Learning from their miscues: differences across reading ability and text difficulty.
This study examines the semantic, syntactic and grapho-phonic systems used during oral reading by an Australian sample of young readers. Patterns across these systems can reveal how young readers approach the task, and have implications for reading instruction. Reading research is yet to identify every skill required in reading or exactly how children learn to read, although it is clear that three language cueing systems, semantic (meaning), syntactic (sentence structure) and grapho-phonic (visual and sound), contribute to the reading process (Hempenstall, 2003; Adams, 1998; Goodman, Watson & Burke, 1987; Clay, 1982; Buettner, 2002; Mitchell, 1982). Farrington (2007) believes that we should teach children to act as 'reading detectives, looking for clues to work out the meaning of a text' (p. 8) across the three cueing systems.
The semantic system relates to the meaning conveyed by language, including the reader's prior knowledge about the world and about language (Parker, 1985). The semantic cueing system is employed when readers ask themselves whether what they read is making sense.
The syntactic system refers to the structure of a sentence, including the parts of speech, and grammar (Goodman et al., 1987). Readers use this cueing system to check that the text 'sounds like language' (Goodman & Watson, 1998, p. 122) and that what they are reading follows the rules of language.
The grapho-phonic system describes the combined use of letters (orthographic system) and sounds (phonological system) (Goodman et al., 1987). This system may help a beginning reader to decode a word by assessing its visual and oral properties. If it is a regular word, such as bed, where there is a direct one-to-one correspondence between the letters, b-e-d and their sounds, this may lead to successful decoding. The debate
The exact contribution that each system makes towards the acquisition of
reading has been the subject of popular debate, often characterised as the whole-language versus skills-based approaches. The whole-language (or top-down) approach characterises reading as a process of active meaningmaking (Department of Education and Early Childhood Development (Victoria), 2007). Whole-language proponents such as Goodman and Goodman (2004) argue that all three language cueing systems work together to help a reader make sense of what they are reading, although these are believed to operate at differing levels of importance. Goodman et al. (1987) consider that semantic cues are deep structure processes, while grapho-phonic cues are surface structure processes. Smith (2004) explains that deep structure processes involve knowledge and meaning, while surface structure processes are the physical characteristics of text such as the visual and sound properties. Goodman et al. (1987) put the most emphasis on semantic followed by syntactic cues in the reading process. They suggested that grapho-phonic cues are utilised only when the former systems are unavailable. Whole-language advocates believe that skilled readers are more likely to depend on meaning and grammatical cues and are less likely to use grapho-phonic cues than are less skilled readers (Stanovich, 2000). Goodman (1979) claimed that the semantic acceptability of a reader's miscues prior to correction is the greatest predictor of reading ability. Consequently a focus on meaning is expected to lead to skilled reading (Robinson & McKenna, 2008).
The skills-based (or bottom-up) approach focuses on reading as the breakdown of a whole into parts so that it can be understood. Skills-based theorists place a greater emphasis on the role of the grapho-phonic system. They propose that proficient readers rely initially on visual and sound properties of text, and then on semantic and syntactic cues (Wren, 2008). Less proficient readers tend to rely on semantic and syntactic cues because they do not have sufficient skills to use the graphic and phonic features of text (Nicholson, 1993).
The main difference between these two approaches is the emphasis placed on the strategies employed by skilled readers as they process text (Robinson & McKenna, 2008). Whole-language proponents value semantic and syntactic cues while skills-based supporters value grapho-phonic cues. Westwood (2004) argued that if the whole-language approach is correct, reading instruction should focus on the use of semantic and syntactic strategies; if the skills- based viewpoint is correct, teaching should emphasise visual and sound decoding strategies. Although the theoretical standpoint adopted by educators will influence the instructional methods that are employed in classrooms, it is clear that many classroom practitioners do not take an all or nothing approach, but instead combine aspects of each perspective.
Goodman (1997) introduced the concept of Miscue Analysis in 1965 as a method of examining the reading process. Goodman (2008) assumed that the underlying processes that enable successful reading are identical to the processes that contribute to miscues. Goodman (1979) defines a miscue as 'an actual observed response in oral reading which does not match the expected response' (p. 5). Miscue Analysis provides a comprehensive assessment procedure that examines readers' miscues to see the extent to which the semantic, syntactic and grapho-phonic systems are utilised. It provides information about the reading process (Fitzharris, Blake Jones & Crawford, 2008) through identification of the number of reading errors as well as the hypothesised underlying processes that may have contributed to them. It also provides details about a child's reading ability.
Miscues are common to every reader irrespective of their reading ability (Davenport, 2002). At the individual level, Miscue Analysis can provide information about whether a reader is guessing based on grapho-phonic cues, if the reader substitutes incorrect words that retain the meaning of the expected response, or if the reader understands the text (Campbell & Green, 2006). At the group level we can examine possible developmental patterns for readers of the same age and reading ability.
Since the period 1965 to 1974, when numerous studies investigated children's miscues during oral reading, few Miscue Analysis studies have been reported. Given the high usage of Running Records in Australian primary schools, it seems probable that teachers and researchers have access to valuable information about early reading processes. In a position paper by McKenna and Cournoyer Picard (2006), the authors discuss the relevance of Miscue Analysis in the current educational climate and argue that the assessment of the three language cueing systems is useful for instruction. Research conducted by Long (1986) concluded that although Miscue Analysis is complex and time consuming, it provides a comprehensive picture about the reader and their control over the cueing systems.
