QJER logo 2
[ Contents Vol 14, 1998 ] [ QJER Home ]

Does Reading Recovery improve phonological skills?

Claire M. Fletcher-Flinn
The University of Auckland
Campbell Yahya White
The University of Queensland
Tom Nicholson
The University of Auckland
Reading Recovery (RR) is an intervention programme designed to help children catch-up to their normal progress peers through intensive one-to-one tutoring. A 'whole language' view of reading with an emphasis on meaning comprises its philosophical base and therefore phonological processing skills are not its main focus. This is contrary to much empirical evidence linking reading to underlying phonological skills. This study examined individual differences in the acquisition of phonemic awareness and phonological decoding in 13 New Zealand children in the RR programme. The results showed that after 50 lessons most of the children were not performing within normative levels in reading and at least 30 percent were performing quite poorly on the phonological measures. This indicates that additional resources that target the development of phonological skills should be added to the programme in order to make it maximally effective.

Once hailed as the jewel in the crown of the New Zealand Education system, Reading Recovery (RR) has become that country's greatest educational export. It is claimed by its many supporters to be a 'catch-all' programme that can accelerate the reading progress of slow readers of many kinds. In Queensland, leading government politicians have appeared on television, claiming that their government is addressing the problems of literacy that have captured the public's attention, by increasing the number of RR teachers. The programme is now used extensively in many parts of the English-speaking world, with a tremendous potential influence on education and literacy levels.

RR is an intensive programme aimed at accelerating readers, who after a year of regular schooling are still not performing at the level of their peers. Developed by Professor Marie Clay (Clay, 1985, 1991, 1993a, 1993b), the programme consists of a series of daily, half-hour lessons, in which instruction is given by qualified teachers on a one-to-one basis. Ordinarily, the programme is supposed to raise the child to a reading level at or above the class average, with the reasonable expectation that the child will continue to progress at pace with his or her classmates thereafter. The timeframe for improvement is within a maximum period of twenty weeks (100 lessons), with most children being discontinued in half that amount of time or 'released but not discontinued' (Clay, 1985; Iverson & Tunmer, 1993).

Clay specifies that the children selected should be the lowest readers in their school, regardless of other variables such as intelligence, emotional problems or school history. According to Clay (1991) reading is essentially a 'top-down' process in which many different cues may be integrated to form an understanding of written text. This 'meaning-based' view of the reading process implies that each delayed reader will have varying reasons for slow progress and that weaknesses in one or two specific skills can be compensated for by strengths in others. She claims that the best way to remediate the particular weaknesses identified by the tutor is through the guided reading of continuous text. Therefore, the focus of the programme is the progression through a series of sequential 'book levels'. The child is introduced to a new book every day, which is read initially at the end of each session and reviewed at the beginning of the next. The contents of the day's book are used to develop skills and to remediate any deficiencies that the child might be showing. According to Clay, teaching specific skills out of context is a serious threat to the acceleration process and should be avoided as much as possible.

Other components in the programme include: letter identification, the 'making and breaking' of words (a recent addition to the programme), use of 'sound boxes' (an Elkonin technique) and arranging a previously-read story that has been cut into pieces by the teacher (all described by Clay, 1993a). These skills are intended to develop decoding, phonemic awareness and syntactic awareness, respectively. None of these 'metalinguistic' skills are regarded as essential in themselves to reading; they are only heuristic cues that may help identify the meaning of a word in text. The weight given to any one component of the RR programme is up to the discretion of the RR teacher. Most of the components are expected to be covered in each 30-minute session.

The lack of emphasis on metalinguistic skills stands in stark contrast to a large body of evidence showing that metalinguistic (phonemic) knowledge is crucial for reading acquisition (for example, Byrne, Fielding-Barnsley & Ashley, 1996; Liberman, 1973, 1983; Liberman & Shankweiler, 1985; Tunmer, 1991; Tunmer, Herriman & Nesdale, 1988; Tunmer & Hoover, 1993; Perfetti, 1991). 'Phonemic awareness' refers to an understanding that words 'are composed of sequences of meaningless and somewhat distinct sounds' (Juel, 1988, p. 437), whereas 'phonological recoding' (sometimes referred to as phonological decoding) is the ability to translate letters and clusters of letters into phonological forms and to use this knowledge to decode new words (Iverson & Tumner, 1993). These skills are among the best predictors of reading achievement and conversely, poor phonological processing skills are a common cognitive deficit found in virtually all poor readers (for example, Badian, 1994; Juel, 1988; Snowling, Goulandis & Defty, 1996; Stanovich & Siegel, 1993; Vellutino, Scanlon, Sipay, Small, Pratt, Chen & Denckla, 1996).

Despite the lack of emphasis on phonological skills in the programme, evaluations of RR have been generally favorable, at least over the short-term with effect sizes diminishing over time (for example, Iverson & Tunmer, 1993; Pinnell, Lyons, DeFord, Bryk & Seltzer, 1994; Wasik & Slavin, 1993). In an American-based attempt to improve the programme by the inclusion of a modified version in which children received explicit training in phonological recoding, Iverson and Tunmer (1993) found that children receiving the modified and standard forms of RR were virtually identical at discontinuation on all measures. These included book level, word recognition, phoneme segmentation and deletion and phonological decoding. However, the most significant finding was that the children who received the modified programme took less time to complete it. They argued that the modified version was more cost effective, surely a serious consid eration at a time when educational resources are tight.

