Top Lang Disorders Vol. 30, No. 4, Pp. 275–287 Summarizing .

LWW/TLDTLD3004-05October 28, 20100:58Char Count 0Summarizing Expository TextsSummarizing is more than retelling; it involves analyzing information, distinguishingimportant from unimportant elements andtranslating large chunks of information intoa few short cohesive sentences. Many students, however, do not know how to summarize and appear not to understand thenature and purpose of a summary. Consequently, they use inefficient strategies whenproducing written summaries, such as writingdown everything, writing too much, not writing enough, or copying word-for-word, andthey make no attempt to synthesize the information (Jones, 2007). They may be ableto read and summarize short texts with familiar content fairly well. However, as textbecomes more difficult, increases in length,and is less coherent, and if students have insufficient background knowledge, their comprehension falters and they exhibit moredifficulty in producing coherent, meaningful summaries. Struggling readers may needinstruction and practice in summarizing paragraphs; other students may need instruction and practice in summarizing largerchunks of information. However, if the material becomes more difficult to comprehend,students who previously could summarizemulti-paragraph sections may need to returnto summarizing at the paragraph level andwork up to section summarization and summarization of whole chapters.Instruction in summarization should teachreaders to integrate ideas and generalize fromthe text information. When students learn togenerate summaries, they must identify themost important information in a text, condense the information into a very brief form,and then restate the information in their ownwords. Brown and colleagues (Brown, Campione, & Day, 1981; Brown & Day, 1983) wereamong the first educational researchers tosuggest teaching students a rule-based summary strategy. It was based on the van Dijkand Kintsch (1977) concept of macrorules(processes of deletion, generalization, and integration) that are used to operate on themacrostructure when summarizing. Brownand colleagues identified five macrorules for277summarization: Delete material that is trivial. Delete material that, although important,is redundant. Substitute a superordinate term or eventfor a list of items or actions, for example,if a text lists cats, dogs, gerbils, and parrots, one can substitute the word pets; orintegrate events or concepts by substituting a superordinate action for a list of subcomponents of that action. Select a topic sentence. If there is no topic sentence, invent yourown.The task of summarizing is cognitively complex and taxes working memory. To use thesesummarization strategies, students must analyze information at a fairly deep level whilesimultaneously manipulating the text. Beingaware of the explicit structure of the text canbe an aid to summarizing information. Themore students are aware of this explicit structure the better able they are to summarize(Armbruster, Anderson, & Ostertag, 1987).DEVELOPMENT OFSUMMARIZATION SKILLSBrown and Day (1983) investigated howstudents in fifth, seventh, tenth grades andcollege read and summarized passages thatwere designed to enable use of their summarization macrorules. All age groups usedthe deletion rules effectively. For instanceswhere superordination could be used, students’ options were to (1) delete the unit,(2) repeat it exactly, (3) use a superordinate inefficiently, or (4) use a superordinateefficiently. Fifth graders typically just deletedlists. Seventh graders tended to repeat the unitexactly or attempted an inefficient subordination. Tenth graders and college students typically used superordination effectively; theyrarely repeated units or used inefficient subordination. Use of topic sentences increasedwith age. The ability to invent explicit topicsentences to state the implicit main idea ofparagraphs is difficult and develops with age.Fifth-grade students rarely attempted to inventCopyright 2010 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

