Flashcards Revisited: Training Poor Readers to Read · PDF fileFLASHCARDS REVISITED 277 they found no improvement in the reading abilities of poor readers after they received word

Journal of Educational Psychology1997, Vol. 89, No. 2, 276-288

Copyright 1997 by the American Psychological Association, Inc.0022-O663/97/S3.0O

Flashcards Revisited: Training Poor Readers to Read Words FasterImproves Their Comprehension of Text

Annette Tan and Tom NicholsonUniversity of Auckland

Forty-two below-average readers, between 7 and 10 years of age, were given single-wordtraining, phrase training, or no training. Trained children learned to decode target wordsquickly and accurately, using flashcards; untrained children only discussed the target wordsand read them once. Trained and untrained children read aloud passages containing targetwords and were tested on their comprehension. Trained children had better comprehensionthan did the untrained children when questioned about passages and asked to retell them.Results suggest that an emphasis on rapid word recognition benefits poor readers.

Reading is a multicomponent skill whereby the reader hasto use a number of different cognitive processes involvingword recognition, access of word meanings, parsing ofsentences, semantic analysis of sentences, and interpretationof the overall text. Many of these linguistic processes arealready automatic in that they demand little or no cognitiveeffort for the native speaker, inasmuch as they are part ofgeneral language comprehension. One process that is notautomatic, and that has to be taught to beginning readers, isword recognition. This skill takes several years to learn, andeven then most pupils will not have the speed and fluency ofskilled adult readers (Gough & Hiilinger, 1980). The au-tomatization of word recognition usually comes after a longperiod of extensive reading practice.

In designing the present study, we reasoned that a focuson the development of automatic word recognition skillswould be most likely to have a positive effect on poorreaders' comprehension, as word recognition is one aspectof the reading process that is very difficult for them (Adams,1990; Juel, 1988, 1994; Nicholson, 1991; Stanovich, 1986).

The theoretical rationale for the study comes partly fromLaBerge and Samuels's (1974) model of automatic infor-mation processing in reading. They argued that to developreading proficiency, conscious attention is first needed inorder to recognize words. Attention-free (i.e., automatic)word recognition is later achieved through learning andpractice. One implication of their model is that if wordrecognition consumes too much mental attention, then theextra effort taken in recognizing words will detract fromcomprehension at sentence, paragraph, and text levels. In-dividuals possess limited attentional resources, and the tasksof decoding and comprehending text both require these

Annette Tan and Tom Nicholson, Department of Education,University of Auckland, Auckland, New Zealand.

We are sincerely grateful to the children who participated in thestudy, as well as to their parents, their teachers, and their schoolprincipal, who kindly allowed us to carry out this study.

Correspondence concerning this article should be addressed toTom Nicholson, Department of Education, University of Auck-land, Private Bag 92019, Auckland, New Zealand. Electronic mailmay be sent to [email protected].

limited resources. Hence, by putting pressure on attentionalcapacity in order to decode words, less attention will beavailable to process the meaning of the text.

The theoretical rationale for the study is also linked to the"bottleneck hypothesis" put forth by Perfetti (1977) andPerfetti and Hogaboam (1975) to explain the relationshipbetween word recognition (i.e., decoding), speed of reading,and reading comprehension. According to this hypothesis,inefficient decoding processes will take an excessive shareof mental resources available for comprehension, and willproduce less efficient comprehension. This proposal hasbeen further elaborated by Perfetti and Lesgold (1977) andPerfetti (1985) as "verbal efficiency theory," which empha-sizes rapid automatic decoding as a primary factor in read-ing comprehension.

The relationship between fast decoding and comprehen-sion is not simply that the faster you read the better youcomprehend. Carver (1990) has shown that, in fact, indi-viduals increase their comprehension when they are givenmore time to read. This takes into account the influence oftext difficulty. Some texts will take more or less time tocomprehend. The notion of verbal efficiency, however, ap-plies only to word recognition, which needs to be automatic,enabling full use of one's cognitive resources for compre-hension, which may involve varying amounts of time, de-pending on the difficulty of the text.

It has been argued (Gough & Hiilinger, 1980) that readingconsists of decoding skill and linguistic comprehension.Gough and Tunmer (1986) have called this the "simpleview" of reading. In this view, decoding skill is independentof linguistic comprehension, in that decoding involves theability to read pseudowords such as "poon" or "cosnuv,"which have no meaning at all. Likewise, linguistic compre-hension involves processes that are independent of decod-ing, in that they are essentially the same processes thatoperate when one is listening.

However, there is evidence to suggest that decoding is notindependent of linguistic comprehension (Cromer, 1970;Cramer & Wiener, 1966; Oaken, Wiener, & Cromer, 1978).According to Oaken et al., "a high level of identificationskills may not be a sufficient condition for the occurrence ofadequate reading comprehension" (p. 72). In their study,

276

FLASHCARDS REVISITED 277

they found no improvement in the reading abilities of poorreaders after they received word identification training in-dependent of context. Cromer (1970) has argued that an-other requirement is necessary, that learning to read must beaugmented by the ability to group words meaningfully intopatterns (e.g., phrases, sentences). In this way, pupils aremade aware that words exist in larger language units andthat words are not to be learned as unrelated lists that areunconnected with one another.

A simple test of the importance of decoding as an inde-pendent skill would be to train students to recognize wordsmore rapidly and to see if this leads to improvements inreading comprehension. This has been done in a number ofstudies, with inconclusive results. In an often-cited study,Fleisher, Jenkins, and Pany (1979) trained fourth- and fifth-grade poor readers in rapid decoding by using flashcards. Inone experiment, they used single-word training; in another,they used phrases. After training, the pupils read a passagecontaining the key words. The training did improve wordrecognition speed and accuracy for the trained words, butthis did not transfer to improvements in readingcomprehension.

The Fleisher et al. (1979) results have been questioned byBlanchard (1980), Blanchard and McNinch (1980), andHolt-Ochsner and Manis (1992). Yet, inconclusive resultssimilar to those of Fleisher et al. have been obtained bySamuels, Dahl, and Archwamety (1974), using similarsingle-word training procedures. Yuill and Oakhill (1988,1991) have also reported inconclusive results of training inrapid decoding. Their findings suggested that the problemsof poor comprehenders were not a result of poor decoding.Statistically nonsignificant results of rapid decoding prac-tice have also been reported by Van Den Bosch, Van Bon,and Schreuder (1995), although the training in their studyinvolved pseudowords rather than real words, and the com-prehension task involved semantic verification of true anduntrue sentences rather than reading of passages.

The failure of these training studies to improve poorreaders* comprehension of text has led to some negativeconclusions in textbooks on the psychology of reading.Rayner and Pollatsek (1989), for example, stated that"Merely training a child to say words quickly will notnecessarily result in improved comprehension" (p. 391). Asimilar conclusion was reached by Just and Carpenter(1987), who commented that "Efficient word recognition isnot sufficient for good reading. A number of training studiesthat improved the word-recognition speed of poor readersdid not find commensurate increases in their reading level"(pp. 458-459).

However, we recently replicated the Fleisher et al. (1979).study with a small sample of adult English-as-a-second-language learners (N = 4), using a multiple-baseline re-peated measures design, and were able to obtain gains incomprehension, although the gains did not continue in thereversal phase (Tan, Moore, Dixon, & Nicholson, 1994).These data suggested that further research was warranted.