Because children in both of the RR groups in the Iverson and Tunmer (1993) study performed as well or better than a group of normal progress reading controls on measures of phoneme awareness and phonological decoding, it might be reasonable to conclude that even the small amount of phonological training in the programme is sufficient to enhance phonological ability and hence reading. However, Iverson and Tunmer point out that the 100 percent discontinuation rate is inconsistent with most studies where the rate is lower, including New Zealand studies (Glynn, Crooks, Bethune, Ballard & Smith, 1989). They suggested that the difference in programme efficacy might be in its implementation, perhaps due to the American teachers either having a more advanced academic background which included reading specialisation leading to greater sensitivity to particular problems, or having different methodological orientations about the reading process prior to becoming RR tutors, which carried over into their RR instruction. Because all of the children in the study were deficient in phonological processes, either factor would contribute to a greater teaching emphasis on phonological skills in both the modified and standard version of RR than would occur in other teaching contexts.

An interesting effect of the RR programme that did not receive comment from Iverson and Tunmer (1993) was the very large post-test gain on the Dolch Word List by the RR groups compared with the standard intervention group. The Dolch List contains high frequency words commonly found in basal readers. While the RR teachers did not emphasise these words in teaching, the children might have learned them incidentally through the story reading that they were required to do on a daily basis. Because the children were at such a low level in reading, they would have been exposed to American books that contained much word repetition. Knowledge of these high frequency words may have enabled the children in RR to bootstrap themselves into reading in a way that is less apparent in other studies of RR, where this does not occur. Support for the efficacy of teaching children to read words that they are likely to come across in text has been shown to improve reading accuracy and comprehension among poor readers (Nicholson, 1998; Nicholson & Tan, 1999; Tan & Nicholson, 1997). These unique aspects of the Iverson and Tunmer (1993) study may explain why the American children receiving RR improved considerably in reading, whereas this has been less apparent in other studies (for example, Centre et al., 1995; Glynn et al., 1980).

The generally positive effect of RR does not mean that it helps all those who receive it. Even Clay (1985) reported a 5-6 percent withdrawal rate because of failure to make progress. However, recent evaluations suggest that the figure may be much higher. Reports by an Ohio State group cited in Wasik and Slavin (1993), suggested that almost 30 percent of those who entered RR were still below a comparison group of low achievers two years after leaving the programme, despite having received more than 60 tutoring sessions. A recent study of the effectiveness of RR conducted in New South Wales by Center, Wheldall, Freeman, Outhred and McNaught (1995) measured the effects of the programme 12 months after discontinuation. Although they found that RR did increase mean levels of reading performance of tutored children relative to controls, they estimated that only 35 percent of the children directly benefited, with an equal proportion still performing at low enough levels to require further remediation. The remaining 30 percent who did improve would probably have done so even without the extensive intervention, judged by the variance in the control group.

If a third of the children in RR are failing as suggested by these studies, then questions of efficacy need to be raised, in particular, regarding the identification of these children and reasons for lack of progress. Given the extensive literature on the crucial role of phonological processes in reading, the lack of emphasis on this component is the most likely reason for lack of progress and may constitute a serious deficit in the programme. The aim of the present study was to examine rates of improvement on various phonological skills during the RR programme and the effects of these skills on various measures of reading performance. Although a control group is normally used, other normative groups are provided that allow comparisons to be made. Based on the research reviewed, phonemic awareness and phonological decoding were expected to be the highest predictors of reading performance, in terms of word recognition, accuracy in text reading and reading comprehension. It was hypothesised that the greatest gains made in reading performance would be made by those children who developed these skills.

One of the advantages of RR frequently cited by Clay (1993a) and others (Smith & Elley,1994) is that the programme can be individualised to the needs of the child. If this happens in practice, then those children who are particularly weak in a specific component of reading should receive extra targeting in that skill. Presumably, this would lead to a larger relative gain in their scores on measures of that skill, compared with those who were initially better at that skill. In terms of the present study, this should lead to an interaction effect on skills such as phonemic awareness and phonological recoding, when rates of improvement between those initially high and those initially low on these skills are considered. The absence of such an effect would provide evidence that specific weaknesses were not being targeted sufficiently.



The participants at the beginning of the study reported here were 18 children from eight primary schools in East Auckland, New Zealand who were selected by their school for the RR programme. This is predominantly a lower middle to middle socioeconomic (SES) area of Auckland. However, the data from five children were dropped from the analyses for varying reasons. One child had been removed from the programme because of lack of progress and referred to the Special Education Service. Another child left the school he was attending after only 21 RR lessons and was not given a place in the programme at his new school. Three further children, all from the same school, were not included because they had begun the programme later than expected and consequently had fewer than 22 lessons each. The final sample consisted of 13 children: 9 boys and 4 girls. The mean age of the children at first testing was 6.18 years (ranging from 5.96 to 6.42 years). Eight of the children were classified by their teachers as NZ European, two as NZ Maori, one as Tongan, one as Filipino and one as Indian. All spoke English fluently except for one child who was learning English as a second language.