LWW/TLDTLD3004-05278October 28, 20100:58Char Count 0TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2010a sentence, and even college students did itonly in half the cases where it would be appropriate to do so.Kintsch (1990) investigated the type of organization and kinds of inferences made bysixth grade, tenth grade, and college studentsin their summarizations of texts. Students readtexts with 7 paragraphs that compared twodeveloping countries in terms of their futurepotential. The topic information was stated inthe initial paragraph and at the conclusion ofthe passage. The three major subtopics—theattributes of the countries—were never explicitly mentioned and had to be inferred bythe reader.The summaries were evaluated in twoways: (1) for the total number of text propositions and several categories of inferences, and(2) for the macrostructure level of each statement. The following types of inferences werecounted: Generalizations are inferences that reduce the number of text propositions.They are inferences about the overall gistor meaning of the text. Elaborations are inferences that are notdirectly implied by the text; they originate from the reader’s own knowledgeabout the content or related information. Reorderings are inferences that rearrange text content in an order that is different from the original text. Connectives are the words that expressbridging inferences and that function toprovide coherence between expressedideas.Each statement in the summary was ratedfor its macrostructure level. Level 1 statements consisted of topicstatements, either a label (e.g., “It’s aboutPeru and Argentina”)or a more elaboratedtopic statement (e.g., “that the two countries are compared”). Level 2 statements were composed of inferred subtopics (e.g., geography, economy, and society) on which the countrieswere being compared. Level 3 consisted of other text-basedmacropropositions (e.g., government, ed-ucation, farming) that could functionas subheadings for groups of detailedstatements. Level 4 statements were concrete details,not specifically tied to a macrostructureelement.All three groups of students produced summaries of similar length (number of propositions stated). A significant developmentaleffect was observed for all types of inferences and use of connectives. The sixth-gradestudents produced very few generalizations,elaborations, reorderings, and bridging inferences in the form of connectives. The 10thgrade students produced significantly moreof all types of inferences compared to thesixth-grade students, and the college studentsproduced significantly more generalizations,elaborations, and reorderings than the 10thgrade students. The limited use of inferred information in summaries of the younger students indicates that their summaries werecomposed largely of information selected sequentially from the original text. In contrast,older students, particularly college students,used inferences to formulate their summariesand would reorder the information to highlight the main points of the passage.All groups produced a similar number ofLevel 4 details. Use of the other macrolevelpropositions increased significantly with age,so there was a relative decrease in the prominence of text details. This indicates that olderstudents were more aware of the major ideasin the text and the overall organization of thetext.By sixth grade, students had a good understanding of what summarizing means, butthey had difficulty with identifying the important ideas and especially trying to formulate main point statements of their own.The developmental lag in macroprocessing of expository text shows up in othertasks such as study strategies, note taking,and rating the importance of ideas (Brown,Bransford, Ferrara, & Campione, 1983; Brown& Smiley, 1977), comprehension monitoring(Scardamalia & Bereiter, 1984), and essay writing (Bereiter & Scardamalia, 1987). WithoutCopyright 2010 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

LWW/TLDTLD3004-05October 28, 20100:58Char Count 0Summarizing Expository Textssupport for the summarization process, it appears that sixth-grade students have limitedability to produce true summaries.By late elementary school, students arereading an increasing number of expositorytexts, and test data clearly shows that they areexperiencing difficulty in comprehendingsuch texts (Dubravac & Dalle, 2002; Saenz &Fuchs, 2002). This led us to ask: Could summarization skills be introduced in elementaryschool as a way to promote comprehensionof expository texts, and if so, how shouldthey be introduced? What types of summaries can children in late elementary schoolproduce?EVALUATING SUMMARIZATION IN THEUPPER ELEMENTARY GRADES AFTERINSTRUCTION IN TEXTMACROSTRUCTUREProject ARC (Achievement in Reading andContent Learning) was a 3-year professionaldevelopment program with five school districts in Utah that was designed to supportteachers of fourth- and fifth-grade students inimproving reading comprehension. Its goalwas to raise students’ reading scores (particularly of those in the lowest 20%) by educating teachers in current reading strategies.To evaluate the results of the main project,a delayed intervention design was employed.In each district, classrooms were selected toserve as control and treatment classes. In thisdesign, classes that initially served as controlslater served as treatment classrooms. Teachers in the treatment classrooms received professional development days in which theywere presented with a wide range of strategies to promote fluency, vocabulary, and comprehension development. In addition, literacy specialists conducted monthly planningand evaluation meetings with the teachersand modeled teaching strategies in the classrooms. Yearly pre- and postmeasures of students’ reading skills were the primary meansof evaluating effectiveness of the training inthe main study. These assessments included279the Gates MacGinitie Reading Test (MacGinitie, MacGinitie, Maria, & Dreyer, 2000), oralreading rate (words correct per minute), subtests from the Developmental Reading Assessment (Beaver & Carter, 2002), and the UtahCriterion Reference Test.The current investigation was based on asecondary data set drawn from the largerproject. This article reports on activities during the final semester of the project with students from two school districts whose teachers learned how to develop comprehensionby teaching about expository text macrostructures. It was theorized that helping educators develop awareness of text micro- andmacrostructure patterns could guide themtoward the specific content of instructionneeded by students.Students in these treatment classrooms received training in identifying text structuresby using graphic organizers. As part of the assessment activities, students wrote summariesof short expository texts. Using these writtensamples as the data set for the current investigation, two primary questions were asked: What microstructure and macrostructuredifferences do fourth- and fifth-grade students exhibit in their written expositorytext summaries? What effect does teaching of text structure have on the microstructures andmacrostructures exhibited by students intheir written summaries of expositorytexts?MethodsParticipantsIn the ARC component reported here,fourth- and fifth-grade students from twoschool districts were participants. They weredistributed across grade level as 240 fourthgrade students (155 treatment [7 classrooms]and 85 control [5 classrooms]) and 254fifth students (168 treatment [7 classrooms]and 86 control [5 classrooms]). Pretestingindicated that the students in the treatmentand control classrooms were not significantlyCopyright 2010 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

