Use of Top-Level Structure in Text: Key for Reading Comprehension of Ninth-Grade Students Author(s): Bonnie J. F. Meyer, David M. Brandt and George J. Bluth Source: Reading Research Quarterly, Vol. 16, No. 1 (1980), pp. 72-103 Published by: Wiley on behalf of the International Reading Association Stable URL: http://www.jstor.org/stable/747349 . Accessed: 22/06/2014 07:27 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact
[email protected].
.
Wiley and International Reading Association are collaborating with JSTOR to digitize, preserve and extend access to Reading Research Quarterly.
http://www.jstor.org
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
72
Use of top-level structure in text: Key for reading comprehension of ninth-grade students' BONNIEJ.F. MEYER DAVID M. BRANDT GEORGEJ. BLUTH ArizonaState University
THE STUDY
INVESTIGATES ninth-grade students' use of a reading
strategy (the structure strategy) which focuses on following the organizational structure of text in order to determine what is important to remember. Texts read were well organized with problem/solution or comparison structures; signaling varied the saliency of these structures. Signaling effects were expected to interact with mastery of the structure strategy. Regardless of signaling, good comprehenders on the Stanford Achievement Test were expected to follow the structure strategy while poor comprehenders were not. However, comprehension underachievers (vocabulary substantially above comprehension test scores) were expected to follow the structure strategy only when signaling was present. Most predictions were supported; the structure strategy appeared to be a particularly effective retrieval mneumonic. Its development with age across different discourse types is discussed. Also characterized is the approach to reading and retelling of ninthgrade students who do not employ this structure strategy.
Usagede structurede texte de niveausupedrieur, cld de comprehensionde lecturechez des edlvesde troisieme. l'usage d'une strat6gie de lecture (la strat6gie de structure) d'616vesde troisieme qui se concentre sur une structure d'organisation suivie du texte dans le but de determinerce dont il est important de se souvenir. Les textes lus 6taient bien organis6s avec des structures de comparaison ou de probl6me/solution; la signalisation a chang6 la mise en 6vidence de ces structures.On avait prevu que les effets de signalisation r6agiraientavec la maltrisede la strat6gie de structure.On avait anticip6 que les candidats comprenant bien le test d'accomplissement Stanford suivraient la strat6gie de structure tandis que le autres non, ceci sans consid6rer la signalisation. " Cependent, on avait anticipe que les candidats difficulte de comau-dessus des r6sultats substantiellement prehension (vocabulaire de test de comprehension) suivraient la strat6gie de structure seulement en presence de signalisation. La plupart des predictions etaient soutenues; la strat6gie de structure semblait &treun recours mn6CETTE ETUDE ANALYSE
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
73
monique particulierement efficace. Son d6veloppement avec l'age a travers diff6rents types de discours est discut6. On a 6galement caract6ris6 l'approche de lecture et de r6p6tition d'616vesde troisieme qui n'emploient pas cette strat6gie de structure.
El uso de estructuras de alto nivel en el texto, clave para la comprensidn de lectura por estudiantes de grado noveno la utilizacion por alumnos de grado 9 de una estrategia de lectura (estrategia estructural), que centra enseguir la estructura organizacional del texto para determinar qu6 es lo importante de recordar. Los textos leidos estaban bien organizados en terminos de problema-soluci6n o estructuras de comparaci6n; sefializaci6n vari6 la prominencia de estas estructuras. Se anticip6 que los efectos de sefializaci6n actuarian reciprocamente sobre la maestria de la estrategia estructural. A pesar de la sefializaci6n, se esperaba que individuos que mostraran buena comprensi6n en el "Stanford Achievement Test" (Test de Suficiencia Stanford) seguirian la estrategia estructural, pero no los que demostraran deficiente comprensi6n. No obstante, se esperaba que los alumnos de comprensi6n deficiente (con vocabulario superior a los resultados de examenes de comprensi6n) seguirian la estrategia estructural s61o cuando existiera sefializaci6n. La mayoria de las predicciones se confirmaron; la estrategia estructuralresult6 ser una mnem6nica especialmente efectiva. Se discute su desarrollo con avance de edad a trav6s de diferentes tipos de conversaci6n. Tambi6n se describe la t6cnica de lectura y recontamiento de los alumnos de grado 9 que no utilizan la estrategia estructural.
ESTE ESTUDIO INVESTIGA
Since most of the knowledge acquired in schools is gained via written prose, an important educational goal is to help students more efficiently acquire information from their reading. Reading programs at the upper elementary through high school levels stress the development of reading comprehension; a component of reading comprehension is skill in following the organization of a passage (Carroll, 1972; Davis, 1941). A number of available materials (e.g., Dechant, 1970; McGuire & Bumpus, 1971; Niles, 1974; Sack & Yourman, 1972) encourage readers to look for the author's organization in a text in order to increase their retention. This recommendation for reading teachers and their students to identify and use the author's organization in prose is based on common sense notions. The present study investigates the validity of this recommendation with a systematic, theoretically based procedure for identifying the author's organization of a passage and the organization used by the reader to remember the passage (Meyer, 1975a, 1975b; Meyer
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
74
* Number1, 1980 READING RESEARCH QUARTERLY
XVI/ 1
& Freedle, Note 1). This study investigates whether or not following the organization of text relates to identification of ninth-grade students as good, average, and poor in reading comprehension and to the amount of information they can remember from their reading. The prose passages which they read were well organized; saliency of this organization was varied through signaling.
Background A number of recent investigations (Beaugrande, 1980; Brown & Smiley, 1978; Frase, 1969; Kintsch & Keenan, 1973; Meyer, 1975a, 1975b, 1977, 1979; Meyer & McConkie, 1973; Smiley, Oakley, Worthen, Campione, & Brown, 1977; Taylor, 1979, 1980; Beiger & Dunn, Note 2) have examined the effects of the structure among the ideas presented in a text on what the reader learns and retains from the text. The structureof text specifies the logical connections among ideas in text as well as subordination of some ideas to others. Specifying the structure of text provides several benefits for conducting reading research. First, aspects of text structure provide significant dimensions along which passages may be evaluated as to their similarities and differences. Second, specifying the text structure allows the researcherto identify the amount and type of information which readers remember from text. Third, it allows identification of variations which arise between text and a reader's understanding of the text. For example, Meyer's prose analysis system has been successfully used to objectively identify and classify different types of top-level organizational patterns in expository text (Bartlett, 1978; Meyer, 1977, in press-a, Note 3; Meyer & Freedle, Note 1). Five basic types of patterns are problem/solution, comparison, antecedent/ consequent, description, and collection (including sequence) (Meyer, in press-a, Note 3). These top-level structures are equivalent to the major schemata used by authors to organize their texts (Anderson, 1977; D'Angelo, 1979; Kintsch & van Dijk, 1978; Rumelhart & Ortony, 1977; Meyer, Note 4). The prose analysis system can also be applied to a recall protocol written by a student after reading a passage to ascertain the superordinate schema used by the reader to recall the text. Thus, use of this system allows for comparison between the top-level structures or schemata in text and in students' written retellings of text. Without models of text structure, reading researchers would be confined to looking at task variables such as adding prequestions, without any way to specify their interaction with the text materials. In
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
75
fact, the lack of tools for specifying this text variable led most psychologiststo avoid researchwith prose and confine their investigations of learning and memory processes to the learning of nonsense syllables and words. However, in the 1970s, psychologists (Crothers, 1973; Frederiksen, 1975a;Kintsch, 1974; Meyer, 1975a, 1975b)applied work in linguistics (Fillmore, 1968; Grimes, 1975; Halliday, 1968; Simmons, 1968) to the problem of specifying the type and organization of information in text. Knowledge about the structure of text has made it possible to predict quite adequately which ideas will be recalled from text by college students and how long they will need to study the text. Specifying the structure of text permits theorizing about how readers process and understand text. Current research on prose comprehension is attempting to integrate task, text, and reader(such as prior knowledge, purpose, perspective, and cognitive style) variables to better understand the reading process (Meyer, in press-a; Meyer & Rice, in press; Reder, 1980).
