerably more-or perhaps less. The fact that we cannotsay shows how much more we need toknow about investing in students-andhow much more seriously educators andtheir allies might take the idea of im-proving their productivity.OI

References

Aitkin, M; Anderson, D., and Hindc. J."Modeling of Data on Teaching Styles (withDiscussion)." Journal of the Royal StatisticalSociety, Series A 144 (1981): 419-461

Aitkin, M.; Bennett, S. N . and Hesketh,J. "Teaching Styles and Pupil Progress: ARe-Analysis." British Joumal of EducationalPsychology 51 (1984), in press.

Bennett, S.N. Teaching Styles and PupilProgress. London: Open Books, 1976.

Gallup, G.H. "The 15th Annual GallupPoll of the Public's Attitudes Toward theSchools." Phi Delta Kappan 65, 1 (1983):33-47.

Giaconia, R.M., and Hedges, L.V. Iden-tifying Features of Open Education. Stan-ford, Calif.: Stanford Center for EducationalResearch, 1982

Hedges, L.V; Giaconia, R.M.; andGage, N. L. Meta-Analysis of the Effects ofOpen and Traditional Instruction Stanford,Calif: Stanford University Program onTeaching Effectiveness, 1981.

Hom, A. and Walberg, H.J. "Achieve-ment and Attitude as Functions of Quantityand Quality of Instruction. " Journal of Edu-cational Research (1984), in press.

National Commission on Excellence inEducation. A Nation at Risk. Washington,D.C.: U.S. Government Printing Office,1983.

Raven, J. "The Most Important Problemin Education is to Come to Terms WithValues." Oxford Review of Education 7(1981): 253-272.

Samson, G.; Graue, M.E.; Weinstein,T; and Walberg, H.J. "Academic and Oc-

26 EDUCATIONAL LEADERSHIP

26 EDUCATIONAL LEADERSHIP

cupational Performance: A QuantitatiseSynthesis." American Educational Researchlournal (1984, in press)

Schiller. D P. and Walbcrg, H J. "Japan:The Learning Socich-." Educational Leader-ship 39 (March 19821 411-413.

Stevenson, H Comparisons ot lapanese.Taiwanese, and Amencan MathematicsAchievement. Stanford, Calif: Center forAdvanced Studs iml the Behaxioral Sciences.1983.

Walberg, 1- "A Psychological Thecor ofEducational Producti-ih." In Psychologyand Education Edited b% F H. Farley andN Gordon. Bcrkclce. Calif: MlcCutchan.1981

Walberg, 11J "Scientific litcrac! andEcononlic Productiviht in Intcrinationial Per-spcctic." DIaedalus 112 119831: 1-28.

Walberg, H.J. "Familics as Partnelrs inEducational Productivit." Phi Delta Kap-pan (19 84a), in press.

Walberg. H.J. "Quantification Reconsid-ered " In Review of Research in Education.Edited b! E. Gordon. Washington, D.C.:American Educational Research Associa-tion. 1984b.

Walberg, H.l. "Synthesis of Research onTeaching." In Handbook of Research onTeaching. Edited b! M.C. Wittroclk. ash-ington., DC.: American Educational Re-search Association. 1984c.

Walberg. H.J., and Genova. W .I. "Staff.School, and Workshop Influences onKno-lcedge Use im Educational Impro-e-ment Efforts." Journal of Educational Re-search ,2 (1982): 69-80.

gWalberg. H.J.; Hamisch. D.; and Tsai.S.-L "High School Productiits- in TsrhveCountries." Journal of Educational Research(1984). in press.

Walberg. H.I.; Schiller. D.P.; and Haer-tel, G. D. "The Quiet Resoluton in Educa-tional Research.' Phi Delta Kappan 61. g(1979): 179-182.

Walberg. H. .. and Shanahan. T. -HighSchool Effects on Indiv-idual Students.Educational Researcher 12. -7 1983t: 4-9.

Waxman. H.C.. and Walberg. H.J. TheRelation of Teaching and Learning: A Re-view of Reviev's." Contemporary EducationReview 2 (1982): 103-120.

MAY 1984

A Guide to Educational Trouble-Shooting

Walberg's work shows that no single factor by itselfdetermines learning, but that a few key factors in

combination do. Educators can use this knowledge toexamine their own school programs.

RALPH W. TYLER

ans of us who arc concerned statistical significance with substantive from a relatively few factors or s-anriableswith teaching and learning or social significance; hc seeks to explain a numerical approximation to the manyhavc had difficulty in making interactions among variables in com- particular quantities reported as data.

constructive use of reports of education- mon-sense terms; and he examines and Earlv macro studies in the fields ofal research. particularly those dealing reports both macro studies and micro agricultural economics illustrate the na-with large aggregates of data. Herbert studies. ture of macro studies in the social sci-Walbcrg has done a superior intcrpreta- A macro study in the social sciences ences. The investigators sought to devel-tion of one of the most massive collec- deals with large bodies of data aggregat- op an equation that would produce antions of data on school learning. He ed over a large number and variety of approximation to the actual reportedavoids the common weaknesses of many phenomena. It seeks to develop a math- agricultural production in the leadingstatistical reports; he recognizes the ematical equation that will produce Western nations. The original equationcomplexity of much human learning that was developed predicted the quanti-and does not trn to reduce it to a tv of a product produced by the nationsimplistic model; he discusses the mean- Ralph W. Tyler is Director Emeritus. from the amount of land devoted toing of the data as well as indicating the Center for Advanced Study of the Behav- producing the product, the number ofquantitative results; he does not confuse ioral Sciences, Palo Alto, California. persons employed in the production.

Copyright © 1984 by the Association for Supervision and Curriculum Development. All rights reserved.