Journal of Educational Psychology
2011, Vol. 103, No. 1, 190 –205
© 2011 American Psychological Association
0022-0663/11/$12.00 DOI: 10.1037/a0021621
Orthographic Analogies and Early Reading:
Evidence From a Multiple Clue Word Paradigm
Robert S. Savage, Louise Deault, Julia Daki, and Julie Aouad
McGill University
Two experiments using a variation of the clue word analogy task (Goswami, 1986) explored whether
children can make orthographic analogies when given multiple clue words, beyond the known effects of
purely phonological activation. In Experiment 1, 42 children (mean age 6 years and 8 months) were first
taught 3 “clue” words (e.g., fail, mail, jail) and then shown target words sharing orthographic and
phonological rimes (e.g., hail), phonological rimes (e.g., veil), orthographic and phonological vowel
digraphs (e.g., wait), phonological vowel digraphs (e.g., vein), or unrelated controls (e.g., bard). All word
types were advantaged at posttest over unrelated controls. A small additional advantage for orthographic
and phonological rimes over phonological rimes was evident in by-participant analysis. Finally, regression analysis showed a specific relationship between onset-rime phonological awareness and orthographic rime clue word task transfer. Experiment 2 replicated Experiment 1 with 30 children (M age ⫽
7 years, 0 months) and added a distinct group of children taught multiple clue words sharing vowel
digraphs (e.g., gait, maim, maid). Results showed advantages for all words over unrelated controls and
a small additional advantage for orthographic and phonological vowel digraphs over phonological vowel
digraphs in the by-participant analysis. Overall, results suggest that some young children do have the
ability to make orthographic analogies when given multiple exemplars but that most improvement in
target word reading reflects purely phonological activation. Practical steps for identifying genuine
analogy use in a subset of children are thus described.
Keywords: reading, analogy, children, phonology, orthography
sentation of spelling knowledge of GPCs in words (Ehri, 1992,
2005; Savage & Stuart, 2006; Stuart & Coltheart, 1988).
Rime-based models of early reading acquisition (Goswami,
1993, 1999; Goswami & Bryant, 1990) emphasize analysis of a
syllable into the initial consonant or consonants (sometimes referred to as “onsets”) and “rimes” (the subsyllabic unit containing
the vowel and terminal consonant[s] of a syllable). In the interactive analogy model first proposed by Goswami, (1993), onset-rime
phonological awareness is held to drive early onset-rime-based
orthographic analogy use in beginner readers. In analogy tasks,
children are shown a word such as beak, are told its pronunciation,
and then are asked to read words such peak or bean.1 On the basis
For some time now, cognitive– developmental models of early
reading have rested on the assumption that accurate word recognition is a crucial corequisite to skilled reading comprehension
(see e.g., Ehri, 2005; Perfetti, 2003, 2007; Snowling & Hulme,
2005; for reviews). In turn, phonological awareness, the ability to
reflect on or manipulate the sound elements of the language,
appears to be necessary but not sufficient for the acquisition of
word recognition skills in English (Byrne, 1998; Goswami &
Bryant, 1990; Snowling, & Hulme, 2005) and many, if not all,
other languages (Caravolas, Volı́n, & Hulme, 2005; McBrideChang et al., 2005; Perfetti, 2007; Share, 2008; Ziegler & Goswami, 2005).
Beyond this general consensus, two sorts of phonology-based
models of reading acquisition are discernible in the literature in
English. Some models of word reading acquisition are based on
small grapheme-to-phoneme correspondences (GPCs). These
models emphasize analysis of all constituent phonemes of a syllable (e.g., back segmented into b–a–ck; Coltheart, Curtis, Atkins,
& Haller, 1993). Some GPC-based theories emphasize that word
reading acquisition is marked by qualitative changes in the repre-
1
The term “analogy” usually refers to approaches to reasoning wherein
the relationships between sets of concepts are shared but where the content
of the concepts involved is not. It might thus be argued that the process of
deriving a pronunciation of a word where both the relation and the content
(e.g., spelling and pronunciation of eak in beak and peak) are shared across
words should not be referred to as an analogy. The term “analogy” was first
used to describe the speeded reading of words and pseudowords by literate
adults (Glushko, 1979; Kay & Marcel, 1981). Soon after, studies indentifying children’s apparent sensitivity to orthographic rimes in readingaccuracy tasks were reported and early rime analogy by children was seen
as the developmental precursor of adult rime analogy shown in priming
effects in speeded word-reading tasks (e.g., Goswami, 1986). No fully
articulated conceptual model bridging these adult and developmental usages of the term “analogy” has yet been advanced beyond the shared role
of larger rime units. Despite these various conceptual issues surrounding
the term, “analogy” is now widespread in all of these literatures and for this
reason alone is retained here.
This article was published Online First January 24, 2011.
Robert S. Savage, Louise Deault, Julia Daki, and Julie Aouad, Department of Educational and Counselling Psychology, Faculty of Education,
McGill University, Montreal, Quebec, Canada.
Correspondence concerning this article should be addressed to Robert S.
Savage, Department of Educational and Counselling Psychology, Faculty
of Education, McGill University, Montreal, Quebec, Canada H3A 1Y2.
E-mail: [email protected]
190
ANALOGIES FROM MULTIPLE CLUE WORDS
of this and other evidence (e.g., Zorzi, Houghton, & Butterworth,
1998a 1998b), Goswami (1999) argued that onset-rime phonological awareness comes to represent a distinct pathway to reading in
English independent of spelling knowledge or use of GPC rules.
In more recent rime accounts, language-specific influences of
orthographic rimes have been posited. Ziegler and Goswami
(2005) reviewed evidence that rimes are more consistently represented than are GPCs in English. GPCs are, however, excellent
guides to pronunciation in other “transparent” spelling systems,
such as German or Italian. On this basis, they advanced a psycholinguistic grain size hypothesis, which argues, in essence, that
small grain-sizes reflecting GPCs will be most useful for all
readers in orthographies, such as German and Italian, containing
many words that can be pronounced accurately using such GPCs.
In English, which allows many exceptions to pronunciations derived from GPCs, other larger grains of correspondence between
clusters of letters and phonological abilities at the level of onsetrime, alongside the use of analogies based on orthographic rimes
to learn to read, may also be beneficial. Goswami, Ziegler, and
Richardson (2005) make the strong claim that differences in
spelling-to-sound consistency between English and other transparent orthographies result in the use of different orthographic grain
sizes by beginning readers.
There is some evidence that children’s use of rime analogy is
indeed limited in transparent orthographies such as Finnish or
Dutch (Holopainen, Ahonen, & Lyttinnen, 2002; van Daal,
Reitsma, & van der Leij, 1994). Goswami et al. (2005) claimed
that early sensitivity to orthographic rimes is also reflected in
performance in orally administered phonological awareness tasks.
In their study, English and German children in kindergarten (considered prereaders) and Grade 1 (considered readers) were asked to
detect the odd word out in sequences such as bank, tank, pink,
where the shared rime (-ank) is always represented consistently in
spelling, compared with word sets such as good, hood, loud, where
the rime (-ood) can be spelled in more than one way (e.g., good vs.
food). English readers in Grade 1 were the only group to show
advantages for orthographically consistent oddity words. This
study was executed in a cross-sectional manner. With four distinct
groups, matching on extraneous pupil variables is crucial. However, little matching on important pupil background was reported,
and the German and English children also differed in age. The best
test of the hypothesis would be with two groups of age-matched
children followed longitudinally as they move from preliteracy to
becoming readers. Furthermore, the two experiments in the study
did not in fact find the crucial three-way Consistency (consistent
vs. inconsistent oddity stimuli) ⫻ Group (prereader vs. reader) ⫻
Language (English vs. German) interaction to reach statistical
significance. There was also no direct assessment of the word
reading ability for the oddity task items themselves, so the assumption that orthographic differences underlie phonological task differences cannot be verified. Together the evidence from this study
should probably be treated very cautiously until data from wellexecuted replications are available.
However, even this broad dichotomy between rime and
phoneme-based models of word reading acquisition is an oversimplification for several reasons. First, neither the rime nor GPCbased models are presented as exclusive accounts of word reading
acquisition. The two models differ in the developmental sequence
of children’s mastery of the two types of linguistic units. Rime-
191
based models admit an important role for GPC units in learning to
read once children have been taught and learned them, and similar
distinct contributions of phoneme-to-grapheme correspondence
knowledge in spelling that come to influence reading (Goswami,
1999). Many phoneme-based models admit the converse for later
rime use, albeit after reliable GPC decoding has first been established (Duncan, Seymour, & Hill, 1997; Ehri & Robbins, 1992;
Seymour, Duncan, & Bolik, 1999). The rime-based models have
been termed large-units-first models, and the GPC-based models
have been termed small-units-first accounts (G. D. A. Brown &
Deavers, 1999).
Most small-units-first accounts argue that children become
aware of the consistencies between spellings and sound at the level
of orthographic rimes only after significant amounts of exposure to
the spelling system (Treiman, Mullennix, Bijeljac-Babic, &
Richmond-Welty, 1995). A good deal of evidence is consistent
with this hypothesis (e.g., Bowey & Hansen, 1994; Bowey &
Underwood, 1996; G. D. A. Brown & Deavers, 1999; Stuart,
Masterson, Dixon, & Quinlan, 1999).