Issues of Present Study The content structure, resulting from Meyer's (1975a, 1975b) approach to prose analysis, shows the text's overall organization and the interrelationships between its ideas and their relative importance from the text's perspective (see Figure 1). These are the dimensions of text which are of primary interest in the present study and are also the strengths of Meyer's system over other available approaches to prose analysis (Meyer & Rice, in press; Cofer, Scott, & Watkins, Note 5; Meyer, Note 6). The passages read in the present study were well organized in (Meyer, press-a, Note 3) with clearly identifiable top-level structures. It was hypothesized that good readers at the ninth-grade level would employ top-level structures in their recall protocols of the same type as those found in the well-organized texts, while poor readers would not. The top-level structures in these texts were expected to provide readers with a systematic, organized strategy for encoding information from text and retrieving it from memory. We expected good readers to approach text with knowledge about how texts are conventionally organized and a strategy to seek and use the top-level structure in a particulartext. Use of the top-level structure was expected to relate to the amount of information remembered; organization has been shown to be a crucial variable in learning and memory (Bower, Clark, Lesgold, & Winzenz, 1969; Kintsch, 1977, Chapter 5). In contrast to good readers, poor readers were expected to approach text without knowledge of texts'
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
* Number1, 1980 READING RESEARCH QUARTERLY
76
XVI/1
t-? _
0
V)
>
"
0
J:. cl
ol l 0
.13 z 0d
C4
0H
OH
a
•
_X
oE0 0
•
z
00
U0
4)d
o3
_
_
u
u~ Q
Ur
C
gj3
.-
"
U)
4)
os
0
C14
:-a
~n
o
__N
.zl
aa
k
vltl
4)
l
0
4)
0
"
v
-
.b~ lc(
vcl
0
U
U
ct
0`
4)
u
c~U
4.
vl*
.
v
~d
C.
U
o
x
0
(/2<
4)
HC
3h
10
I
c-, cr*t
C0
?rl~~
a~~
.dOvlv
k-o4)
~~n
3
a)
Ulo
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
o
4)
C
Use of top-level structure
MEYER,BRANDT,& BLUTH
77
organization and no effective strategy for utilizing the top-level structure of a passage. In addition, a group of readers (labeled comprehension underachievers) were identified who could plausibly fall into Flavell's (1977) production deficiency category; that is, readers who could use the top-level structure in text, but would not without explicit prodding. These readers had vocabulary test scores closer to those of the good readers, but comprehension text scores closer to those of the poor readers. They were considered above the poor readers on decoding skills and thus probably had sufficient available short-term memory capacity (Kintsch & van Dijk, 1978; Kintsch & Vipond, 1979) to interrelate ideas, but were deficient in this skill. For this group, we investigated whether or not signals provided by an author explicitly stating the top-level structure of text would facilitate use of this structure in their written retellings and the amount of information recalled. Signaling has been defined as information in text which does not add new content on a topic, but which gives emphasis to certain aspects of the semantic content or points out aspects of the structure of the content (Meyer, 1975a). The types of signaling identified included explicit statement of the structure of relations in the text structure, preview statements, summary statements, and pointer words or evaluative signaling. Signaling of the first type, the structure of relationships, was the main type manipulated in this study; no summary statements were presented. In addition, relationships at the top half of the content structure were signaled rather than subordinate relationships. The relationships which occur at the superordinatelevel in the structure are rhetorical relationships: comparison, causal, description, collection, and problem/solution (response). (For examples, refer to Figure 1.) Signaling of these relationships explicitly points them out to readers; examples of signaling for comparison relationships include "in contrast," "however," "but," and "on the other hand" and for causal relationships, they include "therefore,""as a result," "so that," "in order to," and "because." (See Appendix for examples.) If signaling is not provided by a writer, then the reader must infer an appropriate logical relationship among propositions. The primary type of signaling investigated in this study is parallel to Halliday and Hasan's (1976) conjunction cohesion; however, it was examined at the macroproposition level where it interrelates groups of sentences and paragraphsrather than clauses and sentences at the microproposition level of text structure.
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
78
READING RESEARCH QUARTERLY *
Number 1, 1980
XVI/1I
Signaling was expected to affect the groups of ninth-grade students in different ways due to differences in their dominant reading strategies for determining what is important to remember from text. Figure 2 depicts reader strategies and their interaction with signaling. The dominant strategy of good comprehenders on the prose learning task was hypothesized to be the structure strategy, the strategy utilized by most college students (Meyer, 1979, Note 7). These students were expected to follow the text's superordinate relational structure (see Figure 1) and focus on the text's message and how it relates to supporting major details. Processing activities hypothesized for the structure strategy are similar to explanations given for the levels effect (information superordinate in a hierarchical text structure remembered better than low level information) by Anderson (1976) and Kintsch and van Dijk (1978); only primary emphasis is placed on a search for interrelationships among chunks of complex propositions in text. These processing activities for the structurestrategy focus on a search for major text-based relationships among propositions. That is, there is a search for relationships which can subsume all or large chunks of this information and tie it into a summarized comprehensible whole. Readers employing the structure strategy are hypothesized to approach text looking for patterns which will tie together the propositions contained in the text; in addition, they search for the author's primary thesis which will provide the content to be bound by these patterns or schemata. Then, they search for relationships among this primary thesis and supporting details. For example, when reading the supertanker text (see Appendix and Figure 1), readersemploying this strategy recognize in the first sentence that the propositions may fit their problem schema. Each new proposition is related back to the problem, making the problem of oil spills from supertankerscontinually selected for retention in the short-term memory buffer for interpreting the new propositions. Previous knowledge about problems keeps the readers searching for causal relationships among descriptiveinformationabout the problem, such as why it is a problem and what caused it. Also, prior experience with problems leads readers to anticipate and search for solutions, solutions which must satisfy most of the previously stated causes of the problem. Thus, the problem and its causes are retrieved continually from long-term storage to the short-term memory buffer for relating to the subsequent propositions in the passage. This additional processing of these superordinate propositions and their interrelationships increases the depth with which they are processed (Craik & Lockhart, 1972) and the ease with which they can be retrieved.
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
& BLUTH MEYER, BRANDT,
Use of top-level structure
79
= ,,.o
•
,.d
U.
_ J
I
,I
o
~~~
U
0a
o,
z• ~.
cl
k
r .,
'
,
?-
"•0
';
a
cnUHCl) ! ~Cl)
0
~(e
•
...
Ht
c
"
H
\)[L-
•
bo
o Cl Ar
0(Cf
In
?n .,"
L
.••s [-,'
Cl) tt
*U 0
T
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
80
READING RESEARCH QUARTERLY *
Number 1, 1980
XVI/1I
The top-level structure and major interrelationships are also employed by these students (structurestrategy I in Figure 2) to guide retrievaland production of the recall protocol; it is hypothesized to a top-down retrieval search guided by the structure of relationships. These students are assumed to construct memory representations of text propositions which are similar in terms of their hierarchical relationships to the content structuredepicted in Figure 1;when recalling the text, they begin their retrievalsearch with the top-level structureand systematically work from the superordinate relationships and content downward. As seen in Figure 2, under predicted free recall outcome, the organization of a protocol written by a student employing the structure strategy matches the top-level structure of the text. In contrast to good comprehenders, poor comprehenders were hypothesized to follow the default/list strategy (see strategy II in Figure 2). While the structure strategy is a systematic plan for processing text, the default strategy is not. The reader has no focus and simply tries to remember something from the text. As seen in Figure 2, the protocol organization predicted for readers with the default strategy is different from the top-level structure of the hierarchically organized text; it is a list-like 'collection of descriptions about the passage topic with no attempt to interrelate them. An alternative strategy to the structure and default strategies is that of a reader who provides a conceptual structure of relationships for understanding the text different from that given in a well-organized text (Meyer & Freedle, Note 1). Although this strategy is of theoretical and practical interest, it was not expected to emerge in this study since it has been exhibited rarely by college students in prose learning studies with well-organized text (Meyer, 1979). Different effects of signaling were expected for the two dominant reader strategies. Good comprehenders were expected to follow the solid lines under structurestrategy in Figure 2 regardless of the presence of signaling. The basis for this prediction was data on university students who were also hypothesized to follow the structure strategy. Signaling in studies with university students has had no effect on amount of recall (Marshall & Glock, 1978-1979;Meyer, 1975a, 1979, in press-a), protocol organization (Meyer, 1979), type of information in recall (Meyer, 1979), ability to answer questions tapping signaled structural relationships (Meyer, Note 7), nor reading speed (Meyer, 1975a;Britton, Meyer, Glynn, & Penland, Note 8). However, signaling does appear to affect cognitive capacity as measured by reaction time to a secondary task (reactions to random clicks while reading text with and without
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
81
signaling, Britton, Meyer, Glynn, & Penland, Note 8). Good comprehenders at the ninth-grade level were expected to have no difficulties figuring out the structure without signaling as indicated by the solid line in Figure 2. The dominant strategy of poor comprehenders and underachievers in reading comprehension was expected to be the default strategy. The poor comprehenders were expected to follow the dashed path in Figure 2. That is, signaling was expected to have no effect because the structure strategy was not an available strategy for these readers. However, the reading comprehension underachievers, hypothesized to fall within the productive deficiency category (Flavell, 1977) on the structure strategy, were expected to benefit from signaling. For this group, signaling was expected to enable their use of the organized structure strategy for learning and remembering text which they ordinarily could not impose on their own; as seen in Figure 2 (yes responses under default strategy), signaling was expected to affect protocol organization and total recall. Marshall and Glock (1978-1979) found improved recall for junior college students with the presence of signaled text relationships.
