GEORGE HnLocKs, JR

Synthesis of Research onTeaching Writing

Twenty years of research in writing, includingboth the composing process and teachingmethods, reveal that writing involves stop-review-start-again processes that teachersneed to recognize in their assignments.

esearch in the field of writingover the past two decades or so

R has focused on the process ofcomposing as well as on the teachingof composition My recent review ofabout 2,000 studies (Hillocks 1986a)indicates that the findings of researchon the composing process stronglysupport the findings of research onclassroom teaching. The results pro-vide some answers to an importantquestion: what types of knowledge dowriters need for effective writing?Those answers provide a guide fordeveloping more effective writing cur-riculums. This article examines re-search on the composing process, re-search in teaching composition, andthe implications of their results forcurriculum development.

The Composing ProcessResearch on the composing processindicates that writing is an enormouslycomplex task, demanding the use of atleast four types of knowledge: knowl-edge of the content to be writtenabout; procedural knowledge that en-ables the manipulation of content;knowledge of discourse structures, in-cluding the schemata underlying vari-ous types of writing (e.g., story, argu-ment), syntactic forms, and theconventions of punctuation and usage;and the procedural knowledge thatenables the production of a piece ofwriting of a particular type (Hillocks1986b). The research of Bereiter, Scar-

damalia. and their colleagues (Ber-eiter 1980, Bereiter and Scardamalia1982. Scardamalia and Bereiter 1983,Scardamalia et al. 1982) and of Flowerand Haves (1980, 1981a. 1981b), andothers strongly indicates that the pro-cesses and subprocesses of composingare hierarchically related and recur-sive. That is to sav. the act of writingany set of words in a compositionrequires a review of what has alreadchbeen written

Purposes and constraints. Availableresearch suggests the existence of thehierarchical levels in Figure 1, whichis adapted from Bereiter (1980). Toillustrate, let us assume that a directorof curriculum has been cal,cd upon bya board of education to investigate theneed for reform of the mathematicscurriculum in the school system andto present a position paper concern-ing it by a particular date. The astutecurriculum specialist will recognize.even at the outset, the presence ofseveral purposes and constraints inthis task and will begin work in light ofthose. These purposes have to do pri-marily with content. The astute direc-tor will also recognize that she will bewriting for several audiences: theboard of education, members of theadministration. mathematics teachers,and perhaps some interested parents.Such problems are the purposes andconstraints that appear to control thecomposing process and which, there-

MAY 1987

Ir

-1

fore, appear in the top trapezoid of thetriangle. Because the process is recur-sive, however, the purposes and con-straints may change as a result ofrmnking and writing at lower levels.Th- is, for example, the operation ofwsrat Bereiter calls the contentproces-sar may provide new insight into thepurposes related to content and audi-ence. These new purposes would in-fluence the levels below.

Content knowledge and processes.The second trapezoid is divided intotwo closely linked parts that appear toinfluence each othcr, (1) contentknowledge and processes and (2) dis-course knowledge and processes. Thefirst of these has to do with recallingand transforming content Research byScardamalia, Bereiter, and Goelman(1982) indicates that while childrenhave no difficulty in recalling informa-

tion when prompted by a conversa-tional partner, the simple act of recall-ing data systematically for writing,when no such partner is present, is aprocess that must be learned. Re-searchers asked youngsters in gradesfour and six to write as much as theycould on a topic. When they askedeach youngster individually to writeeven more, without suggesting what towrite, the children did write more.When contentless prompts seemed tohave exhausted what children had tosay, specific questioning revealed thatthe children had far more content thanthey had previously revealed. Bereiterand Scardamalia hypothesize that chil-dren write briefly, not for lack ofknowledge, but for lack of adequatemeans for tapping the knowledge theydo have. To use Bereiter and Scarda-malia's terminology, children need tolearn to conduct a memory search.

Anderson, Bereiter, and Smart(1980) conducted classroom experi-ments to help students learn to per-form memory searches. Over a periodof 12 writing sessions, they asked stu-dents to write a list of all single wordsthat might be important to a topicbefore they began writing. This proce-dure provided children with themeans for conducting a memorysearch, for they wrote longer cqmposi-tions at the end of training eveh whenthey were not requested to write outthe list of single words. Other tech-niques such as brainstorming, "clus-tering," and "mapping" (Buckley andBoyle 1983) may have similar effects.Such techniques may well have theeffect its proponents claim for them.

Content processing also appears toinclude collecting and transformingnew data. Our curriculum directorwould not simply collect and presentmathematics achievement raw scores.She would transform those scoresthrough various statistical proceduresSeveral studies (Fichtenau 1969; Hill-ocks 1979, 1982; Widvey 1971) provideinstruction in transforming data: moving beyond the level of objects, ac-tions, and their details to comparison(including metaphor), generalization,definition, hypothesis, and argument.Practice in these skills has a powerfulimpact on subsequent writing Weshall examine them again later underthe heading of inquiry

Discourse knowledge and processesResearch by Flower and Hayes (1981a)suggests that the processing of contentmay be closely intertwined with dis-course knowledge and processes. It isclear that knowledge of form (schema-ta) enables young children to produceelementary stories (Stein and Trabasso1982). Emig (1971) argues that formlearned in school (e.g., the five-para-graph theme) enables high school stu-dents to produce compositions quick-ly and easily. At the same time, Flowerand Hayes (1981a) report that writerswho first focus on the final form ofdiscourse and try to produce its partsshort-circuit the normal generatingprocesses and become mired in anunmanageable task that blocks writing.Current writing textbooks tend to fo-cus on learning the characteristics of agiven type of writing (e.g., description,

72 EDUCATIONAL LEADERSHIP

4!