Semantic cueing system
Research has illustrated that miscue patterns which maintain the meaning of a text vary across reading ability. In a study conducted with 120 children in their third year of schooling, Clay (1982) explored whether skilled readers were constructing meaning to a greater degree than their less skilled peers. Clay defined reading as a process in which a child 'can extract a sequence of cues from printed texts and relate these, one to another, so that he understands the precise message of the text (1979, p. 13). Based on data from the children's reading of five levelled texts, students were placed into three groups (low, average and high reading accuracy) according to the number of reading errors made. Students in the high reading accuracy group made miscues that maintained text meaning an average of 80% of the time compared with the low reading accuracy group who, on average, demonstrated meaning construction in 54% of their miscues. The more proficient readers relied on the semantic cueing system to a greater extent than less proficient readers when faced with unfamiliar text.
Alexander (2001) attempted to replicate Goodman's (1965) finding that the use of contextual cues is positively associated with reading proficiency. With a sample of 42 students from Grades One to Three, Alexander concluded that the idea that the semantic cueing system increased with reading proficiency was only weakly supported.
Nicholson, Lillas and Rzoska (1988) reported on a small group of 32 young children who read lists of words (no context cues) or short stories (context cues). Their results suggested that six-year-olds' and less skilled eight-year- olds' reading accuracy was improved by contextual cues, while skilled eight-year-old readers were not aided by context. Therefore, skilled eight-year-old readers were less likely to make use of semantic cues than their less skilled counterparts. These findings, which were not subjected to significance testing, are inconsistent with Clay's (1982) and Alexander's (2001) results, suggesting that the evidence concerning use of semantic cueing systems is mixed.
Syntactic cueing system
Goodman (1976) cited developmental findings based on the readings of six young children who were followed for seven years from 1966 to 1972. Goodman noted that as children's reading advances, their tendency to produce syntactically acceptable miscues also increases. Wilde (2000) added that more skilled readers will attempt to make corrections when a sentence fails to sound like a sentence, while a less skilled reader may continue reading without making any amendments.
In 1963, Clay (1982) began a longitudinal study in New Zealand based on weekly observations of 100 five-year-old children reading. On average, it was found that 72% of these beginning readers' errors were substitutions that were classified as syntactically acceptable, with a range between 64% made by the less skilled readers to 76% by the most skilled readers in the group. An average of 41% of the reader's errors were grapho-phonically similar. This research suggests that beginner readers make greater use of the syntactic system than the grapho-phonic system when faced with words that are unfamiliar. Clay also found that 80% of the substitutions made by a high reading accuracy group were completely or partially grammatically acceptable, compared with only 62% of a low reading accuracy group. On the basis of this evidence, it can be assumed that reliance on syntactic cues increases with reading ability.
Grapho-phonic cueing system
A key point made by Goodman et al. (1987) was that skilled readers only use the grapho-phonic system when they cannot gather enough cues from the syntactic and semantic systems. Proficient readers are more inclined to use a mixture of the three cueing systems, while less proficient readers may rely too much on the grapho-phonic cueing system (Wilde, 2000; Tompkins, 2006). Other researchers (Clay, 1982; Laing, 2002) have found little difference in the use of graphic and phonic information across reading ability. However, the generalisability of their findings is constrained by small sample sizes.
Westwood (2004) claimed that the ability to decode a word based on its sound properties is a prerequisite for the employment of the grapho-phonic cueing system during reading. Goodman et al. (1987) argued that greater utilisation of the grapho-phonic system is characteristic of less proficient readers. On the other hand, Clay (1982) and Adams (1990) argued that the ability to use graphic and phonic cues is one of the greatest differentiators between successful, independent readers and struggling readers. This view is held by numerous reading researchers (see National Inquiry into the Teaching of Literacy, 2005 for a comprehensive literature review). For example, Eldredge (2005) found that phonics plays an important role in recognising words, which in turn is important for reading fluency. Research from children with learning difficulties has found that struggling readers often have difficulty grasping symbol-sound correspondence rules (Coltheart & Prior, 2006; Vellutino, Fletcher, Snowling & Scanlon, 2004). Therefore, children with poor letter-sound awareness are more likely to be reading below age and grade expectations. This body of research has provided support for the idea that there is greater reliance on the grapho-phonic system by more successful young readers.