All children had been receiving reading instruction that can best be described as 'whole language' (sometimes referred to as a 'psycholinguistic or book experience' approach) from their classroom teachers, where the major emphasis is placed on encouraging children to use contextual cues from the sentence to identify unknown words. Similar to RR, some attention is given to the Elkonin technique of hearing sounds in words and this is called 'process writing'. Meaning rather than letter-sound information is given the most priority in confirming language predictions. (See Thompson, 1993, for more information about early reading instruction in New Zealand schools.) The main point is that the RR instruction the children received was essentially a more intense version of what they had in the classroom. This is also the conclusion of Tunmer, Chapman & Prochnow (1998), in their explanation of the RR instruction in New Zealand. The RR teachers in this study were well-trained in all aspects of the RR programme.


Several intelligence, language, reading and phonological tests were administered to the children both before (prete st) and after RR training (post-test). Standard procedures for administration and scoring as described in the test manuals were followed for all of the standardised norm-referenced tests. These tests are described in the text below and a summary provided in Table 1.

Table 1: Description of tests

TestsMeasuresConceptual Meaning
Wechsler Preschool and Primary Scale of Intelligence - Revised (WPPSI-R)Block design

General intelligence (ability)
Verbal intelligence (ability)
Metropolitan Readiness Tests (MRT)School language
Language skills
General language
Verbal understanding
Both of the above
Burt Word Reading Test - NZ Revision (Burt)
Word recognition
Book Level (BL)
Reading in text
Neale Analysis of Reading Ability (Neale)Accuracy
Word recognition
Story comprehension
Roper/Sneider Phonemic Awareness TestPhoneme segmentation, blending, deletion and substitutionPhoneme (phonological) awareness
Bryant Test of Basic Decoding Skills (Bryant)
Phonological decoding

Intelligence and language

Standard scores on the block design and vocabulary subtests of the Wechsler Preschool and Primary Scale of Intelligence - Revised (WPPSI-R) (Wechsler, 1989) were used to provide an estimate of general and verbal ability, respectively. In addition, the language skill subtests of the Metropolitan Readiness Tests (MRT) (Nurss & McGauvran, 1986) were used as measures of general language and listening ability. The school language subtest was designed to measure the understanding of basic cognitive concepts as well as grammatical structure. It consists of nine verbal statements which are read to the child and the child is asked to choose the picture which best describes each statement from a multi-choice set of three. The listening test is similar in nature except that there are four choices in each set. It was designed to assess the ability to integrate and reorganise information, to draw inferences and evaluate material that is presented orally. The language skills measure was an aggregate of both raw scores, as described in the manual (p. 8).


The Burt Word Reading Test - New Zealand Revision (Burt) (Gilmore, Croft & Reid, 1981) was used as a measure of word recognition. It consists of 110 words graded in approximate order of difficulty. Scoring consisted of number of words correct. The Neale Analysis of Reading Ability - Revised, Form 2 (Neal) (Neale, 1988) measured the children's accuracy in passage reading and their level of comprehension. It was only administered at post-test because it was thought to be too difficult for the reading age of most of the children at pretest and therefore would not discriminate between them.

Although not a standardised test, a measure of book level (The Observation Survey, Clay, 1993b) was also taken. This was obtained by taking a running record of the errors the child made on a previously unseen book selected by the RR teacher and classified by the New Zealand Department of Education on a scale of difficulty from 1 to 20. The book level for the purposes of this study was also the instructional level, that is where the child was at least 90 percent accurate. If the child was not performing at the level selected by the teacher, an easier text was administered until the child could read it at 90 percent accuracy or above.

Phonological ability

The Roper/Schneider Phonemic Awareness Test (Roper/Schneider, 1984) was used as a measure of phonemic awareness. This test consisted of six subtests, each measuring a different aspect of phonemic awareness. These were phoneme segmentation, phoneme blending, deletion of initial phoneme, substitution of initial phoneme and substitution of final phoneme. Each subtest consisted of seven items, giving a maximum score of 42.

The Bryant Test of Basic Decoding Skills (Bryant, 1975) was used to measure the ability to read pseudowords. It consisted of a list of 50 pseudowords that the child is expected to read in accordance with the usual sound-letter rules of the English language. Before presentation of the list, the child was told, ' I am going to show you a list of funny words. They are not real words, but I want to see if you can read them. The child was then presented with the list and told to skip any words that he or she could not read.


The children were tested in two sessions at both pre-test and post-test. For the pretests, the first session took place either in the week before each child was planned to commence RR (14 children) or in the first week of RR (4 children). All children were tested individually and for the most part this testing took place in the usual RR environment at the child's school, although three children at one school were tested in the staffroom, which was unoccupied.

The child was introduced to the researcher by the RR teacher, who then left the room. The child was told by the researcher that the teacher wanted to play a few games involving sounds and words, as well as to listen to him/her reading. The tests were then administered in the following order: WPPSI - R (block design; vocabulary), the Burt, the Roper/Schneider phonemic awareness test, the MRT (school language; listening) and the Bryant decoding test. A break of 30-60 minutes was given between the first three tests and the last three tests, often coinciding with the school's lunch break. The child was praised during and after each test, dependent on their level of attention and not their performance. At the end of the session, the child was offered a lollipop and a sticker for 'trying their best'.

The post-test sessions took place after approximately 10 weeks or 50 lessons of RR. The same procedure and order of testing was followed for the administration of the language and reading tests except for inclusion of the Neale at the beginning and the measure of book level at the end.