Method Subjects A representative sample of 102 ninth-grade students attending a junior high school in Mesa, Arizona participated in the study. The students were divided into groups of good, average, and poor comprehenders on the basis of their performance on the reading comprehension scale of the Stanford Achievement Test and a district reading achievement test. The mean grade equivalent on the comprehension scale of the Stanford Achievement Test for ninth-grade students in this school was within one month of the National mean grade equivalent for grade nine students. In addition to test performance, reading and English teachers rated each student as good, average, or poor in reading performance; only those students whose standardized test scores and reading performance appraisal matched in terms of the good, average, and poor categories participated in the study. The median percentile scores on comprehension from the Stanford Achievement Test administered by the district two months prior to our study were 84, 58, and 32 for the good, average, and poor groups, respectively. Signaling of the author's top-level structure was hypothesized to be particularly helpful for students with adequate word knowledge
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
82
READINGRESEARCH QUARTERLY* Number 1, 1980
XVI/1I
and word attack skills, but poorer reading comprehension (i.e., underachievers). Twenty-six of the 102 students were identified as belonging to this group. The stanine score of each of these students on the reading comprehension scale was at least one stanine below their stanine score on the vocabulary scale of the Stanford Achievement Test. In addition, in order for students to be selected for this subgroup, their stanine scores on the vocabulary scale had to be four or greater. The median percentile score on the reading comprehension scale for this group was 43, while their median percentile score on the vocabulary scale was 60.
Materials Two well-organized passages of expository text with clearly identifiable top-level structures and appropriate reading levels for ninthgrade students were selected for the study. Passages with comparison and problem/solution top-level structures were selected for study; these structures are found frequently in the text of school materials. The toplevel structure for each passage was identified by analyzing the passage with the prose analysis procedure described by Meyer (1975a, 1975b). The comparison (adversative) passage dealt with dehydration and compared two contrasting views; versions of this passage and its structure can be found in Meyer (in press-b) and Meyer and Freedle (Note 1). The problem/solution passage dealt with supertankersand was a modified version of an article with the same title appearing in Read magazine (1975), a magazine for reading and English at the junior high level. The text is reproduced in the Appendix and its structurecan be seen in Figure 1. A with-signaling and without-signaling (Meyer, 1975a) version of each passage was written. In the with-signaling version, the top-level structurewas explicitly stated; in the without-signaling version, it was not. For example, the with-signaling version of the supertanker passage began with "A problem of vital concern is the prevention of oil spills from supertankers," while the without-signaling version did not include the words "problem"and "solution" and began with "Prevention is needed of oil spills from supertankers." In addition, in the withsignaling version, the three-fold solution was explicitly pointed out to the reader (see Appendix, underlined words of signaling in the second paragraphof the supertankerpassage). In the without-signaling version, the underlined words in the Appendix were deleted so that the relationships of problem/collection of solutions and comparison to an incorrect solution were no longer explicit and highlighted. For example,
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER, BRANDT, & BLUTH
83
the without-signaling version read "An immediate halt of the use of tankers on the ocean is not possible since about 80 percent of the world's oil supply is carried by supertankers. Officers of supertankers must get top training in how to run and maneuver their ships. Tankers should be built with several propellers for extra control and backup boilers for emergency power. Ground control stations should be...." There were 242 words in the with-signaling version including the title, "Supertankers,"and 216 words in the without-signaling version which excluded the title. The same procedure was followed for the passage on the loss of body water; however, the comparison structure required few words of signaling. The with-signaling version differed from the without-signaling version primarily by its inclusion of the phrase, "In contrast to the action taken by coaches," and the title, "Views Class on Loss of Body Water." The with-signaling version contained 187 words, while the withoutsignaling version contained 169 words. A recognition test consisted of sentences that represented verbatim statements, paraphrases, inferences, topically related intrusions, and statements unrelated to the topic. The paraphrases were writtenusing Anderson's(1972) proceduresfor writingcomprehensiontest questions. The inference sentences represented inferences that we judged necessary to understand the passages. The intrusions dealt with the same topic as the information in the passage but were neither discussed nor implied. The unrelated sentences addressed completely different topics and were included to see if the students were attending to the task. Procedure The students were stratified on comprehension level and sex, and then a stratified random assignment procedure was used to assign them to with- or without-signaling versions of the passages and order of presentation of the two passages. The study was conducted during the regular English period with students from all experimental conditions represented in each classroom. Each student received a booklet consisting of two passages, that had been assigned to the student through the random procedure prior to the experimental session, and lined pages for recall. Each student read one version of the dehydration passage and one version of the supertanker passage. After reading a passage and placing it out of sight in a large envelope, the students wrote down all they could recall from the passage using words remembered from the passage or their own words. One week later, the students were again asked to write down all they could remember from the passage. After recalling
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
84
READINGRESEARCH QUARTERLY * Number 1, 1980
XVI/1I
each passage a week after reading it, each student then took a recognition test for each passage. Students were asked if they recognized the statements as sentences from the passages read on the prior week and to give a confidence rating on this task. Thus, a student could rate a sentence as "yes, certain (5)," "yes, think so (4)," "yes, guessing (3)," "no, guessing (2)," "no, think so (1)," or "no, certain (0)." Scoring The recall protocols were scored using Meyer's (1975a) scoring procedure; reliability coefficients among three independent scorers were above .95. In addition to the total number of idea units recalled from the passage (total recall score), an index was given for the degree to which the student recalled the most central ideas of the passage. Idea units recalled corresponding to the top-level rhetorical predicates and related content at the highest level of the content structure of the passages were credited two points and major idea units at the second and third level were credited one point; none of the other lower levels (seven levels for both passages) were credited. This score was called the central idea score. A weighted recall score consisted of the sum of a student's total recall score and central idea score. The content structures of the passages were divided into thirds to examine any differences in processing different types of information between students who did and did not utilize the top-level structure of text. Levels one and two in the content structures of the passages were labeled the message. The supporting major details were located in levels three and four in the content structureand the minor details were located in levels five and lower. These three types of information are shown on the supertanker passage in the Appendix. The top-level structure of each protocol was analyzed to determine whether or not its top-level rhetorical structure was the same as that used by the author of the text. An example of the scoring system for the supertanker passage will be given to clarify the procedure. If the protocol was organized into two related clusters of ideas, one related to problems of supertankers and the other related to solutions for these problems, the protocol's content structure would be organized with a problem/solution top-level structure and classified as using the same organization as the author. For those protocols organized with the same structure as the author's, seven to ten points could be assigned. Ten points were given if the previous requirements were met and the student used the words "problem" and "solution" in his or her protocol. Nine points were given if only "solution" was explicitly stated; eight points
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER, BRANDT, & BLUTH
85
were earned if only "problem"was stated and seven points were given if the above requirements were met, but neither the word "problem" nor "solution" appeared. Five points were given for a protocol which stated explicitly or implicitly in one sentence that there was a problem and solution, but simply listed ideas from the passage without organizing the problems and solutions in related clusters. Three points were given if some other conceptually related hierarchical structure was used besides that found in the original text (list-like collections of descriptions were excluded from this category); as expected very few (4 of the 408 protocols) fit into this classification. Two points were given for a protocol where no explicit nor implicit mention was given to problems of supertankers or proposed solutions and the ideas were organized into a collection of descriptions about supertankers. One point was assigned to a protocol that implied no problems nor solutions and presented a random listing of ideas from the passage which were not collected under supertankers or any one topic. Protocols scored with five to one points were classified as not using the author's organization. Reliability coefficients were .95, .97, and .98 among the three independent scorers for this measure on 18 randomly selected protocols.
