Rwieut of Educatiqnal ResearchWinter 1976,VoL. tt6, No. t, Pp. 14lt

Components of a Psycholory of Instruction:Toward a Science of Design

Robert Glaser

Unio erity of P itts burg h

It is a well-known historicai fact that two major areas ofscientific psychology, psychometries and general experimentalpsycholory, came out of different traditions and have developedin different ways. Psychometrics has become a major teehnologi-cal appiication of psycholory, with primary effort being devotedto practical techniques and less effort to theoretical concerns. Ineontrast, the experimental psychology of learning and cognitionhas been almost exclusively a theoretical endeavor, with littleeffort devoted to application and the design of practicalteehniques for assisting in the conduct of human affairs. Al-though practical work has been carried out in educationalpsycholory, industrial psyehology, and human engineerin& nointegrated body of special technique of application has emerged.In recent years, however, there has been increasing interest inand social pressure for the development of professionaltechniques for the application of rvhat knowiedge there is of

This paper is a revised vercion of a seminar talk presented at EducationalTesting Senrice, Princeton, New Jersey, June 19?4. It also served as the basis fora State of lhe Scieace AddresE at the annual meeting of the Eastera Psyehologi-cal Association, New Yorh April 1975. lYork on this pap€r was carried oui at theLeertring Researeh and Developmenl Center, University of Pittsburgh, and waszupponed in pari by funds from the National Instiuute of Education (NIE),United States Departmen! of Health, Eduealion, and Weifare. The opinionsexpressed do not necessariiy reflecl lhe posicioo or poiicy of NIE, and no officialendorsernent should be iafered.

1

RE\IEW OF EDUCATIONAL RESEARCH Yol.4{1, No. 1

iearning, cog?itive processes, and human development. It ap-pears that some linking of theory and practice needs to takeplace.

It is of interest to note in this regard that John Dewey, in hispresidential address before the American Psychological Associa-tion in 1899, expressed concern about developing a linking sci-ence between psychological theory and practical worl<. Deweysaid the following:

"Do we not lay a special linking science everlmrhereelse betw6en the theory and praCtical work? We haveengineering between physics and the practical work-ingmen in the miils; we have a scientific medicinebetween the natural science and the physieian."t Thesentences suggest... that the real essence of the prob-lem is found in . .. [a] connection between the two ex-treme terms-between the theorist and the practicalworker-through the medium of the linking science. Thedecisive matter is the extent to which the ideas of thetheorist actually project themselves, through the kindoffrces of the middleman, into the consciousness of thepractitioner. It is the participation by the practical manin the theory, through the agency of the linking science,that determines at once the effectiveness of the workdone, and the moral freedom and personal developmentof the one engag'ed in it. (1900, pp. 110-111)

It is the lteacher's] inability to regard, upon oecasion,both himself and the chiid as just objects working uponeach other in speciflc ways that compels him to resort topurely arbitrary measures, to fall back upon mereroutine traditions of school teaching, or to fly to thelatest fad of pedagogical theorists-the latest panaceapeddled out in school journals or teachers' institutes-just as the old physician relied upon his magic formula.(pp. 112-113)

In this paper, my concern is similar to Dewey's, and Iwouid like to speculate on the nature of a "linking science"-apsychology of instruction-between the scientific knowiedge oflearning (including human cognition and development) andeducational applications.

As a further historical note, I refer to Edward L. Thorndike'sbooh published ia L922, entitled The Psgchology of Arithmetic.

rDewey is quoting Eugo Munsterberg, Psgchologg and life, p. 138. (New York:Eoughton, Mifflin & Co., 1899.)

,

GLASER COMPONENTS OF A PSYCHOLOGY OT INSTRUCTION

In the preface, Thorndike wrote as follows:

Within recent years there have been three lines ofadvance in psychology which are of notable significancefor teaching. The first is the new point of view concerningthe general proeess of iearning. We now understand thatlearning is essentially the formation of eonnections orbonds between situations and responses, that the satisfy-ingness of the result is the chief foree that forms them,and that habit rules in the realm of thought as truly andas fully as in the realm of action.

The second is the great increase in knowledge of theamount, rate, and conditions of irnprovement in thoseorganized g"oups of hierarchies of habits which we callabilities, such as ability to add or ability to read. Practiceand improvemenf are no long:er vagf,re generalities, butconcern changes which ar.e definable and measurable bystandard tests and scales,

The third is the better understanding of the so-called"higher processes" of analysis, abstraction, the forma-tion ofgeneral notions, and reasoning. The older view ofa mental chemistry whereby sensations were com-pounded into percepts, percepts were duplicated by im-ag€s, percepts and imagss were amalgamated intoebstractions and coacepts, and these lvere manipulatedby reasoning, has given way to the understanding ot'thelaws of response to elements or aspects of situations. . . .

This book presents the applications of this nelverdynamie psychologT to the teaching of arithmetic. (pF:v-tri)

In this boolq Thorndike appiied his theory and findings aboutlearning directly to the teaching process. The theory ofstimulus-response bonds thal made up complex chains of be-havior was applied to the analysis of arithmetie tasks; the task ofadding integers, for example, was carefully analyzed in terms ofS-R bonds that could be taught and observed by the teacher.Thorndike also applied the results of his experimental work ontransfer of training and reward in suggesting practical teachingtechniques. He rejected the old notion of lraining general facul-ties and aecepted the fact that training needed to be carried outin more specific eontexts. lle injected his notions of reinforce-ment by indicating that students should rvork on problemswhere, as a result ofcarrying out a successful response, a studentcould see the utility of his behavior.

There is an important differenee betrveen Dewey andThorndike, in terms of the publications I have cited, with respect

3

REVIEIV OF EDUCATIONAI RESEABCE YoL.16, No. 1

to what it takes to translate science into practice. Dewey pressedfor some kind of intermediate linking science. fle conceived of aspecisl structure that interrened belween scientific lheory andpractical application. Thorndike, on the other hand, was eon-cerned with the rtrore direct applieation of what he knew aboutlearning and psychological method to teaehing practice. In addi-tion to his generai theory of iearning, he brought to educationaitopicr a scientific approach which involved carefui analysis of thenature of the task, the design of teaching techniques as afunction of his experimental findinp, and measurement of whatthe task analysis indicated were the cornponenls of the pertbr-mance being learned.

Thorndike's approach set a very special pattern: the combina-tion in one person ot' the theoretical scientist and the appliedscientist interested in designing instructional procedures. Andsince that time, for major advanees in the psychologa of instnrc-tion, we have come to look for individuals interested in bothfields, particularly someone trained in the science of psycholorywho is rnotivated to look at problems in edueation. Such a tactic,howevet, has its shortcomings. It is a highly individualistic,noncumulative kind of venture which does not necessarily lead tothe development of a linking science in which knowledge can beaccumulated into a body of techniques and procedures for practi-cal application by a professional. In contrast, my coneern in thispaper is with the possibilities for the development of a linkingstructure which, because of its own cumulative strength as abody of theory and practice, would be less dependent upon thesporadic interests and insights of individuals.