These small-units-first models can be contrasted with early rime
use models that have argued that children’s very early sensitivity
to phonological rimes, and a purported causal link between early
phonological rime awareness and word reading, drives the use of
orthographic rime at the beginning of word reading acquisition
(e.g., Goswami, 1986, 1993; Goswami & Bryant, 1990). As noted
earlier, the psycholinguistic grain size hypothesis appears to assume that both early and late rime strategies contribute to the
emergence of sensitivity to orthographic rime consistency in English readers (e.g., Goswami et al., 2005; Ziegler & Goswami,
2005). There is some evidence that explicit onset-rime awareness
in children aged 4 –5 years is associated with their later sensitivity
to rime units, when aged 8 –9 years (Seymour et al., 2000).
However, early and late rime-use strategies clearly reflect distinct
cognitive mechanisms. Early orthographic rime use is based on
very young children’s phonological sensitivity to rimes in the
absence of much knowledge of the characteristics of English
spelling system. Later rime use is based on sensitivity to orthographic rime consistency after substantial exposure to the orthography. It is the claim that children have the ability to use orthographic rimes early that is investigated here.
The Strategic Use of Rime in Early Reading
Although most models of reading described above differ substantially, they are all developmental accounts in the sense that
they relate use of different phonological grains at the onset-rime
and phoneme level for reading to chronological age and reading
experience. Some subsequent models of early reading have instead
emphasized a strategic approach to reading tasks by young children that is more dependent on the exigencies of the task than
previous models assumed. G. D. A. Brown and Deavers (1999)
argued that children’s early reading is characterized by a flexibleunit-size strategy. They argue that children experience a mismatch
between strong rime phonological skills in early word reading
alongside emerging letter-level spelling knowledge. As a result,
children seek to develop both levels of representation in the mental
lexicon. Consequently, if the demands of a particular experimental
task allow children to use their phonological onset-rime skills
more readily, they will operate at this grain, but if a task allows
192
SAVAGE, DEAULT, DAKI, AND AOUAD
children to use their orthographic letter-sound skills more readily,
they will instead operate at this grain. Their approach to experimental tasks is in this sense strategic in that children choose
flexibly how best to approach a given reading task with the
phonological and orthographic resources they have available at
that time. Consistent with this view, Brown and Deavers showed
that when children are given a pseudoword, such as dalk, in
isolation, pronunciation is more likely to reflect use of GPCs,
whereas when cued by a concurrent rime-based analogy word,
such as talk, pseudoword pronunciations of dalk will often rime
with the word talk.
Some studies of pseudoword processing reporting both rime
consistency and word regularity effects in reading are also broadly
consistent with these claims of parallel grains of representation
(Stuart et al., 1999). Recent studies of children’s self-reported
strategy use in analogy tasks and isolated word reading tasks
(Farrington-Flint, Coyne, Stiller, & Heath, 2008; Farrington-Flint
& Wood, 2007) and studies showing differences in approaches to
decoding following distinct forms of classroom instruction
(Deavers, Solity, & Kerfoot, 2000) also suggest flexibility in
phonological rime or phoneme unit size use in reading tasks. On
the other hand, much evidence exists that on more explicit phonological tasks, prereading children show advantages for phonemes over rime (e.g., Cassady & Smith, 2004; Geudens, Sandra,
& Martensen, 2005; Savage, Blair, & Rvachew, 2006), suggesting
that explicit phoneme and letter knowledge skills may not be
mismatched in prereaders. Furthermore, as G. D. A. Brown and
Deavers (1999) acknowledged, in experimental tasks where the
clue word (e.g., talk) is available while children are asked to
pronounce a target (e.g., dalk), the rime-based pronunciations
reported might entirely reflect activation of purely phonological
rime information (e.g., Savage, 1997, 2001).
Evidence for the Early Use of Rime in Reading
Acquisition
One of the strongest sources of evidence for the early use of
rime-based orthographic analogy in early reading comes from
Goswami’s clue word analogy task. Here, children are first shown
a set of analogy target words, such as beak and bean, and control
words, such as lake. In a subsequent posttest, children are shown
the pronunciations of clue words, such as peak, and reread the
target words. Sometimes particular advantages in reading rimeanalogous target words have been found at posttest, possibly
suggesting a privileged status for onset and rime-based analogies
(e.g., Goswami, 1986, 1988). For example, in a study by Goswami
(1993, Experiments 2 and 3), prereaders and beginning readers
read words sharing the rime (e.g., wink-pink), the onset (trim-trot),
the head (trim-trip), or a vowel digraph meat-heap). Children
made more analogies to words sharing onsets and rimes than to
words sharing other letter clusters. Such results support the interactive analogy model of early onset-rime use (Goswami, 1993), in
which phonologically underpinned onset-rime representations
emerge first in reading development. Subsequently, several studies
have reported equal transfer for rime and nonrime units in young
children (Bowey, Vaughn, & Hansen, 1998; Duncan et al., 1997;
Savage, 2001; Savage & Stuart, 1998), which does not so readily
support the interactive analogy model.
Phonological Priming Effects in Clue Word
Analogy Studies
Other research has suggested that variation in the clue word task
influences evidence of clue word facilitation (G. D. A. Brown &
Deavers, 1999; Muter, Snowling, & Taylor, 1994; Savage, 1997).
The availability of clue word information at posttest appears to be
associated with facilitation at posttest. The possibility that clue
word information may act in some way to activate or prime only
phonological information and not spelling information among
words has been the focus of particular attention in the literature.
Phonological priming was first reported in the adult reading literature (e.g., Tanenhaus, Flanigan, & Seidenberg, 1980) and refers
to the activation of related pronunciations following exposure to a
word pronunciation. In the developmental literature on analogy,
“phonological priming” has come to be used as a general term to
describe any evidence of the activation of rime phonology in the
absence of relevant orthographic information in rime tasks that
purport to measure orthographically based reading processes.
Rime phonological priming is thus clearly indicated in the clue
word task when (a) children are provided with only phonological
clue word information at posttest and yet still show improvement
in pronouncing target words (e.g., Savage, 1997) or (b) when
showing transfer to target words that share pronunciations but not
spelling patterns (e.g., beak-seek; e.g., Bowey, 1999).
Many studies to date have reported very strong phonological
rime-priming effects in the clue word task (Bowey, 1999; Bowey
et al., 1998; Nation, Allen, & Hulme, 2001; Roberts & McDougall,
2003; Savage, 1997; Savage & Stuart, 1998; Wood, 2002; Wood
& Farrington-Flint, 2001). This suggests that maybe very little or
even none of the rime analogy effect reflects any learning about
spelling patterns in words (cf. Farrington-Flint & Wood, 2007).
Although aspects of methodology and interpretation of some reported effects have been disputed (e.g., Goswami, 1999), the
findings and procedures used have been replicated and generalize
across variations in clue word paradigms (Bowey, 1999; Nation et
al., 2001; Savage, 2001). Perhaps decisively, Roberts and McDougall (2003) used an analogy task derived closely from one of
Goswami’s own studies (Goswami, 1990). Here children were
invited to read words sharing orthographic rimes (e.g., most-post),
phonological rimes (e.g., most-toast), or ambiguous words (e.g.,
most-cost). Significant advantages for words sharing orthographic
rimes over the other two word types would suggest that the
analogy effect does have an orthographic component. Roberts and
McDougall found, however, that phonological rimes were equivalent to orthographic rimes at posttest. This suggests strongly that
improvement in target word reading in the clue word analogy task
reflects the activation of only phonological information.
Studies of Multiple Clue Words
Given that phonological priming effects may explain all or a
very large part of the pattern of improvement in target word
reading in traditional clue word tasks, Savage and Stuart (1998,
Experiment 2) sought to explore the effects of teaching multiple
clue words and then asking children to read target words in a
separate session where clues were not present. This approach may
be a good way to encourage the use of genuine orthographic
analogy as evidence from studies of children’s use of analogy in
ANALOGIES FROM MULTIPLE CLUE WORDS
inductive reasoning tasks suggests that multiple examples may
increase analogy use (A. L. Brown & Kane, 1988; FarringtonFlint, Wood, Canobi, & Faulkner, 2004; Goswami, 1993). The
presence of multiple clue word exemplars is also likely to encourage orthographic rime use. The use of multiple clues also more
closely approximates the approaches that educators might use in
classroom reading interventions using rime family approaches
(e.g., Christensen & Bowey, 2005; Comaskey, Savage, & Abrami,
2009; Johnston, 1999). In sum, there are several reasons to believe
that experiments with multiple clue words will have an increased
sensitivity for detecting the use of orthographic rimes over experiments with single clue words.
Savage and Stuart (1998, 2001) pretaught children families of
three riming words such as bark, mark, and park, and then asked
to read words such as dark presented in isolation a few minutes
later. Children showed that they could use pretaught clue word
information to read analogous target words but were no better at
this for rime-analogous words than vowel-analogous words (e.g.,
children were pretaught bark, mark, dark, and then shown the
untaught word part). Savage and Stuart (2006) explored the longitudinal importance of the analogy strategy in a sample of 6-yearolds. Results demonstrated that vowel-based, but not rime-based
analogy at age 6 predicted reading skills at age 8. It is important to
note, however, that no control was provided for the possible effects
of phonological rime priming in any of these studies. It is arguable
that controls for a phonological prime are required to establish
whether orthographic learning has occurred, even in this multiple
clue word design, as priming effects may affect results in any
context where rhyming words are presented over relatively short
periods of time (Farrington-Flint & Wood, 2007; Roberts &
McDougall, 2003). Study 1 therefore investigates this important
unresolved issue contrasting target words sharing both orthography and phonology (e.g., seek-reek) and those sharing only phonology (e.g., seek-beak).