Results Immediate and Delayed Recall: Analysis of Variance Separate five-factor analyses of variance were conducted for each of the passages since they differed on structure type, content, amount of signaling, and number of idea units. The five factors were signaling, comprehension level, time of recall, sex, and order of passage presentation. For the three dependent variables of total recall, central idea score, and weighted recall for both passages, the main effects of comprehension level and time of recall were significant at the .0001 level, but neither the effects of signaling, sex, nor order were statistically significant. More specifically, for both passages and all dependent measures, the free recall data clearly supported the groupings of students into good, average, and poor comprehension groups. For the supertanker passage, percentages of the total number of idea units recalled were 34%, 25%, and 15% summed over time of recall for the good, average, and poor readers, respectively, F(2,78) = 18.02. Percentages for the central idea score for this passage were 51%, 38%, and 17% for the three comprehension groups, F(2,78) = 13.07. Percentages recalled for the total recall score from the dehydration passage were 41%, 30%, and 17%,F(2,77) = 22.06; for the central idea
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
86
* READINGRESEARCH QUARTERLY
Number1, 1980
XVI/ 1
score, they were 70%, 46%, and 27%, F(2,77) = 17.58, for the good, average, and poor groups of comprehenders, respectively. In addition, for both passages, significantly more information was recalled immediately from the passages than one week later (supertanker immediate total recall M = 31%, delay M = 17%, F(1,78) = 198.09; dehydration immediate M = 39%,delay M = 18%,F(1,77) = 240.78). For the supertanker passage, there was a statistically significant time x signaling interaction which will be discussed later. Other significant interactions were not consistent among dependent variables, passages, nor logically interpretable, and will not be discussed further.2 Chi square analyses were utilized to determine whether or not there were differences in use of the text's top-level structure by good and poor comprehenders. As shown in Table 1, the majority of good comprehenders used the same type of top-level structure for organizing their recalls as the author of the text, while most low comprehenders did not. In recalls where the text's top-level structurewas not employed, most (over 99%) were organized into collections of descriptions, the structure expected for students employing the default strategy (Figure 2). Table1 Numberof good, average,and poor comprehenderswho used and did not use the top-leveltext structurein theirimmediateanddelayed recallsof two passages Supertankers
Passage
Recall's TopLevel Structure Comprehension Level from Achievement Test
Same as Author
Different from Author
Dehydration Recall's TopLevel Structure
Same as Author
Different from Author
Immediate Free Recall Good Average Poor
Good Average Poor
22 10 18 17 26 9 72= 12.65, p < .002 Delayed Free Recall 13 19 22 13 4 30 2 = 11.76, p < .003
25 15 4 = 2
30.32p<
7 20 31 .0001
15 9 2
17 26 32 =2 14.91,p < .001
Table 2 presents the mean total recall scores and standard deviations obtained by good, average, and poor comprehenders who used and did not use the text's top-level structure. For all comprehension
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
87
groups, recall conditions, and passages read, the number of ideas recalled was much greater for students using top-level structures like that of the text. Table2 Mean total recall scores and standarddeviationsof good, average, and poor ninth-gradecomprehenderswho used and did not use the text's top-levelstructurein their immediateand delayedfree recalls Passage
Comprehension Level from Achievement Test
Supertankers
Dehydration
Recall's TopLevel Structure
Recall's TopLevel Structure
Same as Author
Different from Author
Same as Author
Different from Author
57.5 13.5 52.9 16.6 55.0 16.8
44.4 21.5 34.2 11.6 22.7 11.1
43.8 10.5 39.1 10.8 40.5 6.4
18.2 13.7 11.8 9.1 7.2 8.3
Immediate Free Recall Good Average Poor
M SD M SD M SD
60.0 19.4 49.7 13.9 41.8 10.4
43.9 13.4 33.2 15.7 22.5 12.3
Delayed Free Recall Good Average Poor
M SD M SD M SD
41.5 18.0 36.5 15.6 24.8 10.8
26.0 18.6 16.1 11.2 9.8 9.5
In addition, students who utilized the text's top-level structure in their written retellings recalled significantly more message units, major details, and minor details for both passages at both testing times (p < .05 for all comparisons). Figures 3 and 4 depict these data for the two passages. Use of the text's top-level structure was related particularly to recall of the message; for both passages there was a significant information type x use of structure interaction (p< .0001). It is plausible that students who used the top-level structureof the text on the immediate recall task but not on the delayed task would surpass the performance of those who used it consistently due to the fact that the former group integrated the information with different prior knowledge structures. However, the data from both passages did not support this view. Students who used the top-level structureimmediately but not a week later performed very similarly on the immediate free recall test to the group who consistently used the top-level structure; in
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
READING RESEARCH QUARTERLY *
88
Number 1, 1980
XVI/1
Figure 3. The effects of the use of the top-level structure immediately and a week after reading the dehydration passage on the type of information recalled. .8 USE OF TOP-LEVEL STRUCTURE
SN .7
'
N = 22 Use Immediate and Delay = 22 Use Immediate, - Not Delay
N = 53 Neveruse I-111[[[IfIl-1[[[141
Immediate, Nor Delay
"- .5 "Immediate
.,.4
o
Immediate
o
0 o
.3 Delay Immediate
.2-
Delay
Delay I
I
Message
Major Details Type of Information
I Minor Details
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
89
MEYER,BRANDT,& BLUTH
Figure 4.
The effects of the use of the top-level structure immediately and a week after reading the supertanker passage on the type of information recalled. .8 USE OF TOP-LEVEL STRUCTURE N = 27 Use and N = 22 Use Not
--------.7
Immediate Delay Immediate Delay
N= 43 NeverUse WII4 44444.IHII4I4I Immediate Nor Delay
.6
"
. .