In the sense described above, B. F. Skinner continued in thepattern of E. L. Thorndike, and most of those who becameinterested in programmed learning and teaching machines con-tinued to work in this mode. As the field beeame popular,however, it took on a superficial momenturn that separated itfrom the impiicit theory that generated it; no substantial struc-ture was built up into which new data, parameters of appiication,and boundary conditions could be piaced.

In the late 1950's and early 1960's, as part of a generalZeitgeist, the notion of a linking science was being nurtured.Bruner (1964) contrasted the nature of a theory of instructionwith a theory of learning. He pointed out that a theory oflearning is descriptive, whereas a theory of instruction is pre-scriptive in the sense that it seis forth rules speeifying the mosteffective way to achieve knowledge or mastery of skills. A theoryof learning describes, after the fact, the conditions under whichsome competence is acquired. -{. theory of instruction is a norma'tive theory in that it sets up criteria of performance and thenspecifies the conditions required for meeting them. Skinner, too,

4

GI.A,SEB COMPONENTS OF A PSYCEOLOGY OF INSTRUCTION

had made this point in the course of his interest in the technoloryof teaching, since the aature of his approach to the study ofbehavior makes the development of procedures for prescribingconditions for learning almost indistinguishable from a theoreti-cal description of learning. Most approaches to psychoiogicallsowledge emphasize both the theoretical and ernpirical description of learning, they have not been concerned with the problemsof prescriptive science. There is, however, at the present time agrowing feeling that a strong test of the adequ acy of descriptivetheory in the behavioral and social sciences can be made throughattempts at application based upon the developrnent of prescrip-tive theory for the design of'social policy and social institutions,including education.

The Actiuity of DesignThe general characteristics of a prescriptive seience of desigrr

have been discussed recently by Herbert Simon in his book,TheSciences of the Artifictal (1969). Simon's ideas on this matter areworth noting here. Ile points out that it traditionaily has beenthe task of the sciences and other disciplines in the university todescribe how things are and how they work, and it has been thetask of professional schools to teach how to design and makethings. The intellectual activity of design is involved not only inproducing material artifacts as in engineering, but aiso in pre.scribing remedies for a sick patient, devising a sales plan for acompany, constructing:a new social welfare policy for a state, anddesigning a program of instruction for a school system. Simonwrites:

Design, so constnred is the core of all professional train-ing; it is ihe principai mark that distinguishes the pro-fessions from the sciences. Schoois of engineering, as wellas schools of architecture, business, education, law, andmedicine, are all centrally coneerned with the process ofdesign. (pp.55-56)

In view of the key role of design in professional activity, it isironic, Simon argues, that prescriptive design sciences are lessprominent in professionai school curricuia than they might be.

Engineering schools have become schools of physics andmathematics; medical schools have become schoois ofbiological sciences; business schools have become schoolsof finite mathematics. The use of adjectives iike "applied"coneeals, but does not change, the fact. (p. 56)

Currieulum topics are seiected from disciplines that are thoughtto be most relevant to professional practice; but desig:r, as

D

RE]NENI OF EDUCATIONAL RESEABCII Vol.46, No. I

distinguished from descriptive analysis, is not necessarilytaughL To some ertent, this phenomenon is a function of theprofessional schools being absorbed into the general culture ofthe university and hankering after respectability in terms.of theprevailing norms of academic respectability. Descriptive theoryand analysis is intellectually tough and prestigiously teachable.Design and application has generally appeared to be more intei-lectually soft, intuitive, and "cookbooky." (I have before used thee:ryression "by-the-aumbers"; Simon's word "cookbooky" ismuch better.) This certainly seems to be the existing state ofaffairs with respect to the application of psychologT lo the designot'insarrction.

In an effort to explore the possibilities for design theory inpsychology and education, a lead can be taken from certainintellectually rigorous practices that have been developed inother frelds. The essence of design is to devise courses of actionaimed at changing existing situations into preferred ones, andteehniques called "optimization methods" haye been developedin statistical decision theory, manag:ement science, and engineer-ingdesign that are concerned with deeiding upon optimal coursesof action. In very general terms, the technique is this: Given aset of alternative goals or possibilities for action, certain fr-redparameten and constraints of the situation, and a function thatdescibes the relationship between these factors, find a set ofvalues that provides the best means of attaining possible out-comes. .{. stock application of this paradigm, described by Simon,is to the so-ealled "diet.problem." Given the goal of losing acertain number of pounds; given parameters and eonstraintssuch as food prices and nutritional content; and given therelationship betrveen the eost of a diel, calories per day, andminirnum needs for nutritionai requirements; flrnd the kinds andquantities of food necessary to maximize utiiity-for example, nomore than 2,000 calories ner day with proper nutritional re-quirements. Once such a problem can be formalized in terms of aquantitative functionai relationship, then standard mathemsti-cal techniques can be applied to maximize ihe outcome subject tothe given constraints. On the basis of this solution, a course ofaction can be decided upon.

In an e.rploratory wzy, the forrnal apparatus of optimizationmethods has been introduced into instructional design. RichardAtkinson and his students (Atkinson & Paulson,.L972; Groen &Atkinson, 1966) have described procedures for optimizingpaired-associale list-learning of lhe kind found in inilial readingtasks or in learning the vocabulary of a seconci language. Thiswork makes it eiear, however, that at the present time. thedeterminalion of opcirnal alternatives is a relativeiy easy matteronly in "irivial" eases. AEkinson is careful to poinl out thac

o

GLTISEB COMPONENTS OF A PSYCEOLOGY OF INSTRUCTION

formal uses of optimization routines developing out of linearprogramming theory, dynamic progtamming, and control theoryare of little help for the complex performances and instructionalprocedures of most interest in education. Ilowever, his work todate on simple eases might help clarify some of the steps involvedin devising and testing optimal instructional strategies.

A significant problem in using optimization methods is therequirement for a formal description of the functional relation-ships involved. If one can employ a formal model like statisticallearning theory, then standard optimization methods can beapplied. However, such formal descriptions are not readily forth-coming for the complex cognitive tasks and instructional proce-dures that are ofcentral interest to educators. For progress no$r,on the basis of our current knowledge and ability to model anddescribe the learning'process, new kinds of prescriptive methodsare requtred. But still, descriptive theory of some kind is anecessary prerequisite for prescriptive theory if the design pro-cedures we will use in the design of instruction are to be at all likethe procedures used in other professions. Of signifrcant interestis that instructional design-the development of instructionalprocedures and methods--can also become a strong way of test-ing deseriptive theory.

