Memory & Cognition1981, Vol. 9 (6), 541-549

Topicalization effects in cued recallof technical prose

DAVID E. KIERASUniversity ofArizona, Tucson, Arizona 85721

The perceived topic of a passage should determine what information is given priority instorage effort for later recall. The topic should also determine how effective a later recall cueshould be, in that recall should be best if the cue is the same as the passage topic. These issueswere studied by investigating cued recall of passages that contained information about twocandidate topic items, either of which could be marked as the passage topic by initial mentionor the sentence surface subject position. The recall cue was either the item marked as the topicor the nontopic item. If the cue was the topic, recall about the topic item was greater than recallabout the nontopic item. If the cue was the nontopic item, recall about the two items wasroughly the same, unaffected by the topic marking. But the topic and nontopic cues producedthe same overall level of recall. In contrast to the original hypothesis, the results are interpretedas the topic marking and recall cue acting as instructions for what information the subject shouldemphasize in recall. It is argued that the two-topic passages used in this work are processeddifferently from the usual one-topic passages in prose memory studies.

Some of the information in a piece of prose is moreimportant than the rest of the information. If one isasked to study a passage, one should remember the moreimportant information better than the less importantinformation, This fact is well documented under thename of the "levels effect," in which information at ahigh level of importance in a propositional analysis ofa passage is better recalled than information at lowlevels (see Johnson, 1970; Kintsch & Keenan, 1973;Kintsch, Kozminsky, Streby, McKoon, & Keenan,1975; Kozminsky, 1977; Meyer, 1977). The importancelevel is defined in terms of the hierarchical semanticstructure of the passage information.

The role that importance level plays in current theoryof prose memory is best represented by the macrostruc­ture theory of comprehension (Kintsch, 1977; Kintsch& van Dijk, 1978; van Dijk, 1977). According to thistheory, a reader selects or constructs a set of macro­structure propositions that represents the gist or impor­tant content of the passage, and which is then givenpriority for storage into memory. The levels effectappears because propositions higher in the passagestructure will usually be important macropropositions,or closely related to the macropropositions, and so arelikely to be stored and later recalled by the reader.

Although the semantic content of a passage is themain determinant of which propositions are important,

This research was supported by the Personnel and TrainingResearch Programs, Office of Naval Research, under Con­tract N00014-78'{:-Q509, Contract Authority IdentificationNo. NR 157-423. Thanks are due to Gary Foulke, Susan Bovair,Mark Sternpski, Jim Lentz, and Kathy Warrickfor their assistancein collecting and analyzing the data. Requests for reprintsshould be sent to the author at the Department of Psychology,University of Arizona, Tucson, Arizona 85721.

there also exist many features of textual surface struc­ture that signal what content is important, or thematic(van Dijk, 1979), and so should influence what isremembered from a passage. One such signal is a title forthe passage. Kozminsky (1977), using passageswith twoalternate themes, found that recall was biased in favor ofpropositions relevant to the theme mentioned in thetitle. Another important surface-level signal is initialmention; items or ideas appearing first in a passage tendto be viewed as thematically important simply byvirtue of their position (Kieras, 1980). Better recall ofinitially appearing information has been reported (e.g.,Kintsch et al., 1975; Meyer, 1977), but it has usuallybeen attributed to better recall of important informa­tion, which in conformity with a linguistic convention,usually appears first. However, such studies have usuallyconfounded structural level of importance with initialmention. Another important surface-level signal tothematic content is the topic-comment assignment ofitems mentioned in individual sentences (perfetti &Goldman, 1974, 1975; van Dijk, 1979). The simplestfonn of this signal is simply which item appears as thesurface subject of the passage sentences. Kieras (1981)found that a particular candidate topic was chosen asthe passage topic more often if it appeared as the surfacesubject of the sentences. Clements (1979), using sen­tence topic-comment assignment and other aspects ofassigning thematic importance, found that materialmarked as important was recalled better than unmarkedmaterial.

Such signals should govern what is stored. Nowconsider how readers might retrieve information froma previously studied passage when given a cue to thecontent. The obvious application of macrostructuretheory is that the macrostructure of the passage deter-

Copyright 1981 Psychonomic Society, Inc. 541 0090-502X/8l /060541-09$01.15/0

542 KIERAS

mines not only what is to be stored, but also what theretrieval routes for that information might be. That is,the reader selects the main item, or main referent, of thepassage, around which the macrostructure should beorganized (see Kieras, 1980; van Dijk, 1979), and usesthis perceived topic as a sort of "address" for storage.When given a cue for recall, the reader looks first underthe corresponding "address" in memory for the passageinformation. If the desired information is found, recallproceeds smoothly. If not, the retrieval attempt will bedisrupted and a different retrieval strategy must be used,resulting in poorer recall. Hence the originally perceivedtopic of a passage will influence the effectiveness of arecall cue; a cue that matches the topic should be superiorto one that mismatches. Such an effect would be aprose memory analogue to the "encoding-specificity"principle studied heavily in verbal learning.