THIRD NATIONAL THINKING SKILLS CONFERENCE

"DIMENSIONS OF THINKING"JUNE 15-17, 1987CINCINNATI, OHIO

· Demonstrations, symposia, and workshops on thinking skills and processes, criticaland creative thinking, instructional techniques, curriculum development, high risk stu-dents, computers and thinking, and more.

* Presentations by John Barell, Jonathon Baron, Sandra Black, Ron Brandt, JohnBransford, Arthur Costa, Reuven Feuerstein, Carolyn Hughes, Beau Fly Jones, HerbertKlausmeier, Bob Marzano, William Maxwell, Ray Nickerson, Richard Paul, David Per-kins, Barbara Presseisen, Stuart Rankin, Charles Wales, Debbie Walsh, Arthur Whim-bey, and others.

* For school and college teachers. administrators, curriculum developers, supervisors,principals, textbook writers, college students, business and community leaders.

* Inspired by Dimensions of Thinking, a new framework for curriculum and instruc-tion to be published as an ASCD comprehensive member book in early fall 1987.

* Sponsored by the Cincinnati Public Schools Cosponsored by ASCD, the North Cen-tral Regional Educational Laboratory (Chicago), and others.

To register, write or call Zulfi Ahmad, Conference Director, Cincinnati Public Schools,230 E. 9th St., Cincinnati, OH 45202. Telephone (513) 369-4090

narrative, argument, or "the para-graph"). At the sarne time, however,they ignore, the procedures for gener-ating such writing. The result is thatstudents may know that a "good" para-graph should have'a topic sentence(though few actually do) and supportor that arguments consist of proposi-tions supported by reasons and rele-vant data. But those same students maynot be able to generate a "good"paragraph or a developed argument. Itis one thing to identify the characteris-tics of a piece of writing but quiteanother to produce an example of thetype. Recent research at the Universityof Chicago indicates that sixth-gradersrank ordered arguments of varyingquality in the same way as adult ex-perts But they were unable to pro-duce arguments of high quality.Knowledge of discourse, then, appearsto have two dimensions: declarativeknowledge, which enables identifica-tion of characteristics, and proceduralknowledge, which enablesproduction

Gist units Research by Hayes andFlower (1980) suggests that as a resultof thinking about purposes, con-straints, content, and form, writersproduce abstract chunks of discourse,what Bereiter (1980) calls "gist" units.A gist unit is a generally circumscribedunit of content that has not been laidout in any detail but for which thewriter probably has notions of formand purpose.

Semantic, Vtrbatim, andgrapbemic'units. Research by Matsuhashi (1981)and Bereiter, Fine, and Gartshore(1979) suggests that the composing ofwritten sentences involves three fairlydistinct stages. To begin with, writersappear to have a general notion ofwhat is to be written (semantic units)and proceed to work out the specificlexical items to produce what Bereitercalls a verbatim unit, a sequence ofwords not yet recorded, but which thewriter can state upon request. Writingthese words produces graphemicunits, which are often different fromthe verbatim units announced orally.The limits of working memory (sevenwords plus or minus three) severelyrestrain the number of words that canbe planned in advance. Matsuhashi'sresearch indicates that in developing alongsentence, writers work with se-

mantic units, which become clauses orphrases. They work out the firstphrases or clauses explicitly, havingonly a general idea of what will follow.After recording the first unit, theypause to plan the remainder. The se-mantic units appear to include keys tothe kind of structure to be produced,though not the specific lexical items.That is, while the semarftic unit maynecessitate that a certain kind of rela-tionship be established (e.g., tempo-ral, cause/effect), it allows for choiceamong a variety of specific syntacticand lexical structures. At this level awriter may review alternative con-structions and choose one in light ofits appropriateness to purposes, con-tent, and form.

Verbatim units are the lexical stringsthat writers hold in mind as they re-cord what theyv wish to write. Graphe-mic units are the recorded versions ofthem. Verbatim units differ slightly1from graphemic units. According toBereiter, Fine, and Gartshore (1979),the differences between the two are inthe direction of "correctness." Thus,writers do not simply record verbatim

units but edit them during recording,sometimes omitting, adding, or chang-ing a word, sometimes making "cor-rections" in usage.

Editing. Editing appears at the verytip of the triangle indicating a stagethat follows the production of graphe-mic units: the correction of spelling orusage, the addition and deletion ofwords or phrases, the restructuring ofsyntax, and so forth.