Hence, the present study took another look at the effectsof training in rapid decoding by repeating once again workdone by Fleisher et al. (1979). Our study differed from the

original, however, in several ways. First, there was morecontrast between the single-word training and the phrasetraining, in that the phrase training used short sentences aswell as phrases. Second, pupils were given more extensivetraining, involving more passages. Third, the test passageswere less difficult to understand. Fourth, the training ofisolated words was accompanied by some explanation of themeanings of the words when it was apparent that pupils didnot know what the words meant. As a result, pupils in thisstudy may have had a better understanding of the meaningsof the trained words than in the original study.

In other respects, the present study was similar to Fleisheret al. (1979) in that the training involved poor readers andflashcards, and the assessment of reading comprehensioninvolved the reading of passages, answering questions, andrecalling what each passage was about. It was expected thatsingle-word training may not have a positive effect oncomprehension, which was the finding of previous research-ers. However, it was expected that phrase training, whichemphasized context, might have positive effects.

Method

Participants

There were 42 pupils in the study (24 boys and 18 girls),including twelve 7-year-olds, twelve 8-year-olds, nine 9-year-olds,and nine 10-year-olds. The pupils were all from the one school,which was located in a low-income suburb in Auckland, NewZealand, and was attended by pupils from a range of culturalgroups. The sample of pupils selected for the study came from agroup of 54 children who were identified by the school as below-average readers, based on their performance on an informal proseinventory, which was made up of a series of graded passages(Department of Education, 1983). As a result of further screeningwith other reading tests, 12 pupils were dropped from the originalsample, leaving a final sample of 42.

Screening Measures

The main pretest screening measure was the Neale Analysis ofReading Ability (Neale, 1989), which was used to measure readingcomprehension, reading rate, and reading accuracy levels. The testconsisted of six graded passages for use in assessing pupils in the6-12-year age range. In this test each passage was read orally bythe pupil. This task was followed by a set of questions to assesscomprehension. According to the test manual, the Neale test hasgood reliability ranging from ,81 to .93, at each age level.

Pupils were also assessed for phonemic awareness, using theRoper test of phonemic awareness (Roper, 1984), which has 42items. An easy item on this test involves segmentation (e.g., Say"at." What are the 2 sounds in "at"?). A hard item on the testinvolves deletion and substitution (e.g., Say "run." Instead of "n,"end the word with "g"). According to Juel, Griffith, and Gough(1986), this test has reliabilities greater than .70.

The final screening measure was the Bryant Test of BasicDecoding Skills (Bryant, 1975), which has 50 items. The first 20words are simple consonant-vowel-consonant words (e.g., "fev");the next 20 are more complex single-syllable words (e.g., "blor").The final 10 words are multisyllable (e.g., "uncabeness"). Accord-ing to Juel (1988), reliabilities range from .90 to .96.

278 TAN AND NICHOLSON

Design

The purpose of the screening tests was to select three matchedgroups. The selection was done in two phases. First, pupils withineach chronological age group who had similar comprehensionscores, varying by no more than 5 months in reading age, wereassigned to matched triplets. Each pupil in the triplet was thenrandomly assigned to one of three conditions: single-word training,sentence training, or a control condition. The other two screeningmeasures acted as a check on the initial selection. In each of threetraining conditions of 14 children, there were four 7-year-olds,four 8-year-olds, three 9-year-olds, and three 10-year-olds. Each ofthe pupils in any one condition was matched with another ofsimilar chronological and reading age, located in each of the othergroups. In short, each condition included the same number ofpupils at each chronological age, who were each in turn matchedwith pupils of a similar reading level in the other conditions.

Training Program

Each pupil in each training condition was given five trainingsessions; each session lasted about 20 min. Each pupil was trainedand assessed individually. There was no group training or assess-ment. Each training session focused on words from a single story.The five sessions thus covered 5 stories. Because there was a rangeof ability and age levels within each group, 25 different storieswere used, selected from the New Zealand School Journal for theyears 1975-1987 (Ministry of Education, 1989), as well as fromthree commercial reading programs: Literacy 2000, LiteracyLinks, and Jelly Bean Books. The stories varied in length accord-ing to the reading level of pupils in each group. At the 6-yearreading age level, the length was 100-300 words; at the 7-yearreading age level, it was 200-300 words; at the 8-year level, it was300-500 words; and at the 9-year level, it was 350-450 words.

For each pupil, only stories that were suitable for their readingage were given to them, based on grading provided by the SchoolJournal Catalogue (Ministry of Education, 1989) and manualsaccompanying the commercial materials. Although a number ofdifferent stories were used, the matched triplets in each trainingcondition were each given the same stories, or parts of stories, toread. In this way, the effect of story content and difficulty wascontrolled.

The words to be trained varied in number. The general rule wasto select 7-8% of the total number of words in the passage. Forexample, a passage with 120 words had 10 words selected from itto form the word list, whereas another passage of 300 words had20 words selected. The list words ranged in number from 10 to 25.The words were selected as a result of pilot trials of the passageswith several pupils who were not part of the study. These pupilswere asked to read the passages orally. Their misreadings of wordswere used as a guide to the selection of target words. Otherwords were selected by us because they seemed as if they mightprovide some difficulty.

To assess comprehension, 12 questions were devised for eachpassage; 8 were factual (explicit), and 4 were inferential (implicit).The explicit questions could be answered simply by rememberingwhat was in the passage. The implicit questions required the pupilsto combine text details with their own general knowledge, so as tomake inferences and draw conclusions (Nicholson & Imlach,1981; Pearson & Johnson, 1978). These questions were designedto quiz pupils on what was in the text, without focusing too muchon prior knowledge, as prior knowledge questions can often beanswered without reading the text at all (Pearson & Johnson, 1978;Raphael & Wonnacott, 1985). All of the questions were checked

by two other persons to ensure that they fit the explicit-implicitcategories.

The process of preparing the explicit and implicit questions wasdone independently of the selection of the target training words,but sometimes the answers to questions did include training words.The extent to which this happened varied from passage to passage.In some passages, only one or two target words were part of theexpected answers, whereas in other passages, up to five or sixtarget words were part of the expected answers.

Procedure

In the single-word training condition, each pupil was taught torecognize each target word by using flashcards. Each word, printedon an index card, was shown to the pupil to be read aloud. If thepupil could not say the word, it was then shown in a two-wordphrase, which was written on the back of the index card, to helpunderstand its meaning (e.g., if the word was "lemonade," thetwo-word phrase was "lemonade drink"). These phrase contextswere not taken from the test passages, but were different contextsaltogether, so as not to give pupils any cues as to the meaning ofthe test passages. In addition to this minimal context, the moving-thumb-across-the-word method was also used, in which the pupilread aloud one bit of the word at a time, so as to pronounce theword correctly. These short context cues were given only when thepupil had trouble with the meaning of the word during the first trialof list words. Once this was clear, the pupil was trained to readeach word on his or her own, without the extra context. The phrasecontexts took up only a small proportion of the training time. Thevast majority of training time involved the use of single-wordflashcards.

The flashcard training continued until the pupil could recognizethe word on each card in approximately 1 s. This was usuallyachieved within a 20-min time frame, which was the time allocatedto each training session. After the training, a randomly orderedword list, made up of the target words, was given to the pupil, toread as fast and as accurately as possible. Every effort was madeto have pupils read each target word in less than a second. Thecriterion rate in Fleisher et al. (1979) was 90 words per minute orless, with at least 95% accuracy. In our study, if pupils did notmeet this criterion, training would continue, along with retestingon a differently ordered list until the criterion was approximated,or until training time ran out, whichever occurred first. If the pupilcould read within one word per second, this was acceptable. Thetraining on the word lists was kept to within approximately 20 minfor each session, which meant that there was only a limited amountof time to bring about an increase in speed. As a result, the Fleisheret al. criterion was not necessarily achieved for each pupil.