. Questions for Instn*ctional PsychotogyFor the development of an instructional psychology, there are

two questions that need to be asked, the first methodological andthe second substantive. The f?rst is: What can be learned abouttechniques to be used in the application of psychologieal knorvl-edge to the design of instruction from the strategies bf designused in other fields? One answer to this question recogrizes thefact that an effective design strategy incorporates procedures foridentifying admissible alternatives and then prbceeds to makedeeisions about the most satisfactory of these alternatives. Inthis regard, a main lesson to be learned from the rvork to date isthat design is not merely assembling a problern solution fromwhat is known, but is rather a search for the most appropriateassernbling of the components involved. The components of adesiga problem need to be assembled into a number of alterna-tive procedures; exploration of these tentative paths then needsto be pursued so that the rnost promising ones ean be follorved upand the less promising ones given a lorver priority. The designprocess essentially involves the generation of alternatives andthe testing of these alternatives against practicai requirements,constraints, and values. This is not done in a singie generate-and-test cycle, but through an iterative series involving thegeneration of alternatives, testing them (through actual smail-

BE\TIEW OF EDUCATIONAT RESEARCE YoL 48, No. 1

scale studies or through simuiation), describing revised alterna-tives, testing them, and so on. This will take us away from theinttritive, one-shot innovation mode of educational reform to amode of operation in which reforms are seen as actual or simu-lated e:rryeriments, with each experiment providing informationfor successive improvement and refinement of possible alterna-tives.

A second question to be considered is: Given methodologies fordeciding among possible alternatives, what are the substantivecomponents that are required as the data to which thesemethodologies can be applied? This question is a large one forpsychological research, and discussion of it will comprise theremainder of this paper. Regardless of the descriptive theorywith which one wort<s, four components of a prescriptive theoryfor the desiga of instructional environments appear to be essen-tial: (a) analysis of the competence, the state of knowledge andskill, to be achieved; (b) description of the initial state with rvhichlearning begins; (c) eonditions that can be implemented to bringabout change from the initial state of the learner to the statedescribed as the competence; and (d) assessment procedures fordetermining the immediate and long-rang:e outcomes of theconditions that are put into effect to impiement change from theinitial state of competence to further development. These com-ponents of a psychology of instruct.ion comprise theinformalion-the parameters, eonstraints, and functionalrelationships-that is required for employing procedures to op-timize instruction or for deciding betrveen instructional alterna-tives.I shall discuss each of these in turn, but before doing so,letme give you some feeling for the general nature of the kind ofindividual cognitive development with which I am concernedhere and to which the above components refer.

7' he D eaelop ment oJ' C o mpetence

The process of instruction, as distinguished from education ingpneral, is, to a large e.ttent, concerned with the development ofthe behaviors and cognitive structures that differentiate be-tween the no'rice and the competent performer in a particularsubjecl matter and intellectual skill. In attaining this knowiedgeand skill, the learner proceeds through a novitiate stage andthen on to a stage of relative expertise; he or she learns to be agood reader, a competent mathematician, a deep thinker, a quicklearner, a creatiye person, an inquiring individual, and so on.Competence in these aetivities is assessed according to criteria ofe.rpertise established by the school and the communily; morespeciiicaily, it is assessed by subjeet-matter requirements, peer-group expeclations, and the generai social and professional

8

GI.ASEB COMPONENTS-OF A PSYCEOLOGY OF I}ISTR,UCTION

criteria for what constitutes low, average, and high levels ofcompetence. The educational and social community adjusts itse.xpectations to the competence level of the learner so thatinitially awlrrrard and partially correct performances are accept-able, whereas later, they are not.

The changes that take place as an individual progtesses fromignorance to increasing competence are of the following kinds: (a)Yariable, awkward, and erude performanee changes to perfor-mance that is consistent, relatively fast, and precise. Unitaryacts change into larger. response integrations and overallstrategies. (b) The contexts of performance ehange from simplestimulus patterns wtth a great deal of clarity to complex patternsin which information must be abstracted from a eontext of eventsthat are not a1l relevant. (c) Performance becomes increasingiysymbolic, covert, and automatic. The learner responds increas-ingly to internal representations of an event, to internalizedstandards, and to internalized strategies for thinking and prob-lem solving. (d) The beharrior of the competeirt individual be-comes increasingly seif-sustaining in terms of skillful employ-ment of the rules when they are applicable and subtle bending ofthe rules in appropriate situations. Increasing reliance is placedon one's own ability to generate the events by rvhich one lear:rsand the criteria by which one's performanee is judged andvalued. It is the understanding and facilitation of this process ofchange from ignorance to cornpetence, from noviee to expert,that is a major focus of the emerg:ing psychology of instruction.Consider now the components required to facilitate this process"

' Components of a Psyghology of Insttttction

The Analysis of Competent Perform,ance

Central to a concern with instruetional processes is the prob-lem of task analysis; analytic description is required of rvhat it isthat is to be learned. What has a eompetent performer in asubject-matter domain learned that distinguishes him from anovice? What distinguishes a skilled reader from an unskilledone? When a task anaiysis identifies the properties of a eertainclass of performance, then inferenees can be formulated andtested coneerning optimal instructional processes for acquiringthese performance abilities. Analyzing the content of instructionmeans studying tasks considerably more complex than thosetypically studied in the laboratory. It also requires techniquesfor the detaiied analysis of performance in terms of the demandsplaced on cog:ritive processes and on knowledge and skills as-sumed to be in the learaer's repertoire as acquired throughinstruction, development, or self-learning.

o

RTITIEW' OF EDUCATIONAL RESEARCE YoL.t6, No. I

The requirement for the analysis of competenl performance isrelated to the specification of behavioral objectives so stronglyadvocated by many educational psychologists. This salutary ad-vice given by behavioral psychologists is now being taken seri-ousiy by cognitive theorists concerned with the cognitive compo-nents of criterion performance. There seem to be two mainasp€cts to such an analysis. One is the identification of theinformation structures that are required for performance, andthe other is a description of the processes and cognitivestrategie+-heuristics and algoithms-that need to be applied tothis information, and which themselves are part of the informa-tion data base.

As an interesting case in point, consider the work that has beengoing on in the cognitive simulation of expert chess players. Anarticle by Simon and Chase (1973) summarizes differences be-tween novice and average players, and masters and grandmastersin chess. They indicate that the most likely explanation for thee.rtraordinary skiil of the ehess master is that he is acquaintedwiih tens of thousands of familiar patterns of pieces, and heassociates many of these patterns with plausible moves by takingadvantage of the informational features represented by thepatterns. The basic heuristics that guide the search for goodmoves are based upon the perceptuai ability to recog?ize aninformational pattern on the board. "For example," Simon andChase point out, "every chess player of even moderate skill isfarniliar with the advice: 'If there's an open file, put a Rook onit' " (p.402). The pattern of an open file triggers this heuristic andinitiates a move in a heuristic search for the best move. For achess master, hundreds of immediately recognized patterns maybe associated with an algorithmic solulion-i.e., moves that leadto the guaranteed win of a piece or a checkmate-so that a seriesof rnoves may be piayed almost by rote. The key to understandingchess skill iies in understanding the large percepEual vocabularyof piece configurations, the associated algorithms, and the par-ticular perceptual processes involved in this skili.