o 4
a .3 S Immediate
SImmediate .2
--
Delay .1
•Immediate
Delay Delay 0
I Message
Major Details Type of Information
I Minor Details
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
90
READING RESEARCH QUARTERLY * Number 1, 1980
XVI/1
contrast, on the delayed test, these same students performed like the group that never used this strategy (see Figures 3 and 4). This finding is supportive of the argument that use of the top-level structure is particularly important for facilitating a systematic top-down retrieval strategy. Immediate and Delayed Recall: Multiple Regression Analysis Multiple regression and stepwise regression analyses were used to examine the predictive power of use of the text's top-level structure as measured by the ten-point scale described in the scoring section. Other predictors, entered into the equations, were vocabulary scores on the Stanford Achievement Test, reading comprehension scores on the achievement test, sex, and signaling. They were used to predict two recall measures. The first was recall of major details, a measure devoid of the potential bias which could occur between message units and use of structure due to overlap of the problem/solution and comparison relationships. For all passages and all recall conditions, use of the author's top-level structure was the best predictor for recall of major details (see Table 3). For the supertankerpassage on the immediate recall task, use of the author's top-level structure predicted 12%of the variance in recall; vocabulary added an additional 2% with no substantial contribution from the other predictors. On the delayed recall task for this passage, use of the author's structurein the delayed recalls predicted 48% of the variance with comprehension test scores as the next best predictor increasing the predictive power 2%. For the dehydration passage on the immediate free recall test, use of the author's top-level structurepredicted 16%of the variance in major detail recall with sex adding an additional 2%. For this passage a week later, use of the author's structure predicted 60% of the variance in recall and the next best predictor, sex, could add only 1%. These data are congruent with the argument that use of the text's top-level structure is a particularly effective retrieval strategy for relatively long retention intervals. The second set of analyses examined the relationships between a student's use of the text's top-level structure on one passage and the magnitude of recall from another passage. More specifically, use of the author's organization on the first passage a subject read was used to predict total recall on the second passage read. Again, the other predictors were vocabulary, comprehension, sex, and signaling. Use of the author's organization on the dehydration passage (comparison toplevel structure)was the best predictor of total recall from the supertanker
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
91
Table3 Regressioncoefficientsfor recallof majordetails Predictor
Multiple R
R2
R2 Change
Simple R
Beta
F
Supertanker: Immediate Free Recall Structure' Vocabulary Signaling Sex Comprehension
.35 .38 .41 .42 .42
.12 .14 .17 .18 .18
.12 .02 .02 .01 .00
.35 .25 -.20 .12 .25
.244 .097 -.177 .100 .086
5.20 .35 3.06 .98 .28
.628 .260 .079 -.101
67.19 3.62 1.21 .54
.396 -.125 -.109 .129 -.100
12.76 1.82 1.38 .53 .31
.744 .091 .247 -.191 -.048
103.34 1.97 4.12 2.33 .55
Supertanker: Delayed Free Recall Structure Comprehension Sex Vocabulary
.69 .71 .72 .72
.48 .51 .51 .52
.48 .03 .01 .00
.69 .40 .15 .34
Dehydration: Immediate Free Recall Structure Sex Signaling Vocabulary Comprehension
.40 .43 .44 .44 .44
.16 .18 .19 .19 .20
.16 .02 .01 .00 .00
.40 -.13 -.12 .25 .22
Dehydration: Delayed Free Recall Structure Sex Comprehension Vocabulary Signaling
.77 .77 .78 .78 .79
.59 .60 .61 .62 .62
.59 .01 .01 .01 .00
.77 .19 .41 .38 .02
'Useof author'stop-levelstructure.
passage (problem/solution top-level structure)for both recall conditions (see Table 4). On the immediate recall task, use of structure predicted 28%of the variance, with comprehension, the next best predictor, adding 7%. On the delayed recall, use of structurepredicted 37%of the variance, with sex adding an additional 2%. In contrast, vocabulary was the best predictor when the supertanker passage was read first and use of its toplevel structure was employed to predict recall on the subsequently read dehydration passage. On the immediate recall task, vocabulary predicted 20% of the variance with use of author's structure, the next best predictor, contributing an additional 16%. On the delayed recall task, vocabulary again predicted 20% of the variance in total recall of the dehydration passage with use of structure adding 7%. As can be noted from Table 1, a greater percentage of the recalls were organized with the author's top-level structure on the topic of supertankers (42%) than on dehydration (34%). Different ease for utilization of the top-level structure of one passage over the other may be due to differences in the
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
* Number 1, 1980 READINGRESEARCH QUARTERLY
92
XVI/1
Table4 Multipleregressioncoefficientsfor use of author'sorganizationon previouspassageas a predictorof subsequenttotalrecallof a different passage Predictor
Multiple R
R2
R2 Change
Simple R
Beta
F
-.093 .047 .324 .056 .345
.61 .15 4.79 .06 2.60
-.08 .15 .54 .21 -.07
.48 1.63 16.51 .99 .10
-.077 -.096 -.290 -.514 -.056
.43 .62 5.33 4.58 .16
.054 .059 .436 -.074 .302
.18 .23 3.03 .08 5.11
Supertanker: Immediate Free Recall (N = 48) Signaling Sex Structure' Comprehension Vocabulary
.13 .13 .55 .61 .64
.02 .02 .30 .37 .41
.02 .00 .28 .07 .04
-.13 -.02 .54 .53 .56
Supertanker: Delayed Free Recall (N = 48) Signaling Sex Structure Comprehension Vocabulary
.05 .17 .63 .64 .65
.00 .03 .39 .41 .42
.00 .02 .37 .02 .00
-.05 .15 .61 .39 .34
Dehydration: Immediate Free Recall (N = 53) Signaling Sex Structure Vocabulary Comprehension
.12 .13 .42 .61 .61
.01 .02 .18 .18 .38
.01 .01 .16 .20 .00
.12 .03 .40 .54 .47
Dehydration: Delayed Free Recall Signaling Sex Vocabulary Comprehension Structure
.13 .19 .49 .49 .56
.02 .04 .24 .24 .32
.02 .02 .20 .00 .07
.13 .16 .46 .40 .43
'Use of text'stop-levelstructure(measuredby ten-pointscale)on priorrecalltask with differenttext topic.
magnitude of signaling, content factors, such as familiarity, or structural factors, such as developmental differences in acquisition of the structure strategy with different discourse types. Regardless, greaterease in the use of top-level structureon the supertankerpassage probably contributed to its decreased predictive power for recall scores from the dehydration passage. Delayed Recognition Test Use of the top-level text structure also related to the recognition data. As seen in Table 5, students using the text's structure were less likely to claim recognition of intrusions than students who did not use the text's structurefor their delayed free recall protocol (F(1,95) = 9.33, p <.003, supertankerpassage; F(1,95) = 4.05, p < .05, dehydration passage). As shown in Figures 3 and 4, students who use the text's top-
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
MEYER,BRANDT,& BLUTH
Use of top-level structure
0
00
ef.)
cO
"0 c 0
co
00
z
S0
0
%4
4 --0
G
H
cn
4)
co7
0 cdn o~ ~ L~< Cd 'EAl 0 HzL0
o
o 1r
E~
-
--0
C.
=
z
"O
0
c~G) In.7 11
H
co
~" c1
v
0G) Iz Q
K
-0 ce.