LWW/TLDTLD3004-05280October 28, 20100:58Char Count 0TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2010Figure 1. Graphic organizer for wildfires passage. Copyright 2010 by Carol Westby, Barbara Culatta, Barbara Lawrence, and Kendra Hall-Kenyon. Shared by permission of the authors.different in their scores on the GatesMacGinitie Reading Test (MacGinitie et al.,2000).tion strategies or syntactic patterns of complex sentences.AssessmentsProceduresTeachers received two professional development training days. In addition, literacyspecialists conducted monthly planning andevaluation meetings with the teachers. Inthese sessions, teachers identified ways to address students’ literacy needs and increasetheir opportunities to practice using skillsand strategies. Particular attention was placedon providing teachers with ways to supportstudents’ comprehension of expository texts(e.g., by orchestrating discussions around relevant topics, highlighting text structure, representing the organization of texts, and identifying relevant connections among ideas).Using classroom curricular materials, teachers had students identify text structures (e.g.,descriptive, compare-contrast, cause-effect,problem-solution), note signal words for thestructures, complete graphic organizers, anddraw graphic organizers. No one strategy forteaching text structures was employed. Theemphasis was on teaching text structure asa way to promote comprehension. Teachersdid not explicitly teach writing summariza-Summarization assessments were createdthat replicated authentic curricular task demands and were tied to curricular content.Students were asked to read and then summarize three short expository passages (seepassages available as Supplemental DigitalContent 1 at http://links.lww.com/TLD/A2).All three passages were written at thefourth-grade level (according to the DaleChall readability scale) and were basedon the content of fourth-grade curriculumbooks. Two passages were cause-effect andone was compare-contrast. For two of thepassages, students identified the comparecontrast or cause-effect text structure ofexpository texts, filled in a cloze graphic representation, marked signal words on linesrepresenting cohesive links, and wrote a summary of the paragraph from their representation. Figure 1 shows the graphic organizerfor the “Wildfires”passage and Figure 2 showsthe organizer for the “Bear” passage. For the“Decline of the Tigers”passage, students wereasked to generate their own graphic organizerand then write a summary.Copyright 2010 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

LWW/TLDTLD3004-05October 28, 20100:58Char Count 0Summarizing Expository Texts281ture analysis, holistic macrostructure analysisand a six-trait rubric.Microstructure analysisA microstructure analysis was conductedusing Computerized Language Analysis(MacWhinney, 2000), which included meanlength of T-unit (independent clause plus anydependent clauses attached to it), numberof independent clauses, number and type ofconnectives (a variety of temporal and causalconnectives), and number and type of dependent clauses (adjective, adverb, noun). Thefirst author of this article marked all T-unitson the students’ writing sample. Graduatestudents entered the written sample usingthe CHAT format onto computers. The firstauthor rechecked all T-units before codingeach unit.Holistic macrostructure analysis related toworking memoryFigure 2. Graphic organizer for bear passage.Copyright 2010 by Carol Westby, Barbara Culatta,Barbara Lawrence, and Kendra Hall-Kenyon. Sharedby permission of the authors.The ARC summary tasks were simplerthan the tasks used by other researcherswho have investigated summarization because the students in the ARC study wereyounger. Therefore we used shorter passagesand provided more macro and microstructure supports. The graphic organizers displayed the overall macrostructure of the textswith the most important points. The partial details and the presence of some explicit connectors in the organizers providedmicrostructure support.Evaluation of student written summaries included both microstructure and macrostructure analyses. Persons doing the analysis andcoding of the written samples did not knowthe grade level of the sample or if the samplewas from a student in the treatment or controlcondition. Three types of analyses were conducted on the written summaries: microstruc-A holistic macrostructure scoring of thesummaries was designed using Scardamalia’s(1981) concept of working memory. Thisinvolved creating a holistic 0 to 4 pointmacrostructure rubric based on number ofelements of working memory that appearedto have been used in the production ofthe summary (see Supplemental Digital Content 2 at http://links.lww.com/TLD/A2). Scardamalia used a classification system following Pascual-Leone’s (1970) concept of mentalpower (M-power) or mental space (M-space),which involves defining different levels of taskdifficulty according to the number of mental units coordinated simultaneously. Other researchers have termed this executive processing space (Case, 1985).Baddeley and Hitch (2000) discussed thevalue of Pascual-Leone’s approach to working memory in accounting for developmental changes. As children develop, the number of concepts they can manipulate in theirworking memory increases. Expository textsrequire coordination of at least four content units that must simultaneously be integrated: (1) Statements must link to a centraltopic/theme; (2) Statements must be linkedCopyright 2010 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