From an instructional point of vierv, the target behavior ofinterest is that the chess master's performance seerns to involvea buildup in long-term memory of a

vast repertoire ot' patterns and associated plausiblemoves. Early in practice, these move sequences arearrived at by slow, conscious heuristic seatch-"If I takethat pieee, then he takes this piece . . ."-but with prac-tice, the initial condition is seen as a pattern, quickly andunconsciously, and the plausible move eomes almostautomatically. Such a learning process takes !ime-

10

GI"ASEB COMPONENTS OF A PSYCEOLOGY OF INSTRUCTION

years-to build the thousands of familiar chunks neededfor master-level ehess. (Simon & Chase, 1973, p. 408)

It is to be noted further that grandmasters may possessexceptional talents along certain dimensions, but their talentsare chess-speeific. There is no evidence that masters dern-onstrate more than above-average competenee on basic intel-iectual factors. Thus, the acquisition of chess skill depends, inlarge part, on building up speeifie recognition memory for manyfamiliar chess patterns. In a psychology of instruction, this kindof eontrastive analysis of the informational content and skills ofcompetent performers and novices might be prototypic of thekind of research that is espeeially relevant to an understandingof the objectives of instruction.

Consider another example of work on simple arithmetic prob-lems and the nature of competent performance in addition andsubtraction. Studies carried out by Suppes and Groen (1967);Woods, Resnick, and Groen (1975); and Resnick (in press) sugEiestan interesting relationship betrveen what children are taught todo and how they eventually perform efficiently. Young childrenare generally taught to soive a single-disit addition problem suchas 6 + 6 by an algorithm in which they count out six blocks, thencouut eight blocks, and then count to eombine the set. lYithpractice, children perform this smoothly; when the blocks aretaken away, they frequently shift to counting on their fingers,and then eventually shift to internal processing. lYhen thenature of this internal processing is examined, it is found thatmost chiidren car?y out addition by using what has been called a"choiee model." They appear to se! a mental counter to themagnitude of rvhichever number is larger and then inerement bythe smailer number. Some children retain the earlier model usedin instruction-that is, they increment six times, then incrementeight more times, and then read their mental counter. The mostefficient children, however, appear to be able, without directinstruction, to convert a routine that has been taught into adifferent routine-a routine that shows they have discoveredcommutativity and have developed a performance that requiresfewer steps. It is to be noted that the inirial teaehing procedurereflected the rational "union of sets" definition of addition, andthus is a mathematically correct procedure that represents thesubjeet matter clearly and provides a routine that is easy todemonstrate and learn. For an effieient performer, however, theroutine is awkqrard and slow. Thus, the routine derived byrational anaiysis of the subject-matter structure is transformedto a performance routine that refleets a more sophisticateddefinition of the subject malter.

11

REYIEW OF EDUCATIONAL RESEA.R,CE YoL 46, No. 1

lYhat are the implications of this analysis? On the face of it, itsrould seem that we ought to abandon the algorithm suggestedby direct analysis of tasks in favor of analysis of skilled perfor-ms.nee. We can argue thar the rational analysis of tasks may noEmatch skilled performance and that it therefore should not beused as a basis for instruction. It would seem best to carry outdetaiied empirical analyses of skilled performanee on subject-matter tasks and teach the routines uncovered by such analyses.Ilowever, in discussing her work, Resnick (in press) points outthat such a conciusion couid be in error, since it rests on theassuanption that effieient instruction is necessarily direct in-strr:ction in skilled performance strategies rather lhan instruc.tion in routines thar put learners in a good position to invent orderive efficienl strategies for lhemseives. So, it is impiied thatthe teaching routines in elementary arithmecic rvere not poorones that inhibited the acquisition of efficienl performance, butrnay have been good ones lhat fostered the invention of moreeffrcient algorithms.

As suggested by the above examples, the work on the analysisof comperent performance that is going on at the present time isol two kinds: the characterization of the information structuresand cognitive processes of the skilled performer, and behavior-ally oriente<i work on rational task analysis. Such analyses ofhuman competence and subject-matter tasks may aliow us ro dotwo thing: regarding the optimization of instruction: (a) Specify-ing the slructures and processes by which competent individualsmight be performing a task may put us in a position to try toteach these processes to individual learners. (b) Knowing thal atask is performed efficiently in one way rather than in anothermight enable us to design instruction so that the perfornoancelearned allows individuals lo direcriy or indirectly transfer lo themore efficient method.

It would be a serious omission to leave the topic oi task analysiswithout referring to the inr'iuential rvork of R,oberl Gagn6 onlearning hierarchies (1962, 1970). This rheory continues to beffidely accepted as a framervork for investigating instructionalprocesses and for designing educalional procedures and cur-ricula in valious subject matters. Gagn6 has presented us rvich asystem for rational task analysis based upon a cumuiacive learn-ing model ihat states that there are different types of learning,w'ith the simpler types being prerequisite states for learning themore complex t3pes. For example, problem soiving, a eomplexhigherorder type ot' learning, requires rule learning, a iorver-order task, as a prerequisite; and rule learning, since rulesconsist of relationships between concepls, requires conceptlearning as a prerequisite; and so torth. In generai. rhe lorver-orCer task is defrneC as being prerequisiie co a higher-order iask

19

GLASER COMPONENTS OF A PSYCHOLOCY OF INSTRUCTION

when competence in the simpler task facilitates positiv'e transferin learning the more eomplex task.

In addition to a clear<ut transfer relationship, bhere are,however, several possible reiationships that might exist betweenprerequisite tasks and superordinate tasks. The iower-order taskmight be one of a number of eomponents of the more complextask, each of which can be aequired independently of the others,but all of whieh must be eombined to produce the higher-orderperformance. Alternatively, the lower-order tasks may them-selves be hierarchically related to one another, constituting asequenced progression leading to increasingiy complex perfor-mance. Lower-order tasks may also be eornpetencies which facili-tate the learning of the more complex task, but which drop out inthe more "skillfui" performance. Furthermore, the lorver-ordertasks might function as heuristics for discovering or inventingprocedures for carrying out the more complex task. Researchaiong these lines, i.e., investigating the acquisition of complexperformance on the basis of existing competencies, is especiailyrelevant for instructional psycholory.

Description of Initial State

Instruction begins rvith an initial state of the lea,rner, andinstruction proceeds on this base toward the development ofcompetent performance. There are two approaches to this com-ponent of instructional desigrr: "immediate" and "long-term."The-immediate approach is to take seriously the fact that effec-tive instruction requires carefui assessment of the strengths,,weaknesses, styies, and background interests and talents of'individual learners. lYhat are the detaiis of rvhat a child knowsand does not know at particular points in his or her learning?What are the details of the skills that he or she is deveioping?!;t/hat needs to be improved? lYhat strengths can be capitalizedon? lYhat do various developmentai leveis and various culturalbackgrounds mean for what shouid be taughr and how it shouidbe taught? Educational practices need to be designed so thatanswers to these kinds of questions are possible for ali indiviciu-als attending schooi. Teachers and stucients need to be in aposition to obtain and utilize this kind of information; with it,teachers can prescribe the instruction required, and studentscan assess their own abilities and select appropriate instruciion.