The Contribution of Rhyme and Phoneme Awareness
to Analogy Use and Reading
Goswami and Bryant’s (1990) original model of reading acquisition makes two claims: First, children’s ability to recognize rimes
in words is causally linked to the development of reading skills;
second, that rime-based analogy represents the mechanism by
which the rime-reading association is operationalized. Both questions have been addressed by exploring individual differences in
reading and phonological skills. Regarding the first assertion,
several studies demonstrated that rime-detection ability is indeed a
unique predictor of reading (e.g., Bryant, Maclean, Bradley, &
Crossland, 1990). More recently, there has been lively debate
about the appropriate methodology for establishing specific rimereading associations (e.g., Bryant, 1998, 2002; Castles &
Coltheart, 2004; Hulme et al., 2002; MacMillan, 2002; Savage,
2001). It has also been argued by some that rime and phoneme
skills are part of a single phonological ability, rather than distinct
abilities (Anthony & Lonigan, 2004; Papadopoulos, Spanoudis, &
Kendeou, 2009), although other researchers have found different
patterns (Savage & Rvachew, 2006).
One specific methodological issue that has emerged from this
debate is that, to be comparable, measures of phoneme and rime
awareness must control for extraneous task demands. That is, all
193
aspects of the task apart from the size of the phonological unit
examined should remain equal, (Bryant, 1998). Under such conditions, Hulme et al. (2002) and Savage and Carless (2005) have
shown that measures of phonemic awareness were good longitudinal predictors of reading skills, whereas onset-rime awareness
was a weaker or nonsignificant predictor. This specific methodological issue also applies to Goswami and Bryant’s (1990) second
claim that onset-rime phonological awareness helps children to
make predictions about consistency of letter strings. Although
several studies support Goswami’s assertion that there is a specific
link between individual differences in children’s rime sensitivity
and their early analogy use (e.g., Goswami, 1990b; Goswami &
Mead, 1992), several others studies have found that phoneme-level
skills predict analogy use in the clue word task (e.g., Savage &
Stuart, 2001; Walton, 1995; Wood, 2000). All such studies are,
however, limited by the fact that they used phonological tasks that
were not equivalent in terms of the extraneous task demands. Most
often, rime oddity tasks in which children judge the odd one out in
a spoken set, such as pin, win, sit, fin, and in which they are not
required to articulate or manipulate the shared phonological unit
have been contrasted with tasks requiring articulation or manipulation of a phonemic unit, for example, in blending or segmenting
phonemes (e.g., segmenting cut to make c-u-t) or in deleting a
phoneme from within a word to produce a word or pseudoword
(What is beak without the b?). Task demands and phonological
unit size are thus perfectly confounded, rendering all results hard
to interpret. In one of the few studies to use equivalent onset-rime
and phoneme tasks (Roberts & McDougall, 2003), phoneme but
not rime awareness predicted improvement in target word reading
at posttest in the clue word task. Roberts and McDougall (2003)
used a version of Goswami’s original clue word task and showed
that the effects there reflected phonological priming. Thus, despite
much research, it is still not clear what the relationship is between
well-matched phonological tasks assessing rime- and phonemegrains of representation and the use of genuinely orthographic
analogy by young children. The present study therefore seeks to
explore this issue.
The Present Research
This article thus addressed three questions:
1.
Is there evidence for spontaneous orthographic rime
analogy use, controlling for phonological priming effects?
2.
Is there an advantage at posttest for target words sharing
orthographic rimes, controlling for effects of priming
and compared to target words sharing vowels?
3.
What is the association between phoneme awareness,
onset-rime awareness and vowel and improvements in
target word reading at posttest?
Experiment 1 addresses Questions 1 and 3 by exploring the
effects of teaching families of words sharing rimes on spontaneous
reading of words sharing orthographic or phonological rimes or
vowel digraphs. The study also explores the specificity of the link
between individual differences in improvement in reading of target
words sharing rimes and rime phonological awareness. Experiment
SAVAGE, DEAULT, DAKI, AND AOUAD
194
2 explores Question 2 by investigating the effects of teaching
families of words sharing either (a) rimes or (b) vowel digraphs on
reading of words sharing orthographic or phonological rimes or
vowel digraphs.
Experiment 1
Method
Participants. Approximately 80 consent request letters were
sent out initially to request participation in the study. Forty-two
first grade students (23 girls, 19 boys) from local suburban elementary schools took part in the study following parental informed
consent. Their mean chronological age was 6 years, 8 months
(range ⫽ 6 years, 1 month to 7 years, 4 months). All students
whose parents granted consent were included in the study; thus,
students were not preselected or screened prior to commencing the
study. Students were assessed on their reading ability, vocabulary,
and letter-sound knowledge to establish the typicality of the sample.
Reading ability was assessed using the Wide Range Achievement Test (WRAT-3; Wilkinson, 1993), which is a test of single
word reading. The mean WRAT-3 standard score was 97.21
(SD ⫽ 11.04). The Letter Sound Knowledge task is an experimental task that requires students to identify the sound that is associated with each letter of the alphabet. Scoring is based on the total
number of letter-sounds identified, for a maximum of 26 points. In
the present sample, students identified a mean of 20.45 lettersounds (SD ⫽ 5.05). In addition, students were assessed on their
receptive vocabulary ability using the Peabody Picture Vocabulary
Test (PPVT-III; Dunn, & Dunn, 1996). The mean PPVT standard
score was 98.19 (SD ⫽ 10.63). These scores all confirm that a
group of children had been sampled that was quite typical on these
measures.
Design and procedure.
The study used a fully withinsubjects design, with all students participating in each condition of
the experiment. Students participated in a total of five 20 –30-min
sessions as part of the experimental paradigm. In Session 1,
students completed the three background ability measures described above. In Sessions 2– 4, students engaged in the analogy
tasks, with two different word sets presented per session. The three
analogy task sessions were counterbalanced by order across students to vary the presentation of word sets. In Session 5, students
completed two phonological awareness tests involving rime production and final consonant (coda) production. The rime task and
the coda task were also counterbalanced with respect to order of
presentation within sessions.
Administration of the analogy tasks followed a carefully standardized procedure, originally developed in a modified version of
the clue word analogy task (Goswami, 1986), which was based on
the presentation of a single and often concurrently available clue
word. The modified task was based on one described by Savage
and Stuart, (1998, 2001). This approach allowed children to learn
multiple clue word exemplars that were not made available when
target words were re-presented. In the pretest phase, the first word
set was administered to students in random order. Students were
asked to read the 16 target and filler words out loud one at a time.
Students’ responses were recorded verbatim on the scoring sheet to
analyze reading errors. In the teaching phase, students were asked
to read the three clue words sharing rimes (e.g., fail, mail, jail) one
at a time. If students were unable to read the word, then they were
told the pronunciation of the word by the examiner, with no
additional explanation or guidance, apart from general encouragement. Students read the clue words in random order until reaching
a threshold of three correct pronunciations of each word. Examiners recorded the number of trials needed to learn the clue words
to this criterion. After accomplishing three correct pronunciations
of each clue word, the examiner then placed all three words
together and read them out loud before putting them away. After
the teaching phase, students took a quick break by engaging in an
unrelated visual search task for 2–3 min to separate the two phases
of the experiment. Following this, students were posttested on the
16 target and filler words one at a time, and their responses were
recorded. Finally, students were asked to read the three clue words
one more time to check whether they remembered them and could
have used them to make inferences at posttest. This entire procedure was then repeated again with the second word set to complete
the session.
Two phonological awareness tasks were also administered to
students in Session 5 using the materials and following the procedure described in Savage et al. (2006). In both common unit tasks,
students were given two words and asked to identify the sound that
was the same in the two words. In the rime-production task,
students were given words such as heart and cart and had to
produce the rime unit, art, as the shared sound. In the codaproduction task, students were given words such as dog and bag
and had to produce the coda unit, g, as the shared sound. Both tasks
were administered using pictures of each word to help students
represent and remember the words. For each task, students were
given three practice trials before administering the 10 test items.
Feedback was given during the practice trials to ensure that students knew the type of response for which the examiner was
looking; however, no feedback was given during the test phase.
For each item, students were given multiple prompts (pointing to
the picture, pausing, repeating the words) until they provided a
response.
Materials and stimuli. As shown in Table 1, a set of six clue
words and corresponding target words was developed. These
words were carefully selected with the intention of choosing words
that were comparable in terms of reading difficulty, word length
and frequency in the English language. For each word set, there
were 16 corresponding target words: two orthographic and phonological rime-primed words, two phonological rime-primed
words, two orthographic and phonological rime-primed words,
two phonological vowel-primed words, two control words, and six
filler words. The two orthographic and phonological rime-primed
target words shared the same rime sound and spelling patterns with
the clue words (e.g., seek-reek), whereas the phonological rimeprimed words shared only the rime sound but not the spelling (e.g.,
seek-beak). The orthographic and phonological vowel-primed
words shared the same vowel sound and spelling with the
clue words (e.g., seek-deem), whereas the phonological vowelprimed words shared only the same vowel sound with the clue
words (e.g., seek-heat). Control words shared neither rime nor
vowel sound patterns, nor grapheme-phoneme correspondence
with the target words (e.g., seek-tart). They provided a control
against general improvements in word reading ability following
the teaching phase. The six filler words were also unrelated and
ANALOGIES FROM MULTIPLE CLUE WORDS
195
Table 1
Word Set Used in Experiment 1 With Word Frequencies From Stuart et al. (2010)
Target words
Word set
Clue words
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
1
seek
peek
geek
cool
pool
spool
skirt
shirt
squirt
scoop
whoop
snoop
fail
mail
jail
float
bloat
goat
deem 0
heel 0
heat 51
lean 5
reek 0
meek 5
beak 14
teak 0
tart 11
fowl 3
root 11
boot 38
rude 16
nude 0
tool 5
fool 19
rule 24
yule 0
yarn 0
germ 0
firm 0
whirl 16
curd 3
churn 0
dirt 76
flirt 0
curt 0
blurt 3
torn 27
fleet 0
brood 0
proof 14
crude 0
prude 0
troop 19
droop 5
group 59
croup 0
sport 30
clout 0
waif 0
bait 5
vein 3
cane 5
hail 0
pail 5
veil 0
bale 0
dork 0
faun 0
soak 14
moan 3
tone 0
cone 14
moat 8
coat 176
dote 0
rote 0
worn 30
void 0
2
3
4
5
6
Controls
Note. O ⫹ P ⫽ orthographic and phonological; P ⫽ phonological.
served only to make sure that there were an equal number of words
sharing letters and sounds with clue words, as there were words not
sharing these patterns.