After the word list, the pupil was given the passage from whichthe target words had been taken and was asked to read it aloud. Thepupil was told that some questions would be asked about thepassage after it was read. The time to read the passage and thenumber of errors made were both recorded. If an error occurredwhile the pupil was reading aloud, the pupil was not usually toldthe correct word, although some helpful prompting was given ifthe pupil halted on a word while reading. Prompting was done ingeneral terms (e.g., "try to work it out if you can"), without givingclues that would directly assist word recognition. Words recog-nized correctly with prompting were scored as correct. The pupilwas then asked orally 12 comprehension questions relating to diestory. The pupil would respond orally rather than in writing. Allresponses were recorded. The session ended by asking the pupil toretell the story orally. To help start recall, standardized promptswere used (e.g., "If you go back to class, and the teacher asks you


what the story was about, what will you say? It was about.. .? Andthen . . . ? What happened next . . .? After that.. .?")• If theseprompts failed, other prompts were as follows: "What happened inthe first part of the story? The middle part? The last part?" Theresponses were recorded for later analysis.

In the phrase-training condition, pupils were shown mostlyphrase cards, but sometimes sentence cards, containing the targetwords from the passage they were later to read aloud. The phrasesor sentences were designed to show the meanings of each targetword, taking care that none resembled phrases or sentences fromthe test passages or from the questions following the passages. Thetarget words were not underlined or highlighted. The pupil wasinstead shown the target word in each phrase or sentence and toldthat it was an important word to focus on. For example, the targetword "lemonade" was taught as part of a long phrase written on theflashcard: "A cool lemonade drink." This was not a phrase thatoccurred in the test passages. In the case of the target word"raspberry," this was taught as part of a sentence written on theflashcard: "I like raspberry jam on bread." As in the single-wordtraining, pupils were given help with pronunciation if necessary.Practice in reading these phrase or sentence flashcards continuedthroughout the 20-min training session, with the aim of achievingthe preset criterion rate of 90 words per minute. At the end of the20-min session, the phrase-trained pupils were given the list oftarget words to read. As in the single-word training, time to readthe list and number of errors made were both recorded. The pupilswere then given the passage corresponding to the target words,asked to read it aloud, answer comprehension questions, andfinally retell the passage in their own words.

In the control condition, pupils were trained for the sameamount of time as for the other conditions, although they did notsee the written forms of target words until they were tested on arandom list at the end of the training period. The target words wereeach read aloud to the pupil, who was then asked to explain whatthe words meant (e.g., "What does lemonade mean to you? Canyou say it in a sentence? If you can't, I'll say it in a sentence andyou tell me what it means in my sentence"). After listening to anddiscussing the words for about 20 min, the pupil was given arandomly ordered list of the words that had been discussed andwas then told to read them as quickly and as accurately as possible.Only one list reading was given to each pupil. As was done for theother groups, reading time and number of errors made were re-corded. Each pupil read the corresponding passage, answeredquestions, and retold what he or she could remember of it. Thecontrol children were thus exposed to the same words, taught forthe same amount of time, and were able to discuss the meanings ofthe words, as was the case for the other groups. However, theywere not given the same training in rapid word recognition.

The training sessions for all three groups were approximately 20min in duration. As in the Fleisher et al. (1979) study, pupils whodid not meet the 90-words-per-minute criterion on the first listtesting were given more training and then were retested, using asecond, randomly ordered list of the same target words. This wasdone only if the first list testing was completed within the 20-mintime limit, in which there was time allowed for additional training.If pupils met the 90-words-per-minute criterion on the first listtesting, they were not retested. The control children, however,were given only the one testing on each list. Unlike pupils in theother conditions, they were not expected to reach a preset criterion.The criterion of 90 words per minute was not necessarily achievedby each pupil, inasmuch as there was a 20-min time limit to thetraining sessions.

At the end of each of the five training sessions, each child in thestudy was given a questionnaire that had a Likert scale format.

There were four questions, including "Did you like the lesson?"and "Would you like to come to another lesson?" After eachquestion, the children were shown five different pictures thatillustrated a dog in various postures ranging from excited to sad.The children were asked to put a circle around the picture of the"puppydog" that matched their feelings. They then took the ques-tionnaire back to class to share with their teacher. The purpose ofthe questionnaire was to ensure that the pupils were enjoying thesessions and were not getting bored or dispirited as a result of thespecial teaching, especially the flashcard practice. Teacherschecked the questionnaires after each session to ensure that pupilswere not getting bored or dispirited. Pupils responded well to thetraining sessions. They were happy to compete against themselves,in a one-to-one situation, to see how many list words they couldread at the end of the training session.

The 12 comprehension questions (8 explicit, 4 implicit) werescored in two ways. In the first way, the answer had to be exactlycorrect. This was the strict scoring criterion. In the second way,answers were scored as correct if they were fairly close. This wasthe lenient criterion. These two scoring methods were intended topick up precise as well as reasonable comprehension. The passagerecall responses were scored out of a total of 8, with separatescores for details (up to 4 points given) and for overall gist ormessage (up to 4 points).

A detailed marking scale was constructed for both the compre-hension questions and the recall task. The marking scheme wasused as a guide for all of the children's responses. For strictscoring, the pupil's response needed to be very close to the literaltext. For example, in one story with imaginary events, childrenwere asked, "What was the hat made of?" (in the passage, it wasmade of orange peel). With the strict scoring criterion, only "or-ange peel" was acceptable. An incorrect answer would be, forexample, a response such as "banana skin." With the lenientscoring criterion, a reasonably close answer was acceptable. Forexample, another question was "How did I get up 10 stories high?"The words in the passage were, "I was allowed to choose somespecial shoes which walked on walls and ceilings. I went up likea fly, 10 stories high. It was a fantastic feeling." The correctanswer was "because of my special shoes." But if the pupil said,"with the sticky bit on the bottom," this was accepted in terms ofthe lenient criterion. The answer was plausible, although it wasinformation that was not stated in the passage and had beeninferred by the pupil.

The marking scheme for recall responses was also very explicit.For gist scoring, there were four main points that were scored foreach passage. Pupils got a mark for each point. The method forscoring children's recall of details was to give points for number ofdetails, from a score of 1 for one detail, to a score of 4 for morethan six details.

All responses were scored by the researcher and another marker,with very few instances of disagreement. The second marker wasuninformed with respect to the training conditions. The answers tothe comprehension questions were explicitly set out in the markingscheme. The answers expected for gist recall were also set out inthe marking scheme. Cases in which differences of opinion oc-curred were resolved through discussion until there was 100%agreement.

Results

The results of the screening tests are shown in Table 1.The data for reading rate, accuracy, and comprehension areexpressed as reading ages. Several pupils did not achieveraw scores high enough to be assigned a reading age, and so


Table 1Means and Standard Deviations of Pretest Measures

Condition Analysis of variance

Measure

Single word Phrase Control

M SD M SD M SD

Chronological age7 years8 years9 years

10 yearsNeale reading rate3

7 years8 years9 years

10 yearsNeale accuracy6


10 yearsNeale comprehension


10 yearsBryant Test of Basic

Decoding Skills'"7 years8 years9 years

10 yearsRoper Phonemic

Awareness Testd


10 yearsNote. Age distribution in*N = 36, t e r r o r = 24. b

*p< .05. **p < .01.