The use of procedures for providing this kind of information forteaching requires the adoption of an artitude that looks upon theinformation obtained as information for improving instruetion,and not simpiy as a test for evaiuating and ciassifying students.For this purpose, it has been uset'ul to provide teachers withhierarchies of increasing compeience in various school subjects

told

BEYIEW OF EDUCATIONAL BESEARCE YoL {6, No. 1

QResnick, W'ang, & Kaplan, 19?3). These take the form of "st::uc-hrred maps" into which a teacher can place a chiid and therebydirect attention to prerequisite skills that might need to belearned or advanced skills that the child might explore. Thehierarchical map serves as a guide upon which both lhe teacherand the child can impose additional judgments. The provision ofprocedures for identifying the current competenee and talents ofthe learner in a way that provides a basis for instruction isgenerally not done in cu:reht. educational methods at a levei ofdetail necessary for the effeetive guidance of individual learners.The implementalion of such procedures is not only a matter ofresearch, but also largeiy a matter of administrative change andthe design of appropriate materials.

The mora long-term approach derives from the fact that ap-titude and intelligence tests are the prevalent methods forassessing initial states that are, to some extent, predictive ofeventual educational suceess, but these measures do not providesufficient information about instruetional processes (Glaser,L97U. Having been devised primarily for purposes of selection,these measures do not provide a basis for deciding:horv instruc-tion might be designed to make the attainment of suceessfulperformance more probable. The significant requirement in thisregard for a psycholory of instruction is to describe the initialstate of the learner in terms of processes involved in achievingcompetent performance. This would then allow us to influencelearning in two ways: (a) to design instructional alternatives thatadapt to these processes, and (b) to altempt to improve anindividual's competence in these processes so thal he is morelikely to profit from the instructionaI procedures available. Thereis, at the present time, a spum of interesting research devoted toanalyzing the underlying eognitive proeesses thai contribute tointeiligence and aptitude-like performance. Three illustrativeexamples will be presented.

In a recent series of studies by Hunt, Frost, and Lunneborg(1973), students 'rere classiJied into high- and iorv.oerbal abilitygtoups and into high- and lorvgztontitatiuri abilicy groups on thebasis of a battery of tests used for selection lbr college entranceat the University of lYashington. The individuals in each of thesegroups were then given a series of tasks employed in laboratorye.xperiments on the experimental analysis of information process-ing models of memory. In this w8y, the characteristics ofhigh-verbal ability and high-quantitative abiiity studencs, asdefined by aptirude tests, were exarnined in terms of cogrritiveprocesses, as defined by tasks used !o investigate partieulartheories of cognition. The conciusions trom the studies tenta-tiveiy indicate ihat there is a relacionship betrveen verbai abilicyand the rapiciity and efficiency of data manipularion in short-

14

GI.ASER COMPONENTS OF A PSYCHOLOGY OT INSTRUCTION

term memory, and between quantitativa ability and resistance todistraction while consolidating information in short-term mern-ory.

It is thus suggested that verbal and mathematical aptitude isrelated to the nature of information processing in memory, andthe interesting question for an instructional psychoiory iswhether we can proceed further and identify situations wherethe speed and other properties of such processing will be predic-tive of school achievement. Such an endeavor could have rnoresignificant implieations than present correlationally derived re-lations between aptitude tests and school success because clueswould perhaps be available about how verbal and mathematicalability processes might be modified or employed for learning.

In a very receRt paper, Estes (1974) discusses the digit-spantest that appears on the Stanford-Binet. At year ten, the sub.ject's task is to repeat a sequence of random digits after theyhave been read aloud by the examiner. The test correlatessatisfactorily with the usual validation criteria, but the interest-ing instruetional question is: If an individuai scores low on thistest, what instructional procedure should rve expect to be usefulin improving this performance, performance that we know iscorrelated with academie accomplishment? Estes describes re-cent researeh and theory dealing with short-term memory forsequences of items that indicate that the dig:it-span task appearsto involve a hierarchical structure of representations in memory.A quote gives the gist and flavor of this:

On presentation of the digil sequence of 6914?2, theindividual is conceived to subgroup the sequence into twochunks, assigzring a code to each which he maintains inmemory, and within each ehunk relating the items of thesequence to the ordinal numbers 1, 2, and 3. On a requestto recall the string, the individual brings into memory hiscoded representations ofthe two chunks; each ofthese inturn activates recall of the individual digits and theirassociated serial positions. lYhile this process goes on,the indi:riduai must hold the partially reconstructedsequence in an output response buffer by an inhibitoryproeess until the decoding is complete and then emit thedigits in the proper order. (p. ?aB)

Estes points out that such an analysis of performance on thedigit-span task may have implications for assessing individualdifferences. Young or mentaily retarded children might fail thetest because of insufficient familiarity rvith the sequence ofordinal numbers or beeause of inexperience in ordering materialswith the number sequence. An individual may not perform weii

t<

REVIEW OF EDUCATIONAL RESEARCE VoL 46, No. 1

because he has not developed an appropriate strategT of group-ing (although he might utilize grouping when prompted by theexaminer), is unable to accomplish the coding process necessaryto take advantage of chunking, or lacks the capacity for selectiveinhibition in buffer storage necessary to order his output prop-erly. Estes writes:

Clearly, it would be possible with the advantage of addedtheoretical insight to augment the standard digit spantest in such a way as to loealize the souree of diffrculty foran indinidual who fails under the standard procedure.This augmentation would quite likely do little to improvethe predictive value of the test, but it might be ofconsiderable help in indicating how deficient perfor-mance in this and related tasks might be remedied. (p.744)

Elolzman (1975) has studied letter series completion problemsof the sort used by the Thurstones (1941) in their factor anaiyticstudies of intelligence. Letter series consist of a sequence ofalphabetic characters running in a consistent pattern. In anyone test item, usually about a dozen of these patterned lettersare presented to the examinee followed by four blank spaces. Theindividual must fill in the four blanks with letters that areconsistent with the pattern exhibited by the previously pre-sented letters of that series. For example, the individual mightsee the problem "defgefghfghi- - - -"and be asked to flll in theblanks. W'ork on analyzing this task has been carried out bySimon and Kotovsky (Simon & Kotovsky, 1963; Kotovsky &Simon, 19?3), rvho have obtained protocols of adolescents andaduits solving these sorts of problems; then, based on theseobservations, they wrote computer prog"ams to simulate hu-mans' solution routines. Four basic component routines arenecessary for the simulation of eorrect solution. The first routineis the detection of relations betrveen letters: Are letters identical,sequential, or sequential in reverse order? The second routine orsubskill is the discovery of periodicitg in a series. This involvesnoticing that letter relations repeat themselves at regular, pre-dietabie intervals. A third routine, called pattern descriptio'n,assembles knowledge of letter relations and knowledge ofperiodicity into a rule that generates the series. The final routinerequired is extrapolo,tion, This invoiyes remembering the patterndescription and using this rule to generate the appropriateletters for the blanks.