The target word sets were initially matched across word type for
overall orthographic structure. That is, wherever possible, words
with consonant-vowel, digraph-consonant (CVC) structures (e.g.,
beak) were used in all conditions of the experiment. Sometimes it
was not possible to find CVCs to fit all experimental categories of
stimuli. Where the construction of sets required more complex
structures be admitted (e.g., whirl or brood), we sought as far as
possible to use equal numbers of items with such structures across
word types. The target word types were initially matched using the
Children’s Printed Word Database (Masterson, Stuart, Dixon,
Lovejoy, & Lovejoy, 2003; Stuart, Dixon, Masterson, & Gray,
2003). This measure was based on all words in 685 books found in
the reading records of a sample of children aged 5–7 years in
schools in the United Kingdom. The written word measure provides a sensitive measure of children’s orthographic knowledge
that is closely associated with performance on word reading tasks
in a range of experimental and intervention studies (see e.g.,
Savage, Carless, & Stuart, 2003; Stuart et al., 1999). The measure
was preferred for this aspect of its validity, alongside its relative
recency compared with other more dated but still frequently
used measures (e.g., Carroll, Davies, & Richman, 1971). A
recent enlarged and updated database based on N ⫽ 1,011
books has been created (Masterson, Stuart, Dixon, & Lovejoy,
2010). Stimulus items were also matched using this most recent
database. The mean frequency of target words per million in the
2010 word set was 12.17 (SD ⫽ 26.40). Preliminary analyses of
variance (ANOVAs) also confirmed that there were no significant differences between the five experimental word types on
this measure of frequency, F(4, 44) ⫽ 1.09, MSE ⫽ 711.64, p ⫽
.38, ns, ␩2 ⫽ .09.
Results
Following standard practices, by-item and by-participant databases were first constructed for pre- and posttest target word
reading scores,2 clue word reading scores, and phonological
awareness measures. As in previous studies (e.g., Goswami, 1988;
Farrington-Flint & Wood, 2007; Savage & Stuart, 1998, 2001), the
measure of target word reading at pretest and posttest was
the correct articulation of the appropriate unit in target words (e.g.,
the number of rimes correctly articulated in rime analogous words
and the number of vowels correctly articulated in vowel analogous
words). This provides a more sensitive measure of improvement
that allows units of different sizes, such as rimes and vowels, to be
more fairly compared. Preliminary analyses using standard procedures (Tabachnick & Fidell, 2007), confirmed that neither data set
showed any marked deviations from normality or evidence of
outliers, and they were suitable for general linear modeling of
repeated measures with ANOVA. The means and standard deviations for the target word reading for the by-participants data set are
presented in Table 2. Inspection of these means suggested an
improvement in reading of all four analogous target words at
posttest, with little evidence of improvement for control words.
There was some evidence of differential improvement within this
set of words, with the largest improvement evident for orthographic and phonological rime-primed words.
Preliminary analyses indicated that there was no interaction
between target word reading and the fully counterbalanced order
of target word teaching sessions, confirming that there were no
2
By-item and by-participant analyses have complementary strengths:
By-item analyses add confidence that overall results do not reflect the
disproportionate impacts of a subset of participants, and by-participant
analyses add confidence that overall results do not reflect the disproportionate impacts of a subset of items.
order effects. Data were therefore analyzed using a 2 (Test: pretest,
posttest) ⫻ 5 (Word Type: orthographic and phonological rime
primed, phonological rime primed, orthographic and phonological
vowel primed, phonological vowel primed, control) ANOVA with
repeated measures on test and word type. There was evidence of a
significant chi-squared for the Mauchly Sphericity Test in the
word type and Test ⫻ Wordtype interaction in the by-participants
analysis only, so the corresponding Greenhouse-Geisser statistics
are reported for these analyses.3 In all other cases, the sphericityassumed statistics of analyses are reported. A preliminary analysis
confirmed that there was a significant Test ⫻ Wordtype interaction
by-participants, F(3, 138) ⫽ 12.74, MSE ⫽ 1.47, p ⫽ .001, ␩2 ⫽
24; by items, F(4, 44) ⫽ 12.94, MSE ⫽ 4.26, p ⫽ .001, ␩2 ⫽ .54.
Post hoc analyses using Tukey’s honestly significant difference
(HSD) test revealed that there was a significant difference between
means at pretest as well as at posttest. In both the by-participant
and by-items analyses, the pretest differences were due to the
words in the orthographic and phonological rime-primed condition
being read significantly more often than all or most other word
types at pretest.
To better explore the improvement in reading by word type at
posttest, analyses of covariance (ANCOVAs) were run on the
posttest scores with repeated measures on word type and with the
five corresponding pretest word type measures as covariates for
both by-participant and by-item analyses. There was again evidence of a significant chi-squared for the Mauchly Sphericity Test
for word type in the by-participants analysis only, so the corresponding Greenhouse-Geisser statistics are reported for these particular results. The main analysis revealed a main effect of Word
Type ⫻ Participants, F(3, 115) ⫽ 9.90, MSE ⫽ 2.71, p ⫽ .001,
␩2 ⫽ .22; by items, F(4, 24) ⫽ 3.30, MSE ⫽ 9.84, p ⫽ .03, ␩2 ⫽
.36. There were no other significant effects. The adjusted means
are shown in Figure 1. Post hoc contrasts of the adjusted posttest
means by word type using Tukey’s HSD test revealed that there
were significant advantages for all target words over the control
Table 2
Means, Standard Deviations, and Cohen’s Effect Sizes for
Words in Experiment 1
Pretest
Target words
M
SD
Posttest
M
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
Controls
By-participants
1.98 2.36
4.02
1.60 2.01
3.50
2.50 3.09
4.81
1.69 2.09
3.64
1.55 1.90
1.69
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
Controls
By-items
4.17 14.00
4.42 12.25
3.91 16.75
2.73 12.83
4.03
5.92
SD
Cohen’s effect
size
2.94
2.45
3.20
2.90
2.04
0.87
0.81
0.99
0.84
0.06
Mean number of analogies made
SAVAGE, DEAULT, DAKI, AND AOUAD
196
6.08
4.54
4.37
3.86
5.12
1.79
1.68
2.02
1.77
0.13
Note. O ⫹ P ⫽ orthographic and phonological; P ⫽ phonological.
Maximum n ⫽ 12 (by-participants analyses), maximum n ⫽ 26 (by-items
analyses).
vowel-primed
rime-clued
rime-primed
controls
Wordtype
Figure 1. Adjusted posttest means in Experiment 1 (by participants).
words. There was also a significant advantage for the orthographic
and phonological rime-primed words over the phonological rimeprimed and both types of vowel-primed words in the byparticipants analysis. A very similar pattern was evident to that
reported above in the by-items analysis with one important exception: The mean difference between the phonological rime-primed
words and orthographic and phonological rime-primed words escaped conventional significance in this analysis. Together, these
results show that there are significant effects of purely phonological rime-priming, even in the multiple clue word task used here.
It is important, however, that there was a significant difference
between the orthographic and phonological rime-primed words
and both the phonological rime-primed words and both of the
vowel-primed word types at posttest even after controls for pretests differences. This finding is consistent with the idea that
children can make genuine orthographic rime analogies.
Effect sizes for the means in these analyses were also calculated
using Cohen’s standard equation for mean differences (Cohen,
1988), based on the formula (posttest ⫺ pretest) / (pooled pretest
standard deviation). These effect sizes are presented below the
word type means in Table 2. Cohen (1988) has described effect
sizes as “small” when d ⫽ 0.2, “medium” when d ⫽ 0.5, and
“large,” when d ⫽ 0.8. Cohen (1988) and others have also cautioned against an overly rigid interpretation of these scores across
different behavioral contexts; nevertheless, such analyses are
widespread in the literacy-acquisition literature in establishing the
practical importance of significant findings. Interpretation of the
means presented in Table 2 using Cohen’s broad criteria show that
the improvement for all target words at posttest, except the controls, was large in all cases. The value added difference in effect
sizes between orthographic and phonological rime-primed and
3
6.92
5.58
8.75
5.83
5.42
vowel-clued
6
5
4
3
2
1
0
Sphericity tests assess the equality or symmetry of the variances
between levels of factors in repeated measures designs in ANOVA. Results
from ANOVAs violating this assumption of equality cannot be trusted, as
they undermine the validity of the analysis. Violations of sphericity can be
detected using Mauchly’s W test. Where the W test is significant, it is
standard practice to make corrections for violations of sphericity using the
Greenhouse-Geisser procedure. The Greenhouse-Geisser procedure alters
the degrees of freedom, thereby altering the significance value of the F
ratio, and allows legitimate interpretation of main effects and interactions
(e.g., Howell, 1997).
ANALOGIES FROM MULTIPLE CLUE WORDS
phonological rime-primed words and orthographic and phonological vowel-primed and phonological vowel-primed target words,
respectively, at posttest is of particular theoretical interest, as this
directly indicates the size of the orthographic rime analogy effect,
above and beyond the effects of phonological rime priming. These
effects calculated from the means in Table 2 are .15 and .06,
respectively, (by participants) and .25 and .11 (by items). These
effects might thus be considered small across all analyses.