ConditionF(2, 30)

AgeF(3, 30)

InteractionF(6, 30)

7.448.299.31

10.14

6.756.006.477.56

5.755.946.567.44

5.336.276.787.72

1.255.25

23.3326.67

25.5028.7537.6738.67

0.190.200.100.24

0.980.301.250.59

0.360.590.341.72

0.350.890.211.47

1.894.118.15

17.90

6.616.551.533.51

7.468.279.50

10.08

6.046.357.867.61

5.586.237.617.03

5.446.336.647.64

8.255.25

21.6716.00

32.2526.7531.3333.33

0.340.300.080.08

0.881.781.660.71

0.431.120.840.39

0.461.060.591.30

8.625.74

18.883.61

6.451.897.101.53

7.258.259.28

10.31

6.945.946.647.81

5.566.286.786.81

5.236.386.727.17

3.256.00

22.6724.33

23.7529.7535.6734.67

0.290.220.260.26

1.200.051.971.97

0.130.351.561.29

0.320.700.241.13

5.857.12

13.2817.47

8.465.564.933.22

0.21

0.13

0.29

0.15

0.07

0.44

281.30**

2.35

5.67**

13.06**

10.23**

6.39**

0.75

0.41

0.49

0.13

0.47

1.50

•olds, N = 4; 8-year-olds, N = 4; 9-year-olds, N = 3; 10-year-olds, N — 3,27. c Maximum score for Bryant test is 50. d Maximum score for Roper test is 42.

each group: 7-year-N = 39, df error =

they were dropped from the analysis. For reading rate therewere 2 such pupils in each condition; for reading accuracythere was 1 pupil in each condition. This reduced the sampleto 36 for reading rate and 39 for reading accuracy.

Two-way analyses of variance (ANOVAs) were carriedout for each of the screening measures, looking at the effectsof conditions, age, and their interaction. There were nostatistical differences among the three training conditions onany of the measures, although there were differences amongthe age groups on all of the measures except rate of reading(p = .10). There were no interactions for any of the mea-sures, indicating that reading levels within age groups wereevenly spread across the training conditions.

These screening results suggested that the matching pro-cedures in allocating pupils to the three training conditionshad produced three groups of children with similar readinglevels. The results also confirmed the status of these chil-dren as poor readers, in that the reading ages of children ineach group as measured by the Neale Analysis of ReadingAbility were well below their chronological ages.

The Bryant Test of Basic Decoding Skills, which is not a

standardized test, has 50 items. The results showed very lowscores for children in all conditions. For example, Juel et al.(1986), using survey data from a sample of 80 second-gradechildren, reported a mean score of 30.6 for the Bryant test atthe end of second grade. In contrast, mean scores for chil-dren in this study at the 7-year-old chronological age levelwere 1.25 for the single-word condition, 8.25 for the phrasecondition, and 3.25 for the control condition. The meanscores of older children in this study were also in line withthe mean scores of poor readers reported by Juel (1988).

The phonemic awareness test, which also is not standard-ized, has 42 items. Results showed that the pupils in thisstudy were approaching ceiling at the 9- and 10-year-oldchronological age level. However, the mean scores of 7- and8-year-old pupils in this study were still well below themean of 36.3 for end of second grade, as reported by Juel etal. (1986).

In this study, we used a randomized block design, involv-ing 42 pupils assigned in matched triplets to three condi-tions. This meant that the triplets could be incorporated intoeach analysis as a blocking factor. The design thus involved


14 blocks, three conditions within blocks, and five repeatedmeasures corresponding either to lists or passages. Becausethe triplet information was not relevant to the goals of thisstudy, we will report only the main effect of conditions, themeasures linear trend component, and their interaction. Forall analyses, significance probabilities of p < .05 andp <.01 are used.

The results for speed and accuracy are shown in Table 2.A series of repeated measures ANOVAs were carried outfor each dependent measure, that is, passage and list speed(reported as words per minute), overall passage accuracy(reported as percentages), and proportion of correctly readlist words, also read correctly in passages (reported aspercentages). Means for accuracy of reading words in listsare reported in the text. No analyses were carried outinasmuch as there were only perfect scores (i.e., no vari-ance) in the single-word and phrase conditions. (The corre-sponding mean level of correct list-word reading in thecontrol condition was about 50%.)

There was no difference among the three conditions forspeed of reading in context (i.e., the passages read aloud).Trie nonsignificant conditions effect appears unusual but isunderstandable if we consider that the trained words repre-sented no more than 8% of the passage words. Thus, any

effects of the training would be washed out by the largenumber of untrained words that had to be read aloud in thetest passages. Although the linear component of the passagemain effect was not statistically significant (p = .07), theinteraction revealed that the speed of children in the single-word condition increased more from Passage 1 to Passage 5relative to tnat of children in the control condition, f(13) =2.61, p < .025 (see Figure 1).

The results for speed of list reading showed only a con-ditions main effect. Follow-up contrasts, using Fisher's leastsignificant difference (LSD) procedure (Levin, Serlin, &Seaman, 1994), showed that the single-word and phrase-training groups had statistically higher scores than did thecontrol group. There were no passage or interaction effects.

The mean scores for reading accuracy of list words inpassages (not presented in Table 2) showed an advantagefor the training conditions over the control condition. Thetraining conditions had perfect scores (100%) for all of thelist words in all five passages, whereas the control conditiondid less well (Passage 1: M = 52.43, SD = 32.56; Passage2: M = 60.36, SD = 30.07; Passage 3: M = 46.54, SD =29.93; Passage 4: M = AIM, SD - 32.43; and Passage 5:M = 48.70, SD = 29.65).

The results for accuracy of passage reading produced a

Table 2Means and Standard Deviations of Speed and Accuracy

Measure

Speed (wpm)Passage

12345Across passages

List12345Across lists

Accuracy (%)Passage

12345Across passages

List/passage

2345Across list/passage

Single

M

55.8259.1660.7761.0666.7360.71

54.4354.5852.3252.9953.8553.64

93.3893.8293.0794.2495.9394.09

90.6492.5795.0094.7194.7993.54

word

SD

15.7316.4918.0417.8417.3414.65

24.5918.5312.5212.5011.7114.57

4.083.323.913.372.482.83

7.958.635.455.734.612.43

Condition

Phrase

M

69.3572.4467.3771.1873.6270.79

62.0373.6060.8664.0166.1565.33

93.5495.5794.1895.3197.2095.16

90.7997.3695.4397.0796.3695.40

SD

31.3734.7535.4634.7031.2932.97

35.4038.1723.4430.0628.4129.77

4.933.144.183.521.733.10

12.056.029.115.617.984.76

Control _ . . .Cnnrlifinn

M

57.3857.1951.3754.6457.1755.55

24.6227.5025.3426.2927.0526.16

88.2490.5686.6491.4991.3489.65

78.7979.5777.2995.2183.5782.89

SD F(2,26)

2.3726.6325.4322.3922.8426.5524.35

29.51**13.9212.018.878.29

10.7110.09

9.52**8.337.578.435.025.066.43

6.67*29.3227.7230.117.99

28.2215.84

Analysis of variance

Linear passageF(l, 13)

3.89

0.00

9.33**

5.62*

Condition X LinearPassage F(2, 26)

4.64*

0.22

0.18

0.84

Note, wpm = words per minute.*p< .05. **p< .01.


80

70

50

40

• Phrase

B Single word

• Control

Passage

Figure 1. Speed of passage reading (wpm = words per minute) across Passages 1 to 5 for thesingle-word, phrase, and control conditions.

conditions effect. Follow-up LSD contrasts showed thatboth groups of trained pupils were more accurate than thecontrol pupils. There was improvement across passages, andthis did not interact with conditions.