Using this information about the possible cognitive processesinvolved, Holzman taught elementary school children to be veryproficient in the deteetion of relations and the discovery ofperiodicity. As a result of their training, children were able to

16

GLASER COMPONENTS OF A PSYCHOLOGY OF INSTR,UCTION

show substantial pretest to posttest gains on a typical letterseries completion test. lVlost strikingly, the children were sig-nificantly more able than controi subjects to demonstrate perfectposttest solutions to the types of problems which they founddifficult on the pretest. Both the controi subjects, as a result ofrepeated testing, and experimental subjects were able to makegains on ea:ry problems, but the children trained on componentsubskills seem to have acquired an information managementstrategy that allowed them frequently to reach perfect solutioneven on difficult problems. The skills taught to the children inthis study were quite speeific; however, the question is raisedabout the possibilities for the analysis of abilities that are moregeneral than these and that might provide a basis for trulygenerative intellectual abilities.

Studies like those I have just described raise the possibilitythat measures of intelligence and aptitude, analyzed in terms ofcognitive processes, will, as llunt and his colleagues (19?3) write,'tnove many psychometric predictions from static statementsabout the probability of sueeess to dynamie statements aboutwhat can be done to increase the likelihood of success" (p. 118).And furthermore, "Ilopefuily [this] new viewpoint . . . will lead tomeasuring instnrments which are diagnostic, in the sense thatthey tell us how the institution should adjust to the person,instead of simply telling us which people already are adjusted tothe institution" (p. 120).

Cond;itions Thot Foster the Acquisition of Competenee

This third eomponent of instructional design-the conditionsthat can be implernented to foster the acquisition ofcompetenee<ssentially invoives the procedures that assistlearning and the techniques and materials that are designed intothe environment in which learning occurs. In this regard, weshouid recognize that the little we do know about iearning isla:own in terms of deseriptive science. Littie investigation hastaken place from the point of view of utilizing this informationfor designing the conditions of instrucrion. Excepiions to this arethe work on behavior modification, the work of Gagn6, and thelimited work referred to earlier on optimization models forpaired-assoeiate forms of learning. However, for the most part,these enterprises have not considered complex cognitive perfor-mance in'any intensive way. What is required is that research oninstnrction be cast into the mold of a design science that at-tempts to maximize the outcomes of learning for different indi-viduals. A new form of experimentation would be called for wherethe tactic is not to develop models of learning and performance,but to test existing models by using them for maximizing learn-

II

RSVIEW OF EDUCATIONAL RESEARCH Vol.46, No. 1

ing under various conditions. For this purpose, we need a theoryof the acquisition of competent performance. Such a theorywould be conceraed with how an individuai acquires increasinglycomplex performances by assembling the present components ofhis repertoire, by manipulating the conditions and events aroundhim, and by employing his knowledge of how he learns. With thedeveiopment of sueh a theory in mind, some very brief prelimi-nary obser.rations can be made on knowledge structures inmemory, on g:eneralized abilities for learning to learn, and on thenature of reinforcement.

I{noutledge st 1)ctures. Some recent work on the semantic struc-ture of information in memory (e.g., Greeno, in press) has beenconcerned with the sernantic networks and information process-ing mechanisms that are available at different levels of subject-matter competenee. If, at various levels of learning or stagesof competence, the kinds of knowledge we wish to create inthe minds of students can be speci.fied in this way, then someinteresting implications are suggested for the relationshipamong: subject-matter structure, curriculum content, and in-stnretional design. One such reiationship can be seen by distin-guishing betrveen the structure of a subject-matter domain as itis organized by scholars studying that domain and the structurethat is devised for teaching it (Glaser, 1973). The structure of asubject-matter discipline, as employed foi' the purpose of ad-vanced scholarship, consists of theories, eoncepts, and definitionsthat serve to make the domain manipulable for the work ofsubject-matter experts. Ilowever, the structures employed forthis purpose are not necessariiy the most useful for faciiitatingthe learninb of an individual at a less advanced level of develop-ment or subject-matter sophistication. Good theory for thescholar may not be gnod pedagogical theory; what leads to knowl-edge for the expert may neither lead to knowledge for the novicenor help him to develop competence. It follows that a significanbconsideration for instructional design is the organization ofcurriculum sequences that provide knowledge structures opti-mally organized for morring the novice toward expertise. Appro-priately designed structures for learning: can reduce the amounrof information that mtrst be held in mind to comprehend the sub-ject matter; forexample, a verbal label, a conceptual formulation,or a rule or principle may help to organize and surnm anize a largenumber of observations. The rule can be thought of as a struc-ture or representation by rvhich an individual is directed ordirects himself to look at the relevant features of what mightotherwise be an unorganized task situation. As a consequence, astudent can generalize across the superficial detaiis of the iim-ited set of experiences encountered in instruction (Gilbert,1962). Some ways of organizing information may permit better

18

GLASIER COMPONENTS OF A PSYCHOLOCY OF INSTRUCTION

memory retrieval than other ways and, as a result, facilitate thelearner's capacity to learn new things on the basis of what he hasaiready learned and to access information readily for thinkingand problem solving. The organization of subject-matter contentcan do for the learner what advanced theory does for the expert.Such organizations, however, are not readily available; they aresometimes devised by ingenious teachers and built by them intoinstructional procedures. I would further suggest that the na-ture and the design ofthese organizations or pedagogical struc-tures are a unique province of study for a psychology of instrue-tion. ,

T e ac hing g en er aliz e d I e anring - t o-l e ant, abili ti e s. I n the acquis i-tion of eompetence, a signifrcant instructional consideration isthe way in rvhich individuais use their current cornpetence andcomponents of their repertoire for learning' new higher-orderperformance or for solving problerns that lead to learning thishigher-order pertormance. Thus, an appropriate concern forinstnrction is the possibiiity for teaching general strateg:ies thatwill help individuals learn on their own and be less dependent onthe instructor's elegance in presenting particular tasks. Aninterest in teaching sueh general "learning to learn" abilities hasbeen widely expressed by educators and psychologists, but at thepresent time, there is little seientific basis for such instruction.One possible basis can come from the studies already describedon the process analyses of aptitude-like skills. Still anotherpotential basis for such instruction might be provided by thegrowing number of information processing analyses of problem-solving tasks.

In a recent paper, Besnick and Glaser (in press) argue that theproeesses involved in certain kinds of problern solving are proba-bly similar to the processes involved in learning in the absence ofdirect or complete instruetion, and that instruction in theseprocesses might constitute a means of increasing an individual'sgeneralized learning-lo-learn abiiities" A rnodel of problem solv-ing was developed in which three interacting phases rvere iden-tified: (a) problem detection, in rvhich the inapplicability of"usual routines" for solving a problem is noted and a problem orgoal is formulated; (b) feature deteetion, in whieh the taskenvironment (the externai situation, which includes both physi-cal and social features) is scanned for cues that might lead toappropriate actions; and (c) goal analysis, in rvhich goals aresuceessively reformulated, partly on the basis of external taskcues, in order to yield solubie subgoals that contribute eventuallyto solution of the problem as presented. A study by Schadler andPellegrino (Note 1) has shown that requiring subjects to ver'baiize their goals and strategies in each of these phases, beforemaking overt moves torvard solution, greatiy enhanees lhe

19

REVIE]Y OF EDUCATIONT.L RESEARC!{ Vol.46, No. I

likelihood of problem solution. Along these lines, it seemsreasonable to antieipate that ways can be found to make indi-viduals more conscious of the role of environmental cues inproblem solving. Individuals might be taught strategies of-fea-ture scanning and analysis that will enhance the likelihood oftheir noticing cues that prompt effective actions while somehow"deactivatingi'those cues that prompt ineffective actions. Suchself-regulation could be a major characteristic of sucbessfulselflearning and problem solving. The specifre suggestions thatcan be offered at this time for instrugtion of such generalizedlearning abilities are limited, since relativeiy little has been doneon developing task analyses that eharacterize these generalprocesses in instructable terms; but work on problem soiving isespecially relevant to this important goal of instruction. Relatedto this is work on reinforcement effects to which I now turn.