The term revision has not beenused in the triangle, even though theresearch indicates that major changescan and do take place even at thehighest levels of planning. Flower andHaves (1981a), for example, show ayoung man change his purpose, form,and content after reconsidering hisaudience. In addition, as already indi-cated. minor revision or editing takesplace as verbatim units become gra-

*phemic units and after graphemicunits have been recorded. Revision isused here to refer to the reexamina-tion of a whole discourse or somefairly extensive part of it, after tbefirstrrsion of tWar discourse bas been

MAY 1987

completed. Revision in this sense in-volves the reexamination of the wholeproduct in light of purposes, content,and form. Bridwell's (1980) highschool seniors, for example, averaged61 revisions per piece of writing. Thevast majoriry of those revisions werecosmetic and mechanical. About 19.6percent were revisions at the sentenceor multisentence level She found norevisions at the level of the wholecomposition. No students changed ba-sic purposes, content, or form

We should not be shocked y-v thesefindings. How many of us, after pro-ducing a first draft, scrap the entiremanuscript? We may do it occasional-

Iy, but not often. What provokes largerscale revisions has not been studied inany detail, hut we can surmise thatmajor revisions come about becausethe writer recalls information forgot-ten during the initial writing, receivesnew information, or thinks of an alter-native way to organize. The sources ofsuch new information are most likelyto be some experience related to thetopic at hand (eg, reading a relatedarticle or book) or to feedback froman audience

Research indicates that the relationships among the parts of Figure 1 arenot only hierarchical but recursive

of Rraeaid' an theTad..l of Wi,

TiW d.Oh~ of sx methods of teaching writing vries widel.T1std o traditional schomol grammar (i.e., the definitio

of pin a ahntg m etc.) has no effect an raising~6I ~imtdelwrldn Ev=eryl "_ : focus of in ctio examnind•i'Antbi'iW - M.voreover, heavy empass on mechans an

-S-' umidn8 everyerror) resuts in sigifiantlosses in overall

. The pres~ent~atrion of good piece of writing as models is signifi-

ca! amie usefid thdn the study of grammar. At the same time, treat-/UWI~*w setestudy of models almost exclusiveY are considerably

i tn ote an vuilaien techniques from{ Ia cmbibW T°her phactice of building complex sentences from

_·a, a be.. show" to-be effective in a large number of experi-t d r - '4 ,aesew._ sow sentence combining, on the average,las~eiaoseIeehctIarr e afree writing a a means of enhancing- - -m 'L t mstude-t -- ', t-

/ ieS: Sult wa-,speif ..c quetions that students app .to.il 'kow oR i5 wrin ae' , a powerful effect on enhancing quality.

/.,...~l ,,,5 - o ei ve stau.. de-,.nts:: appe.ar to imternlize~~~:reia tokPI k 1*LCi~ 11* ms~lU even when they

th11e m 11 to b eo w n g f e at i g n w m t/-A Criteria infront of tm./JIM .,_q-" d, se student attntin on strategies for transmform-I.:i aW. lmFor .;p spent Might fnd and sta specific dails

r r . eda1sc vvidly' examine sets of data to developIt ~IF-" '.. ~ 1 v~y-. o-r analy.e situations that pre-

p~~~~~~ h a i~~bot those situations. onlm.,WaP~*e~ tremw-- 5 ar@ o. times- more effectivead Vover :, times are mor effective than the

* odlPieces of witingIr t . focis ask-s s-tudents, to write eely abot tevera n, nstri:ction. technique, fre writing is more

/Wun m~ ,'-"~~-_._the q!uality of stude-.t writing~~~~~ tt~~Is as focumn of instruction examned

W W - for k" rioi. ~ db· Ipiffer greytly from each/ ~'~dilimm ~ m~~~tretmnt d, _~..., us - .,m~I th~~~Yjd ~ soew in the wurrlt~I'curricmlum, Indeed, 9en-

the " ""=. C,~ " omodeluse modl.ad iniuiry ma Occasional useof ods,donot e.""-ize the study of vnodes 01clsivelY

e'S in wfch writer' lot down al of their ideas on awintegraed with other technique a

bot nsauuoiy seachI~ _eorge ll~locks, Jr.

".... children writebriefly, not for lackof knowledge, butfor lack of adequatemeans for tappingthe knowledge theydo have."

(Flower and Hayes 1980. 1981a). Thatis, writers do not set goals once andproceed to content and discourseknowledge They do not think aboutcontent and form once and proceed togist units The evidence strongly sug-gests that writers continually reconstruct goals. plans, and content (Scar-damalia et al 1982, Flower and Haves1981a) The) apparently must do sobecause the space available in workingmemory requires that larger plans beheld in long-term memory whileshort term memory focuses on thegeneration of relatively brief graphe-mic units Each reconstruction ofgoals, plans, and gist units affords op-portunities to assess and change whathas been written At the same time, thenecessity for such continuous reconstruction suggests why writing is sodifficult for so many people

Classroom InstructionThe complexity and difficulty of thecomposition pr(cess also indicates theinadequacy of current school practices Applebee (1981) found that theaverage preparation for writingamounts to about three minutes, thatmost writing assignments in schoolsask students to supply short answers ofone word to a sentence, and that themost students are likely to write is aparagraph-call it 150-200 wordsSuch conditions may account for thebrevity of writing reported by re-searchers (Emig 1971). Under suchconditions, students mav not be ableto develop their capacities to conductmemory searches, construct and re-

74 E MEDUCATIONAL LEADERSH:P

construct complex plans, transformdata, process much more than theymight produce in an extended conver-sational turn, or revise in more than amechanical fashion.