As shown in Table 2, the children's scaled scores were in the average range on both block design (measuring general ability) and vocabulary (measuring verbal ability). In contrast, they performed poorly on all of the other linguistic, phonological and reading pretests. The mean language score on the MRT for the group placed them in the lowest 10 percent of the population of 6-year-olds according to the test norms. The mean score on phonemic awareness was only 8.23 correct from a possible score of 42. Four children did not score at all. On the Burt, all of the children scored well below the minimum equivalent age band in reading (a raw score of 20) and 11 of the 13 children were unable to pronounce a single pseudoword.

As a group the children made significant and substantial gains on all of the replicated measures between the initial and final testing as shown by a series of matched-pairs t-tests and effect sizes for correlated designs (Dunlap, Cortina, Vaslow & Burke, 1996). The largest effect sizes were on the MRT and phonemic awareness. However, despite these gains, the children were still below average on all of the norm referenced reading and language tests (that is, Burt, Neale and MRT). The average score on the Neale accuracy was equivalent to a reading age of about 5 years 10 months (extrapolating 1 month per raw score), while comprehension was about 6 years 2 months. Similarly, the extrapolated reading age on the Burt was 5 years 9 months at post-test. All are less than the average chronological age of 6 years 5 months for our sample. The children remained low on language skills performing in th e lowest 30 percent of the six-year-old norming population and four children in the sample still could not pronounce any pseudowords. There were also large standard deviations, indicating large individual differences on the tests.

Figure 1 shows a scatterplot of pre-test and post-test scores on phonemic awareness. Three separate groups can be identified: (1) four children who scored poorly on phonemic awareness at pretest and remained low at post-test, (2) four children who scored in the low to moderate range at pretest and remained there relative to the other children and (3) a third group whose pretest scores were 'mixed', but who all scored highly at post-test. At pretest, three out of five of the children in this group scored in the moderate range. One child was initially high and remained high, while another scored zero on the pretest and made a gain of 30 points.

Table 2: Tests of significant differences between pre-test and post-test means

MeanSDMean SD(df=12)
Chronological Age (years)
Vocabulary (standard score)10.543.62N/AN/AN/A
Block Design (standard score)9.232.24N/AN/AN/A
MRT8.233.17 11.384.1913.86**2.13
Burt (raw score)3.38 5.7417.466.976.61**0.76
Book Level2.92 2.8412.004.585.02**0.63
Phonemic Awareness (max=42)8.23 7.9623.4610.7111.29**2.66
Phonological Decoding (max=50) 0.461.393.624.113.31*0.81
Neale (Accuracy)N/A10.796.60N/AN/A
Neale (Comprehension)N/A3.772.45N/AN/A
Note. N/A indicates not applicable
*p<.01, **p<.001

Because we were interested in the relative gains for children who came into the RR programme with weak phoneme awareness skills compared with those with better skills, a two-way analysis of variance (ANOVA) was used to examine the gains in pre-test to post-test performance. This involved separating the sample on the basis of pre-test scores on phonemic awareness. This gave a low group (n = 8) who scored 10 or below and a high group who scored 11 or above (n = 5). The high group remained significantly and substantially higher than the low group at final testing (F(1,22) = 19.49, MSE = 1002.35, p < .001, dc = .84). The results also verified that the sample as a whole had improved significantly and substantially at final testing (F(1,22) = 29.32, MSE = 1507.85, p < .001, dc = .90). However, there was no significant interaction between groups (F(1,22) = .03, p = .86). This shows that the rate of gain was similar between the two groups.

Figure 1

Figure 1: Plot of individual scores of phonemic awareness at pre-test and posttest

The relationships between the dependent variables were measured using Pearson correlation coefficients. Table 3 shows the correlations between the variables on the pretests. It is important to note that performance on phonemic awareness was moderately (and significantly) correlated with all of the reading measures at pretest. Because a child's general ability or initial reading progress could affect these relationships, partial correlation coefficients were calculated. When general ability (block design) was held constant, the relationships between phonemic awareness and performance on phonological decoding (pr = .84) as well as between phonemic awareness and the Burt (pr = .58) remained the same or were higher. When book level was controlled for, the relationships between phonemic awareness and phonological decoding decreased but remained highly correlated (pr = .61). Thus, relationships between phonemic awareness and the other reading related variables persisted regardless of general ability or reading level.

The Burt was also highly correlated with book level (pr = .91). This suggests that the Burt and book level are essentially measuring the same thing. There was also a high correlation between vocabulary (WISC-R) and the MRT (pr = .89), which remained virtually the same (pr = .88) even after initial book level was held constant.

Table 3: Pre-test correlation coefficients (r)

Vocabulary BDMRTBurtBLPA
Block Design (BD).42

MRT.89 .37

Burt.34 .15.43

Book level (BL).32 .07.43.91

Phonemic Awareness (PA).34 .
Phonological Decoding (PD)-.20.02-.

Table 4 shows the correlations between the variables on the post-tests. Similar to the pattern on the pretests, phonemic awareness was moderate to highly correlated with all of the other reading related variables, including the Neale accuracy and comprehension. With either general ability or book level partialed out, all of these correlations remained high (partial correlation coefficients ranged between .61 and .73). Except for a low and negative relationship with phonological decoding, scores on the MRT were moderately related to the other variables.