The two studies reported here address the role oftopicalization in cued recall using brief technical pas­sages. The question is whether marking an item as thepassage topic and cuing recall with either the topic itemor a nontopic item affects recall for propositions aboutpassage items. Experiment 1 demonstrates effects usinginitial mention as the topicalization marker. Experi­ment 2 follows up by exploring these effects in moredetail, using initial mention and surface subject assign­ment as topicalization markers and examining recall forpassages in which the topic was determined by semanticfactors rather than textual surface structure. An impor­tant feature of these experiments is that the topicmarking was done independently of the specific passagecontent. Thus recall of exactly the same material can becompared when it is marked as a topic and when it isunmarked. This was achieved by using passages chosenfrom earlier studies (Kieras, 1980, Note 1) that each hadtwo candidate topics, either of which could be markedas the passage topic. Recall was compared for cases inwhich the different candidate topics were marked as thepassage topic and cases in which the different candidatetopics served as the recall cue.

EXPERIMENT 1

In this experiment, several two-topic passages wereused, with the topic marking produced by the candi­date topic that was described first in the passage (seeKieras, 1980). An example passage is shown in Table 1,in which the two candidate topic items, labeled A and B,are "Biotransformation" and "The Liver." The passageconsists of two sentences about Item A, two sentencesabout Item B, and a linking sentence that connectsthem. The A-first version consists of the two A sen­tences, followed by the linking sentence, followed bythe two B sentences. In the B-first version, the B sen­tences appear first, then the linking sentence, and theA sentences. The passages were prepared to be reason­ably comprehensible in both versions.

Table IA-Pirst and B-Pirst Versions of an Example Two-Topic Passage

A-Pirst VersionBiotransformation is the chemical transformation that causes

the inactivation of drugs, the detoxification of environmentalpollutants, and the deactivation of chemicals that can causecancer. Biotransformation of harmful agents involves an oxida­tion reaction which is mediated by complex enzymes, and if thisprocess does not take place, a drug entering the body may actindefinitely, Biotransformation defends the body against theeffects of toxins and is carried out in the liver. The liver, weigh­ing three pounds in the human adult, is the largest organ in thebody and performs diverse functions. Through the large portalvein of the liver passes all the blood that has absorbed digestedfood and other substances from the intestines.

B-Pirst VersionThe liver, weighing three pounds in the human adult, is

the largest organ in the body and performs diverse functions.Through the large portal vein of the liver passes all the bloodthat has absorbed digested food and other substances from theintestines. Biotransformation defends the body against theeffects of toxins and is carried out in the liver. Biotransforma­tion is the chemical transformation that causes the inactivationof drugs, the detoxification of environmental pollutants, and thedeactivation of chemicals that can cause cancer. Biotransforma­tion of harmful agents involves an oxidation reaction which ismediated by complex enzymes, and if this process does not takeplace, a drug entering the body may act indefinitely.

MethodMaterials. Eight passages were selected from those reported in

Kieras (Note I, Experiment 4) that showed good effects of theinitial-mention variable. These passages would thus be expectedto produce different apparent topics when presented in the twoorders. For each passage, two cues, one or two words long, weredevised that named each candidate topic of the passages. Thepassages were prepared as small booklets, one for each subject,with one passage per page.

Design. The experiment was within subjects, with fourconditions, corresponding to the two passage versions, A-firstor B-first, combined with the two possible cues, the A cue andthe B cue. A series of 8 by 8 random Latin squares was used todetermine the assignment of passages to version and cue condi­tions for each subject, ensuring that for multiples of eightsubjects, each passage would appear twice in each conditionand each subject would see each condition twice and eachpassage once. Another series of Latin squares was used tobalance the order of appearance in the booklets.

Subjects. Forty-eight subjects of both sexes participatedfor extra credit in an introductory psychology course. Thesession required about I h.

Procedure. The subjects were run in groups of up to 20 insize. The subjects studied each passage in their booklets; theywere paced by the experimenter at I min/passage. After readingall eight passages, subjects immediately began working on adistractor task consisting of a large sheet of arithmetic problems.After 8 min had elapsed, subjects were instructed on the recallprocedure. Each subject was given a booklet consisting of eightpages, with a cue word for one of the passages at the top of eachpage. The recall cues were arranged in the same order as the pas­sages for each subject. The subjects were instructed to write asmuch as they could remember, in their own words, of thepassage corresponding to the cue word. They were paced at3 min for each passage recall attempt. The amount of timeallowed was observed to be adequate.

ResultsA propositional analysis was performed on the

passages and used as a scoring key, along the linesdescribed by Kintsch (1974), Bovair and Kieras (Note 2),and Turner and Greene (Note 3). The recall protocolswere scored blind, without knowledge of the originalpassage version condition, for reproductions of the pas­sage propositions. The scoring criterion was fairly strict;a proposition was counted as recalled only if the subjectsupplied all of the same terms in the proposition, withsynonymous terms or references allowed. No credit wasgiven for partial recall of a proposition or for generalsimilarity of meaning. The strict scoring, together withthe delayed-recall procedure, produced a fairly low levelof recall, around 11% on the average. Since such lowlevels of recall are difficult to score, it was desirable tohave some check on the reliability of the scoring. Twoindependent raters scored the recall protocols, eachusing his own propositional analysis, and then the raterwas included as a factor in the analysis. Recalled prop­ositions from the sentences about each candidate topicitem were then counted, but the propositions from thelinking sentence were not included. Hence the analyseswere done only on the recall of propositions from thetwo sentences about each item. The amount of recallwas converted to proportions to eliminate the slightdifferences in the number of propositions contained inthe sentences, averaged over passages, and subjected toan ANOVA. The factors in the analysis were the scoringrater, the item of recall (the topicalized item or the non­topicalized item), the recall cue (topic or non topic),and the passage version (A or B). In these experiments,the materials were not randomly sampled, but theywere carefully constructed and selected to produce theintended topic manipulation. Hence passages were nottreated as a random factor (see Clark, Cohen, Smith,& Keppel, 1976; Wike & Church, 1976).