000 enE
wl
4-A
abD
00
00*-*
o
co)G0
P
00
cto CdV Orr
CoQ)
5
'CG CC
)
V-
) cE
coQ~ CA Cdc CA \cd cIO
>1 0 .-'-,
W)00 Q CdO ONW) & CISO
' CA i
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
93
94
READINGRESEARCH QUARTERLY* Number 1, 1980
XVI/I
level structure remembered more of the passage's message at the delayed testing. Thus, they had a much better criterion on the delayed recognition test for judging whether or not an item on the same topic as the text had actually been stated in the text. However, Table 5 displays no significant relationship between use of structure and recognition of verbatim statements. On the basis of past research (Bransford & Franks, 1971; Frederiksen, 1975b;Kintsch, 1974), differences were not expected among the recognition scores of verbatim statements, paraphrases, and inferences one week after reading the passages; recognition scores on these sentences were expected to be higher than recognition scores for intrusions and unrelated sentences. Only the latter prediction was supported by the data. For both passages on the delayed recognition test, verbatim sentences were recognized better than any other type of sentences (p < .001). These data, summarized in Table 5, are more supportive of Hayes-Roth and Hayes-Roth (1977), who maintain that the studied wordings of passages are functional components of their memory representations. There were consistent differences in the recognition of inference sentences by the comprehension groups on the two passages. Good and average comprehenders were less confident in their recognition of inferences as sentences from the passages than poor comprehenders, F(2,95) = 4.19, p < .02, supertanker passage; F(2,95) = 3.00, p < .05, dehydration passage. There were no consistent differences among the comprehension groups on the verbatim statements or paraphrases for the two passages. Analysis of Signaling Effects Effects of signaling were examined through analyses of variance of total recall scores and structure scores. No signaling effects were expected for good and poor comprehenders. Signaling effects were expected for underachieversin reading comprehension; such effects were also likely for the average comprehenders since the majority of underachievers fell into this group. Table 6 presents the findings. For the dehydration passage, signaling had no effect on the performance of good, average, and poor comprehenders. The supertanker passage contained one-third more signaling; for this passage, use of the author's top-level structure and recall scores on the immediate recall task appear to be influenced by signaling. Use of the author's structure as measured by the ten-point scale, F(1,78) = 5.6, p < .02, and
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
95
MEYER,BRANDT,& BLUTH
Table6 Meanrecallof idea unitsfromtext withand withoutsignalsfor good, average,poor, and underachievingstudents Passage Comprehension Level from Achievement Test
Supertankers WithWithoutSignals Signals
Dehydration WithWithoutSignals Signals
Immediate Free Recall Good
M 53.4 SD 14.9 N 16
56.5 23.0 16
54.1 13.8 16
55.2 18.6 16
Average
M 46.2 SD 19.7 N 20
35.7 9.6 15
42.4 17.8 20
41.9 15.5 15
Poor
M 31.5 SD 13.4 N 14
24.7 15.2 20
29.2 12.6 14
25.1 17.6 20
Underachievers
M 48.5 SD 13.2 N 13
33.0 14.4 13
43.8 14.2 13
39.9 20.3 13
Good
M 32.4 SD 19.7 N 16
38.0 19.6 16
27.6 18.8 16
32.8 19.6 16
Average
M 23.4 SD 17.7 N 20
24.1 14.5 15
19.6 16.2 20
17.9 14.5 15
Poor
M 10.9 SD 9.4 N 14
13.1 11.6 20
6.1 9.3 14
11.6 12.4 20
Underachievers
M 23.3 SD 16.9 N 13
22.3 15 13
18.8 16.9 13
20.9 15 13
Delayed Free Recall
total recall, F(1,78) = 7.05, p < .01, were greater for the with-signaling condition than the without-signaling condition on the immediate recall task. Mean differences in the total recall scores between the withsignaling and without-signaling conditions for the supertanker passage on the immediate recall test were -3.06, 10.48, and 6.55 for the good, average, and poor comprehenders, respectively. However, this aid of signaling was not maintained a week later on the delayed test. The interaction between time, comprehension group, and signaling was not statistically significant. For the 26 students identified as underachievers in reading comprehension, the presence of signaling in the supertanker passage
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
96
* Number 1, 1980 READINGRESEARCH QUARTERLY
XVI/I
increased their recall on the immediate free recall task but not one week later. For total recall scores, the presence of signaling approached significant levels, F(1,24) = 2.37, p <.14; there was a significant time x signaling interaction, F(1,24) - 9.33,p <.005. As can be seen in Table 6, the mean total recall score for the with-signaling condition immediately after reading the passage was 48, while it was 33 for the without-signaling condition; on the delayed free recall, differences were minimal. The model's (Figure 2) predictions were aimed at the immediate free recall task; different findings at the two recall times suggest that the structure strategy may be particularly important as a retrieval strategy. Increased use of the text's top-level structure with signaling was in the predicted direction, but not statistically significant, time x signaling interaction, F(1,24) = 2.03, p < .17. Dichotomizing structure use on the immediate task, nine students who read with-signaling versions used the author's structure, while four did not. Without-signaling, five used the structure and eight did not, X2 = 1.39, p < .24, n.s. For the dehydration passage, signaling had no effect, F(1,24) = .02, p < .89. In summary, an author's explicit statement of the text's toplevel structure does not affect the retrieval strategy employed by ninthgrade students a week after reading a passage. However, there is some indication that greater amounts of signaling assist the learning and immediate retrieval of students with deficient comprehension skills who cannot employ the structure strategy without assistance.
Discussion Four important findings related to the use of the text's toplevel structureresulted from this study. First, less than 50%of the ninthgrade students sampled utilized this strategy at least once in their reading and recall tasks, and only 22% utilized it consistently on the four protocols written. Second, most ninth-grade students, rated by their teacher and standardized tests as high in reading comprehension skills, used the same top-level structure for organizing their recall protocols as the author of the passage, while most students with low reading comprehension skills did not. Third, students who employed this strategy of using the text's top-level structure recalled much more information from the passage than those who did not. Fourth, students who used this strategy could discriminate better between information consistent with the semantics of the passage and intruded information on the same topic than students who did not employ this strategy. The data show a strong relationship between comprehension skills and use of the top-level structure in text. In addition, use of the
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
97
text's top-level structure in organizing one's recall of text is highly correlated with the amount of information recalled. Two recent dissertations (Bartlett, 1978; McDonald, 1978) support the contention that this relationship is not only correlative, but causative. McDonald found that teaching the top-level structure of stories facilitated recall of poor comprehenders in the primary grades. Bartlett taught ninth-grade students the expository discourse types of problem/solution, comparison, antecedent/consequent, and description (Meyer, 1975a, 1977, in press-a, Note 3; Meyer & Freedle, Note 1). This instruction increased their ability to identify and use the text's top-level structure and nearly doubled the amount of information remembered. Instruction effects appeared durable over an extended period for readers scoring above the 19th percentile on the vocabulary test of the Stanford Achievement Test. The ability to utilize the text's top-level structureappears to be an important organizational strategy for remembering information in text. A systematic study is needed of its use by different types of readers with different discourse types. At present, we have looked at this factor with some adult readers. In a sample of junior college students with text organized by problem/solution top-level structures (Meyer, Bartlett, Woods, & Rice, Note 9), more than 50% of the students used this same type of top-level structure in their recalls; those students who employed this strategy remembered more than those who did not. Also with these problem/solution texts, 80% of a sample of 130 college undergraduates used the same top-level structure as found in these texts (Meyer, 1975a). In samples of graduate students and college graduates in young, middle, and old age ranges, most (80-100%) used the same type of top-level structures as they read in well-organized passages (problem/solution, comparison, antecedent/consequent, description) (Meyer & Freedle, Note 1; Meyer, Rice, Knight, & Jessen, Note 10). However, they appear to be able to employ alternate top-level structures, as evidenced by a group of graduate students who rejected a proposed solution in a passage written with a problem/solution structure and employed comparison or antecedent/consequent top-level structures in their recall protocols (Meyer & Freedle, Note 1). Taylor (1980) has looked at use of a text's top-level structure fourthand sixth-grade students and adults with a descriptive passage by (general statements followed by specific examples). Use of the top-level structure in recalls written immediately after reading the text was 82%, 47%, and 12%for adults, good sixth-grade comprehenders, good fourthgrade comprehenders, respectively (N = 17 for each group). The good comprehenders in Taylor's study were reading on grade level and are
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
98
* Number1, 1980 READING RESEARCH QUARTERLY
XVI/1
more comparable in comprehension test scores to the average comprehenders in our study than our good comprehenders. Use of structure by our average comprehenders was 51% (problem/solutionsupertankers) and 43% (comparison-dehydration). The ability to employ the structure strategy (see Figure 2) may develop with age and schooling. Freedle (Freedle & Hale, 1979; Freedle, Note 11) suggests that competence with story structure precedes competence with expository structure. Use of the structure strategy may progress in the following sequence with different discourse types: stories, description, antecendent/consequent, problem/solution, and comparison (argumentative text). In the present study, fewer ninth-grade students used the structure strategy with the comparison passage than the problem/ solution passage. However, this finding could be attributed to the uncontrolled factors of passage topic and amount of signaling. In our study, mastery of the structure strategy probably has been attained by the ninth-grade students who organized their four recall protocols with the texts' top-level structure. Others may be approaching its mastery, but have greaterfacility with certain structuresor topics. The other students may consist of groups related to Flavell's (1977) production-mediation deficiency hypothesis. Some may not have used the strategy because they are unaware of it (default strategy, change = No, Figure 2). Others may have knowledge of the strategy but be unable to put it into production without assistance (default strategy, change = Yes, Figure 2). This latter group may consist of underachievers in reading comprehension as well as those average and poor comprehenders who benefit from the clear and repeated signaling of the top-level structure. The signaling may provide them with enough assistance to produce this strategy for the immediate recall task; however, when this assistance is not available on the delayed task, they cannot employ the structure strategy on their own to organize retrieval. The predictions of the model depicting the interaction between signaling and dominant reader strategies received some support from the data on the supertanker (problem/solution) passage. As predicted, signaling had no effect on recall nor structure use of good and poor comprehenders, but it did increase recall of comprehension underachieversand tended to increase their use of text structure. For the dehydration (comparison) passage, the predictions for good and poor comprehenders held, but signaling had no effect on the comprehension underachievers. Regardless of the addition of signaling to this passage, few comprehension underachievers employed the structure strategy.