Similarly, the Burt and the phonological decoding were also moderately to highly correlated with all of the variables except for the MRT. All these correlations remained moderate to high (and statistically significant) when general ability and initial book level were held constant for the Burt and phonological decoding (partial correlation coefficients ranged between .56 to .96), except for the relation between phonological decoding and the Neale accuracy, which dropped (pr = .33). The important point from these analyses is that both phonemic awareness and phonological decoding were related to reading comprehension regardless of the child's book level or general ability.

Table 4: Post-test correlation coefficients (r)


Book level (BL).45 .94

Phonemic Awareness (PA).50 .71.73

Phonological Decoding (PD)-.

Neale (Accuracy).37 .
Neale (Comprehension).37 .

Two hierarchical multiple regression analyses were carried out to determine the amount of unique variance accounted for by phoneme awareness and phonological decoding, with regard to reading comprehension, once the effect of intelligence (general and verbal) was partialed out. In the first analysis the order of entry was: block design, vocabulary, phonological awareness and phonological decoding. Only phonemic awareness accounted for a moderate amount of the variance (adjusted R2 = .33). When phonological decoding was entered before phoneme awareness, it also accounted for a moderate proportion of the variance (adjusted R2 = .45). A scatterplot of post-test scores on the two tests showed that many children scored well on phonemic awareness and poorly on phonological decoding. Of the five children who scored highest on phoneme awareness (30 or above), only one child scored poorly on phonological decoding.


As a group, the children made substantial gains on all literacy and language measures after an average of 50 lessons on RR and this finding is consistent with other research (Clay, 1985; Iverson & Tumner, 1993; Wasik & Slavin, 1993). Since many of the children in our study were able to read very little at the start of the RR programme, the increase in reading performance appears impressive. However, when judged against New Zealand reading norms, most of the children were still reading below the level of their age cohort and only two children had been discontinued (15 percent). It is possible, though, that with more RR lessons (50 - 100 lessons are recommended), the rest of the children would have been brought up to the reading level of their peers and would have continued to improve.

A better-controlled longitudinal study that argues against this possibility has been reported by Chapman et al. (1998). They observed the progress of 26 New Zealand children who received RR and compared their progress over time both with another group of poor readers who received no training and a group of normally developing readers. They found that, both immediately after the training and one year later, the RR trained children were no better at reading than the control group of poor readers and were significantly worse than the normally developing readers on all language, literacy and self-concept tests.

Some of the largest gains were on the MRT, measuring school language and listening skills and this may be a nice spin-off effect of the programme, which is most likely due to the linguistic focus of RR lessons and the one-to-one delivery format. Nicholson (1997) also found that low SES children in RR with low scores on a picture vocabulary test improved significantly, but middle SES children with average scores did not change much from pre-test to post-test. This suggests a regression to the mean effect, which may also be problematic in the interpretation of the MRT gains in our study because the children's scores were so low at pretest. Lack of a control group which did not receive RR also precludes attributing both the reading and MRT gains entirely to RR.

The main focus of this study was to determine the relationship between the development of phonological skills and progress in RR. The results showed that the children were very low on both phonemic awareness and phonological decoding at initial testing. This finding is similar to that of Chapman et al. (1998) who reported that from a cohort of 152 New Zealand children, those selected for RR without exception had severe difficulties with phonological processing skills. The finding that children participating in RR have difficulty in relating letters to sounds also agrees with most of the existing literature on poor readers (for example, Iverson & Tunmer, 1993; Juel, 1988; Siegel, 1993).

The phonemic awareness of the children improved considerably between pre-test and post-test, showing one of the largest effect sizes. Their average scores (23.46) were similar to the mid-year mean of a sample of 129 American first grade children tested by Juel, Griffith and Gough (1986) who had a modal score of 25.9 correct on the same test. This gain in phonemic awareness is also compatible with previous research on RR (Iverson & Tumner, 1993; Chapman et al., 1998). However, Castle, Riach & Nicholson (1994) reported comparable gains for a Year 1 no-intervention control group of New Zealand 5-year-old children (pre-test = 8.02, post-test = 22.5) without any special instruction and so did Chapman et al. (1998) for both no-treatment controls (pretest = 8.30, post-test = 16.41) and RR children (pretest = 5.24, post-test = 17.48). No differences were found between the scores for the RR children and no-treatment controls at either pre-test or post-test from the latter study.

In our study, half of those who started at a very low level on phonemic awareness tended to remain low relative to the others. This appears contrary to comments by Clay (1993) and Smith and Elley (1994) that the weakest skills of each child are targeted specifically by the teacher in RR. If this had been the case, the phonemic awareness of those lowest on this measure would have improved the most. There was one child who actually made a significant gain of 30 points, but this child knew little English at the start of the programme and may have had difficulty understanding the directions.

Despite the overall gains made on phonemic awareness, there appeared to be few real gains made by the children on phonological decoding. The mean score at post-test on the Bryant test of 3.6 (7 percent correct) remained well below that of the Juel et al. (1986) first grade sample which had a mean score of 13.5 correct on the same test. A comparison with Iverson and Tunmer's (1993) modified RR group shows that the mean of their children on phonological decoding of 8.2 correct after 41.8 lessons was considerably higher than the children in this study. Using New Zealand samples, Nicholson (1997) reported a mean score of 12.6 correct on the Bryant test for a control group of children from schools in a low socioeconomic area and Connelley, Johnston and Thompson (1999) reported an average of 17 percent correct for the reading of one-syllable pseudowords for Years 1 and 2. Some of these scores are not directly comparable to the Bryant test as Iverson and Tumner (1993) devised their own list and Connelly et al. (1999) do not specify their pseudoword list. Nonetheless, their measures of phonological decoding are similar in that both consisted of simple 'consonant-vowel-consonant' pseudowords, as in the Bryant test.