With regard to scoring reliability, the correlation was.861 between mean recall scores of the two raters,using the 384 points corresponding to mean recall ofeach item from each passage for each subject in eachcondition. The second rater produced scores that werevery slightly higher than those of the first rater (10.6%vs. 11.3%), which was marginally significant [F(1,47) =3.25, p < .1]. Hence there was very good agree­ment between the two raters, especially consideringthat they used separate propositional analyses. Butthere was an effect of rater, which will be discussedbelow.

The mean proportions of recall of each item in eachcondition are shown in Table 2. The main effect of i~mof recall was significant [F( 1,47) = 14.82, p < .01] ;propositions from sentences about the topic items wererecalled better overall than those about non topic items(13% vs. 9%). But there was an interaction of item ofrecall with cue, which was strongly significant [F(l,47)=34.18; P < .01]. As shown in Table 2, when the pas­sage topic was cued, recall of the topic item was higherthan recall for the non topic item. However, if the cue

TOPICALIZATION EFFECTS 543

Table 2Mean Proportion of Recall About Each Item

Item ofVersion

Cue Recall A B Mean

Topic Topic .14 .18 .16Nontopic .07 .07 .07

Nontopic Topic .07 .13 .10Nontopic .15 .07 .11

Note- The standard error of thesemeans isabout .01.

was not the topic, there was no difference in recallbetween the topic and non topic items.

The interaction of rater with cue and item was signifi­cant [F(1 ,47) =8.59, p < .01] , but presents no problemin interpretation because the ordering of the cell meansfor each rater show exactly the same pattern. Further­more, this interaction accounts for a very small propor­tion of the variance (eta square =.11%) in contrast tothe Cue by Item interaction (eta square =3.76%).Hence, although there were rater effects, they appear tohave been unimportant.

Finally, there was a pair of interactions with versionthat show that the A items and B items differed in theirmemorability. Version and item interacted [F(I,47) =29.45, P< .001], as did version, cue, and item [F(I,47)= 6.11, P < .051. No other effects approached signifi­cance.

DiscussionThe result that information about an item presented

first in a passage is recalled better than is informationabout an item presented later duplicates the findingreported by Meyer (1977) and others that recall isbetter for the first-mentioned material. However, thisresult is stronger in this study, since order of appearancein the passage was manipulated separately from thesemantic content associated with the item. Thus eventhough the most important content normally appearsfirst in a passage, initial mention influences recall becauseit has topic-signaling value in itself.

The results in Table 2, however, contain a moretheoretically interesting result, that recall about an itemis specific to the relation of the recall cue to the origi­nally perceived passage topic. That is, if the cue is thepassage topic, recall favors the topic, but if the cue isnot the topic, recall of topic and nontopic informationis equal. Marking an item as the passage topic apparentlydoes not govern what the reader stores, since in thenon topic cue case there is no difference in recall of topicand nontopic information. Rather, the original passagetopic seems to determine what the effect of a laterrecall cue will be. Notice that there was no main effectof cue (p> .1), which indicates that the overall level ofrecall does not depend on whether the cue is the topic.This result conflicts with the memory retrieval role ofthe passage topic presented above.

There are two problems with these results that moti­vated the second experiment. The first is that the Item B

544 KIERAS

propositions are better recalled overall than Item Apropositions (13.3% vs. 8.8%), which produced theinteractions with version. In contrast, as presented inmore detail in Kieras (1980, Note 1), in these passagesthe A items are generally preferred to the B items aspassage topics. A possible explanation for the differencebetween the items is that the B items showed a tendencyto be more specific than the A items and the factsstated about the B items tended to be more specific andconcrete. An example of this difference is the items inthe Table 1 passage. Although the A·B label is clearlyarbitrary, in practice the passages were composed bystarting with an A item and trying to devise a suitableB item; this process apparently resulted in differences initem generality and concreteness. It could be that themore general candidate topic makes the best choice forthe passage topic, since it logically subsumes the otherpossible passage topic, but highly specific facts about aspecific item are easier to recall in this type of memoryexperiment. This problem was explored in the secondexperiment by using passages for which normative topicchoice data were available and comparing passages thatshowed either a strong bias in favor of one item's beingthe topic or essentially no such bias.

The second problem is that since the topic-markingmanipulation involved a serial position manipulation,the observed effects may be due to simple serial positioneffects rather than to topic marking. Thus the resultscould be explained as follows: In the topic cue condi­tion, subjects are able to start their recall with the cueditem and then recall the passage sentences in the sameorder as originally presented. The unmarked iteminformation is recalled poorly because it is recalled lastand, so, suffers from output interference. Hence a resultlike Deese and Kaufman's (1957) primacy-only serialposition curve appears. In the non topic cue condition,the cue encourages the subject to start in the middleof the passage, thus disrupting the normal serial orderof recall, leading to poorer performance and making theoriginal version of the passage irrelevant.