Levelled texts are series of reading books graded in difficulty and are often used to assess reading progress for benchmarks (Brabham & Villaume, 2002). Hatcher (2000) found that the semantic, syntactic and phonic features of 200 Reading Recovery levelled texts could predict the approximate difficulty level of a book 69% of the time. Notwithstanding the gap in understanding what might account for the remaining 31% variance, this suggests that context, grammar and sound demands increase with text level. Clay (2000) uses the categories of Easy (95-100% accuracy), Instructional (90-94% accuracy), and Hard (80-89% accuracy) to depict the degree of ease or difficulty with which an individual child reads a text accurately. Wilde (2000) argues that when a text is too difficult for a child to read, there can be negative emotional effects and a loss of story understanding. It is difficult for a child to maintain meaning when reading each word is a complex task. Treptow, Burns and McComas (2007) investigated the impact of text difficulty on the comprehension of three students who read two passages at three text difficulty levels. The Independent level comprised 100% known words, Instructional level contained 93-97% known words, and the Frustration level contained 80-90% known words. For all three students, comprehension was the highest at the Independent level, followed by the Instructional level and lowest at the Frustration level. Treptow et al.'s study suggests that when a text is too difficult, young readers are unable to monitor meaning. These issues highlight the role played by fluency in reading. Fluent, as opposed to skilled, readers bring both language comprehension and text decoding skills to bear on the reading comprehension task. There are many definitions presented for fluency (see for example, Hoover & Gough, 1990), with some implying that decoding will be effortless, that oral or silent reading will follow the rhyme and rhythm of language, and that comprehension is occurring, all in balance (Abraham, 2000). In other words, fluent reading is at a higher performance level than skilled reading. Others see fluency as a component of skilled reading combining rate (words per minute) and accuracy (number of words correctly identified) of reading (eg. Stanovich, 1991), or decompose fluency into different components (eg. Pressley, Gaskins & Fingeret, 2006). Consequently the definitions of skilled reader and fluent reader appear to be amalgamated and sometimes confused in the research literature.
As text difficulty increases, decoding automaticity can decrease and readers devote most of their attention to the individual word and less to maintaining of meaning (Rasinski & Hoffman, 2003). When a reader is unable to construct meaning, they tend to focus on the superficial aspects of text (Goodman, 1997). Menosky (1976) gave readers stories of increasing difficulty to read and found that as text levels increased, readers relied more on the grapho-phonic system compared to the syntactic and semantic system. Therefore visual and sound features of text are utilised irrespective of reading ability when text difficulty increases.
In this study, miscues made by an Australian sample of young readers, particularly in their use of the three language cueing systems (semantic, syntactic and grapho-phonic), are examined, with a focus on differences across students of varying reading abilities (Below Average, Average, and Above Average) and on different text levels (Easy and Difficult).
On the basis of the evidence from both whole-language and skills-based approaches, it is hypothesised that:
1. A greater percentage of Above Average readers' miscues will maintain meaning and grammatical relationships than those of Average readers, who in turn will maintain these semantic and syntactic relationships more than Below Average readers.
2. A greater percentage of Above Average readers' miscues will maintain graphic and phonic similarity to the correct response than those of Average readers, who in turn will maintain these more than will Below Average readers.
On the basis of evidence on the impact of text difficulty on reading miscues, it is hypothesised that:
3. A higher percentage of miscues that maintain meaning and grammatical relationships will be made when reading at a lower text level, than a higher text level.
4. A higher percentage of miscues that maintain graphic and phonic similarities to the correct response will be made when reading a more difficult text than an easier text.
One hundred students were sampled by random stratification based on reading level from a larger sample of 259 students. The students came from nine urban Australian schools across the first three grades of their education. The mean age of participants was 6.3 years (SD = .72) with ages ranging from five to eight years.
Two levelled text series were used in this study. These were the Catholic Education Commission Victoria Benchmark Kit (referred to in this paper as the 'old' set) and its successor, the AlphaAssess Benchmarking Kit ('new' set). Each kit included a text at each level from 1 to 28. Running Record sheets were provided for each text. Word count ranges in the old set varied from 24 to 163 words per text. In accordance with the benchmarking protocol associated with the old set, not all texts were read in their entirety due to long word lengths and time limitations. Short story texts at levels 1, 2, 3, 5, 6, 7, 15, 18 and 21 were read from beginning to end, as was the case for those texts in the new set where word counts ranged from 28 to 201 per text. Where texts in the new set were longer than this, shorter but meaningful sections were read.
Students were asked to read two matched pairs of texts (two old and two new benchmarking texts), resulting in the collection of data from four texts for each student. Fieldworkers administered the old text first, based on the teacher's estimate of the student's instructional level. If the student did not read at instructional level (90-94% accuracy), a more appropriate levelled text from the old set was selected. Once a reading level was determined the student then read a new text that matched, that is a text identified at the same numbered level of difficulty (1 to 28). The next text pair was then read at the 'hard' level (80-89% accuracy) as determined by the old text set. Some students read all four texts in one sitting, while others were assessed over a number of sittings.
Running Records of all four texts for each student were analysed for miscues. Errors included in the Miscue Analysis were substitutions, omissions, insertions and self-corrections. Repeats of errors were not included. If the reader was told the word during reading by the fieldworker, this was not included as an error because miscue procedures do not provide for analyses of this type. Goodman et al.'s (1987) procedure was used to code each miscue made by each reader and to analyse the context in which the miscue occurred using the following criteria:
1. Syntactic acceptability: does the miscue occur in a structure that is syntactically acceptable in English? Responses coded yes, no, or partial. For partial coding, the miscue is either syntactically acceptable within the initial or last section of the sentence. Alternatively, the miscue is syntactically acceptable within the sentence but not the whole text.
2. Semantic acceptability: does the miscue occur in a structure that is semantically acceptable in English? Responses coded yes, no, or partial. For partial coding, the miscue is either semantically acceptable within the initial or last section of the sentence. Alternatively, the miscue is semantically acceptable within the sentence but not the whole text. Semantic acceptability cannot be coded higher than syntactic acceptability.