The final set of analyses sought to explore the relationship
between individual differences in transfer in the current multiple
clue word task and different levels of phonological awareness
ability and clue word knowledge. Stepwise linear regression models were used for each of the clue words types except the control
words that had shown no improvement in the analysis above. The
number of target words articulated at posttest was the dependent
variable in each case. Following the procedure in previous studies
(Goswami & Mead, 1992; Roberts & McDougall, 2003; Savage &
Stuart, 2001), the respective number of words read at the pretest
for each target word dependent variable was entered at Step 1,
followed by either the common unit rime or common unit coda
tests at Step 2. Preliminary inspection of analyses showed uniformly high statistical tolerance scores (all tolerance scores in
excess of .95), indicating that there was no evidence of colinearity
in these analyses. The results of these analyses are depicted in
Table 3. The results are very clear in showing that although pretest
target word reading score was, as expected, a strong predictor of
posttest target word reading, coda common unit test performance
was never a significant predictor of improvement in target word
reading between pre- and posttest when entered at Step 2. In
contrast, as shown in Table 3C, rime common unit test performance was a significant and unique predictor of improvement in
target word reading between pre- and posttest, specifically for the
orthographic and phonological rime-primed target words. This
pattern of significant association was not evident for the phonological rime-primed target words in Table 3D.
As a further check on the specificity of these associations, all
regression analyses in Table 3 were rerun with WRAT wordreading entered at Step 2, prior to entering phonological awareness
measures at Step 3. Although WRAT word reading was a significant predictor of growth in target word reading at posttest, the
results for phonological awareness measures were essentially un-
197
changed by this additional analysis: The rime common unit test
still explained a significant unique 4% in improvement in reading
of orthographic and phonological rime-primed words. This pattern
of significant and highly specific associations in Table 3 is thus
consistent with the idea of a specific link between individual
differences in rime phonological awareness and use of orthographic rime analogy task (Goswami & Mead, 1992). Rime common unit test performance was also uncorrelated with coda common unit task performance (r ⫽ .19, p ⫽ .23, ns), consistent with
the idea that the two tasks tap into different forms of variance in
phonological ability (Savage & Rvachew, 2006).
Last, inspection of the number of clue words read after the
posttest was complete showed that children read a mean of 14.93
(SD ⫽ 3.17) clue words from a maximum of 18, showing that, for
the most part, children had learned and had stored in their mental
lexicons up to three relevant clue words for each analogous target
word during the posttest. Hierarchical regression analyses were run
to explore whether there was a significant relationship between
pre- to posttest improvement in reading for each word type in the
clue word task and the effectiveness of clue word learning. As with
the analyses above, for the common unit tasks, posttest target word
reading was the dependent variable, with respective pretest target
word reading and then clue words known at posttest as the dependent variables at Steps 1 and 2, respectively. These analyses
showed that clue word learning was associated with posttests
improvements in target word reading in the rime-clued condition
only, explaining a unique 4.5% of the variance in target word
reading at Step 2.
Discussion
Experiment 1 sought to explore whether children can make
genuine orthographic analogies following exposure to families of
rime-analogous clue words, above and beyond the known effects
of purely phonological priming. Results of both by-participant and
by-item analyses provided clear evidence that children were able
to make analogies to target words between pre- and posttest when
using a multiple clue word paradigm in which the clue words were
not present at posttest, replicating Savage and Stuart (1998, 2001).
Beyond this, the results showed that children made inferences
about pronunciations, even when the target words shared no or-
Table 3
Regression Analyses Exploring Phonological Awareness as a Predictor of Clue Word Transfer
Independent variable
Step
Dependent variable
␤
⌬r2
A. Orthographic and phonological vowel-primed target words (posttest)
1
2
2
1
2
2
1
2
2
1
2
2
Vowel-primed target words (pretest)
Rime common unit test
Coda common unit test
Vowel-primed target words (pretest)
Rime common unit test
Coda common unit test
Rime-primed target words (pretest)
Rime common unit test
Coda common unit test
Rime-primed target words (pretest)
Rime common unit test
Coda common unit test
.67
.12
.03
.65
.12
.12
.67
.26
.00
.62
.18
.00
.44ⴱⴱ
.00
.00
.43ⴱⴱ
.00
.00
.53ⴱⴱ
.06ⴱ
.00
.43ⴱⴱ
.02
.00
B. Phonological vowel-primed target words (posttest)
C. Orthographic and phonological rime-primed target words (posttest)
D. Phonological rime-primed target words (posttest)
ⴱ
p ⬍ .01.
ⴱⴱ
p ⬍ .001.
SAVAGE, DEAULT, DAKI, AND AOUAD
198
thographic relationships. This effect is shown in the significant
improvement in both the phonological vowel-primed and phonological rime-primed conditions in ANOVA, as well as in the robust
effect sizes for improvement between pre- and posttest for both
phonological vowel- and phonological rime-primed words. However, the finding that there was a modest but significant additional
effect for orthographic and phonological rime-primed over phonological rime-primed words in the by-participants analyses is consistent with the idea that children also showed some significant
capacity to make genuine orthographic inferences. In contextualizing this last result, analysis relevant to the real world importance
of these findings showed that the Cohen’s effect sizes for orthographic and phonological rime-primed words were only marginally greater than that for phonological rime-primed target words,
an additional value-added effect size that was small in size using
Cohen’s (1988) standard criteria for measuring effect sizes.
Experiment 1 also sought to explore the relationship between
individual differences in analogy use in the current multiple clue
word task and different levels of phonological awareness ability,
following the regression-based approaches reported in several existing studies (Goswami, 1990a; Goswami & Mead, 1992; Roberts
& McDougall, 2003; Savage & Stuart, 2001). Results showed that
the only significant association was between rime common unit
task performance and improvement in reading of orthographic and
phonological rime-primed words between pre- and posttest. This
association is exactly the one first hypothesized by Goswami and
colleagues (Goswami & Bryant, 1990; Goswami & Mead, 1992).
This is, however, the first time that such a link has been demonstrated empirically using comparable rime- and phonemeawareness tasks and in a clue word task using multiple clue words,
controlling for the effects of phonological priming.
Experiment 1 also showed that children learned and had available a substantial amount of recently taught and relevant clue word
knowledge. Results also showed that there was a significant association between individual differences in the extent of retention of
clue word learning, as measured by the number of clue words read
correctly after the posttest, and individual differences in improvements in reading in the orthographic and phonological rimeprimed condition at posttest. This arguably provides further evidence that children used orthographic knowledge from taught clue
words to read target words in this condition.
Overall, Experiment 1 has found evidence of rime priming and
some limited but significant orthographic rime inference. One
possible limitation of the experiment is that, throughout the study,
children were only taught families of rimes, so this may have
directed them toward rimes, possibly at the expense of other units
such as shared vowel digraphs. Indeed, such a strategic approach
is predicted from the flexible-unit-size hypothesis (G. D. A. Brown
& Deavers, 1999; Farrington-Flint & Wood, 2007). Experiment 2
therefore sought to explore the patterns when children were exposed to families of clue words sharing vowel digraphs in comparison with exposure to families of clue words sharing rimes.
students participated in the second study (M age ⫽ 7 years, 0
months; range ⫽ 6 years, 5 months to 7 years, 7 months). Students
completed the same background measures as in Experiment 1 to
assess their reading, vocabulary, and letter-sound knowledge. Two
children completed the pretesting but were unavailable for portions
of the clue word testing sessions, with less than 30% of their data
missing. Both of these children had typical reading and vocabulary
scores. The missing clue word analogy data for these children was
imputed using the SPSS missing value analysis. A preliminary
analysis confirmed that data was missing completely at random
(MCAR, Little’s MCAR), ␹2(2) ⫽ 0.29, p ⫽ .86, ns, confirming
the suitability of data for imputation. Data were then imputed
using the expectancy maximization (EM) procedure with five
iterations. The mean reading level of the 30 participants who made
up the final sample was 96.50 (SD ⫽ 14.58), as measured with the
single word reading test from the WRAT-3 (Wilkinson, 1993).
Students were also given the PPVT (Dunn & Dunn, 1996) and
obtained a mean vocabulary standard score of 98.03 (SD ⫽ 12.56).
In a test of letter-sound knowledge, students identified a mean of
21.40 letter sounds (SD ⫽ 3.24).
Independent of these preliminary assessments, students were
randomly allocated within each classroom into one of two groups:
a rime-taught group and a vowel-taught group. ANOVAs were
then run to confirm that students’ reading, vocabulary, and letterknowledge levels were the same across the two groups. A chisquared analysis was run to confirm that there were equal proportions of boys and girls in both groups. The means for these groups
and gender frequencies are depicted in Table 4. Independent t test
analyses confirmed that there were no significant differences between rime-taught and vowel-taught groups on WRAT word reading, t(28) ⫽ 1.39, p ⫽ .18, ns; PPVT vocabulary, t(28) ⫽ 0.47, p ⫽
.164, ns; chronological age, t(28) ⫽ 1.16, p ⫽ .26, ns; or lettersound knowledge, t(28) ⫽ 1.13, p ⫽ .27, ns. A chi-squared
analysis of gender frequency confirmed that there were no differences in the gender balance of these groups, ␹2(1) ⫽ .14, p ⫽ .71,
ns.