In addition, researchers find thatmost student writing is produced onlyfor teachers to report what informa-tion has been learned (Britton et al.1975). While such writing tasks may bea necessary part of education, theycannot provide practice in developinggoals and plans and in transformingcontent.

What kinds of instruction have thegreatest impact in enhancing students'abilities to deal with a wide variety ofcomposing problems? Part of my re-view mentioned at the beginning ofthis article included a meta-analysis ofexperimental treatments or interven-tions in classrooms My colleagues andI screened several hundred studiesagainst a set of criteria (see Hillocks1986a, 108-110) This process resultedin the selection of 60 well-designedstudies with 72 experimental treat-ments and their control treatments,each of which was coded for a varietyof variables

The meta-analysis was based on thetechniques developed by Glass (1978)and particularly on important develop-ments by Hedges (1981, 1982a,1982b) These combined techniquesenable us to compare the effectivenessof treatments across studies

Instructional Focuswe identified six instructional focuses,each of which was examined in threeor more treatments These includegrammar, sentence combining, modelcompositions, scales and guided revi-sion, inquiry, and free writing. In eachcase the treatment was classified ashaving a particular focus if that focusappeared to consume a major portionof classroom time. The results of theanalysis for focus of instruction appearin Figure 2 They are reported in effectsizes that answer the question, "Whatis the difference between the experi-mental groups' gains and the controlgroups' gains in studies having a par-ticular focus in common among theexperimental groups?" Effect sizes arereported in standard deviations.

Grammar. In treatments with a fo-cus on grammar, students are taught

parts of speech, parts of sentences,kinds of clauses, kinds of sentences.and so forth. The intent of such pro-grams is to help students understand"how the English language works."Many teachers assume that suchknowledge is essential to clear andeffective writing, even though linguistshave argued that such grammar doesnot adequately describe language.

In 1963, Braddock, Llovd-Jones, andSchoer, in light of the review of re-search to that time, concluded that thestudy of grammar had no effect on thequality of student writing. The studiesI (Hillocks 1986a) examined force thesame conclusions. Four studies withfive treatments are represented in Fig-ure 2. The findings indicate that resultsfor the control groups are superior tothose for groups studying grammar

In New Zealand, Ellev and his col-leagues (1976) conducted a very thor-ough, carefully designed study' thatcompared both traditional and genera-tive grammar treatments to a no-gram-mar treatment. These were taught torandomly assigned students over athree-year period with extensive test-ing at the end of even' year of instruc-

tion and at the end of a fourth vearduring which treatments were sus-pended. Ellev and colleagues foundno statistically significant differencesin writing quality and none in sub-scores for writing mechanics amongthe three groups of students at anytime over the four years. Many otherstudies reviewed but not included inthe meta-analvsis support the sameconclusion: the study of grammar doesnot contribute to growth in the qualityof student writing

Given the findings of research onprocess. we cannot expect grammarstudv to contribute much to the qualityof writing. In Figure 1, the earliestpoint in the process at which knowl-edge of grammar might enter is at theverbatim/graphemic level. Not surpris-inglv. no researchers report that writ-ers make use of even a peremptory-grammatical analysis of ,what they aredoing. Certainly, no one consciouslylas's out the grammatical scheme for asentence: "Let's see, subject with mod-ifiers, verb, object, plus iestrictiveclause. Now, how will I fill that in?" Ofcourse, writers use knowledge ofgrammar, but it is intuitive knowledge

-5sMAY 1987

Unilqe seguential curriculums that expand anddevelop your students creative and cognitivetIhnking skills JUST THIK programs have beenproven successful during six years ot testing wilthU S military children

Each IST THIlK program provides a completeyears curriculum with easy-lo-oliow dilrectionand worksheets to*· eve0lo oral Panguge a ,'d gredtrve design xpl,,* teaeh Crmprehenson and rlllcl rhInk nI* adail o, a wide range tf ePaners

igfled learnlng 1dsaDled eIl· aralliel mull- ,,ci. , requ ierenl.

Fornd lo be In commDi anrce a ' 'e rvI,ew ianel andhereore mpe, ne m pIa roon na err iljrm en! r ,rEducal,on Crde Se'0Pnq 600r40 60044 531P& ,I ;A

YI ThiNk (Ages 3 & 4) Si5 S 0

Ju st lm (Kindergar en ) S15 00iJ i T Two (ARes F. 7i 525 00Je Thli Thrn (Ages 7 i S25 0riJol TId Fur (Ages 9) S5 Or)0Jt Thn Fro IAAges 9 1 ) 525 00J Tluk Sl (Ages I 11 12) 525 00Jst ink Sl (Age, 11 ?12 111) 525 0r

* Stidrtk Think o S15 io(Ages 5-7 Graddes 1 7)

* Stnre Think Two 35 g 0(Ages 7 10 Grades 2 3 4i

*· Stbol 'tink Tnr S 5 (,(Ages 11 14 Grades 5 F 7 e3

· These are llUti-Al IIntl and o'Cmalimt CDlm print

Make Checks payable to:homas Geale Publications Inc.

P.O. Box 370540Montara, CA 94037 U.S. A

(213) 276-6394

hckol Purcklu Oers Are Accelted.