Although still below average in reading and phonological decoding, the results on phonemic awareness suggest that more than half of the children in this study have been brought up to a level that is comparable with normal progress children of the same age. However, it is not clear if this might have occurred even without RR. Moreover, the phonological intervention i ntroduced by Castle et al. (1994) produced gains in phoneme awareness that were higher than those in our study after only 15 lessons, taught in small group format rather than one-to-one as in RR.

Of more importance, there still remained another 30 percent of the children in our study who had gained only minimal phoneme awareness skills and 78 percent who had poor phonological decoding skills. The figure for phoneme awareness is similar to that reported by others (Center et al., 1995; Wasik & Slavin, 1993). The most probable explanation for the lack of progress in these children is that there are inherent problems in the programme. Chapman et al. (1999) argued that the poor results for RR are due to the lack of attention to the teaching of decoding and too much emphasis on prediction skills. Similarly, Castle et al. (1994) suggested that RR children are unable to make effective gains in phonological skills because of the limited time available for this aspect of skill development in the RR programme, with the integration of so many components working against effective learning (Castle et al., 1994).

Discounting RR, the most successful remediation programmes combine the teaching of reading with that of phoneme awareness (Bradley & Bryant, 1983; 1985; Hatcher, Hulme & Ellis, 1994). However, RR combines phoneme awareness with spelling. The use of 'sound boxes' involves placing counters in boxes below the word to be studied. As each phoneme is pronounced, a counter is placed in each consecutive box. Eventually, the counters are replaced with letters, so the child is essentially spelling the word that is being pronounced. Thompson and Fletcher-Flinn (1993) and Thompson, Cottrell and Fletcher-Flinn (1996) have demonstrated that procedures used for spelling in New Zealand beginning readers taught by a language or book experience approach do not transfer to reading. Therefore, another possibility is that low progress children would benefit more by systematic instruction in phoneme awareness combined with reading, as in many of the new 'phonics' programmes.

In addition, New Zealand research (Fletcher-Flinn & Thompson, 1998, Thompson & Fletcher-Flinn, 1993) has shown that children who are poor readers have difficulty in using induced sublexical relations (ISRs) to generate reading responses. ISRs are implicit relations between orthographic components of words and corresponding phonological components in memory that are induced from lexical experience and are learned nonconsciously. They form the basis of the child's computational or procedural knowledge about words and can be used to generate reading responses to both words and nonwords.

The induction and updating of ISRs is a dynamic process that is contingent upon the reading and learning of words to a point where they are sufficiently established in memory. The formation of ISRs from the learning of high frequency words in the Iverson and Tunmer (1993) study is likely to have contributed to the children's ability to read pseudowords in both RR groups. Therefore, not only are high frequency words the 'glue words' of text reading but they have additional spinoffs with regard to the formation of ISRs. Because poor readers do not develop ISRs to the same extent as normal progress readers even when matched on reading ability, it would seem sensible to teach high frequency words and to use texts that contain a great deal of word repetition.

Another aim of this study was to identify the effects of phonological skills on reading performance. The results showed that phonemic awareness and phonological decoding were moderately to highly correlated with all measures of reading ability, both at pre-test and post-test. These correlations remained even after general ability was held constant for all subjects. Multiple regression analyses demonstrated that phonological measures were robust and useful predictors of reading comprehension. Phoneme awareness is a measure of the explicit knowledge of phoneme identities. This knowledge is necessary to gain insight into the systematic mapping of letters and sounds in words. Phonemic awareness is therefore an important indicator of children's progress in the understanding of this alphabetic principle. Phonological decoding, however, can be based on either explicit or implicit phonological knowledge. Since tests of each are measuring similar aspects of processing, it is not surprising that better readers, even in this study, have higher levels of phoneme awareness and more well-developed phonological recoding skills. These results are consistent with that of others (Iverson & Tunmer, 1993; Juel et al., 1986).

An interesting finding of the present study was the high level of intercorrelations between word recognition (Burt), book level, reading accuracy and reading comprehension. At final testing, these variables seemed to form a cluster which could be described as 'reading ability'. This provides support for the validity of book level as an accurate measure of reading performance. Although time could be saved by using only one measure of reading, standardised tests, such as the Burt and the Neale, can be used to verify book levels as Chapman et al. (1998) found that book levels were highly inflated by RR teachers. Standardised tests also provide reading ages which is additional useful information.

The results of the present study were consistent with others in showing that RR children show improvements in phonological ability and reading (Chapman et al., 1998; Clay, 1985; Iverson & Tumner, 1993; Wasik & Slavin, 1993). However, the gains in reading were insufficient to bring most of the children up to the level of normal progress children of the same age in New Zealand. Moreover, after an average of 10 weeks or 50 lessons in RR most of the children were still having difficulty decoding a simple 3-letter pseudoword and some showed little evidence of developing phoneme awareness. RR has been criticised for not targeting phonological skills and the results of this study indicate that the RR programme in its present form does not meet the needs of all children.