Design and procedure. Unless otherwise stated, the method
followed the same procedure as in Experiment 1. However, Experiment 2 had an important difference in that two randomized
groups were created to teach students two different sets of clue
words. In the rime-taught group, students were exposed to the
Table 4
Means and Standard Deviations of Vowel-Taught and
Rime-Taught Groups on Reading, Vocabulary, Letter-Sound
Knowledge, and Gender
Group
Rime-taught
Measure
M
SD
Vowel-taught
n
SD
n
Method
WRAT-3 word-reading task
PPVT receptive vocabulary task
Letter sound knowledge task
Gender (boys)
Participants. Around 90 consent request letters were initially
sent out to request participation in the study. Thirty Grade 1
Note. WRAT-3 ⫽ Wide Range Achievement Test (Wilkinson, 1993);
PPVT ⫽ Peabody Picture Vocabulary Test (Dunn, & Dunn, 1996).
Experiment 2
92.87 12.58
99.13 12.38
20.73 2.84
M
100.13 15.94
96.93 13.07
22.07 3.58
9
10
ANALOGIES FROM MULTIPLE CLUE WORDS
same material as in Experiment 1; that is, they learned clue words
that shared the same rime unit. However, in the vowel-taught
group, students were taught a new set of clue words that shared
only the same vowel unit. These new words are shown in Table 5.
Otherwise, for both groups, testing followed the same format as in
Experiment 1. The presence of two groups each of 15 children that
were taught different clue words precluded attempts to run regression analyses of individual differences in analogy use.
Materials and stimuli. The clue word sets used in Experiment 2 are depicted in Table 5. All stimuli for the rime-taught
group remained the same; however, a new set of clue words was
constructed for the vowel-taught group. These clue words had
common vowel digraphs (e.g., beef, sheep, weed) but were not
themselves rhyming words. Clue words for the vowel-taught group
were carefully selected to be comparable to the words in the
rime-taught condition with respect to word frequency and reading
difficulty. As in Experiment 1, we again sought as far as possible
to use equal numbers of items with CVVC structures across the
both clue word sets. The mean frequencies of rime-taught and
vowel-taught clue words per million in the present word set on the
most recent version of the Children’s Printed Word Database
(Masterson et al., 2010) were 31.06 (SD ⫽ 53.48) and 20.33 (SD ⫽
38.88), respectively. There were no significant differences between these clue word groups on this measure of frequency,
t(17) ⫽ .645, p ⫽ .53, ns.
Results
The preliminary data analyses were the same as in Experiment
1 and produced similar results, confirming the suitability of the
data for mixed design ANOVA analyses. The means and standard
deviations for the target word reading for the by-participants data
set are presented in Table 6. Inspection of these means suggests an
improvement in reading of all target word types at posttest, except
the control words. The largest improvement evident was for the
orthographic and phonological rime primed words.
Table 5
Vowel-Based Clue Word Sets Used for the Vowel-Taught Group
in Experiment 2
Word set
Clue words
1
beef
sheep
weed
zoom
spoon
hoop
chirp
mirth
smirk
gloom
spook
shoot
gait
maim
maid
roam
foal
hoax
2
3
4
5
6
199
Table 6
Means, Standard Deviations, and Cohen’s Effect Sizes for
Words in Experiment 2
Pretest
Target words
M
SD
Posttest
M
SD
Cohen’s effect
size
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
Controls
By-participants, rime-taught
3.92
3.64
5.63
3.16
2.43
2.72
4.64
2.98
3.47
3.14
6.19
3.12
2.59
2.45
5.03
2.20
3.34
3.22
3.26
2.99
0.51
0.64
0.80
0.72
⫺0.02
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
Controls
By-participants, vowel-taught
4.20
3.61
6.33
3.35
3.47
3.66
5.07
3.49
4.87
3.98
6.60
4.00
4.13
3.58
5.73
3.52
3.73
3.62
3.60
3.40
0.63
0.47
0.51
0.47
⫺0.04
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
Controls
By-items, rime-taught
4.00
2.34
5.67
2.92
2.61
4.92
4.00
2.09
6.92
3.17
1.75
5.50
3.17
2.25
3.58
2.10
3.06
1.83
1.83
2.43
0.70
0.84
1.22
0.98
0.17
O ⫹ P Vowel-primed
P Vowel-primed
O ⫹ P Rime-primed
P Rime-primed
Controls
By-items, vowel-taught
5.17
2.76
8.00
4.17
2.33
6.42
6.08
2.43
8.00
5.17
1.53
6.67
4.67
2.27
4.67
2.98
2.31
2.04
2.23
2.74
1.19
0.94
0.81
0.63
0
Note. O ⫹ P ⫽ orthographic and phonological; P ⫽ phonological.
Maximum n ⫽ 12 (by-participants analyses), maximum n ⫽ 26 (by-items
analyses).
Preliminary analyses confirmed that there was no interaction
between target word reading and the order of target word teaching
sessions, confirming that there were no order effects. Data were
therefore initially analyzed using a 2 (Test: pretest, posttest) ⫻
Teaching Group (rime-taught, vowel-taught) ⫻ 5 (Word Type:
orthographic and phonological rime primed, phonological rime
primed, orthographic and phonological vowel primed, phonological vowel primed, control) ⫻ 2 (Teaching Group: rime-taught,
vowel-taught group) mixed design ANOVA with repeated measures on test and word type. In this study, there was evidence of a
significant chi-squred for the Mauchly Sphericity Test in the word
type and Test ⫻ Word Type interaction in the by-items analysis
only, so the corresponding Greenhouse-Geisser statistics are reported for these analyses. In all other cases, the sphericity-assumed
statistics of analyses are reported.
In the by-participants analysis only, this ANOVA analysis confirmed that although there was a significant Test ⫻ Word Type
interaction, F(4, 112) ⫽ 9.40, MSE ⫽ 1.47, p ⫽ .001, ␩2 ⫽ 25,
post hoc analyses using Tukey’s honestly significant difference
(HSD) test revealed there were also significant difference between
means at pretest as well as at posttest. This time, the pretest
differences were due to the words in the phonological vowelprimed and phonological rime-primed conditions being read significantly less often than the orthographic and phonological vowel-
SAVAGE, DEAULT, DAKI, AND AOUAD
200
Mean number of analogies made
primed word types at pretest. To better explore the improvement in
reading by word type at posttest, ANCOVAs were run on the
by-participants posttest scores with repeated measures on word
type and with the five corresponding pretest word type measures as
covariates. There was again evidence of a significant chi-squared
for the Mauchly Sphericity Test in the word type and Test ⫻
Wordtype interaction in the by-participants analysis only, so the
corresponding Greenhouse-Geisser statistics are reported for these
analyses. The main analysis revealed a main effect of word type,
F(3, 58) ⫽ 5.53, MSE ⫽ 4.14, p ⫽ .001, ␩2 ⫽ .19. There were no
other significant main effects. The adjusted means are depicted in
Figure 2. Post hoc tests (Tukey’s HSD) showed that the main
effect of word type was due to a significant difference between all
target words and the control words and an advantage for the
orthographic and phonological vowel-primed words over the phonological vowel-primed words.
Analyses of the by-items data using ANOVA revealed a main
effect of test, F(1, 22) ⫽ 75.82, MSE ⫽ 2.52, p ⫽ .001, ␩2 ⫽ .78,
and a main effect of word type, F(4, 54) ⫽ 4.83, MSE ⫽ 12.58,
p ⫽ .001, ␩2 ⫽ .18. There was also a main effect of teaching
group, F(1, 22) ⫽ 8.63, p ⫽ .008, ␩2 ⫽ .28. In addition to these
main effects, there was a significant Test ⫻ Word Type interaction, F(4, 51) ⫽ 7.75, MSE ⫽ 2.17, p ⫽ .001, ␩2 ⫽ .26. There
were no other significant interactions. Post hoc analyses using
Tukey’s HSD test revealed that the control words again differed
significantly from all other word types at posttest.
Effect sizes for the means in these analyses were again calculated. Interpreting results using Cohen’s (1988) criteria, the improvement for all target words at posttest except the controls was
medium to large in all cases for the vowel- and rime-primed and
clued words but were small for the control words. These are
depicted in Table 6. The value added difference in effect sizes
between orthographic and phonological rime-primed and phonological rime-primed words and orthographic and phonological
vowel-primed and phonological vowel-primed target words derived from these mean scores was again small, d ⫽ –.13 and .08
versus .16 and .04 for orthographic and phonological vowelprimed and rime-primed words in the rime-taught versus the
vowel-taught groups, respectively, in the by-participants analysis.
For the same contrasts in the by-items analysis, the corresponding
figures were –.14, and .24, versus .25 and .18.
Finally, to confirm that no confounds were introduced in the
teaching of the vowel-taught and rime-taught words, independent
vowel-clued
7
6
5
4
3
2
1
0
vowel-primed
rime-clued
rime-primed
controls
Wordtype
Figure 2.