If the miscues are both syntactically and semantically acceptable, then the following question can be asked.
3. Meaning change: does the miscue result in a change of meaning? Responses coded yes, no, or partial. For partial, there is inconsistency, loss, or meaning change of a minor idea, incident, character, fact, sequence or concept.
4. Correction: Is the miscue corrected? Responses coded yes, no, or partial. For partial coding, there is either an unsuccessful attempt to correct, or the correct response is read and then discarded.
5. Graphic (visual) similarity: how much does the miscue look like the text? Responses coded as High graphic similarity, Some similarity, and None. An example of high graphic similarity between the correct response and a miscue is reading 'cloth' instead of 'clothes'; of some similarity is 'making' instead of 'changing'.
6. Sound similarity: how much does the miscue sound like the expected response? Responses coded as High sound similarity, Some similarity, and None. An example of high sound similarity between the correct response and a miscue is reading 'a' instead of 'the'; an example of some sound similarity between the correct response and a miscue is reading 'fear' instead of 'phone'.
Once all miscues were coded as above, they were summarised according to Goodman et al.'s (1987) procedures, in order to examine patterns of interrelationships. Patterns of Meaning Construction refer to the readers' level of concern with making sense of the text. The categories reflect the degree to which a reader's miscues had an effect on the construction of meaning. Patterns of Grammatical Relationship refer to the reader's capacity to read sentences that are syntactically acceptable, semantically acceptable and/or corrected. Once all miscues for a text were summarised, Miscues Per Hundred Words (MPHW) and percentages for each code within each variable were calculated.
Analysis and results
Reading data was obtained from a total of four texts for each student, a text pair at a lower reading level and a text pair at a higher reading level. Paired sample t-tests of the mean differences between the MPHW made between each text pair (lower and higher level) revealed no significant differences. Therefore, each levelled text pair was combined by calculating their mean average. Once the pairs were combined to form unified variables, another paired sample t-test was conducted to test for any differences between the MPHW on the lower text level versus the higher text level. Students made significantly more MPHW on the texts at the higher level (M = 9.93, SD = 4.37), compared with those at the lower level (M = 7.67, SD = 5.09), t(99) = 4.37, p < .001). As it can be assumed that text levels increase with difficulty, the lower text level was labelled 'Easy' and the higher text level, 'Difficult'.
Students were grouped according to reading ability across Prep and Grades 1and 2 (these were combined due to the small number of Grade 2 children), into three categories (Below Average, Average, Above Average) in accordance with grade level expectations. Based on minimum standards and targets established by the Catholic Education Office (2007), the cut-offs for reading ability levels, based on the reading of an Easy text, are outlined in Table 1 together with the frequency distributions of participating students in Prep and Grades 1 and 2.
Outliers were dealt with by bringing them closer to the distribution, maintaining their position ranking (Tabachnick & Fidell, 2007). Inspection of the dependent variables (semantic, syntactic, graphic and phonological cueing systems for Easy and Difficult texts) revealed some non-normal distributions which were normalised through a square root transformation. The data were also tested for assumptions of multivariate analysis of variance and found to be consistent with these.
To examine differences across reading ability and the three language cueing systems, a repeated-measures MANOVA was performed with reading ability (Below Average, Average, Above Average) as the independent variable and the language cueing systems as the dependent variables. Significant multivariate main effects were found between subjects for reading ability [Pillai's trace = .31, F(8, 188) = 4.34, p < .001, [[eta].sup.2] = .16] and within subjects for language cueing systems [Pillai's trace = .20, F(4, 93) = 5.64, p < .001, [[eta].sup.2] = .20].
Contrary to the first hypothesis that Above Average readers will score the highest mean percentage of miscues that are semantically and syntactically acceptable, there were no significant differences between reading ability groups in the mean percentage of miscues made. Figure 1 shows that Below Average and Average readers were maintaining meaning an average of 25% of the time when they miscued, compared with 33% for Above Average readers. Results were mixed for the syntactic cueing system. The ability to use semantic and syntactic cues does not distinguish clearly between reading proficiency levels.
However, consistent with the second hypothesis, the between-subjects multivariate effects of reading ability revealed significant differences across the mean percentages of graphically similar miscues [F(2, 96) = 18.31, p < .001, [[eta].sup.2] = .28] and phonologically similar miscues [F(2, 96) = 12.48, p < .001, [[eta].sup.2] = .21]. Bonferroni post-hoc comparisons for the graphic cueing system showed significant mean differences between the Below Average and Average groups [p < .001] and the Below Average and Above Average groups [p < .001] (Figure 1). The Average and Above Average readers appeared to be relying on the visual properties of text approximately 58-59% of the time when faced with unfamiliar words, compared with Below Average readers who appeared to be employing visual cues an average of 35% of the time when they made miscues.
[FIGURE 1 OMITTED]
For the phonological cueing system, Bonferroni post-hoc tests illustrated significant mean differences between Below Average and Average groups (p < .001) and the Below Average and Above Average (p < .001) groups (Figure 1). This demonstrates that Below Average readers were only relying on the sound features of text on average 28% of the time when faced with unfamiliar words. This was substantially different from Average and Above Average readers who showed greater control over phonic cues, using them in approximately 47% of their miscues.