LWW/TLDTLD3004-05282October 28, 20100:58Char Count 0TOPICS IN LANGUAGE DISORDERS/OCTOBER–DECEMBER 2010to one another; (3) The nature of the link between statements can be explicit (because, asa result, if. . .then) rather than general (and,then, so); and (4) statements are simultaneously linked to the central topic and to eachother according to the discourse genre. In theARC holistic rubric, the primary defining characteristics of each level in terms of workingmemory were as follows: Level 0: Child produces random statements; not related to the passage/topic. Level 1: Child appears to be operatingon short-term memory, not using workingmemory; reporting isolated, unrelated details. Level 2: Child appears to be able to holdand manipulate two concepts in working memory—chaining or centering—but not both simultaneously. In chaining,statements (ideas) are related in a temporal or causal relationship to one another.In centering, statements are related to acentral topic, but not necessarily to oneanother. Level 3: Child appears to be able to holdand manipulate three concepts in working memory—so that chaining and centering are integrated simultaneously. Level 4: Child shows ability to usesimultaneous chaining and centeringwith explicit connectives in clausesthat make the relationships betweenideas explicit (e.g., before, because,consequently, when, if. . .then, but, incontrast, similarly).One hundred of the written samples werecoded independently by both the first andthird authors. These two coders then discussed each of the samples. The two coderswere in agreement for nearly all the samples. The primary source of disagreement occurred on samples that were scored as eithera high 2 or a low 3. The coders discussedthe guidelines for the samples until they arrived at agreement for the samples on whichthey had differed. Guidelines were modifiedto improve reliability of coding summaries atmacrostructure levels 2 and 3.Six-trait rubricMembers of the research team who wereeducators expressed concern about themacrostructure scoring because they felt itwould be difficult for teachers to understandand use. They recommended the development of a scoring rubric, noting that teachersare familiar with trait rubrics used in teachingwriting or scoring students’ portfolios. Theresearchers believed that the trait rubricwould enable teachers better to understandwhat specific skills or content they needed toaddress for each student.A second reason for developing thesix-trait rubric was to provide moredifferentiation among the students’ scores.The holistic macrostructure score providesa range of only 0 to 4 points, and the majority of students in the study obtained ascore of 2. The six-trait rubric permitted arange of 0 to 24 points because studentsreceived a score of 0 to 4 in six categories:three microstructure categories (inclusionof accurate, relevant content ideas; use ofsignal words to connect ideas; and sentencestructure); and 3 macrostructure categories(stating the main idea or gist of the passage;text structure/organization marked by logically sequencing ideas; and using the graphicorganizer to guide text summary). Table 1shows this rubric. Analysis in this article wasbased on the students’ total rubric score (thesum of points on all six traits). Five membersof the research team independently scoredthe three samples from 75 students and thencompared and discussed their scoring andmade some modifications in the rubric. Twoteam members independently scored threeof the traits in the six-trait rubric (gist, textstructure, and content ideas) for all of thestudents in this part of the project. Pearsoncorrelations for the three traits were .67, .77,and .70, respectively. Percent of the raters’scores that were with 0 to 1 points were74%, 74% and 54%, respectively; percentageof scores that were within 0 to 2 points were93%, 92%, and 73%, respectively. Team members acknowledge that, in its present form,Copyright 2010 Lippincott Williams & Wilkins. Unauthorized reproduction of this arti

expository texts (Ehren, 2010; Gajria, Jiten-dra, Sood, & Sacks, 2007). Relatively little information is available on development of expository text macrostructure and the rela-tionship between expository text micro- and macrostructures. THE NATURE OF SUMMARIZATION Among a variety of st