It seems clear that the RR programme would benefit from some changes, incorporating systematic phonics and phonemic awareness instruction, the teaching of high frequency words and the use of books with a substantial repetition of a corpus of words. In addition, the inclusion of appropriate tests to measure phonological progress, such as phonemic awareness and phonological decoding should be incorporated into the programme to aid in identifying those children who may need ongoing support in these areas. Although the children in this study showed improvement in literacy skills, further research should include a control group (such as in Center et al., 1995 and Chapman et al., 1998) to provide a comparison with children who received no intervention at all.


Badian, N. A. (1994). Do dyslexic and other poor readers differ in reading-related cognitive skills? Reading and Writing: An Interdisciplinary Journal, 6, 55-63.

Bradley, L. & Bryant, P. E. (1983). Categorizing sounds and learning to read: A causal connection. Nature, 301, 419-421.

Bradley, L. & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. Ann Arbor, MI: University of Michigan Press.

Byrne, B., Fielding-Barnsley, R. & Ashley, L. (1996). What does the child bring to the task of learning to read? A summary of the New England Reading Acquisition Projects. Australian Journal of Psychology, 48, 119-123.

Bryant, N. D. (1975). Diagnostic Test of Basic Decoding Skills. New York: Columbia University Teachers College.

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

Center, Y., Wheldall, K., Freemann, L., Outhred, L. & McNaught, M. (1995). An evaluation of Reading Recovery. Reading Research Quarterly, 30, 240-263.

Chapman, J. W., Tunmer, W. E. & Proc hnow, J. E. (1998). An examination of the effectiveness of Reading Recovery: A longitudinal study. Unpublished report, Massey University.

Clay, M. M. (1985). The early detection of reading difficulties (3rd edition). Auckland: Heinemann.

Clay, M. M. (1991). Becoming literate: The construction of inner control. Auckland: Heinemann.

Clay, M. M. (1993a). Reading Recovery: A guidebook for teachers in training. Auckland: Heinemann.

Clay, M. M. (1993b). An Observation Survey of Early Literacy Achievement. Auckland: Heinemann.

Connelly, V., Johnston, R. S. & Thompson, G. B. (1999). The influence of instructional approaches on reading procedures. In G. B. Thompson & T. Nicholson (Eds), Learning to read: Beyond phonics and whole language (pp. 103-126). New York: Teachers College Press.

Dunlap, W. P., Cortina, J. M., Vaslow, J. B. & Burke, M. J. (1996). Meta-analysis with matched groups or repeated measures designs. Psychological Methods, 1, 170-177.

Fletcher-Flinn, C. M. & Thompson, G. B. (1998). The failure to acquire induced sublexical relations early in developmental reading disability. Manuscript submitted for publication.

Gilmore, A., Croft, C. & Reid, N. (1981). Burt Word Reading Test: New Zealand Revision. Wellington: NZCER.

Glynn, T., Crooks, T., Bethune, N., Ballard, N. & Smith, J. (1989). Reading Recovery in Context. Wellington: Department of Education.

Hatcher, P. J, Hulme, C. & Ellis, A. W. (1994). Ameliorating early reading failure by integrating the teaching of reading and phonological skills: The phonological linkage hypothesis. Child Development, 65, 41-57.

Hiebert, E. H. (1994). Reading Recovery in the United States: What difference does it make to an age cohort? Educational Researcher, 23, 15-25.

Iverson, S. & Tunmer, W. E. (1993). Phonological processing skills and the Reading Recovery program. Journal of Educational Psychology, 85, 112-126.

Juel, C. (1988). Learning to read and write: A longitudinal study of 54 children from first through fourth grades. Journal of Educational Psychology, 80, 437-447.

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

Liberman, I. Y. (1973). Segmentation of the spoken word and reading acquisition. Bulletin of the Orton Society, 23, 65-77.

Liberman, I. Y. (1983). A language-oriented view of reading and its disabilities. In H. J. Myklebust (Ed.), Progress in learning disabilities (Vol. 5, pp. 81-101). New York: Grune & Stratton.

Liberman, I. Y. & Shankweiler, D., (1985). Phonology and beginning reading: A tutorial. In L. Rieben & C. Perfetti (Eds), Learning to read: Basic research and its implications (pp. 3-7). Hillsdale, NJ: Lawrence Erlbaum.

Neale, M. D. (1988). Neale Analysis of Reading Ability (Revised). Hawthorn, Australia: ACER.

Nicholson, T. (1997). Does phonemic awareness training improve the literacy skills of low-SES children? Unpublished report, The University of Auckland.

Nicholson, T. (1998). The flashcard strikes back. The Reading Teacher, 52, 188-192.

Nicholson, T. & Tan, A. (1999). Word identification processes. In G. B. Thompson & T. Nicholson (Eds), Learning to read: Beyond phonics and whole language (pp. 150-173). New York: Teachers College Press.

Nurss, J. R. & McGauvran, M. E. (1987). Metropolitan Readiness Tests. San Antonio, USA: The Psychological Corporation: Harcourt Brace Jovanovich.

Perfetti, C. A. (1991). Representations and awareness in the acquisition of reading competence. In L. Rieben & C. A. Perfetti (Eds), Learning to read: Basic research and its implications (pp. 33-44). Hillsdale, NJ: Lawrence Erlbaum Associates.