The results suggest that knowledge of graphic and phonic representations of text is observed more in Average and Above Average readers than in their less skilled peers. There were no significant differences between Average and Above Average readers for graphic or phonic cue use. Once readers have gained adequate knowledge of letter-sound correspondences, their ability to utilise grapho-phonic cues does not appear to distinguish between Average and Above Average readers. However, the reliance on visual and sound features of a text does determine the ability to read within grade expectations.
The third hypothesis stated that the average percentage of miscues using the semantic and syntactic cueing systems would be higher for the Easy texts than for the Difficult texts. At the univariate level, significant within- subjects differences were found for the syntactic cueing system [F(1, 99) = 10.54, p = .002, [[eta].sup.2] = .10]. Figure 2 shows the mean percentage of miscues employing the cueing systems across the Easy and Difficult texts. Children maintained stronger grammatical relationships in their reading of easier stories (M = 43.96%, SE = 2.19) than of more difficult stories (M = 34.01%, SE = 1.84). There were no significant differences in the use of semantic cues across text levels, although the results were in the predicted direction. This suggests that when children were faced with a more challenging text, they did not rely on meaning cues to a significantly greater degree.
[FIGURE 2 OMITTED]
The fourth hypothesis stated that readers' miscues would demonstrate greater utilisation of the graphic and phonological cueing systems for Difficult texts than for Easy texts. Significant within-subjects differences were found for the phonological cueing system at the univariate level [F(1, 99) = 3.99, p < .05, [[eta].sup.2] = .04]. Students' miscues sounded more similar when Difficult texts were read (M = 43.32%, SE = 1.93) than Easy texts (M = 38.05%, SE = 2.49). This finding illustrates that as the demands and difficulty of a text increase for individuals, they are more likely to make miscues that are phonologically similar to the expected response. Although in the expected direction, these results were not significant. The children did not vary their reliance on the visual properties of text when reading an Easy or Difficult text.
Lack of significant differences between the reading ability groups in their use of the semantic and syntactic systems in this study challenge the whole- language assumptions of Goodman (1979) and Goodman et al. (1987) who emphasise the role of these systems as key determinants of reading ability. In relation to the semantic cueing system, Above Average readers were not maintaining meaning during reading even half as much as the children in Clay's (1982) study. The discrepancy could be explained by the different methods used to create reading ability groups and the grade levels of the children in the samples. For example, Clay's study categorised reading ability based on accuracy scores, whereas this study in addition differentiated reading skill according to achievement of grade expectations. On this basis the utilisation of semantic and syntactic cues was evident to a similar degree regardless of reading skill. It is plausible that some students' attentional capacities were occupied with the utilisation of other reading strategies (Munro, 1995). This could result in less available resources for constructing meaning and maintaining strong grammatical relationships, as is the argument of some fluency researchers who see the decoding process as acting as a bottleneck and inhibiting comprehension.
Whole-language proponents have argued that the grapho-phonic system is more likely to be used by below-average readers (Tompkins, 2006; Wilde, 2000), although there is evidence that students rely on the grapho-phonic cueing system to a similar degree regardless of reading skill (e.g. Laing, 2002; Clay, 1982). Skills-based supporters have proposed that a reliance on the grapho- phonic system is more often observed among more skilled readers (Adams, 1990). Consequently, it was anticipated that the Above Average reading ability groups would reveal the highest usage of the grapho-phonic cueing systems, followed by Average and then Below Average readers. This hypothesis was partially supported with the Below Average group showing least use for both graphic and phonological similarity. In relation to the graphic cueing system, the mean percentage of visually similar miscues was greatest for the Above Average readers, and lowest for the Below Average readers. This suggests that students with Average and Above Average reading ability rely more on the visual aspects of print when reading unfamiliar text more than students with Below Average reading abilities. This pattern was also observed with regards to the phonological cueing system. There were no significant differences between the Average and Above Average readers in their use of grapho-phonic cues. This suggests that the ability to apply letter-sound correspondence rules distinguishes grade appropriate readers from children who are reading below expectations. However, once young children have attained this skill it does not differentiate the Average from the Above Average readers.
Goodman (1976) assumes that the grapho-phonic system is only utilised when semantic and syntactic cues are unavailable, and that less skilled readers spend most of their time relying on the visual and sound properties to decode text. The findings of this study are more consistent with those researchers who have found that poorer readers have difficulty grasping letter-sound correspondence rules (e.g. Coltheart & Prior, 2006; Elderedge, 2005; Vellutino, Fletcher, Snowling & Scanlon, 2004; Westwood, 2004; De Lemos, 2002; Adams, 1990). Therefore, children with poor letter-sound awareness skills are more likely to be reading below grade expectations. This was supported by the finding that children who are not reading in accordance with grade expectations appear to be relying significantly less on the visual and sound properties of text compared with their more skilled peers. It is possible that this is one aspect of their reading that is restricting their progress.