The attack on this problem used in Experiment 2was to include a different passage topic marker, sentencesurface subject assignment. This marker would allowdifferent candidate topic items to be marked as topicalby surface changes in the sentences, without changingthe order of the information in the passage. Also, detaileddata were collected on the order of recall, which wouldallow serial order effects in recall to be directly examined.

EXPERIMENT 2

In this experiment, passages of three types were used:balanced, biased, and surface subject. The balanced andbiased passages appeared in either the A-first or theB-first order. Using topic choice data from earlierstudies, the balanced passages were selected to showlittle or no bias in favor of one of the items, and the

biased were chosen to show a strong bias. The surfacesubject passages appeared in either of two versions:which differed only in which candidate topic appearedas the surface subject of all of the passage sentences. Asin Experiment 1, recall was cued with either of the twocandidate topics.

MethodMaterials. Using the materials and results from Kieras (1980,

1981), six passages were selected, two of each of the three types:balanced, biased, and surface subject. The candidate topicitems of the passages used, together with the proportions ofchoices of each item and the sample size for these proportions,are shown in Table 3. The balanced and biased passages werechosen from those used in Kieras (1980), in which subjectschose one of the two candidate topics in a forced-choice pro­cedure. The surface subject passages were chosen from thoseused in Kieras (1981), in which subjects generated noun phrasestatements of the passage topics; these were scored for whichone of the candidate topics they referred to. Since the state­ments may have referred to something else, these choices are notexhaustive. For ease in reference here, the label Item A wasassigned to the candidate topic that was preferred, even slightly,in each passage.

The balanced passages were chosen to show a symmetricaleffect of initial mention on topic choice, with very little biasin favor of one of the items. As can be seen in Table 3, forboth balanced passages, there is a strong initial-mention effectand the choices of the two items overall are almost of equalfrequency. The biased passages were chosen to show a strongtopic bias. The Table 1 example is a biased passage used in thisexperiment. As shown in Table 1, there is a strong bias in favorof Item A, which in both passages is a very general abstractconcept, whereas Item B is a very specific and concrete item.The effect of topicalizing B by initial mention is only toweaken the preference for Item A. Finally, the surface subjectpassages were chosen to show a substantial effect of whichcandidate topic was marked by appearing as the surface subjectof the passage sentences. Table 4 shows one of these passages;Item A is "vaccines," and Item B is "virus." Due to the dif­ficulty of composing such passages, equating the items onsalience was not attempted for the work in Kieras (1981);the item choices are thus biased. But notice that there is not anobvious difference in the generality of concreteness of the twoitems, as is the case in the biased passages. Thus, in the A­topicalized versions of the passages, A was quite the dominantchoice for the topic, but in the B-topicalized versions, A and Bwere chosen roughly equally often. Thus, overall, there is abias in favor of A.

As in Experiment 1, for the balanced and biased passages, theB-topicalized version consisted simply of the same sentences asthe A-topicalized version in a different order. The surface subjectpassages, however, contained different sentences in the twoversions, which conveyed the same propositional content, butwhich differed in which candidate topic appears in the subjectand which in the predicate. All of the passages were five sen­tences long and were computer justified and printed to occupyabout 14 lines. Each passagewas printed on a separate page, andthe pages were assembled into booklets for each subject. Eachsubject's booklet contained, first, a page of instructions, fol­lowed by six passage pages, followed by a page of arithmeticproblems for a distractor task, followed by a page of recallinstructions and six recall pages, each containing a cue wordthat was one of the two candidate topics for one of the passages.The cue words appeared in the same order as the correspondingpassages. The order of appearance of the passages in the book­lets was separately randomized for each subject's booklet.

Design. The design was between subjects; there were four

TOPICALIZATION EFFECTS 545

Table 3Proportion of Choices of Item A and Item B for Each Passage Type

Choice

Topicalization Version A Topicalization Version B Overall

Balanced (N =24).29 .71 .56 .44.25 .75 .52 .48

Biased (N =48).65 .35 .74 .26.60 .40 .77 .23

Surface Subject (N =11).36 .27 .64 .14.45 .55 .72 .27

Passage Items (A/B) A B

Corona/Solar Wind .83 .17Isomers/Retinene .79 .21

Bioluminescent Light/Flashlight Fish .83 .17Biotransformation/Liver .94 .06

Vaccine/Virus .91 .00Radio Galaxies/Radio Waves 1.00 .00

A B A B

Table 4The A-Topicalized and B-Topicalized Versions of

an Example Surface Subject Passage

A-Topicalized VersionA vaccine is prepared by somehow weakening a virus, often

by inactivating the DNA and using the protein coat which canbe injected to stimulate the immune system of the body. Vac­cines require for their preparation the keeping of a large supplyof virus that is usually grown in systems such as egg or cellcultures. A vaccine is now under test that may soon control thevirus, Hepatitis B, that produces a serious and very debilitatingdisease. Vaccines have become available against many virusesthat once caused diseases which were common and often incur­able although some diseases still cannot be controlled this way.Vaccines were originally discovered and developed by men likeJenner who showed that the virus that causes cowpox, a verymild disease, could prevent infection by the dreaded smallpox.