A possible concern with the results so far is that childrenin the two trained conditions were not only faster but alsomore accurate than were children in the control condition.Thus, any possible effects of the training on reading com-prehension could be attributed to differences in readingaccuracy rather than speed. To take account of this concern,we equated children in all three conditions with respect totheir individual list-reading performance. To do this werestricted our analysis to only those list items that pupils inall conditions read correctly. For each child we calculatedthe number of list items correct and then checked to see howmany of the correctly read list items were also read correctlyin the corresponding passage. We did this for each of the listand passage items. We then calculated the percentage ofcorrect list words that were also read correctly in passages.For example, if for one set of list and passage items, thechild read correctly 10 items out of 20 on the list, and thenread 6 of those 10 items correctly in the correspondingpassage, we would calculate a conditional percentage bydividing the 6 items read correctly in both list and passageby the 10 items read correctly in the list (i.e., 6:10) and thenmultiply by 100 to convert the proportion correct to apercentage correct (60%). Using this metric, we were ableto compare relative accuracy of list reading with relativeaccuracy of passage reading just for those items that eachchild had read correctly in lists.

If the effects of flashcard training were associated onlywith reading accuracy, then we would expect that childrenin all conditions would show a similar pattern of list andpassage percentage scores. If the effects of the trainingextended beyond accuracy, in that children in the two train-ing conditions had achieved a degree of automaticity ofrecognition for the trained items, then the trained children

would be able to transfer these skills to passage reading and,as a result, would be able to recognize correct list itemsmore easily in passages than would children in the controlcondition. The automaticity component of the flashcardtraining would have given children in the two trained con-ditions an edge over children in the control condition. Table2 (List/Passage section) shows results for the percentage ofwords read correctly in lists that were also read correctly inpassages. The ANOVA showed a conditions effect.Follow-up LSD contrasts showed that both training condi-tions had statistically superior scores to the control condi-tion. There was improvement over passages but no interac-tion of passages with conditions. This result revealed thatthe control pupils were less accurate than the trained pupilsin passage reading even for the words they read correctly inthe lists. Thus, it appears that flashcard training effectsextended beyond accuracy.

The trained pupils had learned the list words so well thatthey were able to transfer their list-reading skills to thepassages more effectively than were the control pupils.These findings suggest that even if all three conditions hadachieved similar levels of accuracy in the reading of listitems, the control condition would still not have done aswell as the training conditions in passage accuracy becausethey had not experienced the benefits of overlearning thatwere associated with speed training. Thus, possible effectsof flashcard training on reading comprehension cannot befully explained by the existence of differences in readingaccuracy between the two trained conditions and the controlcondition.

Table 3 shows results for comprehension questions. Forexplicit questions (strict scoring), there was an effect forconditions. LSD contrasts showed that both of the trainedconditions were superior to the control condition. There wasno passage effect, but the linear component of passagesinteracted with conditions. Compared with children in thecontrol condition, children in the single-word condition


Table 3Means and Standard Deviations for Strict and Lenient Scoring Questions

Measure and passage

Explicit12345Across passages

Implicit12345Across passages

Total12345Across passages

Explicit12345Across passages

Implicit12345Across passages

Total12345Across passages

Single

M

5.576.646.437.006.576.44

2.002.642.292.362.712.40

7.579.298.719.369.298.84

6.077.147.367.146.866.91

2.503.072.863.363.002.96

8.5710.2110.2110369.869.84

word

SD

1.021.010.850.780.850.47

0.960.630.831.081.270.47

1.451.271.141.081.490.72

1.070.860.930.770.540.46

1.090.730.950.740.880.47

1.511.191.480.861.030.73

Condition

Phrase

M

6.297.146.576.716.936.73

2.642.932.862.713.142.86

8.9310.079.439.43

10.079.59

6.937.867.436.717.367.26

3.073.073.213.573.213.23

10.0010.9310.6410.2910.5710.49

SD

Control

M SD

Strict scoring

0.830.770.851.071.000.53

0.500.920.660.830.770.33

1.001.071.161.161.330.62

4.144.073.073.793.213.66

1.571.431.291.501.361.43

5.715.504.365.294.575.09

1.561.641.381.531.120.55

1.160.940.831.230.930.55

2.021.611.691.861.450.93

Lenient scoring

0.830.360.851.070.750.43

0.620.830.700.650.800.38

1.040.730.840.831.020.56

4.505.004.294.294.144.44

1.711.791.862.071.791.84

6.216.796.146.365.936.29

1.741.521.201.591.290.60

1.140.800.661.070.800.45

1.891.581.511.821.440.77

ConditionF(2, 26)

205.66**

37.64**

145.49**

160.22**

42.42**

143.30**


Linear passage Condition X LinearF(l, 13) Passage F(2, 26)

0.32 5.42*

2.30 1.33

1.56 6.62**

0.01 2.11

3.93 0.53

0.91 2.36

*p < .05. **p < .01.

improved more over passages, f(13) = 3.20, p < .01. Forimplicit questions (strict scoring), there was a statisticallysignificant conditions effect. Fisher's LSD contrasts showedthat both trained conditions were better than the controlcondition but that the phrase condition also performed betterthan the single-word condition for these implicit questions.There were no passage or interaction effects. When thelenient scoring criterion was used, the results showed asimilar pattern, except that the advantage of the phrasegroup over the single-word group for implicit questions wasnot sustained.

The results for total scores showed significant conditions

effects for both strict and lenient scoring criteria. For strictscoring, Fisher's LSD contrasts showed that the two trainedconditions were superior to the control condition and thatthe phrase condition was also better than the single-wordcondition. For lenient scoring, Fisher's LSD contrastsshowed that the two trained conditions scored better thanthe control condition but that there was no advantage of thephrase condition over the single-word condition. There wasno passage effect for either strict or lenient scoring criterion.However, the linear component of passages interacted withconditions for strict scoring. Fisher's LSD contrasts (p <.05) showed that, compared with children in the control


condition, children in the single-word condition increasedtheir scores more across passages. (Note that control stu-dents actually exhibit a descriptive decrease.) There was noCondition X Linear Passage interaction for lenient scoring.

Table 4 shows the results for passage recall There was asignificant effect for conditions for both gist and detail.Fisher* s LSD contrasts for both recall of gist and recall ofdetails showed an advantage for the two trained conditionsover the control condition. For recall of detail there was alsoa linear passage effect, revealing a general increase in scoresover passages. There were no interaction effects for gist ordetail. The results for the combined gist and detail scores(total scores) showed a somewhat different pattern. Therewas a conditions effect. Contrasts showed that both trainedconditions were better than the control condition. Althoughthe linear component of the passage effect was not statisti-cally significant, its interaction with conditions was. Fish-er's LSD contrasts revealed that both trained conditionsexhibited greater linear increases than did the control con-dition, both ps < .5.

Finally, the results of the puppydog questionnaire wereanalyzed by categorizing them according to whether the doglooked happy, looked so-so, or looked unhappy. For Ques-tion 1 ("Did you like the lesson?"), 100% (n = 42) of thegroup gave "happy" responses. For Question 2 ("Do youfeel good after reading the story?"), 95% (n = 40) werehappy. For Question 3 ("How do you think the teacher feelswhen you read?"), 85% (n = 36) felt happy. For thisquestion, "teacher" refers to the experimenter. For Question4 ("Would you like to come to another session?"), 93% (n =39) indicated they were happy to do so. These resultsindicated that all of the pupils in the study enjoyed the

teaching sessions, although the answer to Question 3 indi-cated that they were a bit uncomfortable about the assess-ment situation at the end of each training session, when theyhad to read a passage aloud. Because they were all poorreaders, this reaction was understandable.