Pinnell, G. S., Lyons, C. A., DeFord, D. E., Bryk, A. S. & Seltzer, M. (1994). Comparing instructional models for the literacy education of high-risk first graders. Reading Research Quarterly, 29, 9-38.

Riach, J. (1993). Effects of Phonemic Awareness Training on Spelling Acquisition. Unpublished Thesis submitted for Master of Arts, University of Auckland.

Roper/Schneider, H. D. W. (1984). Spelling, Word Recognition and Phonemic Awareness Among First Grade Children. Unpublished doctoral dissertation, University of Texas at Austin.

Siegel, L. S. (1993). Phonological processing deficits as the basis of a reading disability. Developmental Review, 13, 246-257.

Smith, J. W. A & Elley, W. B. (1994). Learning to Read in New Zealand. Auckland: Longman Paul.

Stanovich, K. E. & Siegel, L. S. (1994). Phenotype performance profile of children with reading disabilities: A regression-based test of the phonological-core difference model. Journal of Educational Psychology, 86, 24-53.

Tan, A. & Nicholson, T. (1997). Flashcards revisited: Training poor readers to read words faster improves their comprehension of text. Journal of Educational Psychology, 89, 276-288.

Thompson, G. B. (1993). Reading instruction for the initial years in New Zealand schools. In G. B. Thompson, W. E. Tumner & T. Nicholson (Eds), Reading Acquisition Processes (pp. 148-154). Clevedon, UK: Multilingual Matters.

Thompson, G. B., Cottrell, D. S. & Fletcher-Flinn, C. M. (1996). Sublexical orthographic-phonological relations early in the acquisition of reading: The knowledge sources account. Journal of Experimental Child Psychology, 62, 190-222.

Thompson, G. B. & Fletcher-Flinn, C. M. (1993). A theory of knowledge sources and procedures for reading acquisition. In G. B. Thompson, W. E. Tumner & T. Nicholson (Eds), Reading Acquisition Processes (pp. 20-73). Clevedon, UK: Multilingual Matters.

Tunmer, W. E. (1991). Phonological awareness and literacy acquisition. In L. Rieben & C. A. Perfetti (Eds), Learning to read: Basic research and its implications (pp. 105-119). Hillsdale, NJ: Lawrence Erlbaum Associates.

Tunmer, W. E. & Chapman, J. W. (1997). Reading Recovery in relation to metalinguistic abilities, phonological processing skills and literacy achievement. Paper presented at the annual conference of the New Zealand Association for Research in Education, University of Auckland.

Tunmer, W. E., Herriman, M. L. & Nesdale, A. R. (1988). Metalinguistic abilities and beginning reading. Reading Research Quarterly, 23, 134-158.

Tunmer, W. E. & Hoover, W. A. (1993). Language-related factors as source of individual differences in the development of word recognition skills. In G. B. Thompson, W. E. Tunmer & T. Nicholson (Eds), Reading acquisition processes (pp. 123-143). Clevedon, UK: Multilingual Matters.

Wasik, B. A. & Slavin, R. E. (1993). Preventing early reading failure with one-to-one tutoring: A review of five programmes. Reading Research Quarterly, 28, 179-200.

Wechsler, D. (1989). Wechsler Preschool and Primary Scale of Intelligence (Revised). San Antonio, USA: The Psychological Corporation & Harcourt Brace Jovanovich.


We thank the school authorities, Reading Recovery tutors and the children who participated for their cooperation throughout study. We are grateful to Carolyn Sutherland and John Gribben at the University of Auckland for their assistance with tests and statistical analyses. This article was written while the first author was on sabbatical leave at Curtin University of Technology, Perth, Western Australia.

Author details: Claire M. Fletcher-Flinn, Department of Psychology, University of Auckland, Private Bag 92019, Auckland, New Zealand. Claire Fletcher-Flinn received her PhD in psychology from La Trobe University, Bundoora, Victoria and then completed a post-doctoral fellowship in reading processes at the School of Education, Victoria University of Wellington, New Zealand. She is currently a lecturer in child development in the Department of Psychology at the University of Auckland. Email: cm.fletcher-flinn@auckland.ac.nz

Campbell Yahya White, School of Psychology, University of Queensland, Brisbane, QLD, 4072, Australia. Campbell completed an Honours Degree in Psychology at the University of Auckland in 1995, then taught for two years at the International Islamic University, Malaysia. He is presently a PhD candidate in the School of Psychology and tutor in the Graduate School of Education at the University of Queensland. Email: white@psy.uq.edu.au

Tom Nicholson, School of Education, University of Auckland, Private Bag, 92019, Auckland, New Zealand. Tom Nicholson started his career as a school teacher in New South Wales, then worked in the research division of the Education Department of South Australia. He completed his PhD in Education at the University of Minnesota. He is an Associate Professor in the School of Education of the University of Auckland, lecturing in developmental psychology and the reading process. Email: t.nicholson@auckland.ac.nz

Please cite as: Fletcher-Flinn, C. M., White, C. Y. and Nicholson, T. (1998). Does Reading Recovery improve phonological skills? Queensland Journal of Educational Research, 14(2), 4-28. http://education.curtin.edu.au/iier/qjer/qjer14/fletcher-flinn.html

[ Contents Vol 14, 1998 ] [ QJER Home ]
Created 18 Mar 2005. Last revision: 18 Mar 2005.
URL: http://education.curtin.edu.au/iier/qjer/qjer14/fletcher-flinn.html