This study revealed that children who were reading at or above grade expectations were more likely to make use of graphic and phonic cues when they approached an unfamiliar word in a text. As has been suggested (e.g. De Lemos, 2002; Adams, 1990) the ability to decode graphic and phonic information is a reliable determinant of higher reading ability. Therefore, these findings suggest that Goodman and colleagues may have underestimated the role of the grapho-phonic cueing system and overestimated the role of the semantic and syntactic cueing systems in relation to differentiating reading ability.
The third and fourth hypotheses investigated differences among the three language cueing systems when a student is reading with greater or lesser amounts of accuracy. It was proposed that the mean percentage of miscues that maintain meaning and reflect strong grammatical relationships would be greater when children are reading at a lower text difficulty level compared to a higher text difficulty level. This hypothesis was partially supported. Children's miscues demonstrated the maintenance of strong grammatical relationships to a higher degree when they were reading an easier text. Although in the hypothesised direction, this significant difference did not extend to semantically acceptable miscues. This finding is inconsistent with prior research (e.g. Treptow, Burns & McComas, 2007; Goodman, 1976). It is possible that the differences in methodology (assessment of miscues that were maintaining meaning versus asking comprehension questions) contributed to varying results.
Peverly and Kitzen (1998) suggested that if young children have not developed a sound understanding of grammatical rules, then texts with more complex syntactical structures may be more difficult to read. Cunningham et al.'s (2005) research used Reading Recovery texts to examine the abundance of print-based demands, specifically those at the word-, sentence- and discourse- level (somewhat similar to the three language cueing systems). The researchers found a significant difference in sentence-level demands (syntactic system) which were greater for higher text levels. Therefore, as text difficulty increased, it is possible that the syntactic demands of the text were also increasing. This may have made it more difficult for the children to monitor whether what they were reading followed the rules of language, especially since they were simultaneously making use of other language cueing systems. Adams (1990) acknowledged that the ability to maintain strong grammatical relationships when reading may contribute more to reading than previously realised.
The hypothesis that children would make a greater percentage of miscues that were graphically and phonologically similar when reading a more difficult text was also partially supported. Students' miscues were more likely to sound like the expected response when they were reading a Difficult text. Consistent with Menosky (1976), who showed that children rely more on sound properties as a text becomes more challenging, it appears that when students' ability to read a text accurately declines, they tend to rely on the sound properties of unfamiliar words to aid recognition. This should stimulate consideration of Goodman's (2008) assumption that the processes used in successful reading are identical to those that underlie errors.
Strengths, limitations and implications for future research
Miscue Analysis assumes that the underlying processes that contribute to making an unexpected response when reading are equivalent to those that result in an expected response (Goodman, 2008). However, Miscue Analysis cannot capture every process or strategy used by a reader as they attempt to read an unfamiliar word or sentence. This issue is especially pronounced as readers become more skilled, because it is likely that they are selecting cues silently prior to identifying a word. As a result, the miscues that were recorded in this study provide a limited picture of the reading process. Also, this study did not examine the influence of a child's prior knowledge on miscues relating to context of text. Afflerbach (1998) suggested that measurements of prior knowledge might be utilised to rule out the possibility of such bias. Therefore, future research could use a variety of assessment procedures that examine both expected and unexpected responses and content-domain knowledge to investigate the roles of the three language cueing systems during reading.
Many of the studies reported here base their results on very small numbers of participants. This study comprised a sample of 100 children, greater than many of the sample sizes from Goodman's (1976) research. The sample size in this study allowed for more robust empirical investigation with tests of statistical significance.
Goodman et al. (1987) claim that children need to read a whole text from beginning to end to enable an in-depth analysis of at least 25 miscues per text. Due to the nature of this study, not all children read entire stories, although it was ensured that the remainder read meaningful sections of their longer texts. In addition, because children did not read texts that were overly difficult for them, few children made as many as 25 miscues. It would be preferable to examine the impact of different texts and text lengths in accordance with Goodman et al.'s criterion of a minimum of 25 miscues per text. It is possible that semantic and syntactic cues might have been better represented.
Of concern in reading research is the proliferation of definitions and concepts of 'skilled reader'. It appears that definitions change dependent on the goal of the activity or the theoretical perspective taken by the researcher. This complicates comparisons across studies, with consequent problems for evaluation of research in the field.
Implications for intervention and instruction
The emphasis placed on the visual and sound properties of words, as opposed to context, during reading has implications for teaching. This study revealed that students reading at or above grade level expectations produced more grapho-phonic miscues than students reading below expectation. The results suggest that irrespective of reading ability, when a text becomes more difficult, miscues based on sound similarity increase. These findings provide support for the skills-based approach to reading and for the notion that the application of letter-sound correspondence rules is associated with beginning reading skills.
When a text becomes more difficult, syntactic acceptability of miscues decreases. This suggests that reading interventions and instruction should target grammatical relationship skills. Even though many researchers support phonics based approaches, most still argue that meaning-based teaching should also be included. Peverly and Kitzen (1998) and Adams (1991) argue that both approaches could be expected to create more successful results than focussing exclusively on either phonics or context. Rowe (2006) also emphasises that direct letter-sound correspondence skills should be taught first followed by whole-language activities, resulting in the use of both major approaches.