Discussion

The results of the study showed that pupils who receivedword training, whether as single words or as words embed-ded in phrases and sentences, significantly outperformed thecontrol condition on all measures of comprehension. Thesefindings were different than those of Fleisher et al. (1979),who found that reading comprehension scores of a group oftrained poor readers did not benefit from rapid decodingtraining, compared with a group of untrained poor readers,even though the trained students were faster and moreaccurate in their reading of trained words.

In the Fleisher et al. (1979) study, and in the presentstudy, a possible weakness was that the trained and un-trained conditions differed not only in speed of reading, butalso in accuracy. In the present study, it may also have beendifferences in reading accuracy that influenced the compre-hension results, rather than simply the effects of rapiddecoding. Yet our analysis of conditional percentages of listwords read correctly in passages that were also read cor-rectly in lists showed mat children in the two trainingconditions read correct list words better in the passages thandid children in the control condition. Thus, it seems that thetraining emphasis on accurate and quick word reading (andoverlearning) was more important for improved comprehen-

Table 4Means and Standard Deviations of Recall

Measure and passage

Single

M

word

SD

Condition

Phrase

M SD

Control

M SDConditionF(2, 26)


Linear passage Condition X LinearF(l, 13) Passage F(2, 26)

Gist 13.01** 0.011 2.07 1.33 2.00 1.04 1.57 0.762 2.07 1.33 2.07 1.49 1.07 0.923 1.86 1.10 1.57 0.94 0.86 0.954 2.21 1.42 2.07 1.33 0.93 0.735 2.36 1.01 2.29 1.14 1.07 0.83Across passages 2.11 0.73 2.00 0.63 1.10 0.45

Detail 12.58** 10.26**1 1.57 0.94 1.71 0.73 1.00 0.962 2.43 0.51 2.14 1.10 1.36 0.933 2.07 1.00 2.36 0.63 1.50 1.094 2.50 0.94 2.21 1.19 1.57 0.945 2.43 0.94 2.57 1.02 1.00 1.11Across passages 2.20 0.58 2.20 0.44 1.29 0.81

Total 17.97** 4.571 3.64 1.39 3.71 1.14 2.57 1.342 4.50 1.56 4.21 1.53 2.43 1.283 3.93 1.49 3.93 1.07 2.36 1.604 4.71 1.77 4.29 1.54 2.50 1.165 4.79 1.72 4.86 1.61 2.07 1.59Across passages 4.31 1.10 4.20 0.83 2.39 1.34

*p < .05. **p < .01.

1.48

2.82

3.74*


sion than was accuracy on its own. It may seem odd that thecontrol students might misread words in passages that theyhad read correctly in lists. But this phenomenon is some-times reported by teachers; that is, poor or novice readerswill read a word correctly one day but be unable to read itcorrectly another day (Gough & Hilhnger, 1980). In con-trast, the trained pupils had seen the list words many timesand had been able to internalize them in memory so wellthat they were able either to recode them when they en-countered them in passages or recall them in their ortho-graphic forms.

Looking back at the Fleisher et al. (1979) study, we notethat their trained poor reader group was, like ours, bothfaster and more accurate than the control group of poorreaders. Even so, their trained group was not statisticallybetter than the control group in reading comprehension.What is encouraging about the results of our study is that wedid find a statistical advantage in reading comprehension forthe trained poor readers relative to the untrained poor read-ers, in addition to their advantage in both speed andaccuracy.

Why should the present study have found training effectswhen other studies, such as Fleisher et al. (1979), havefailed to do so? One possible explanation is that the poorreaders in our study had received reading instruction in aschool that adhered to a very strong whole-language phi-losophy. When it became clear that our training programwould involve teaching of words in isolation, the schoolbecame very concerned that such instruction would negatewhat they were trying to achieve. In the whole-languagephilosophy, children read words only in context, as in astory. They are encouraged to read for meaning, usingcontext and picture clues, with use of initial letter clues onlyto confirm their predictions or guesses as to the meanings ofwords. Because the poor readers in our study had notreceived explicit phonics instruction, they may have dif-fered from pupils who participated in previous studies of theeffects of flashcard training. The positive results of thepuppydog questionnaire suggest that the pupils in this studymay have been more open to this kind of speeded word-recognition training than was the case in earlier studies, inwhich the training may have seemed less novel. It is alsopossible that speeded word-recognition training, whenadded to the whole-language emphasis on reading for mean-ing, made the effects of the training more pronounced thanin previous studies.

A second possible explanation for our pronounced train-ing effects is that the difficulty levels of passages in ourstudy were pitched closely to pupils' reading levels. A totalof 25 different passages were used to enable this to happen,with passages only one year above the reading levels ofpupils, whereas in Fleisher et al. (1979), there was a 3-yeargap between the reading levels of the passages and thereading levels of the pupils.

A third explanation is that meanings of words were ex-plained as part of the training for all of the pupils, so thatthere would be no question that all of the pupils, includingthose in the control condition, would be able to understandwhat they were decoding. Even pupils in the single-word

training condition were given explanations of the meaningsof words when needed, even though they were trained withflashcards containing just single words.

From a theoretical perspective, the results support thebottleneck hypothesis in that increased decoding efficiencywould necessarily lead to improved comprehension. This isan important result in that our training study shows a causalrelationship between rapid decoding and reading compre-hension, whereas most other studies have only obtained acorrelational relationship (e.g., Hogaboam & Perfetti, 1978;Perfetti, 1985). From a teaching perspective, the findingssuggest that extensive drill and practice of meaningfulwords can be an effective means of relieving the processingbottleneck in reading that occurs when limited attentionaland memory resources become overloaded (Perfetti, 1977).

The training of words in sentence context was sometimesmore effective than single-word training, especially for im-plicit questions. Nevertheless, the effects of each kind oftraining need to be explored further. There was more vari-ability within the phrase- and sentence-training conditionthan in the isolated words condition, which suggests thatthere are individual differences in the effects of each train-ing procedure. It could be that the phrase training is morehelpful for children who are reading at the 8- to 9-year-oldlevel. There is also the possibility that, for some pupils,phrase training is less likely to produce speeded word read-ing than is single-word training.

Because the phrase- and sentence-training condition alsohad similar speed and accuracy of word recognition to thesingle-word condition, it appears that there was enough timein the sentence-training condition to reach the speed andaccuracy criterion, as well as to provide information onword meanings in context. This bonus, however, wouldhave been insufficient without the parallel gains in speedand accuracy of word recognition that came from the flash-card training. This can be seen from the much lower com-prehension scores for children in the control condition, whoalso received instruction in word meanings but were nottrained in rapid decoding.

The effectiveness of combining the concept of overlearn-ing (i.e., extensive drill and practice) with word traininglends support to the arguments of Gilbert, Spring, andSassenrath (1977) and Bloom (1986) that overlearning isnecessary in the development of automaticity. Bloom foundin his study of successful sports persons that athletes whobecame Olympic champions had overlearned particular mo-tor skills. The concept of overlearning also seems applicableto the development of reading skills. We found that pupils'ability to recognize words improved rapidly with practice.These results are in line with those of Gates and Boeker(1923), who found that preschool children learned on aver-age 175% as many words on their fifth session of practice ason their first.