_ _ _ _ _ _ _ _ _ _ _ _

Please send me

c YOUNG THINK[] JUST THINK ONE[ JUST THINK TWOE JUST THINK THREEE JUST THINK FOUR

C JUST THINK FIVEE JUST THINK SIXE JUST THINK SENEIIC STRETCH THINK ONE0 STRETCH THINK TWO[ STETCH THINKll THEEShipping & Handling: Add 5%(S1.50 minimum)CA residents add 6% sales tax

51500S15 00$25 0062500S25 0052500S2500S25 00S35 00S35 00

31500per order

Name

Address

City

State Zip

Amount Enclosed $

that puts itself in the service of thecontent to be rendered.

If formal grammatical knowledge isused at all, it is probably used at thelevel of editing or proofreading, levelsthat can exercise little or no controlover purpose, plans, content, andstyle. In short, we should not expectknowledge of grammar to influencethe quality of writing.

Models. A second traditional andextensively used method of instruc-tion in composition is the presentationof model compositions thought to ex-emplify principles or characteristics ofgood writing. This method dates fromthe classical academies of Greece andRome whose pupils were required torecite orations from memory. The ideawas that students would incorporatethe rhetorical principles involved intotheir own thinking and would havespecific examples to guide their owncomposing of orations This belief inantiquity is not far removed from whatmodern-day cognitive psychologists

,have shown about discourse process-ing-for example, that our processingand production of stories are guidedby bare-bones outlines or schemata ofthe essential elements of stories (Steinand Trabasso 1982).

The effect size for such treatments issmall, .217 standard deviations. This isa somewhat surprising result simplybecause most writing in day-to-daysituations makes use of identifiable

"We can no longeraccept the teachingof grammar as being... conducive toimproving thequality of writing."

"The limits ofworking memory(seven words plus orminus three)severely restrain thenumber of wordsthat can be plannedin advance."

patterns or forms Further, the selec-tion and development of plans appears high in the hierarchy of thewriting process sketched in Figure 1From that alone one might expect thatteaching about discourse patternswould have greater impact on thequality of writing.

However, treatments focusing onmodels tend to emphasize identifying,naming, and perhaps evaluating theparts or features of models. The result-ing knowledge is comparable to whatcognitive psychologists call declarativeknowledge. Then the treatments callupon students to imitate the models:that is, to produce compositions thatmake use of the features studied. How-ever, these treatments do not teach theprocedures for producing a piece ex-hibiting the characteristics studied. Itis one thing to identify a good piece ofwriting and quite another to produceit, just as it is one thing to identify avirtuoso trumpet performance and im-possible for most of us to replicate it

The remaining focuses of instruc-tion are all primarily concerned withprocedural knowledge. All feature ac-tivities that involve students in procedures that seem important for produc-ing discourse. In three cases theyprovide the intervening steps betweendeclarative knowledge and final per-formance-a step that breaks downthe performance task into steps orparts and provides active practice onthe parts or subroutines to make per-formance attainable by students.

76 EDuCATiONAL LEADERSHIP

Sentence combining. Hunt (1965)and others have demonstrated thatsyntactic fluency increases with age.That is, as children move from ele-mentary school through high schooltheir use of more complex syntacticstructures increases dramatically. Anumber of researchers have foundthat direct instruction in producingmore complex syntactic structures re-sults not only in greater syntactic com-plexity, but in increased quality(O'Hare 1973, Faigley 1979, Moren-berg et al. 1978) These procedureshave been called sentence combiningbecause they present students withsets of two or more sentences, requir-ing them to combine the sentencesusing some structure stipulated in thematerials. (Some materials do not stip-ulate structures to be used )

It is important to note that fromO'Hare (1973) onward, sentence com-bining treatments have excluded di-rect instruction in grammar and gram-matical terminology

While the research clearly indicatesthat such work over the course ofseveral weeks results in higher-qualitywriting, the increased quality may not

be due simply to increased complex-ity Rather, as Scardamalia and Bereiter(1983) suggest, it may provide stu-dents with control over an organizedrepertoire of syntactic structures, con-trol that allows them to pick andchoose among a variety of alternativesntactic structures at the verbatim lev-el Such work may also have a positiveeffect in revision, enabling students tosearch more systematically for moreappropriate structures Whatever thecase, the research in Figure 2 indicatesthat sentence combining can have apowerful effect on the quality ofwriting

Scales Equally powerful are the in-structional methods that make director indirect use of scales in instruction.These methods present students withsets of criteria for judging and revisingcompositions In two studies, thescales consist of sets of four composi-tions that illustrate four quality ratings(0-3) along a particular dimensionsuch as elaboration, word choice, ororganization (Sager 1973, Coleman1982) In these studies, the teacherleads students in an evaluation of comrnpositions until students understand

"The focus ofinstruction with thegreatest power is ...inquiry.... [It]involves using setsof data in a structure... to help studentslearn strategies forusing the data intheir writing."

the discriminations illustrated by thescale. Then students receive additionalcomoositior , to rate on their own.Probably more important, when thepieces of writing are not top-rated (3),students receive a set of prompts tohelp them generate ideas for revisingthe piece t

The use of scales and criteria-guid-ed revision appears to help studentsdevelop better control over and un-derstanding of discourse knowledgeAs indicated in Figure 1, instruction ofthis type is likely to affect most aspectsof composing, influencing the selec-tion of content, the development ofplans, the development and arrange-ment of gist units, and the content andstructure of verbatim units.