Reinforcement. Qontingencies of reinforcement pervade theacquisition of competence. However, rvith the strong emerg:enceof cognitive psychologly, and with awareness of the fact that thebulk of our knorvledge about reinforcement is derived frornanimal studies in simple task situations and from human ex-perimentai conterts in which conditions constrain subjects toemploy limited behavioral processes, lve are in some danger ofignoring the potential influenee of reinforcement on complexperformance. There is, on the one hand, a strong suggestion ofdiscontinuity in the operation of reinforcement when movingfrom simple to higher-order behaviors. On the other hand, theview rhat seems best supported ai the moment is that themechanisms of reinforeement are similar at all levels of de-velopment, but variations in response organization result. indifferent phenotypic manifestations (Estes, 1971). As individualsmature, human behavior is organized into higher-order routinesand strategies, and it is these iarge cognitive organizalionswhose probabilities of oecurrence are modifred by reinforeingcontingencies. it is the nature of the unit of response that maydistinguish the mature human learner, rvhereas the operabion ofthe principles of reinforcement may be simiiar for differentspecies and different levels of development and competence.

From the point of view of a theory of insrructional psycholory,we should be further aware that in the naturai settings ofclassrooms reinforcement occurs extensively within a soeial eon-text. This highlights certain dimensions of the nature of rein.forcement that need to be considered in instructional situations(e.9., Bandura, 19?1). One aspect is that people continually ob-serve the behaviors of others as this behavior is rewarded;ignored, or punished; and this observation inl'luences the sub-sequent operation and effect ofreinforcers on the observers. This

20

GLASER COMPONENTS OF A PSYCHOLOGY OF INS"RUCTION

is the phenomenon of modeling and vicarious reinforcement. Asecond aspect is that indiriduals regulate their own actions bymechanisms of self-reinforcement. Seif-generated anticipatoryconsequences allow possible future contingencies to influencepresent behavior, and self-evaluations of the consequences ofone's own actions influence behavior as these eonsequences aremade apparent by classroom reinforcement contingencies.

Assessment of tlte Effeets of Instntctional ImplementationThe fourth component of instructionai design is concerned with

the effects of instructional implementation in the short and inthe long run-effects that occur immediately in the context ofinstruction and effeets that persist in terms of long-terrn trans-fer, generaiized patterns of behavior, and abiiity for furtherlearning. One requirement for this purpose is to break arvay fromthe tradition of norm-referenced measurement to measurementmore concerned with identiff ing the nature of competent per-formance (Glaser, 1963; Glaser & Nitko, 1971). For effectiveinstructional design, tests will have to be criterion referenced inaddition to being norm refereneed. They will have to assessperformance attainments and capabilities that can be matchedto available educational options in more detaiied lvays than canbe ca:ried out with currently used testing and assessment proce-dures. This will be an important part of the development of apsychology of instruction. It is mandatory that testing not standout as evaluative devices that are an extrinsic and externaladjunct of instruetion. Tests need to be interpreted in terms ofperformance criteria so that the learner and the teacher areinforrned about an individuai's progEess relative to developingcompetence. In this way, information is provided for decidingupon appropriate courses of instruction.

The performance measured by tests designed to facilitateinstruction needs to be related to processes identified as compo-nents of competence. For this purpose, some interesting en-deavors can be envisioned. One example is work going on inanalyzing the processes involved in the comprehension of writ-ten language, stimulated by the work in psycholinguistics andcognitive psycholory (e.g:., Caroll & Freedle, 1972). This de-velopment should be juxtaposed with the fact that there hasbeen a great deal of work on the development of tests of readingcomprehension. As we begin to analyze comprehension tasks andrelate them to theories of semantic memory, imagery, and soforth, we should be able to develop tests that provide us withdiagnostic information about component processes that contrib'ute to perforrnance and that can be inr'lueneed through instruc-

D1

BEIrIEW OF EDUCATIONAL RESEAACII VoL 4{i, No. 1

tion. This kind of activity should change the nature of assess-ment procedures and provide us the kind of information requiredfor maximizing instructional outcomes.

Another area of investigation that is beginning to prorridesignificant evaluative information about' the conditions underwhich learning takes place in school contexts should be men-tioned. This is the growing sophistication in the study of thenature of classroom processes. In the past, rve essentially at-tempted to describe school learning by relating the nature ofstudent input to the quality of student output; but the processintervening between the two, the independent variable, was onlygenerally described. Detailed information was rarely obtainedabout differences between effective and less effective classroomprocesses.

There are nolv a number of attempts to research these details.I am especially impressed by the model for such research beingdeveloped by my colleague, William Cooley, in conjunction withPaul Lohnes (Cooley & Lohnes, in press). Their model is derivedfrom Camoll's 1962 model of school learning and consists of sixcomponents: (a) initial ability, rvhich refleets the basic incomingskills and general intellectual development of children in aciassroom; (b) opportunity, which describes the relative propor-tion of classroom aetivities (the dominant elassroom subject-matter themes) that are directly related to the assessed out-comes of instruction; (c) motivation, which reflects a student'stendency to engage in learning activities rvhen the opportunityexists, and operationally defined (in elementary school class-rooms) as the fit between the learning situation and the chiid'sneeds, and the relative incidence of teaehdr praise and encour-agement and their antitheses for particular pupil behaviors; (d)structure and placement, rvhich reflect the extent to rvhich thecurrieulum is structured by specifying objectives, sequences ofinstruction, particular methods used in differentiating studentsor in individualizing instruction, and, in generai, the organiza-tion of instruction and teaching materials; (e) instructionaievents, which reflect the relative incidenee of teacher-pupilinstruptional interaction and observed, for exampie, through theextent of teacher acknowledg:ment of, and feedback rvith respectto, a student's task-related activity; (f) criterion ability, whichreflects end-of-year student performance, for example, on stan-dardized achievement and intellectual ability tests.