B-Topicalized VersionViruses that have somehow been weakened, often by inacti­

vating the DNA and using the protein coat, are used to preparevaccines which can be injected to stimulate the immune systemof the body. Viruses must be kept in large supply in order toprepare vaccines and are usually grown in systems such as eggor cell cultures. The virus, Hepatitis B, that produces a seriousand very debilitating disease may soon be controlled by a vaccinenow under test. Viruses once caused many common and oftenincurable diseases until vaccines against them became available,although some diseases still cannot be controlled this way. Thevirus that causes cowpox, a very mild disease, was shown byJenner, one of the original discoverers and developers of vac­cines, to prevent infection by the dreaded smallpox.

groups, each of 15 subjects, one group for each combination ofpassage version (A topicalized or B topicalized) and recall cue (Aor B). Each subject studied the appropriate version of each ofthe six passages. The design ensured that if a multiple of foursubjects was run, each passage appeared equally often in eachversion-cue combination.

Subjects were assigned to one of the four version-cue com­binations in the order that they appeared for the experiment,with each combination used once in each consecutive group offour subjects.

Subjects. Subjects were 60 University of Arizona undergrad­uates, recruited through campus advertisements, who were paid$2 for participating in the 1-11 experiment.

Procedure. Subjects were run in groups of one to severalpeople. After reading the instructions on the first page of the

booklet, the subjects read each passage in their booklets one at atime. In order to raise the level of recall compared with Experi­ment 1 and to take advantage of the fewer passages being used,subjects were paced at 2 min/passage rather than 1 min/passageas in the first experiment. The instructions asked the subjects totry to remember the facts and ideas in each passage and not toattempt to memorize the exact wording. After completing studyof the last passage, the subjects worked the page of arithmeticproblems for 8 min. Then the subjects read the recall instruc­tions, and then they began to work on the recall pages in thebooklet, which were blank except for a cue word at the top ofeach page. They were instructed to recall in their own words asmuch as they could remember about the passage correspond-

. ing to the cue word, writing in the space provided on the page.They were paced at 3 min per recall page and were not allowedto go back.

ResultsRecall. The passages were analyzed into propositions,

and the recall protocols were scored for reproductionof each passage proposition, as in Experiment 1. Thescoring was done blind with regard to the presentedversion for the balanced and biased passages, and blindwith regard to the recall cue for the surface subjectpassages. Since the scoring reliability was satisfactory inExperiment 1, a more usual scoring procedure wasfollowed here. Scoring was done independently by twojudges, and then a third judge reconciled the two setsof scorings to yield a final single scoring. A strict cri­terion was used as in Experiment 1. In Experiment 1,propositions were classified as being "about" an item ifthey were contained in the sentences about each item. Amore precise rule was used in this experiment to labeleach passage proposition according to whether it wasabout Item A, Item B, or something else. The rule wasthat a proposition was about an item if it either containedthe item (or a synonymous reference to it) as an argu­ment or it embedded or was embedded by a propositionthat did contain the item. The few propositions thatcontained both items were counted as being about bothitems.

The data were then expressed as proportions ofpresented propositions about each item that wererecalled for each subject in each passage, and these

546 KIERAS

were averaged over the two passages within each type.An ANOVA was computed in which the within-subjectsfactors were passage type (balanced, biased, and surfacesubject) and item of recall (topic or nontopic). Thebetween-subjects factors were passage version (A topi­calized or B topicalized) and cue type (topic or non­topic). Table 5 presents the mean proportions of recallfor each combination of passage type, cue, and item ofrecall. Since there were important version effects only inthe biased passages, the table shows averages overversions.

Overall, there was no main effect of version or cue(ps>.l). There was a main effect of passage type:Balanced passages were recalled at .15, biased at .20,and surface subject at .13 [F(2,1l2) = 8.94, p < .01].

The key results correspond to Experiment 1. Prop­ositions about topic items were recalled better thannontopic items [18% vs. 15%; F{1 ,56) =4.25, P <.05].There was an interaction between item of recall and cue[F(1,56) = 19.91, P < .01]. As shown in the last columnof Table 5, for the topic cue, the recall of the topicitem was superior to the nontopic item, and for thenontopic cue, the recall of the two items was similar.Inspection of Table 5 shows that this pattern clearlydiffered between the passage types, but the three-wayinteraction including passage type was not significant[F(2,112) =1.35, p > .1]. However, this could be duesimply to low power. To clarify the results, separateANOVAs were computed for each passage type. For thebalanced and surface subject passages, the results werecomparable to the overall analysis. However, the biasedpassages were rather different from the other two types.The interaction between item of recall and cue was againsignificant [F(1,56) = 8.15, p<.OI], but, as shown inTable 5, it followed a pattern different from that in theother passage types. Inspection of the means for eachversion, shown in Table 6, shows that the dominantfactor in these passages was the difference between theA items, which were the general terms preferred astopic choices, and the specific, but more memorable,B items. Overall, propositions about the B items wererecalled much better than those about the A items

Table SMean Proportion of Recall About Each Item

for Each Passage Type

Passage Type

Item of SurfaceCue Recall Balanced Biased Subject Mean

Topic Topic .19 .23 .17 .20Nontopic .12 .12 .12 .12

Nontopic Topic .15 .17 .12 .15Nontopic .15 .24 .13 .17

Note-The standard error of these means is about .03 forbetween-subjects comparisons and about .02 for within-subjectscomparisons.

Table 6Mean Proportion of Recall About Each Item

for Biased Passages

VersionItem of

Cue Recall A B Mean

Topic Topic .20 .25 .23Nontopic .26 .05 .12

Nontopic Topic .12 .23 .17Nontopic .30 .18 .24

Note-The standard error of these means is about .04 forbetween-subjects comparisons and about .03 for within-subjectscomparisons.