Inquitry The focus of instructionwith the greatest power is what I havecalled inquiry. This approach shouldnot be construed as discovery teachingin which students are presented withproblems or tasks and set free topursue them On0 the contrary. themethod involves using sets of data in astructured fashion to help studentslearn strategies for using the data in

EDtCATIONAL LEADERSHlp

their writing. These strategies rangefrom recording and describing to gen-eralizing, presenting evidence, dis-criminating through the use of criteria,hypothesizing, and so forth.

Teachers using this focus of instruc-tion typically present students withdata (a set of objects, a drawing, a setof scenarios, information about aproblem), designate a task to be per-formed using the data, and provideguidance in performing that task Thisprocedure is repeated using similartasks but different data until studentsbecome proficient in using the strate-gies required by the task. The resultsof these studies indicate that the pro-cess of observing and writing is farmore effective in increasing the qualityof student writing than the traditionalstudy of model paragraphs that illus-trate the use of strategies (Hillocks1979, 1982)

A second example of this focus isthe approach to teaching argumentdeveloped and tested by Trovka(1974) The strategies to be learned inthis treatment include the develop-ment and support of generalizations

on one side of an argument and therecognition and refutation of oppos-ing viewpoints Students receive setsof information relevant to some prob-lem (community pollution, prison re-bellion, purchasing a fleet of taxicabs,etc.) and play the role of one of theparticipants in solving the problem.The necessity of presenting a positionto groups known to hold differentpositions requires students to developand support their own generalizationsand, more important, to recognize, andconfront the positions opposingtheirs The discussion or debateamong those positions leads to writingmore formal arguments. Study resultsindicate that students in the experi-mental program made very large gains(over one standard deviation) in con-trast to control group students whoused the traditional approach of study-ing and imitating model pieces ofwriting

The focus on inquiry, then, appearsto help writers learn strategies fortransforming available data for use inwriting Analysis of data undoubtedlhaffects a writer's plans and is very

I

likely to affect even the highest level ofFigure 1, purposes and constraints.The studies in this group strongly sug-gest that learning to write involves farmore than learning about rhetoric anddiscourse. Writers must learn strate-gies for transforming raw data.

Free writing. The final focus of in-struction examined was free writing,an approach to teaching writing thatasks students to write about whateveris of interest to them in an uninhibitedway Nearly all treatments examined

-combine free writing with sharingideas, peer feedback in small groups,redrafting, and, at some point, teacherfeedback. At the same time these treat-ments eschew the use of grammar,model compositions, criteria for judg-ins writing, and so on, as inhibitingand restrictive (Parker 1979, Ganong1975, Gauntlett 1978) They some-times include prewriting activitiessuch as brainstorming and clustering,which appear to act as aids in search-ing memory for information Such ac-tivities are often grouped together andreferred to as "the process approachto writing.'

Figure 2 indicates that the experimentaL/Control effect for these treat-ments is weak ( 16). Nonetheless, suchtreatments represent a clear advanceover traditional instruction in writingreported by Applebee (1981), instruc-tion that usually provides no prewrit-ing activity, no opportunity for revis-ing, and no feedback until after thewriting is a fait accompli. This tradi-tional instruction (which simply pro-vides an assignment) results in studentwriters who believe that only one draftLS necessary and that Whiteout is thewriter's best friend, permitting the im-mediate elimination of perceived er-rors. The resulting writing may becosmetically more appealing, but it isusually superficial and poorly organ-ized and developed

Clearly, young writers must learnthat effective writing involves a com-plex process that includes prewriting.drafting, feedback from audiences,and revising. At the same time, asFigure 2 suggests, free writing and theattendant process orientation are inad-equate strategies.

Curriculum in Writing-The available research cannot indicatethe optimal curriculum for writing.

.EDUCATIONLAL LEAi)tRttIIP

Further research and theories of dis-course and discourse learning areneeded. However, it does indicatewhat approaches to teaching writingought to be stressed. And it indicatesparts of the rationale for devisingmore effective curriculums. First of all,we know that while declarative knowl-edge is useful, the curriculum needsto stress procedural knowledge formaximum effectiveness.

We can no longer accept the teach-ing of grammar as being in any wayconducive to improving the quality ofwriting. Perhaps some will wish toinclude the study of grammar in thecurriculum for humanistic reasons.Perhaps some grammar must betaught as a means of improving me-chanics and proofreading. However,large doses of traditional school gram-mar do no more good than attentionto mechanical problems when theyarise (see Elley et al. 1976, for exam-ple). But attention to problems as they

arise does appear to have sone effect,especially in contrast to cu-iculumsthat offer little or no attentidn to me-chanics (Bennett 1976)

The procedural knowledge relatedto grammar appears to come out ofsentence combining and construction.Such instruction does not rely on pars-ing and labeling. Rather it stresses theprocedures for developing varieties ofsyntactic structures. The most impor-tant value of sentence combining ap-pears to be in helping students devel-op a repertoire of structures that theycan call upon systematically, both ingenerating sentences and revisingthem.