.A.fter obtaining information on these components of instruc-tion, a multivariate anaiysis procedure is used to determine theregression of criterion ability on the other five components of theinstructional model. This permits an analysis of the total var-iance represented in the criterion variabie that is expiainabie interms of the other components-{a) variance due to incoming

22

CLASER COMPONENTS OF A PSYCHOLOGY OF INSTRUCTION

ability independent of ciassroom process variables, (b) varianceuniquely due to the classroom process variables independent ofinitial ability factors, and (c) variance due to the interaction oroverlap between initiat ability and instructional processes. Inthis way, detailed information is obtained on the kind of class-room implementation of an instructionai system that is effec-tive or ineffeetive in producing school outcomes. What is ofpartieular interest in researeh of this kind is that we can begin torelate the effectiveness of school implementation procedures topsychological dimensions of learning theory and to a theory ofthe acquisition of cornpetence" Each endeavor can reinforce orchallenge the findings of the other.

To conclude: A speculative outline of a psycholory of instruc'tion as a science of desiggr has been presented. Direetions inrvhich it might develop and what some of its substantive compo-nents might be have been suggested. There is much to be done,but many promising leads are now offered for testing fundamen-tal theories of human learning and cognition and for contribut-ing strongly to educational practice.

ReJerence N ote

1. Schadler, M., & Pellegrino, J. W. llrloz;imizing performance in a problem sohringtas&. Unpublished manuscript, Llniversity of Pittsburgh, Learning Researehand Development Center, 19?4.

ReJerencesAtkinson, R. C., & Paulson, J. A. An approach to the psychology of instruction.

P cuchological B u,lletin, L912, 78, 19-67.Bandura, A. Vicarious- and self-reinforeement processes. In R. Glaser (Ed.),The

nature of retnforcemgnl. New York: Academic Press, 19?1.Bnrner, J. S. Some lheorems on instruction illustraled r*ith reference to

mathemalics . Theories of learni,ng and instttt ctio'n. The S iztg'third Y eatbookol the National Society J'r the Stud.y o7 Educatio'n, Part I, 1964, 6.r, 306-335.

Carrotl. J. B.. & Freedie, R. O. (Eds.). Langntage conprehension antd the acquisitionof knoutledge. lYashington, D. C.: V. H. \Yinston & Sons. 19?2.

Cooley, lV. W.. & Lohnes, P. R. Euoluotite inguiry in education New York:Irrrinpon Publishers, in press.

Dewey, J. Psychology and sociai practice. The Psychological Redeto, L90Q,7,106-124.

Estes, 1V. K Reward in human learning: Theoretical issues and strategic choicepoints. In R. Glaser (Ed.),The no,ture of reiniorcement. Nerv York: Academichess, 19?1.

Estes, lY. K- Learning lheory and intelligenee. American Ps'gchologist, t911,29,710-749-

Gagn6, R. M. The acquisition of knowledge. Psychological Redeut. 1962, t9,355-365.

Gagrr6, R. M. ?he conditions oJ'leonting (2nd ed.). New York: llolt, Rinehart &Winston, 19?0.

Gilbert" T. F. )l,athetics: The technology of instructiot:..Jountal o7'ltathetica, 7962,. a+6l. l-lo.

nl,

REYIEW OF EDUCATIONAL RESEARCII Vol.46, No. 1

Glaser, B. Instnrctional technoloCy and the measurement of learning outcomes:Sorne questio as. A merie on P sg c ho lo gi st, 1 963' I 8, 5 1 9' 621.

Glasea B. lndividuals and learning: The new aptitudes. Educatiutol Eeseatchct,L972, l,'O.19.

Glaser, R. EducEtional Prycholo$T and educaiiom. Amtricott Psyehologist, L973,%,5A7.566.

Glaser, R- & Nitko, A- J. Meesuremen! in learaing and insfinrction- In R. L.Tlrorndikc (Ed)b Educatiotul medwremult (2nd ed.). F/ashington, D. C.:Anrerican Council on Education' 19?1.

Greeno, J. G. Cognitivc objectives.of instruction: Theory of kno'r'ledge for rolvingproblems and answering quesiions. In D. Klahr (Ed.), Cogttition ond ia,slrttc',Joz. Eillsdale, N. J.: Lawr"ence Erlbaum Associates, in press-

Groen, G. J. & Atkinson, B. C. Models for optimizing the leerning process'Prychological Bulletin, 1966, 66, 309€20.

Eolzm-an, T. G. Procccs troining as a test of cotnputer eimulation theory.Unpublished master/s thesis, University of Pittsburgh' 19?6'

Ilunq E., Frogt, N., & Lunneborg, C. Individual differencer in cognition:A uew approach to intelligence. In G. E. Bower (8d.), Tha prychology oftear*ing wd motiootiaz (Vol. O. New York: Academic PrErs' 19?3.

Kotovsky, iL & Simon, II. A- Empirical tests of a lheory of humen acquiri'tion of concegls for sequential patterns' Cognitilq Ptychologg' 1913' 4'399424.

Besnick, L. B. Task analyEis in instn:clional design: Some cases- from mathe'maticr. In D. K]ahr(E d.),Cogrtition andinstntclion. Hillsdale, N.J.: LawrenceErlbaurn Associates, in Press.

Resniek, L. 8., & Glaser, B.-Problem solving and intelligence' In L' B' Resnick(Ed), ?he iohlra of i.ntalligea,ce. Ilillsdale, N. J.: Lawrens€ Erlbaum Asso-ciates, in press.

Resnick, L 8., Wang, M. C., & Kaplan, J. Task analysir in curriculum design: A-

hierarchically siquenced introductory mathematics curriculum. Jtutnol ofApplied Behotriet Anolyais, l9?B' d, 679'?10.

Simori,'H. .{- fta cctstuces-of thc artificicl Cambridge, Mass.: MIT Press' 1969'

Simon, Ir. A., & Chase, W. G. SkiU in chess. Americoa Sciqntist, !973, 67,394403.

Siraon, tI. A- & Kotovsky, K. Eumsn acquisition of concepts for sequential pat'terns. Prycrtological Revicu, 1963, 70, 534'546.

Suppes, P., &-Groen-, Q. J.isme-counting mod,els lorfrslq-adg petfotmance data---il'i.rit, a.dd;itiott. /acts (Tech. Rel.90). Also in J. M- Scandura (Ed')' Ez'

ccqrch bt mathemoiice educotion- lYashinSton, D. C.l Nalional Council ofTeaches of Mathematics' 196?.

Thonndike, E.L.The ptychology of orithmetic. New York llacmillan' 1922.

Ttruretone, L. L, A Thurstone, T. G. Faetorial shtdics of i*telligmcc. Chicago:University of Chicago Press, 1941.

Woods, S. S., Resnicls L 8., & Groen' G. J. An experimental tesu of five processmodeis for Eubtractiotr. Journol of Educotioaal Psychology, L916' 67, L7'21.

AUTHOB .ROBERT GLASER. Address: Learaing Research and Development Center,

Univenity of Pittsburgh, Pittsburgh, Pennsylvania 15260. Titls: UniversiiyProfessor; Co-Director, Learning Besearqh and Development Center. Dc'gneer.' B.S., City College of )iew York; M..\., Ph.D.' Indiana University'Speciolizotion' Researeh in learning and cogaitive psychoiogy as it relates to

education.

o/L+