The flashcard training in our study did enable poor read-ers to read words faster but may not have produced auto-maticity of word recognition. The concept of automaticityinvolves not only speed but lack of attention. Lack ofattention implies that there is virtually no mental energyapplied to the task of word recognition and that word-


recognition processes are inflexible in the sense that theyare outside our voluntary control (Jonides, Naveh-Benjamin, & Palmer, 1985). To assess automaticity, wewould need to use a Stroop Color-Word Interference Testtask. One assessment procedure might be to ask pupils toname pictures that have semantically related words printedon them, to see if the printed words interfere with theirability quickly to name the pictures (Ehri, 1987). Anotherprocedure might be a dual attention task in which the pupilis asked to match words according to whether or not theymean the same, while monitoring a tone (Holt-Ochsner &Manis, 1992). We did not do this. Thus, our data apply onlyto the training of one aspect of automaticity.

The results of this study support findings that less skilledand beginning readers do not recognize words as efficientlyas do skilled readers (Gates & Boeker, 1923; Gough, 1993;McCullough, 1955; Samuels & Jeffrey, 1966; Wiley, 1928).Less skilled and beginning readers often perceive words bytrivial details, for example by remembering the dot over theletter "i", or the tail-like appearance of the descender strokeof the letter "y". To compensate for decoding problems,poor readers also rely on context clues to help with wordrecognition (Nicholson, 1991; Stanovich, 1986). In contrast,skilled readers are able to process words completely, with-out relying on context, because of their knowledge of letter-sound relationships.

For many years now, the use of flashcards for instruc-tional purposes has been seen as an ineffective technique forimproving reading skill (e.g., McCullough, 1955). There issome justification for such suspicion, in that pupils who lackdecoding skills, including beginning readers, rely on extra-neous cues for word recognition, such as a thumb print onthe flashcard. Gough (1993) has shown that this does indeedhappen among beginning readers. The use of flashcards,therefore, should be used as a supplement to training in theteaching of phonemic awareness skills and letter-soundrelationships. A number of studies have found that initialtraining in phonemic awareness and the learning of letter-sound rules can get children off to a better start in readingand spelling (e.g., Castle, Riach, & Nicholson, 1994; Ni-cholson, 1994, 1996, 1997). In short, the use of flashcardsalone will not provide the basic skills required to become agood reader, although it is possible to improve speed andaccuracy of specific word recognition (e.g., Lovett, Warren-Chaplin, Ransby, & Borden, 1990). Pupils must first acquirethe ability to decode (Vellutino, 1991). What flashcards cando, once decoding skills are developed, is provide opportu-nities for practice and overlearning, which is necessary tomake progress in reading. Other procedures, such as re-peated reading, can also provide pupils with necessary prac-tice in word-recognition fluency and speed (Rashotte &Torgesen, 1985). Once pupils find that they are improvingin rapid decoding, they will be encouraged to do even morereading, which in turn will provide the very practice thatwill enable them to become good readers (Stanovich, 1986).

Finally, with advances in computer technology, trainingprocedures for speeded word recognition have become moreinteresting and more sophisticated than flashcards (Cohen,Torgesen, & Torgesen, 1988; Torgesen, Walters, Cohen, &

Torgesen, 1988; Van Daal & Reitsma, 1993; Van Daal &Van Der Leij, 1992; Van Den Bosch et al., 1995; Yap,1993). There is also evidence that computer training can beeffective in improving fluency. Reading-disabled pupils,using a computer-based reading system in which they couldget speech feedback on request for words they did not know,were able to improve their ability to recognize those words(Olson, Wise, Conners, Rack, & Fulker, 1989). Rashotteand Torgesen (1985) were able to improve poor readers'fluency by having them reread several times the same storypresented in a computer format. This procedure was wellliked by students, even though it involved repetition. Vander Leij and Van Daal (1989) also found that repeatedpractice in reading of isolated words, using a computerscreen format, improved poor readers' ability to read thosewords. Today, with a computer in nearly every classroom,such programs can provide individual help where needed,without consuming too much of the teacher's time.

This need to explore ways of improving the fluency andaccuracy of children's reading is supported by recent surveyresults, based on more than 1,000 fourth-grade pupils,showing that more than 40% of the sample were unable toread grade-level material, even at a second reading, withadequate fluency and accuracy (Pinnell et al., 1995). Thesedata suggest that many children are not getting the benefitsto reading comprehension that fast and accurate decodingcan bring.

To conclude, the present study has both theoretical andpractical significance. From a theoretical perspective, thestudy provides support for the concept of fast decoding asan important factor, although not the only factor, in explain-ing the difference between good and poor reading compre-hension. From a practical perspective, the results of thisstudy support instructional strategies that offer opportunitiesto develop speeded word recognition.

References

Adams, M. J. (1990). Beginning to read: Thinking and learningabout print. Cambridge, MA: MIT Press.

Blanchard, J. S. (1980). Preliminary investigation of transfer be-tween single-word decoding ability and contextual reading com-prehension by poor readers in grade six. Perceptual and MotorSkills, 51, 1271-1281.

Blanchard, J. S., & McNinch, G. H. (1980). Testing the decodingsufficiency hypothesis: A response to Fleisher, Jenkins, andPany. Reading Research Quarterly, 15, 559-564.

Bloom, B. S. (1986). Automaticity. Educational Leadership, 43,70-77.

Bryant, N. D. (1975). Bryant Test of Basic Decoding Skills. NewYork: Teachers College Press.

Carver, R. P. (1990). Reading rate: A review of research andtheory. San Diego, CA: Academic Press.

Castle, J. M., Riach, J., & Nicholson, T. (1994). Getting off to abetter start in reading and spelling: The effects of phonemicawareness instruction within a whole language program. Journalof Educational Psychology, 86, 350-359.

Cohen, A. L., Torgesen, J. K., & Torgesen, J. L. (1988). Improvingspeed and accuracy of word recognition in reading disabledchildren: An evaluation of two computer program variations.Learning Disability Quarterly, 11, 333-340.


Cromer, W, (1970). The difference model: A new explanation forsome reading difficulties. Journal of Educational Psychology,61, 471-483.

Cromer, W., & Wiener, M. (1966), Idiosyncratic response patternsamong good and poor readers. Journal of Consulting Psychol-ogy, 30, 1-10.

Department of Education. (1983). Informal prose reading inven-tory. Wellington, New Zealand: Author.

Ehri, L. C. (1987). Learning to read and spell words. Journal ofReading Behavior, 19, 5-31.

Fleisher, L., Jenkins, J., & Pany, D. (1979). Effects on poorreaders' comprehension of training in rapid decoding. ReadingResearch Quarterly, 15, 30-48.

Gates, A. I., & Boeker, E. (1923). A study of initial stages inreading by preschool children. Teachers College Record, 24,469-488.

Gilbert, N., Spring, C , & Sassenrath, J. (1977). Effects of over-learning and similarity on transfer in word recognition. Percep-tual and Motor Skills, 44, 591-598.

Gough, P. B. (1993). The beginning of decoding. Reading andWriting: An Interdisciplinary Journal, 5, 181-192.

Gough, P. B., & Hillinger, M. (1980). Learning to read: An un-natural act. Bulletin of the Orton Society, 30, 179-196.

Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, andreading disability. Remedial and Special Education, 7, 6-10.

Hogaboam, T., & Perfetti, C. A. (1978). Reading skill and the roleof verbal skill in decoding. Journal of Educational Psychology,70, 717-729.

Holt-Ochsner, L. K., & Manis, F. R. (1992). Automaticity trainingfor dyslexics: An experimental study. Annals of Dyslexia, 42,222-241.