Writers need knowledge of dis-course structures traditionally provid-ed by the study of model pieces ofwriting. As in the case of traditionalgrammar, such study is, in itself, notenough. But experimental treatmentsthat include the study of types of writ-ing along with procedures for generat-ing them have been much moresuccessful.

Curriculums in the past (e.g., almostany composition textbook) have ig-nored the strategies that writers use totransform raw data for use in writing.The assumptions have been that stu-dents (1) have the knowledge re-quired by the writing tasks suggested,

(2) can gain access to it, and (3) cantransform it for use in the type ofwriting required. For example, manycurriculums prepare students for writ-ing a comparison/contrast composi-tion by presenting and commentingon model compositions of that type.The underlying assumption is thatknowing how others have structuredcomparison/contrasts in writing en-ables one to develop an original com-parison/contrast. However, resultsfrom research indicate that writersneed to learn strategies for recallingand transforming information. Theyneed to practice various techniquesfor searching memory and for manip-ulating what they recall for use inwriting. Techniques such as free writ-ing, brainstorming, and clustering ap-pear to be useful for helping studentsrecall information and to have someimpact on the quality of writing. Signif-icantly more powerful, however, havebeen the kinds of instruction empha-sizing what I have called the strategiesof inquiry-strategies for collectingand transforming data in various ways.

Our initial question was, What typesof knowledge do writers need foreffective writing? The traditional studyof grammar and model compositionsprovides little of the necessary knowl-edge. The most important knowledgeis procedural: general procedures ofthe composing process and specificstrategies for the production of dis-course and the transformation of datafor use in writing. The research indi-cates that when curriculums begin tofocus on such procedural knowledge,they will begin to produce more effec-tive writers.-

Refe

An asterisk after a dissertation entry indi-cates that the dissertation itself was exam-ined rather than rust the abstract fromDisertan AbsraS (DA) or DissertatonAbracs Internatbonal (DAI).

Anderson, V., C Bereiter, and D. Smart."Activation of Semantic Networks inWriting: Teaching Students How to Do ItThemselves." Paper presented at the an-ntaul meeting of AERA, 1980

"Research on thecomposing processindicates thatwriting is anenormouslycomplex task... [inwhich] processesand subprocesses... are

hierarchicallyrelated andrecursive."

Applebee, A N triting in the SecondarmScbool English and be Conten Ares.NCTE Research Report No 21. Urbana,llf: NCTE, 1981.

Bennett, N. Teabdmg Sties and Ppil Pro-gress. Cambridge, Mass: Harvard Uni-versinr Press, 1976.

Bereiter. C. "Development in Writing." InCognitire Processes In Writing, edited byL. W. Gregg and E. R Steinberg. Hills-dale, NJ.: Lawrence Erlbaum Associates,1980

Bereiter. C., J Fine, and S. Gartshore "AnExploratory Study of Microplanning inWriting." Paper presented at the annualmeeting of AERA. 1979.

Berelter, C., and M Scardamalia "FromConversation to Composition: The Roleof Instruction in a Developmental Proc-ess." In Advances in Insrucional Py-chologw', Vol. 2. edited by R GlaserHillsdale, NJ.: Lawrence Erlbaum Asso-ciates, 1982

Braddock, R. R Lloyd-Jones, and L SchoerResetarc in Wr osin Cmpion. Cham-paign, Ill.: NCTE, 1963. ED 003 374.

81MAY 1987

Bridwell, L S "Revising Strategies inTwelfth-Grade Students' TransactionalWriting" Researb in the Teaching ofEnglish 14 (1980): 197-222

Brinnon, J N., T. Burgess, N Martin, A.McLeod, and H Rosen. The Deve/lpmentof Writing AbilIties London: MacmillanEducation Ltd., 1975

Buckley, M. H, and 0. Boyle "Mappingand Composing." In Theory and Prac-tice in the Teaching of CompositionProcessing Distancing and Modeling,edited by M Myers and J. Grav Urbana,11l.: NCTE, 1983

Coleman, D R 'The Effects of Pupil Use ofa Creative Writing Scale as an Evaluativeand Instructional Tool by Primary GiftedStudents " DAI 42 (1982): 3409-A

Elley, W B., I H Barham, H. Lamb, and MWyllie "The Role of Grammar in a Sec-ondary School English Curriculum'Neut Zealand Journal of EducationalStudies 10 1 (May 1975): 26-42 Reprint-ed in Research in the Teaching of Writ-ing 10 (1976): 5-21 ED 112 410

Emig, J The Composing Processes ofTwelfth Graders Urbana. Ill: NCTE,1971 ED 058 205

Faiglev, L L. "The Influence of GenerativeRhetoric on the Svntactic Maturityv andWriting Effectiveness of College Freshmen ' Research in the Teaching of Fn-glish 13 (19'9) 197-206.

Fichtenau, R. L. The Effect of TeachingRhetorical Concepts of Invention, Ar-rangement, and Ss le on the WrittenComposition of Selected ElementarySchool Children in Grades Threethrough Six DAI 30 (1969): 1465 A.

Flower, L S, and J R Haves "The Dynam-ics of Composing: Making Plans andJuggling Constraints In Cognithie Pro-cesses in Writing, edited by L W Greggand E R. Steinberg Hillsdale, NJ Law-rence Erlbaum Associates, 1980

Flower. L S., and J R Hases 'Plans ThatGuide the Composing Process In Writ-ing The Nature, Development, andTeaching of Written Communication,Vol 2. edited by C H Frederiksen andJF Dominic, 39-58. Hillsdale, NJ Law'rence Erlbaum Associates, 1981a.