Adjusted posttest means in Experiment 2 (by participants).
sample t tests were run to confirm that the number of trials to learn
words was the same. The means were 15.29 (SD ⫽ 11.28) and
15.00 (SD ⫽ 9.78), respectively, in the rime-taught and the voweltaught conditions, with no differences evident between the number
of trials to learn the word sets t(28) ⫽ .075, p ⫽ .94, ns. In
addition, the number of clue words recalled after the posttest
across conditions was also analyzed. The means were 16.23 (SD ⫽
1.09) and 15.13 (SD ⫽ 2.48), respectively, in the rime-taught and
the vowel-taught conditions. Results again showed that there were
no significant differences between the number of trials to learn the
word sets, t(28) ⫽ 1.48, p ⫽ .15, ns
Discussion
Experiment 1 found evidence of phonological rime priming and
some limited but significant evidence of orthographic rime inference. One possible limitation of the experiment is that throughout
the study, children were only taught families of rimes, so this may
have directed them toward rimes, possibly at the expense of other
units, such as shared vowel digraphs. Experiment 2 therefore
sought to explore the pattern of improvement in target word
reading when children were exposed to families of clue words
sharing vowel digraphs compared with exposure to families of clue
words sharing rimes. Results of this study showed that there was
again evidence of significant transfer following clue word exposure between pre- and posttest. Inspection of means showed that all
words sharing spellings or pronunciations were advantaged at
posttest over control words. There was no statistically significant
advantage for the orthographic and phonological rime-primed
words over phonological rime-primed words or for orthographic
and phonological vowel-primed over phonological vowel-primed
words in the by-items analysis. There was, however, some evidence of significant advantage for the orthographic and phonological vowel-clued words over the phonological vowel-primed words
in the by-participants analysis only. There was no significant
interaction whatsoever with teaching group (vowel-taught vs.
rime-taught words), suggesting that although the shared units in
multiple taught clue words do have a strong effect on improvements in target word reading at posttest, the presence of shared
rimes versus shared vowels in taught words does not have a large
influence on the relative levels of improvement witnessed at posttest to target words sharing vowels and/or rimes. Together, this
suggests that the largest part of the pattern of improvement in
target word reading might be best understood as reflecting phonological priming effects, rather than genuine orthographic inferences from taught words sharing vowels or rimes. Inspection of
effect sizes confirmed this impression. Here, there were only very
modest and occasional advantages for orthographic and phonological vowel-primed over phonological vowel-primed and orthographic and phonological rime-primed over phonological rimeprimed words, and value added effect sizes were uniformly small.
Beyond this, the modest advantage for orthographic and phonological vowel-primed over phonological vowel-primed words in
some analyses shows that even young children can make orthographic vowel analogies (Savage & Stuart, 2006), and might
suggest that children can be quite flexible in the orthographic unit
used based on what clue word teaching has emphasized (G. D. A.
Brown & Deavers, 1999).
ANALOGIES FROM MULTIPLE CLUE WORDS
General Discussion
This article addressed a series of research questions concerning
the evidence for spontaneous orthographic rime analogy use. The
first question concerned whether children in Grade 1 can show
spontaneous use of orthographic analogy when taught clue words
controlling for the known effects of phonological priming. This
issue is important because although there is much evidence that
children show improvement in target word reading at posttest in
versions of the original clue word analogy task (Goswami, 1986,
1993), this may reflect phonological priming effects rather than
true learning about orthographic rimes (e.g., Bowey et al., 1998;
Roberts & McDougall, 2003; Savage, 1997; Savage & Stuart,
1998; Wood, 2002). The issue is also important because the
influential psycholinguistic grain size hypothesis (Goswami et al.,
2005; Ziegler & Goswami, 2005) suggests that early use of orthographic rime is one of the distinct characteristics of reading acquisition in English compared with German.
In the debate about priming effects in analogy tasks, however,
there has been much less attention to tasks where children learn
multiple clue words that are not present when target words are later
presented. Among the few studies to explore this approach is the
work of Savage and Stuart (1998, 2001, 2006). They were able to
show that children do show signs of improvement in target word
reading under such conditions but that patterns of improvement
were equally strong for vowel-analogous as for rime-analogous
target words. The present results replicated the pattern reported by
Savage and Stuart (1998, 2001), as children again showed improvement in target word reading from pretaught clues not available at the posttest. The present results add an important methodological element to this line of research by being the first to
include controls for phonological priming (the phonological rimeprimed and phonological vowel-primed conditions in Experiments
1 and 2) in the multiple clue word paradigm. Results showed that
children showed significant improvement in target word reading
under these conditions, which was often as strong as when children
were shown target words that shared orthography (the orthographic and phonological rime-primed and orthographic and phonological vowel-clued conditions in Experiments 1 and 2). There
were few clear advantages for orthographic rime-based or vowelbased primes over and above the effects of phonological primes:
Additional effects of orthographic primes were evident in the
by-participants analysis only for rimes in Experiment 1 and the
by-participants analysis only for vowels in Experiment 2. Analyses
of effect sizes using Cohen’s d (Cohen, 1988) also showed that
there were large effects of improvement in target word reading
from pre- to posttest in both experiments for all target word except
controls (d often ⬎ .8 for both by-item and by-participant analyses) and that there were only very modest additional advantages
for orthographic and phonological rime-primed over phonological
rime-primed and orthographic and phonological vowel-primed
over phonological vowel-primed words, which can be termed
small (d ⱕ .25).
One very clear interpretation of the present results is that the
apparent learning of target word pronunciations in the multiple
clue word paradigm largely reflects the same sorts of confounds of
phonological priming witnessed in versions of the clue word task
that involve a single clue word and concurrent clue word prompts,
and which have been reported by many research teams (Bowey et
201
al., 1998; Roberts & McDougall, 2003; Savage, 1997; Savage &
Stuart, 1998; Wood, 2002). Overall, these results provide little
support for the interactive analogy model proposed by Goswami
(1993), which argues that the early use of orthographic rime
analogy is a reading acquisition strategy available to beginner
readers and, in turn, does not provide support for the psycholinguistic grain size hypothesis, which assumes early use of orthographic rimes in reading acquisition in English (Goswami et al.,
2005; Ziegler & Goswami, 2005). The multiple clue word paradigm used here was devised to attempt to maximize the likelihood
of children showing orthographic rime use, to reduce the influence
of phonological priming and opportunistic strategic flexible-unitsize approaches to tasks (G. D. A. Brown & Deavers, 1999;
Farrington-Flint & Wood, 2007), and to more accurately reflect the
ecological demands of real-world use of analogy as a readingacquisition strategy, albeit in a closely controlled experimental
word-learning task.
The current results instead suggest that phonological priming
effects are pervasive in such experiments and make interpretation
of learning in such tasks problematic. Such results might suggest
to some that all versions of clue word learning tasks are thus
flawed as measures of genuine orthographic inference use and that,
as a consequence, researchers need to find new and more authentic
ways to explore orthographic learning where priming artifacts are
less relevant. If so, perhaps the most obvious way to do this is to
direct more attention to training studies that directly compare rimeand phoneme-based teaching. There is, in fact, surprisingly little
well-designed experimental research on this specific question (see
e.g., Torgerson, Brooks, & Hall, 2006; Wyse & Goswami, 2008,
for reviews; Comaskey et al., 2009; Savage, Abrami, Hipps, &
Deault, 2009, for recent experimental intervention studies), and
certainly more attention to intervention might well be warranted.
Although there is clear evidence of phonological priming in the
present designs, a measured and balanced interpretation of the
present results would also draw attention to the fact that there was
some modest additional orthographic learning. These additional
transfer effects were observed when target words shared both
orthographic rimes and vowels in both Experiments 1 and 2 and
sometimes reached significance in post hoc contrasts. The results
of by-participant analyses suggest that where children are taught
families of rime words, they extract orthographic rimes and use
them to make rime analogies (Experiment 1), but where taught
families of word sharing vowel digraphs, they can sometimes
prefer to make vowel analogies (Experiment 2).
These results suggest that at least some children can make
genuine orthographic analogies. A further case-by-case inspection
of improvements made in Experiment 1 revealed that 16 of the 42
participants showed greater transfer for rime-clued than for rimeprimed words. A similar pattern was evident in Experiment 2,
where 13 of the 30 participants showed advantages for vowelclued versus vowel-primed words. Notwithstanding issues surrounding measurement error, this analysis could be cautiously
interpreted to indicate that if spontaneous orthographic analogy is
available at all to children in Grade 1, it is probably available to
only a subset of children.
In some respects, the finding of modest orthographic analogy
effects in word reading outside of the original clue word task is
consistent with other data. Children in Grade 1 report using an
analogy strategy less than 5% of the time when reading isolated
202
SAVAGE, DEAULT, DAKI, AND AOUAD
words, whereas a phoneme-based decoding approach is four times
more frequent in self-reports (Farrington-Flint et al., 2008). Stuart
et al. (2003) and Masterson et al. (2010) analyzed rime frequencies
in each of their corpora of all words in contemporary books for
Grade 1 children. They both report that for the 300 most frequent
CVC syllables in the database, there were 90 and 77 distinct rimes,
respectively. Two thirds of these rimes occurred in only one word
of the 300, suggesting that rime analogy is not likely to be a
productive strategy for deciphering contemporary children’s texts.
Our final question examined the relationship between phonological awareness and analogy strategy use to explore what skills
underpinned individual differences in analogy use. In addition to
occasional main effects of orthographic analogy, Experiment 1
revealed that there was a significant and highly specific association
between rime awareness, as measured using a common-unit task,
and orthographic rime-based transfer. The existence of a strong
link between these two skills has been long been suggested (e.g.,
Goswami, 1990b; Goswami & Bryant, 1990) and may be consistent with Goswami’s view that the use of rime analogy is the link
between rime phonological skills and reading. The present study
used matched explicit rime and phoneme (coda) awareness tasks
for the first time in the multiple clue words task. It should be noted
that explicit rime awareness shows a quite distinct developmental
pattern compared with the implicit rime oddity tasks first used by
Goswami and colleagues. For example, young children find common unit rime tasks harder than coda tasks (e.g., Duncan, Seymour, & Hill, 2000; Seymour et al., 1999), and prereaders generally cannot reliably complete rime common unit tasks at all,
although they can show some modest facility with other units, such
as codas (Savage et al., 2006). In contrast, there is much research
showing prereaders; facility in rime oddity tasks.
Our interpretation of this specific orthographic rime strategyrime phonological awareness link does reflect a genuine association between phonologically underpinned representations of wellspecified orthographic rimes as Goswami (1993) first proposed,
albeit underpinned by an explicit level of phonological awareness.