(26% vs. 14%). Recall of the B item was always fairlyhigh, but recall of the A item was better when the cuewas A, such as with the topic cue for the A versionor with the nontopic cue in the B version, than when thecue was B (19% vs. 9%). Recall of A was also slightlybetter in the A-first version than in the B-first version(16% vs. 12%). Hence the effect of marking producedby initial mention in the biased passages was present, butit was not very important in determining the pattern ofrecall compared with the semantic differences betweenthe items.

Order of recall. The order of recall of passage infor­mation was scored as follows: For each propositionscored as recalled, the serial number of the subject'ssentence containing that proposition was noted. Henceif a subject recalled a certain proposition in his or herthird sentence in the recall protocol, that propositionwas assigned a recall sentence number of 3. Protocolscontaining either no recall or only one sentence ofrecall for a passage were considered to have undefinedorder information, and so, they constituted missingdata points.

The order of recall proved to be closely related to theorder of information in the presented passage. The meansentence number of recall of each proposition wascomputed, and the sentence number in the originalpassage in which the proposition appeared wasnoted. Thisyielded a total of 60 data points, consisting of originalpassage sentence number and mean recall sentencenumber, for the five original passage sentences, threepassage types, two passage versions, and two recall cues.The correlation between sentence of presentation andsentence of recall was high (r = .843, p < .01), suggestingthat subjects recalled the passage propositions in essen­tially the same order as they appeared in the originalpassage.

To explore the recall order data further, a simplemultiple-regression analysis was done to assess the rela­tive importance of version and cue on recall sentenceorder of each item. The mean recall sentence numberswere converted to a proportionate measure by dividingthem by the number of sentences in the recall protocol.A dummy-coded variable, TOPIC, was defined to repre-

sent whether the item of recall was the item marked asthe topic in the passage, and another, CUE, indicatedwhether the item of recall was cued. The data pointswere the mean proportionate sentence of recall for eachitem of recall for each subject in each condition, averagedover passages within types, giving a potential 120 datapoints per passage type. If the order of recall value wasmissing for one of the two passages, the value from theother was used as the average value for both. Cases inwhich the sentence of recall was missing for both pas­sages were not included. A separate forward stepwiseregression analysis was done for each passage type.

For the balanced passages (74 data points), TOPICmade a significant contribution [R2 = .28; F(1,72) =28.35, P < .01]. The contribution of CUE was nonsig­nificant (p> .1). For the biased passages(80 data points),TOPIC was again significant [R2 = .17; F(1 ,78) = 16.33,P < .01], and CUE was again nonsignificant (p > .1).Finally, in the surface subject passages, neither TOPICnor CUE contributed significantly to predicting recallposition, which would be expected, because there wasno manipulation of sentence order in these passages.These results, together with the simple correlation pre­sented above, argue that recall order is determined pri­marily by presentation order in the original passage. Therecall cue has little or no influence on recall order. Thussubjects recalled a passagein roughly the same order as itoriginally appeared in, regardless of what was cued.

Finally, the possibility that some of the recall effectsnoted above could be due to recall order was exploredby computing analyses of covariance for each passagetype, in which the covariate of the recall of each itemwas the mean proportionate sentence of recall of eachitem. The analyses agreed with those reported above onalmost all effects, but there were some differences.However, these discrepancies are almost certainlyartifactual, since in all but one case the covariate wasnot significant. Furthermore, there were many missingdata points, and the discrepancies appeared in the rawmeans of the complete cases used. Thus the effectsobserved on recall appear to be due to the topic and cuefactors and not simply products of modifications inrecall order.

DiscussionThe main results of Experiment 2 agree with those of

Experiment 1: Overall, a topicalized item is recalledbetter than a nontopicalized one, but this effect dependson the recall cue. A topic cue results in the topic item'sbeing recalled better than the nontopic item, but a non­topic cue results in little difference between recall ofthe two items. These effects will be discussed further inthe general discussion below. The remainder of thissection will be devoted to some of the side-productsof Experiment 2.

Distinction between topicality and memorability. Inthe biased passages, the main results did not appear.

TOPICALIZAnON EFFECTS 547

Recall of an item was highest if it was the cued item,regardless of which item was topicalized by the passageversion. However, the item biased as the topic (Item A)was not recalled better than the nontopic item; thecontrary was true. Hence these passages do not pro­duce a cue-topic effect corresponding to the balancedand surface subject passages. The effects observed inExperiment 1 are qualified by the presence of semanticdifferences between the passage items, but the sameeffects appear when these differences are eliminated.But the results for the biased passagesmake the interest­ing point that topicality and memorability of a passageitem are distinct characteristics. The abstract generalterms were heavily favored as passage topics in the topicchoice data, but the specific concrete items subsumedunder these topics were much better recalled. Thisargues that standing in the passage macrostructure is notthe only determinant of recall of a passage proposition.