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
99
Perhaps greater amounts of signaling would have produced the predicted effects. On the other hand, if the structure strategy develops sequentially through the discourse types as ordered above, ninth-grade underachievers in reading comprehension may not be able to change strategies on text with comparison structures without more extensive aids, such as a training program like Bartlett's (1978). The present study suggests that the use of the structure strategy, or following the organization of a passage, is important because it provides ninth-grade students with a systematic learning and retrieval guide. As suggested by the default strategy of the model and supported by the data, the ninth-grade students who did not employ the structure strategy simply tried to list collections of descriptions from the passage without interrelating them. In contrast, those employing the structure strategy compared viewpoints or related solutions to components of the problem, and consequently, developed a rich retrieval network. Use of the top-level structure at retrieval appeared to provide a systematic topdown search of a hierarchy similar in many aspects to that depicted for the text in Figure 1. This top-down retrieval strategy assured recall of the passage's message, but also facilitated retrieval of details which could be linked to this organizational structure. A number of results from the study point to the structure strategy as an effective mneumonic for retrieval. Use of the text's toplevel structure by subjects to organize their immediate recalls, but not their delayed recalls, suggests that the structure strategy was employed for encoding and initial retrieval, but not for retrieval a week later. If the structure strategy minimally affected retrieval processes, then large differences would not be expected in delayed recall scores between subjects who used the structure strategy for all encoding and retrieval tasks and those who used the structurestrategy for encoding and retrieval initially, but not for the second retrieval task given one week later. However, as seen in Figures 3 and 4, these differences were extremely large. In addition, the importance of the structurestrategy for retrievalis supported by the finding that use of the text's top-level structureis highly related to delayed free recall, while not related to the delayed recognition of verbatim sentences from the text. In conclusion, this study has presented a methodology for studying a strategy of learning and recall in which students adhere to the structure provided by the author. Further research is needed concerning how this strategy develops and when and where it will prove to be a useful aid to learning and recall. The usefulness of the top-level structure in reading requires further investigation with different types of discourse,
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
* Number 1, 1980 READINGRESEARCH QUARTERLY
100
XVI/
1
purposes, and time limitations. This line of researchappears to have both practical applications for improving the teaching of reading comprehension and theoretical value in understanding strategy development and selection and how people learn and remember discourse. REFERENCES ANDERSON,J.P. Language, memory, andthought.
DAVIS,F.B. Fundamental factors of comprehen-
Hillsdale, N.J.: Lawrence Erlbaum Associates, 1976. ANDERSON, R.C. How to construct achievement tests to assess comprehension. Review of Educational Research, 1972, 42, 145-170.
DECHANT,E. Improving the teaching of reading.
ANDERSON,R.C. The notion of schemata and the
C.J. The case for case. In E. Bach & FILLMORE,
educational enterprise. In R.C. Anderson, R.J. Spiro, & W.E. Montague (Eds.), Schooling and the acquisition ofknowledge. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1977.
R.T. Harms (Eds.), Universals in linguistic theory. New York: Holt, Rinehart and Winston, 1968.
B.J. Top-level structure as an organiBARTLETT,
zational strategyfor recall ofclassroom text. Unpublished doctoral dissertation, Arizona State University, 1978. BEAUGRANDE, R. Text, discourse, and process. Norwood, N.J.: Ablex Publishing Corporaation, 1980. BOWER, G.H., CLARK, M.C., LESGOLD, A.M., & WINZENZ, D. Hierarchical retrieval schemes
in recall of categorized word lists. Journal of VerbalLearning and VerbalBehavior, 1969, 8, 323-343. BRANSFORD, J.D., & FRANKS,J.J. The abstraction
of linguistic ideas. Cognitive Psychology, 1971, 2, 331-350. BROWN,A.L., & SMILEY,S.S. The development of
strategies for studying texts. Child Development, 1978, 49, 1076-1088.
CARROLL,J.B. Defining
language
comprehen-
sion: Some speculations. In J.B. Carroll & R.O. Freedle (Eds.), Language comprehension and the acquisition of knowledge. Washington, D.C.: V.H. Winston, 1972. CRAIK, F.I.M., & LOCKHART,R.S. Levels of pro-
cessing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 1972, 11, 671-684.
E.J. The psycholinguistic CROTHERS,
structure of
knowledge. In K. Romney & K. Wexler (Eds.), Cognitive organization and psychological processes. Washington, D.C.: National Academy of Sciences, 1973. D'ANGELO,F.J. Paradigms as structural compo-
nents of topoi. In D. McQuade (Ed.), Linguistics, stylistics, and the teaching of composition. Akron, Ohio: University of Akron Press, 1979.
sion in reading. Unpublished doctoral dissertation, Harvard University, 1941. Englewood Cliffs, N.J.: Prentice-Hall, Inc., 1970.
FLAVELL,J.H. Cognitive
development.
Engle-
wood Cliffs, N.J.: Prentice-Hall, 1977. FRASE, L.T. Paragraph
organization
of written
materials: Influences of conceptual clustering upon the level and organization of recall. Journal of Educational Psychology, 1969, 60, 394-401. FREDERIKSEN,C.H. Representing
logical
and
semantic structure of knowledge acquired from discourse. Cognitive Psychology, 1975, 7, 371-458. (a) C.H. Acquisition FREDERIKSEN,
of semantic in-
formation from discourse: Effects of repeated exposures. Journal of Verbal Learning and VerbalBehavior, 1975, 14, 158-169. (b) FREEDLE,R.O., & HALE, G. Acquisition
of new
comprehension schemata for expository prose by transfer of a narrative schema. In R.O. Freedle (Ed.), New directions in discourseprocessing. Norwood, N.J.: Ablex Publishing Corporation, 1979. GRIMES,J.E. The thread ofdiscourse.
The Hague,
Holland: Mouton and Co., 1975. HALLIDAY, M.A.K. Notes
on transitivity
and
theme in English, Part 3. Journal of Linguistics, 1968, 4, 179-215. HALLIDAY,M.A.K., & HASAN, R. Cohesion
in
English. London, Longman, 1976. F. The promiHAYES-ROTH, B., & HAYES-ROTH,
nence of lexical informationin memory.
Journal of Verbal Learning and Verbal Behavior, 1977, 16, 119-136. KINTSCH,W. The representation
of meaning in
memory. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1974. KINTSCH,W. Memoryandcognition.New York: John Wiley & Sons, 1977. KINTSCH,W., & KEENAN,J. Reading rate and
retention as a function of the number of
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
MEYER,BRANDT,&
Use of top-level structure
propositions in the base structure of sentences. Cognitive Psychology, 1973, 5, 257274. KINTSCH,W., & VAN DIJK, T.A. Toward a model
of text comprehension and production. Psychological Review, 1978, 85(5), 363-394. KINTSCH,W., & VIPOND, D. Reading comprehen-
sion and readability in educational practice and psychological theory. In L. GorenNilsson (Ed.), Memory: Process and problems. Hillsdale: Erlbaum, 1979. MARSHALL,N., & GLOCK, M.D. Comprehension
of connected discourse: A study into the relationships between the structure of text and information recalled. Reading Research Quarterly, 1978-1979, 14, 10-56. MCDONALD,G.E. The effects of instruction in the
use of an abstractstructuralschema as an aid to comprehension and recall of written discourse. Unpublished doctoral dissertation, Virginia Polytechnic Institute and State University, 1978. MCGUIRE, M., & BUMPUS, M. The Croft inservice program: Reading comprehension skills. New London, Conn.: Croft Educational Services, 1971. MEYER,B.J.F. The organization
of prose and its
effects on memory. Amsterdam: NorthHolland Publishing Co., 1975. (a) MEYER,B.J.F. Identification
of the structure of
prose and its implications for the study of reading and memory. Journal of Reading Behavior, 1975, 7, 7-47. (b) MEYER,B.J.F. The structure of prose: Effects on
learning and memory and implications for educational practice. In R.C. Anderson, R. Spiro, & W.E. Montague (Eds.), Schooling and the acquisition of knowledge. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1977. MEYER,B.J.F. Organizational patterns in prose and their use in reading. In M.L. Kamil & A.J. Moe (Eds.), Reading research: Studies and applications. Clemson, S.C.: National Reading Conference, Inc., 1979. MEYER,B.J.F. A selected review and discussion of
basic research on prose comprehension. In D.F. Fisher & C.W. Peters (Eds.). Comprehension and the competent reader. New York: Praeger Special Studies, Holt, Rinehart, and Winston, in press. (a) MEYER,B.J.F. Following
the author's top-level
101
BLUTH
structure: An important skill for reading comprehension. In R. Tierney, J. Mitchell, & P. Anders (Eds.), Understandingreaders' understanding. Hillsdale, N.J.: Lawrence Erlbaum Associates, in press. (b) MEYER, B.J.F., & MCCONKIE, G.W. What is recalled after hearing a passage? Journal of Educational Psychology, 1973, 65, 109-117. MEYER,B.J.F., & RICE,G.E. The structure of text.