Through the use of Miscue Analysis, this study has demonstrated that Average and Above Average readers used the grapho-phonic cueing system, or letter-sound correspondence rules, to a greater extent than Below Average readers when faced with unfamiliar text. It has also been demonstrated that students have better control over syntactical cues when the text is easier for them. However, the use of the semantic cueing system was not significantly greater for any one reading ability group or any particular level of text difficulty. These findings add to a growing body of evidence that supports the critical contribution of phonics skills to grade appropriate reading and decoding more difficult texts. It is recommended that practitioners consider how useful it might be to implement skills-based training reading intervention programs alongside whole-language focused instructional approaches.
The data used in this research was collected as part of a Catholic Education Office Melbourne project. The authors wish to acknowledge and thank the participating primary schools, Rosemary McLoughlin and the Literacy Team.
Abraham, P. (2000). Skilled reading: top-down, bottom-up. Field Notes, 10(2), SABES/ World Education.
Adams, M.J. (1998). The three-cueing system. In J. Osborn and F. Lehr (Eds.), Literacy for all: Issues in teaching and learning, (pp. 73-99). New York: The Guilford Press.
Adams, M.J. (1991). Beginning to read: A critique by literacy professionals and a response by Marilyn Jager Adams. The Reading Teacher, 44(6), 370-395.
Adams, M.J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press.
Afflerbach, P. (1998). Reading assessment and learning to read. In O. Jean and F. Lehr (Eds.), Literacy for all: Issues in teaching and learning (pp. 239-263). New York: The Guilford Press.
Alexander, J.C. (2001). Reading skill and context facilitation: A classic study revisited. The Journal of Educational Research, 91(5), 314-318.
Brabham, E.G., & Villaume, S.K. (2002). Leveled text: The good news and the bad news. The Reading Teacher, 55(5), 438-441.
Buettner, E.G. (2002). Sentence by sentence self-monitoring. The Reading Teacher, 56(1), 34-44.
Campbell, R., & Green, D. (Eds.) (2006). Literacies and learners: Current perspectives (3rd ed.). Frenchs Forest: Pearson Education.
Catholic Education Office. (2007). A guide to using P-2 literacy data: Part 1. Melbourne: Catholic Education Office.
Clay M.M. (1979). Reading: The patterning of complex behaviour. Auckland: Heinemann Publishers.
Clay, M.M. (1982). Observing young readers: Selected papers. Portsmouth: Heinemann.
Clay, M.M. (2000). Running Records: For classroom teachers. Auckland: Heinemann Publishers.
Coltheart, M., & Prior, M. (2006). Learning to read in Australia. Australian Journal of Learning Disabilities, 11(4), 157-164.
Cunningham, J.W., Spadorcia, S.A., Erickson, K.A., Koppenhaver, D.A., Sturm, J.M., & Yoder, D.E. (2005). Investigating the instructional supportiveness of levelled texts. Reading Research Quarterly, 40(4), 410-427.
Davenport, M.R. (2002). Miscues not mistakes: Reading assessment in the classroom. Portsmouth: Heinemann Publishers.
De Lemos, M. (2002). Closing the gap between research and practice: Foundations for the acquisition of literacy. Camberwell, Vic: Australian Council for Educational Research.
Department of Education and Early Childhood Development (Victoria). (2007). Evidence-based research for expert literacy teaching (Paper 12). Melbourne: Office for Education Policy and Innovation.
Department of Education, Science and Training (DEST) (2005). Teaching Reading: Report and Recommendations. National Inquiry into the Teaching of Literacy. Canberra: Australian Government.
Eldredge, J.L. (2005). Foundations of fluency: An exploration. Reading Psychology, 26(2), 161-181.
Farrington, P. (2007). Using context cues: children as reading detectives. Literacy Today, 8-9.
Fitzharris, L., Blake Jones, M., & Crawford, A. (2008). Teacher knowledge matters in supporting young readers. The Reading Teacher, 61(5), 384-394.
Goodman, K.S. (2008). Miscue Analysis as scientific realism. In A.D. Flurkey, E.J. Paulson & K.S. Goodman (Eds.), Scientific realism in studies of reading. New York: Lawrence Erlbaum Associates.
Goodman, K.S. (1979). Miscues: Windows on the reading process. In K.S. Goodman (Ed.), Miscue Analysis (pp. 3-14). Illinois: Clearinghouse on Reading and Communication skills.
Goodman, K.S. (1965). A linguistic study of cues and miscues in reading. Elementary English, 42(6), 639-643.
Goodman, Y.M. (1997). Reading diagnosis--qualitative or quantitative? The Reading Teacher, 50(7), 534-538.
Goodman, Y.M. (1976). Developing reading proficiency. In P.D. Allen and D.J. Watson (Eds.), Findings of research in Miscue Analysis: Classroom implications (pp. 113-127). Illinois: Clearinghouse on Reading and Communication skills.
Goodman, Y.M., & Goodman, K.S. (2004). To err is human: Learning about language Processes by analysing miscues. In R.B. Ruddell, & E.J. Unrau (Eds.), Theoretical models and processes of reading. (5th ed.). Newark: International Reading Association.
Goodman, Y., & Watson, D. (1998). A sociolinguistic model of the reading process and reading strategy instruction. In C. Weaver (Ed.), Practicing what we know: Informed reading instruction. Illinois: National Council of Teachers of English.