Simple serial order hypothesis. The hypothesis waspresented above that the topicalization and cue effectsfrom Experiment 1 can be explained in terms of outputinterference resulting from either facilitation or disrup­tion of the subject's strategy of recalling the passageinformation in the same order as it was originally read.This hypothesis can be criticized as follows: First, underthis hypothesis, one would expect that the overalllevel of recall would be worse under the nontopic cuecondition than under the topic condition, since thenontopic cue would always produce disruption of recall,whereas the topic cue would not. However, the averagelevel of recall did not differ between cuing conditions.Second, the same class of effects was observed for thesurface subject passages, in which information about thetwo items was distributed almost uniformly throughoutthe passage. Finally, the order-of-recall data give littlesupport to the disruption hypothesis: Subjects madelittle effort to recall the cued item information first,but, rather, they generally tried to recall in the sameorder as the original passage. Thus the hypotheticaldisruption of recall order has no clear manifestations,and the observed effects should be attributed to altera­tions in the perceived topic of the passages, not todisturbances in the order of recall.

Reproduction of passage order. Apparently, subjectscan retain much of the original order of information in apassage, as shown by the similarity between input andoutput order in these results and as observed before byDeese and Kaufman (1957). However, it can be arguedthat this is not simply verbatim or surface-form memory.For one thing, a strict propositional scoring is verysensitive to near-verbatim recall, because highly para­phrased recall will disagree almost completely with theoriginal text at the level of individual propositions andtheir arguments under scoring rules of the type usedhere. This effect is due to the tendency of highly para­phrased recall to not preserve the original embeddingstructure of the presented propositions. However, if the

548 KIERAS

subject retains the syntactical form of the passagesentences, there will be considerable similarity in theembedding structure between the presented and recalledpropositions. So, if the recall had much verbatim orsurface-form content, the scored level of recall would beconsiderably higher. Thus the low level observed impliesthat there is little surface-form reproduction in the recallprotocols. Further support for this argument is based onwhich item appeared as the surface subject of sentencesin the recall protocols for the surface subject passages,in which one item was the subject of all of the sentences.Only 37% of the recall sentences were judged to have thesame referent for the surface subject as the originalpassage sentences had; 21% had the other candidatetopic as the surface subject; and 42% had some othersurface subject. Thus even this simple and distinctivefeature of sentence surface structure is not very wellretained.

How is the similarity between input and output orderexplained if there is little surface-level retention of thepassage? What readers could actually be remembering isthe order of information in the passage macrostructure.That is, the technical passages used in these experimentscould have some schema-like properties in which infor­mation is presented in some normative order (see Kieras,Note 4, for further discussion). For example, the firstsentence about a topic item was usually a fairly generalstatement and was followed by more detailed informa­tion. If the passage macrostructure had a tree-like formthat was recalled starting with the general information atthe top of the tree and proceeding to the detail infor­mation near the bottom of the tree, then the order ofoutput of the propositions would resemble the order ofthe propositions in the original passage, even withoutretention of the actual surface form.

GENERALDISCUSSION

Both intuition and the memory-location model fortopic and cue effects presented at the beginning of thispaper would predict that recall would be best overallfor topic cues and worst overall for nontopic cues.However, the results contradict this model, becausethere were no main effects of cue. Hence, in theseresults, what is being affected is not how much subjectsrecall from the passage, but rather, what subjects recall.This pattern is clearest in the balanced and surfacesubject passages in Experiment 2. When the recall cuewas the passage topic, the subject favored this topic inhis or her recall, but there was some recall about thenontopic item, as well. This argues that the subject hasmemory access to the item although it was neithermarked as the passage topic nor cued. In the nontopiccue condition, the two items seem to be roughly equallywell recalled in these passages. Since one item wasmarked as topical and the other was cued, perhapssubjects could have similar success in recalling them,

again implying that equal amounts of information aboutthe two items are in memory.

Hence the effect of topicalization in these passagesdoes not seem to be that of altering how much informa­tion was stored about each item. Rather, what could begoing on is that the topic marking is an aspect of thepassage that the subject stores more or less independentlyof the propositions retained. If so, then suppose that thereader is asked to recall the passage about a certain item(the cue) and also remembers that this item was markedas the topic of the passage. Then he or she interpretsthis situation as a task demand that this topical item ismore important to recall than the nontopic item, and sohe or she favors it in the recall effort. But if the cueditem is not the passage topic, the reader feels little needto favor one of the items over the other, and so recallof the two items is roughly equal. Thus the passage topicand recall cue act like instructions to the subject aboutwhat portions of the passage he or she should attemptto recall.

This explanation is in conflict with the macrostruc­ture theory presented above, in which topicalizationwould govern storage priorities. This theory has beensupported by many experiments in which informationhigher in the passage macrostructure was recalled betterthan information lower in the macrostructure. It seemsreasonable to think that a manipulation that alters thetopic of the passage would alter its macrostructure andso should change which information would be remem­bered better. The Britton, Meyer, Simpson, Holdredge,and Curry (1979) and the Kozminsky (1977Y studiesshow just such effects. But there are several differencesbetween the present experiments and those supportingthe macrostructure theory. First of all, the usual prosememory experiment is done with a free recall paradigmor with a cue corresponding to the single, unambiguouspassage topic. The passages usually used probably havea ''tighter'' macrostructure than the two-topic passagesused here, in that the structure would be organizedaround one main item and would have a relatively neattree-like structure. But the two-topic passages used herecould be viewed by subjects as having two distinct, butinterconnected, macrostructures, one organized aroundeach candidate topic, and one of these marked by thetextual surface. structure as being more thematicallyimportant than the other.