In P.D. Pearson (Ed.), Handbook of research in reading. New York: Longmans, in press. NILES,O.S. Organization perceived. In H.H. Herber (Ed.), Perspectives in reading: Developing study skills in secondary schools. Newark, Del.: International Reading Association, 1974. REDER, L.M. The role of elaboration in the comprehension and retention of prose: A critical review. Review of Educational Research, 1980, 50, 5-54. RUMELHART,D., & ORTONY,A. The representa-
tion of knowledge in memory. In R.C. Anderson, R.J. Spiro, & W.E. Montague (Eds.), Schooling and the acquisition of knowledge. Hillsdale, N.J.: Lawrence Erlbaum Associates, 1977. SACK, A., & YOURMAN,J. The Sack- Yourman
developmental reading course. New York: College Skills Center, 1972. SIMMONS,R.F. A semantic analyzer for English
sentences.Mechanical Translationand Corn putational Linquistics, 1968, 11, 1-13. SMILEY,
S.S.,
OAKLEY,
D.D.,
WORTHEN,
D.,
CAMPIONE,J.C., & BROWN, A.L. Recall of
thematically relevant material by adolescent good and poor readers as a function of written versus oral presentation. Journal of Educational Psychology, 1977, 69, 381-387. Supertankers. In J.A. Ball (Ed.), Read: The Magazine for Reading and English, 1975, 24(12), 2-4. TAYLOR, B. Children's
and adults'
recall of
general concepts and details after reading. In M. Kamil & A. Moe (Eds.), Reading research: Studies and applications. Clemson, S.C.: National Reading Conference, 1979. TAYLOR,B.M. Children's memory for expository
text after reading. Reading Research Quarterly, 1980, 15, 399-411.
Reference Notes 1. MEYER,B.J.F., & FREEDLE,R.O. Effects of discourse
type on recall (Prose Learning Series
Research Report No. 6). Tempe, Ariz.: Arizona State University, Summer 1979. 2. BEIGER, G. R., & DUNN, B.R. Sensitivity to developmental
differences in children's recall ofprose:
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
* Number 1, 1980 READING RESEARCH QUARTERLY
102
XVI/1
A comparison of two prose grammars. Paper presented at the annual meeting of the American Educational Research Association, Boston, April 1980. 3. MEYER, B.J.F.Structure of prose: Implications for teachers of reading (Prose Learning Series Research Report No. 3). Tempe, Ariz.: Arizona State University, Spring 1979. 4. MEYER, B.J.F.Research on prose comprehension: Applications for composition teachers (Prose Learning Series Research Report No. 2). Tempe, Ariz.: Arizona State University, Spring 1979. 5. COFER, K.Scoring systemsfor the analysis ofpassage content. Paper C.N.,SCOTT, C., & WATKINS, presented at the American Psychological Association, Toronto, Ontario, Canada, 1978. 6. MEYER, B.J.F.Text structure and its use in the study of reading comprehension across the adult life-span. Paper presented at the annual meeting of the American Educational Research Association, Boston, April 1980. 7. MEYER, B.J.F.Signaling in text and its interaction with reader strategies. Paper presented at the annual meeting of the American Educational Research Association, Boston, April 1980. 8. BRITTON, M. Use ofcognitive capacity in reading text: B.J.F.,GLYNN, B.K., MEYER, S., & PENLAND, Effects of variations in surface features of text with underlying meaning held constant. University of Georgia manuscript submitted for publication, 1980. 9. MEYER, B.J.F.,BARTLETT, V., & RICE,G.E.Facilitation effects of reading passages B.J., WOODS, with the same structure and different content (Prose Learning Series Research Report No. 10). Tempe, Ariz.: Arizona State University, Winter 1979. 10. MEYER, J.L. The effects of different discourse types on B.J.F.,RICE,G.E.,KNIGHT, C.C.,& JESSEN, recall (Prose Learning Series Research Report No. 7). Tempe, Ariz.: Arizona State University, Summer 1979. 11. FREEDLE, R.O. Children's recall of narrative and expository prose: The acquisition of an expository schema. Paper presented at the annual meeting of the American Educational Research Association, Boston, April 1980.
Footnotes 'This research was supported by an Arizona State University Faculty Grant. We would like to thank James DeGracie, Director of Research and Evaluation, Camille Estes, English teacher, and the staff and students at the junior high school in Mesa, Arizona. Requests for reprints should be sent to Bonnie J.F. Meyer, Department of Educational Psychology, Arizona State University, Tempe, Arizona 85281. 2Significantinteractions for the supertanker passage were order x comprehension group F(2,78)= 3.16, p < .05; time x signaling F(1,78) = 7.05,p <.01; time x order x sex F(1,78) = 3.95,p< .05; time x signaling x order x sex F(1,78) = 4.77, p < .03; time x signaling x order x comprehension group x sex F(2,78) = 3.38,p< .04. Significant interactions for the dehydration passage were time x order F(1,77) = 10.6, p< .002; time x sex F(1,77) = 5.17, p< .03.
Appendix Supertankers A PROBLEM OF VITAL CONCERN IS THE PREVENTION OF OIL SPILLS FROM SUPERTANKERS. A typical supertankercarries a half-million tons of oil and is the size of five football fields. A wrecked supertanker spills oil in the ocean; this oil kills animals, birds, and microscopic plant life. For example, when a tanker crashed off the coast of England, more than 200,000 dead seabirds washed ashore. Oil spills also kill microscopic plant life which provide food for sea life and produce 70 percent of the world's oxygen supply. Most wrecks RESULT FROM THE LACK of power and steering equipment to handle emergency situations, such as storms. Supertankers have only one boiler to provide power and one propeller to drive the ship. THE SOLUTION TO THE PROBLEM IS NOT TO IMMEDIATELY HALT THE USE OF TANKERS ON THE OCEAN since about 80 percent of the world's oil supply is carried by supertankers. INSTEAD, THE SOLUTION LIES IN THE TRAINING OF
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
Use of top-level structure
MEYER,BRANDT,& BLUTH
OFFICERS OF SUPERTANKERS, BETTER BUILDING OF TANKERS, AND INSTALLING GROUND CONTROL STATIONS TO GUIDE TANKERS NEAR SHORE. First, OFFICERS OF SUPERTANKERS MUST GET top TRAINING in how to run and maneuver their ships. Second, tankers should be BUILT with several propellersfor extra control and backup boilersfor emergencypower. Third GROUND CONTROL STATIONS SHOULD BE INSTALLED at places where supertankers come close to shore. These stations would act like airplane control towers, guiding tankers along busy shipping lanes and through dangerous channels.
Note. CAPITALIZED = Message; lower case = Major Details; italics = Minor Details; underlined = Signaling.
This content downloaded from 188.72.126.41 on Sun, 22 Jun 2014 07:27:46 AM All use subject to JSTOR Terms and Conditions
103