Goodman, Y.M., Watson, D.J., & Burke, C.L. (1987). Reading miscue inventory: Alternative procedures. New York: Richard C. Owen Publishers, Inc.
Hatcher, P.J. (2000). Predictors of Reading Recovery book levels. Journal of Research in Reading, 23(1), 67-77.
Hempenstall, K. (2003). The Three-Cueing System: Trojan Horse? Australian Journal of Learning Disabilities, 8(2), 15-23.
Hoover, W.A., & Gough, P.B. (1990). The simple view of reading. Reading and Writing: An Interdisciplinary Journal, 2, 127-160.
Laing, S.P. (2002). Miscue Analysis in school-age children. American Journal of SpeechLanguage Pathology, 11, 407-416.
Long, P.C. (1986). The effectiveness of reading miscue instruments. [unpublished research paper, Feb 1986, No. 13, Melbourne College].
McKenna, M.C., & Cournoyer Picard, M. (2006). Revisiting the role of miscue analysis in effective teaching. The Reading Teacher, 60(4), 378-380.
Menosky, D.M. (1976). #1-Grapho-Phonic. In P.D. Allen and D.J. Watson (Eds.), Findings of research in Miscue Analysis: Classroom implications (pp. 75-79). Illinois: ERIC Clearinghouse on Reading and Communication skills.
Mitchell, D.C. (1982). The process of reading: A cognitive analysis of fluent reading and Learning to read. Chichester: John Wiley & Sons Ltd.
Munro, J. (1995). Explaining developmental dyslexia: Orthographic processing difficulties. Australian Journal of Remedial Education, 27(1), 5-15.
Nicholson, T. (1993). Reading without context. In G.B. Thompson, W.E. Tumner and T. Nicholson (Eds.), Reading Acquisition Processes (pp. 105-120). Clevedon: Multilingual Matters.
Nicholson, T., Lillas, C., & Rzoska, M.A. (1988). Have we been misled by miscues? The Reading Teacher, 42(1), 6-10.
Parker, R. (1985). The Reading Process: The practicality of good theory. In L. Unsworth (Ed.), Reading: An Australian Perspective. Melbourne: Thomas Nelson.
Peverly, S.T., & Kitzen, K.R. (1998). Curriculum-based assessment of reading skills: Considerations and caveats for school psychologists. Psychology in the Schools, 35(1), 29-47.
Pressley, M., Gaskins, I.W., & Fingeret, L. (2006). Instruction and Development of Reading Fluency in Struggling Readers. In S.J. Samuels and A.E. Farstrup (Eds.) What Research Has to Say About Fluency Instruction (pp. 47-69). International Reading Association.
Rasinski, T.V., & Hoffman, J.V. (2003). Theory and research into practice: Oral reading in the school literacy curriculum. Reading Research Quarterly, 38(4), 510-522.
Robinson, R.D., & McKenna, M.C. (2008). Issues and Trends in Literacy Education (4th ed.). Boston: Pearson Education, Inc.
Rowe, K. (2006). Effective teaching practices for students with and without learning difficulties: Issues and implications surrounding key findings and recommendations from the National Inquiry into the Teaching of Literacy. Australian Journal of Learning Disabilities, 11(3), 99-115.
Smith, F. (2004). Understanding Reading: A Psycholinguistic Analysis of Reading and Learning to Read. (6th ed.). New Jersey: Lawrence Erlbaum Associates.
Stanovich, K.E. (2000). Progress in understanding reading: Scientific foundations and new frontiers. New York: The Guilford Press.
Stanovich, K.E. (1991). Word recognition: Changing perspectives. In R. Barr, M L. Kamil, P. Mosenthal, & P.D. Pearson (Eds.), Handbook of reading research (Vol. 2, pp. 418-452). New York: Longman.
Tabachnick, B.G., & Fidell, L.S. (2007). Using Multivariate Statistics (5th ed.). Boston: Allyn and Bacon.
Tompkins, G.E. (2006). Literacy for the 21st century: A balanced approach. New Jersey: Pearson Education.
Treptow, M.A., Burns, M.K., & McComas, J.J. (2007). Reading at the frustration, instructional and independent levels: The effects on students' reading comprehension and time on task. School Psychology Review, 36(1), 159-166.
Vellutino, F.R., Fletcher, J.M., Snowling, M.J., & Scanlon, D.M. (2004). Specific reading disability (dyslexia): what have we learned in the past four decades? Journal of Child Psychology and Psychiatry, 45(1), 2-40.
Westwood, P. (2004). Reading and Learning difficulties: Approaches to teaching and assessment. London: David Fulton.
Wilde, S. (2000). Miscue analysis made easy: Building on student strengths. Portsmouth, NH: Heinemann.
Wren, S. (2008). Reading and the three cueing systems. Southwest Educational Development Laboratory. Retrieved September 23, 2008 from http://www.sedl.org/ reading/topics/cueing.html.
Lauren Beatty & Esther Care
The University of Melbourne
Table 1 Frequency Distributions and Reading Level Ranges of Reading Ability Variables Categories n Reading n Reading Level Range Level Range in Prep in Grades One/Two Below Average 18 1-4 8 1-19 Average 37 5-14 12 5-14 Above Average 112 I 15-28 112 15-28