Flower, L S. andJ R Haves "The PregnantPause: An Inquiry into the Nature ofPlanning.' Research in the Teaching ofEnglish 15 (1981b): 229-243

Ganong, F L. Teaching Writing throughthe Use of a Program Based on the Workof Donald M Murray" DAI 35 (1975)4125-A.'

Gauntlen.J F Project WRITE and Its Effecton the Writing (of Iligh Sch(xil StudentsDAI 38 (19'8): '189A'

Glass, G V ''Integrating Findings: TheMeta-Analysis of Research.' In Review ofResearch in Education, Vol. 5, edited byL S Shulman Itasca, Ill. F E Peacock,19'8

". . . writers who firstfocus on the finalform of [a piece ofwriting] and try toproduce its partsshort-circuit thenormal generatingprocess and becomemired in anunmanageable taskthat blocks writing."

Ilayes, J R, and I. S Flower Identifyingthe Organization of Writing ProcessesIn Cognitive Processes in Writing, editedby L W Gregg and E R SteinbergHillsdale, NJ Lawrence Erlbaum Asso-ciates, 198()

Hedges, L v t)istribution Theory forGlass s Estimator of Effect Size and Relat-ed Estimators Journal of EducationalStatiscs 6 (1981): 107-128

1Hedges, L . 'Estimating Effect Size from aSeries of Independent Experiments'P-IVchological Bulletin 92 (1982a): 490-499

Iledges, L V 'Fitting Categorical Models toEffect Sizes from a Series of Experiments ' Journal of Fducational Statustics ' (1982b) 119-137

Hillocks, G, Jr "The Effects of Observational Activities on Student Writing. Research in the Teaching of English 13(19'9) 23 35

lillocks, G ,Jr "The Interaction of Instruc-tion, Teacher Comment, and Revision inTeaching the Composing Processs" Re-search in the Teaching of English 16(1982): 261-278

Ilillocks, G. Jr Research on Written Com-postion: New Directions for TeachingUrbana, Ill ERIC Clearinghouse onReading and Communications Skills andthe National Conference on Research inEnglish, 1986a.

Hillocks, G. Jr 'The Writer's Knowledge:Theory, Research, and Implications for

Practice" In The Teaching of WritingEight- fifth Yearbook of the National So-ciety for the Study of Education, Part 11,edited by Anthony R. Petrosky and DavidBartholomae The National Society forthe Study of Education. Distributed byThe Ulniversity of Chicago Press Chicago, Illinois, 1986h

Hlunt, K W Grammatical Structures Writ-ten at Three Grade Letves NCTE Research Report No 3 Champaign, IllNCTE, 1965 ED) 11 735

Matsuhashi, A. "Pausing and Planning TheTempo of Wrinen Discourse Produc-tion'" Research m the Teaching of En-glish 15 (1981): 113-134

Morenherg, M, D Daiker. and A Kerek"Sentence Combining at the CollegeLevel An Experimental Study" Researchm the Teaching of Fnglish 12 (1978)245-256

0 Hare. F Sentence Combiniing Improing Student Writing Without IlormalGrammar Instruction NCTE Committeeon Research Repoxrt eries, No 15 l r-bana, Ill NCTE, 19'3 ED 073 483

Parker, R P, Jr "From Sputnik to Danmouth Trends in the Teaching of Coimposition EnglishJournal 68, 6 ( i 7

9)32 3"

Sager, C ' Improving the Quality (of WrittenComposition Through, Pupil U se of Rating Scale' DAI 34 (1973): 1496-A Seealso "Sager Writing Scale" E) 091 723 '

Scardamalia, M, and C Bereiler <'The De-velopment of Evaluative, Diagnostic, and'Remedial Capabilities in Children'sComposing" In The Pschology of Written Language A Detelopmental Ap-proach, edited by'M Marnlew LondonJohn Wiley and Sons, 1983

Scardamalia, M., C Bereiter. and II Goelman "The Role of Production Factors inWriting Ability' In What Writers KnowThe Language, Process, and Structure ofWritten Discourse, edited by Martin Nys-trand, 173 210 New York AcademicPress, 1982

Stein. N I, and T Trahasso -What s in aStory: An Approach to Comprehensionand Instruction" In Adtances in Instruc-tional Psychology, edited by R. Glaser.21367 Hillsdale, NJ: Lawrence ErliaumAssociates, 1982

Troyka. I Q "A Study iof the Effect ofSimulation-Gaming on Expository PrdseCompetence of College Remedial En-glish Composition Students"' DAI 34(1974): 4092-A.*

Widvey, L I II "A Study of the rise of aProblem-Solving Approach to Composi-tion in High School English' DAI 32(1971): 2563-A_'

George HUllocks, Jr., is Professor, De-partments of Education and English, UIni-versity of Chicago, 5835 S Kimbark Ave,Chicago,"'L 60637

ED ICAIONAIo LFAI)EKStiIP

Copyright © 1987 by the Association for Supervision and Curriculum Development. All rights reserved.