In processing an ordinary passage, the reader couldselect for storage the most important propositionsabout the single topic. In processing one of the two­topic passages, the reader could select for storage themost important propositions about each of the twocandidate topics and also remember which topic wassupposed to be more important. When asked to recall,the reader of an ordinary passage has no problem; anycue to the topic corresponds to the top level of thepassage information in memory. But the two-topicpassage reader must choose which of the two structures

to emphasize in recall. The relation of the originaltopic to the recall cue would be used to make thechoice.

If this characterization of the present results isaccepted, there are two consequences: One is that themacrostructure theory needs to be modified or expandedto take into account what the reader will do with com­plex technical passages that do not have an obviousstructure organized closely around a single topic. Clearly,some further experiments should be done on the prob­lem. The second consequence is that certain seeminglysimple questions about prose memory may not be easilyanswered. The original question addressed by thesestudies was simply whether the topic of a passagegoverns later recall, content difference being controlled.The two-topic passages seemed a good choice as amethod to allow altering which item was topical with­out making serious changes in the passage content. Butif readers process such passages completely differentlyfrom single-topic passages, the role of the perceivedpassage topic in the storage and retrieval of prose infor­mation cannot be so easily isolated.

REFERENCE NOTES

1. Kieras, D. E. Initial mention as a cue to the main idea andmain item of a technical passage (Tech. Rep. 3). University ofArizona, July 1979.

2. Bovair, S., & Kieras, D. E. A guide to propositional analysisfor research on technical prose (Tech. Rep. 8). University ofArizona, in preparation.

3. Turner, A., & Greene, E. The construction and use of apropositional text base (Tech. Rep. 63). University of Colorado,Institute for the Study of Intellectual Behavior, April 1977.

4. Kieras, D. E. Abstracting main ideasfrom technical prose: Apreliminary study of six passages (Tech. Rep. 5). University ofArizona, August 1980.

REFERENCES

BRITION, B. K., MEYER, B. J. F., SIMPSON, R., HOLDREDGE,T. S., & CURRY, C. Effects of the organization of text onmemory: Tests of two implications of a selective attentionhypothesis. Journal of Experimental Psychology: HumanLearning&:Memory, 1979,5,496-506.

CLARK, H. H., COHEN, J., SMITH, J. E. K., & KEPPEL, G.Discussion of Wike and Church's comments. Journal of VerbalLearning and VerbalBehavior, 1976, 15, 257-265.

TOPICALIZATION EFFECTS 549

CLEMENTS, P. The effects of staging on recall from prose. InR. O. Freedle (Ed.), New directions in discourse processing.Norwood, N.J: Ablex, 1979.

DEESE,J., & KAUFMAN, R. A. Serial effects in recall of unorgan­ized and sequentially organized verbal material. Journal ofExperimental Psychology, 1957,54,180-187.

JOHNSON, R. E. Recall of prose as a function of the structuralimportance of the linguistic units. Journal of Verbal Learningand VerbalBehavior, 1970,9, 12-20.

KIERAS, D. E. Initial mention as a signal to thematic content intechnical passages. Memory &: Cognition, 1980, II, 345-353.

KIERAS, D. E. The role of major referents and sentence topicsin the construction of passage macrostructure. DiscourseProcesses, 1981,4,1-15.

KINTSCH, W. The representationof meaning in memory. Hillsdale,N.J: Erlbaum, 1974.

KINTSCH, W. On recalling stories. In M. Just & P. Carpenter(Eds.), Cognitive processes in comprehension. Hillsdale, N.J:Erlbaum, 1977.

KINTSCH, W., & KEENAN, J. Reading rate and retention as afunction of the number of propositions in the base structureof sentences. Cognitive Psychology, 1973,5,257-274.

KINTSCH, W., KOZMINSKY, E., STREBY, W. J., McKooN, G., &KEENAN, J. M. Comprehension and recall of text as a functionof a content variable. Journal of Verbal Learning and VerbalBehavior, 1975, 14, 196-214.

KINTSCH, W., & VAN DIJK, T. A. Toward a model of discoursecomprehension and production. Psychological Review, 1978,115, 363-394.

KOZMINSKY, E. Altering comprehension: The effect of biasingtitles on text comprehension. Memory &: Cognition, 1977, 5,482-490.

MEYER, B. J. F. What is remembered from prose: A functionof passage structure. In R. O. Freedle (Ed.), Discourse produc­tion and comprehension: Advances in research and theory(Vol. 1). Norwood, N.J: Ablex, 1977.

PERFETTI, C. A., & GOLDMAN, S. R. Thematization and sentenceretrieval. Journal of Verbal Learning and Verbal Behavior,1974,13,70-79.

PERFETTI, C. A., & GOLDMAN, S. R. Discourse functions ofthematization and topicalization. Journal of PsycholinguisticResearch, 1975,4,257-271.

VAN DIJK, T. A. Semantic macro-structures and knowledge framesin discourse comprehension. In M. Just & P. Carpenter (Eds.),Cognitive processes in comprehension. Hillsdale, N.J: Erlbaum,1977.

VAN DIJK, T. A. Relevanceassignment in discourse comprehension.Discourse Processes, 1979,1, 113-126.

WIKE, E. L., & CHURCH, J. D. Comments on Clark's "Thelanguage-as-fixed-effect fallacy." Journal of Verbal Learningand VerbalBehavior, 1976, IS, 249-256.

(Received for publication October IS, 1980;revision accepted May 29, 1981.)