Jonides, J., Naveh-Benjamin, M., & Palmer, J. (1985). Assessingautomaticity. Acta Psychologica, 60, 157-171.

Juel, C. (1988). Learning to read and write: A longitudinal study of54 children from first through fourth grades. Journal of Educa-tional Psychology, 80, 437-447.

Juel, C. (1994). Learning to read and write in one elementaryschool. New York: Springer-Verlag.

Juel, C , Griffith, P. L., & Gough, P.B. (1986). Acquisition ofliteracy: A longitudinal study of children in first and secondgrade. Journal of Educational Psychology, 78, 243-255.

Just, M. A., & Carpenter, P. A. (1987). The psychology of readingand language comprehension, Newton, MA: Allyn & Bacon.

LaBerge, D., & Samuels, S. J. (1974). Toward a theory of auto-matic information processing in reading. Cognitive Psychology,6, 293-323.

Levin, J. R., Serlin, R. C , & Seaman, M. A. (1994). A controlled,powerful multiple-comparison strategy for several situations.Psychological Bulletin, 115, 153-159.

Lovett, M. W., Warren-Chaplin, P. M., Ransby, M. J., & Borden,S. L. (1990). Training the word recognition skills of readingdisabled children: Treatment and transfer effects. Journal ofEducational Psychology, 82, 769 -780.

McCullough, C. (1955). Flash cards—The opiate of the readingprogram? Elementary English, 32, 379-381.

Ministry of Education. (1989). School journal catalogue, 1974-1987. Wellington, New Zealand: Author.

Neale, M. D. (1989). Neale Analysis of Reading Ability. London:National Foundation for Educational Research-Nelson.

Nicholson, T. (1991). Do children read words better in context orin lists? A classic study revisited. Journal of Educational Psy-chology, 83, 444-450.

Nicholson, T. (1994). At the cutting edge: Recent research on

learning to read and spell Wellington: New Zealand Councilfor Educational Research.

Nicholson, T. (1996). The Struggletown project. Unpublishedmanuscript. University of Auckland, Auckland, New Zealand.

Nicholson, T. (1997). Closing the gap on reading failure: Socialbackground, phonemic awareness and learning to read. In B. A.Blachman (Ed.), Foundations of reading acquisition and dys-lexia: Implications for early intervention. Mahwah, NJ:Erlbaum.

Nicholson, T., & Imlach, R. (1981). Where do their answers comefrom? A study of the inferences children make when answeringquestions about narrative stories. Journal of Reading Behavior,13, 111-129.

Oaken, R., Wiener, M., & Cromer, W. (1978). Identification,organization, and reading comprehension for good and poorreaders. Journal of Educational Psychology, 62, 71-78.

Olson, R., Wise, B., Conners, F., Rack., & Fulker, D. (1989).Specific deficits in component reading and language skills:Genetic and environmental influences. Journal of Learning Dis-abilities, 22, 339-348.

Pearson, P. D., & Johnson, D. D. (1978). Teaching reading com-prehension. New York: Holt, Rinehart & Winston.

Perfetti, C. A. (1977). Language comprehension and fast decoding:Some psycholinguistic prerequisites for skilled reading compre-hension. In J. T. Guthrie (Ed.), Cognition, curriculum and com-prehension (pp. 20-41). Newark, DE: International ReadingAssociation.

Perfetti, C. A. (1985). Reading ability. New York: Oxford Univer-sity Press.

Perfetti, C. A., & Hogaboam, T. (1975). The relationship betweensingle word decoding and reading comprehension skill. Journalof Educational Psychology, 67, 461-469.

Perfetti, C. A., & Lesgold, A. M. (1977). Coding and comprehen-sion in skilled reading and implications for reading instruction.In L. B. Resnick & P. A. Weaver (Eds.), Theory and practice ofearly reading (Vol. 1, pp. 57-84). Hillsdale, NJ: Erlbaum.

Pinnell, G. S., Pikulski, J.J., Wixson, K. K., Campbell, J.R.,Gough, P. B., & Beatty, A. S. (1995). Listening to children readaloud: Data from NAEP's integrated reading performancerecord (IRPR) at grade 4. Washington, DC: National Center forEducational Statistics.

Raphael, T. E., & Wonnacott, C. A. (1985). Heightening fourthgrade students' sensitivity to sources of information for answer-ing comprehension questions. Reading Research Quarterly, 20,282-296.

Rashotte, C. A., & Torgesen, J. K. (1985). Repeated reading andreading fluency in learning disabled children. Reading ResearchQuarterly, 20, 180-188.

Rayner, K., & Pollatsek, A. (1989). The psychology of reading.Englewood Cliffs, NJ: Prentice Hall.

Roper, H. D. (1984). Spelling, word recognition and phonemicawareness among first grade children. Unpublished doctoraldissertation, University of Texas, Austin.

Samuels, S. J., Dahl, P., & Archwamety, T. (1974). Effect ofhypothesis/test training on reading skill. Journal of EducationalPsychology, 66, 835-844.

Samuels, S. J., & Jeffrey, W. E. (1966). Discriminability of wordsand letter cues used in learning to read. Journal of EducationalPsychology, 57, 337-340.

Stanovich, K. E. (1986). Matthew effects in reading: Some conse-quences of individual differences in the acquisition of literacy.Reading Research Quarterly, 21, 360-406.

Tan, A., Moore, D. W., Dixon, R. S., & Nicholson, T. (1994).Effects of training in rapid decoding on the reading comprehen-


sion of adult ESL learners. Journal of Behavioral Education, 4,177-189.

Torgesen, J. K., Walters, M. D., Cohen, A. L., & Torgesen, J. L.(1988). Improving sight-word recognition skills in LD children:An evaluation of three computer variations. Learning DisabilityQuarterly, 11, 125-132.

Van Daal, V. H., & Reitsma, P. (1993). The use of speech feed-hack by normal and disabled readers in computer-based readingpractice. Reading and Writing: An interdisciplinary Journal, 5,243-259.

Van Daal, V. H., & Van Der Leij, A. V. (1992). Computer-basedreading and spelling practice for children with learning disabil-ities. Journal of Learning Disabilities, 25, 186-195.

Van Den Bosch, K., Van Bon, W. H., & Schreuder, R. (1995).Poor readers' decoding skills: Effects of training with limitedexposure duration. Reading Research Quarterly, 30, 110-125.

Van Der Leij, A., & Van Daal, V. H. (1989). Repeated reading andsevere reading disability. In H. Mandl, E. DeCorte, N. Bennett,& H. F. Friedrich (Eds.), Learning and instruction (pp. 235-251). Oxford, England: Pergamon.

Vellutino, F. R. (1991). Introduction to three studies on readingacquisition: Convergent findings on theoretical foundations ofcode-oriented versus whole language approaches to reading in-struction. Journal of Educational Psychology, 83, 437-443.

Wiley, W. E. (1928). Difficult words and the beginner. Journal ofEducational Research, 17, 280-289.

Yap, R. L. (1993). Automatic word processing deficits in dyslexia,Amsterdam: Vrije Universiteit.

Yuill, N., & Oakhill, J. (1988). Effects of inference awarenesstraining on poor reading comprehension. Applied Cognitive Psy-chology, 2, 33-45.

Yuill, N., & Oakhill, J. (1991). Children's problems in readingcomprehension. Cambridge, England: Cambridge UniversityPress.

Received February 27, 1995Revision received September 5, 1996

Accepted September 9, 1996

Documents

Flashcards Revisited: Training Poor Readers to Read · PDF fileFLASHCARDS REVISITED 277 they found no improvement in the reading abilities of poor readers after they received word