However, the extent of orthographic analogy use in the present
studies and the documented difficulty young children have with
common unit rime tasks together suggests that at best, this yoked
use of rime analogy and rime common units is a fledgling skill that
is gradually emerging developmentally, rather than a cause of
reading acquisition present in prereaders as Goswami and Bryant
(1990) originally argued. As such, our results are consistent with
the later use of rime units after significant reading experience
proposed by several current models (e.g., Ehri, 2005; Perfetti,
2007; Savage & Stuart, 2006). Finally, the present study also found
performance on the common unit rime and phoneme tasks was
unrelated. This finding is consistent with the idea that the two tasks
are measuring distinct phonological skills. Evidence of the specificity of the impacts of rime- and phoneme-based instruction on
rime- and phoneme- awareness are also consistent with this general view (Comaskey et al., 2009).
Experiment 1 also revealed the existence of a link between the
extent of clue word learning, as measured by the number of taught
clue words correctly read after the posttest, and individual differences in transfer in the rime-clued condition at posttest. This also
provides evidence that children used orthographic knowledge from
taught clue words to read target words sharing rimes. This particular finding bears closer analysis, as it might suggest that the depth
of learning of clue words is crucial to the use of early analogy.
Ensuring depth of learning has been a problem for a long time in
studies of early analogy use. Bruck and Treiman (1992) reported
faster learning of rime pairs in their analogy study but poorer recall
of these word pairs 1 day later. It may be, however, that with
detailed knowledge of clue word orthography, even very young
children might be able to make orthographic analogies. Bruck and
Treiman highlighted the commonalities between word pairs with
colored shared rimes in word pairs. It may be that even more
detailed knowledge of words and perhaps their spellings, too, is
needed before children can use clue words to map the pronunciations of sublexical units in related words they subsequently encounter (e.g., Ehri, 2005; Perfetti, 2007; Savage & Stuart, 2006).
Nevertheless, in a more general sense, there is other evidence
that even young children start to learn the word-specific details of
English orthography. Wright and Ehri (2007) showed recently that
young children in kindergarten and Grade 1 learn words more
effectively when they contain “legal” orthographic elements, such
as doubled letters in coda, over “illegal” onset positions of words
(e.g., JETT vs. RRUG), showing a nascent orthographic ability in
this age group. Further work using more detailed and analytical
clue word teaching may yet be repaid in evidence of early use of
analogy.
Practically, how might educators identify children who do make
genuine spontaneous analogies from those who do not? The present study suggests that the use of careful controls for phonological
primes in the word sets used might be one way to do this effectively. Educators teaching children to read using rime families
might routinely use foil items with shared rime pronunciation but
distinct spelling (e.g., peek–meek with peek– beak) or other items
with shared spelling and distinct pronunciations (e.g., read– bead
and read– bread). The finding of specific associations between
orthographic analogies and both explicit rime awareness and clue
word knowledge after the posttest suggests other ways of accurately identifying children who make genuine orthographic analogies. A further case-by-case analysis showed that 86% of the
children in Experiment 1 who scored zero on the common unit
rime task showed no advantage in reading rime-clued compared
with rime-primed words. Furthermore, of the subset of children
who did show genuine orthographic analogy use in Experiment 1,
all read at least 83% of the 18 taught clue words accurately after
the posttest, whereas only a third of the children who showed no
advantage for rime-clued over rime-primed words achieved this
level of accuracy. Finally, the link between analogy use and
accurate self-reports of analogy use reported by Farrington-Flint
and Wood (2007) may further aid identification. From this view,
children with good metalinguistic rime skills, who also retain
accurate learning of clue words, and who can also articulate
exactly how they read target words may be most likely to show
genuine orthographic analogy use. Assessments of these three
skills might be used together to identify children who can use
orthographic analogy and to avoid misinterpreting phonological
priming effects as orthographic analogy use in Grade 1 children.
Limitations of the Present Study
The present research has several important limitations. First, it is
limited in scope to focus on the early use of rime- and vowel-based
analogies in English and required the oral reading of isolated
ANALOGIES FROM MULTIPLE CLUE WORDS
monosyllables. The study is silent on other languages, or whether
older readers can and do use rime or other forms of analogy.
Similarly, findings reflect research on experimental studies of
spontaneous analogy use and, as such, should be disconnected
from studies explicitly teaching children to use rime analogies.
Although the present results suggest that young children do not
appear to be particularly adept at the spontaneous use of orthographic rime analogy, they may, however, benefit from being
explicitly taught to make rime analogies, perhaps because of
children’s phonological preferences for rimes in some tasks, or
perhaps because rimes are consistent and may be useful patterns in
the ultimate internalization of the English spelling system. Indeed,
our own research (Di Stasio, Savage, & Abrami, 2010) provides
some evidence of long-term advantage for rime- over phonemebased early instruction, for a group of children from low-SES
backgrounds with English as an additional language.
The present study did not feature analysis of the impact of
rime-analogy teaching on spontaneous task performance, and there
is some evidence that a heavy emphasis on rime- or phonemebased approaches influences children’s decoding approaches (e.g.,
Deavers et al., 2000). Phonic approaches did not feature heavily in
our classrooms, where occasional reference to letters and sounds in
a more holistic approach to early literacy was evident, rather than
any systematic phonics teaching. Our results may well reflect these
early educational experiences. Patterns of clue word task performance may however be different in different pedagogical contexts.
Implications of Findings and Future Research
The implications of the present findings for research and practice are several. First, for research, the results probably suggest that
phonological priming effects are pervasive in clue word tasks. As
a result, new research paradigms may be required that emphasize
fuller and more explicit teaching of clue word spellings and
rime-based decoding strategies, probably through well-designed
intervention studies. The main implications for practice are cautionary in nature. Educational professionals should be wary of
children’s apparent facility to generate pronunciations for target
words on the basis of short-term learning of relevant clue words,
as this approach may only reflect phonological priming. Educators
can best verify whether children are using orthographic analogies
appropriately by (a) contrasting pairs (e.g., peek-meek and peekbeak) and (b) exploring children’s self-reported strategies
(Farrington-Flint & Wood, 2007). These approaches to matching
distinct student skills to learning strategies may maximize the
chances of children using genuine spontaneous orthographic analogies in educational settings.
References
Anthony, J. L., & Lonigan, C. J. (2004). The nature of phonological
awareness: Converging evidence from four studies of preschool and
early grade school children. Journal of Educational Psychology, 96,
43–55. doi:10.1037/0022-0663.96.1.43
Bowey, J. A. (1999). The limitations of orthographic rime analogies in
beginners’ word reading: A reply to Goswami. Journal of Experimental
Child Psychology, 72, 220 –231. doi:10.1006/jecp.1998.2484
Bowey, J. A., & Hansen, J. (1994). The development of orthographic rimes
as units of word recognition. Journal of Experimental Child Psychology,
58, 465– 488. doi:10.1006/jecp.1994.1045
203
Bowey, J. A., & Underwood, N. (1996). Further evidence that orthographic
rime usage in non-word reading increases with word level reading
proficiency. Journal of Experimental Child Psychology, 63, 526 –562.
doi:10.1006/jecp.1996.0061
Bowey, J. A., Vaughn, L., & Hansen, J. (1998). Beginning readers’ use of
orthographic analogies in word reading. Journal of Experimental Child
Psychology, 68, 108 –133. doi:10.1006/jecp.1997.2421
Brown, A. L., & Kane, M. J. (1988). Preschool children can learn to
transfer: Learning to learn and learning by example. Cognitive Psychology, 20, 493–523. doi:10.1016/0010-0285(88)90014-X
Brown, G. D. A., & Deavers, R. (1999). Units of analysis in nonword
reading: Evidence from children and adults. Journal of Experimental
Child Psychology, 73, 208 –242. doi:10.1006/jecp.1999.2502
Bruck, M., & Treiman, R. (1992). Learning to pronounce words: The
limitations of analogy. Reading Research Quarterly, 27, 374 –388. doi:
10.2307/747676
Bryant, P. E. (1998). Sensitivity to onset and rime does predict young
children’s reading: A comment on Muter, Hulme, Snowling, & Taylor
(1997). Journal of Experimental Child Psychology, 71, 29 –37. doi:
10.1006/jecp.1998.2455
Bryant, P. E. (2002). It doesn’t matter whether onset and rime predicts
reading better than phoneme awareness or vice versa. Journal of Experimental Child Psychology, 82, 41– 46. doi:10.1006/jecp.2002.2672
Bryant, P. E., Maclean, M., Bradley, L., & Crossland, J. (1990). Rhyme
and alliteration, phoneme detection and learning to read. Developmental
Psychology, 26, 429 – 438. doi:10.1037/0012-1649.26.3.429
Byrne, B. (1998). The foundations of literacy: The child’s acquisition of the
alphabetic principle. Hove, England: Psychology Press.
Caravolas, M., Volı́n, J., & Hulme, C. (2005). Phoneme awareness is a key
component of alphabetic literacy skills in consistent and inconsistent
orthographies: Evidence from Czech and English children. Journal of
Experimental Child Psychology, 92, 107–139. doi:10.1016/j.jecp.2005.04.003
Carroll, J. B., Davies, P., & Richman, B. (1971). The word frequency book.
Boston, MA: Houghton Mifflin.
Cassady, J. C., & Smith, L. L. (2004). Acquisition of blending skills: Comparison among body-coda, onset-rime, and phoneme blending tasks. Reading Psychology, 25, 261–272. doi:10.1080/02702710490512307
Castles, A., & Coltheart, M. (2004). Is there a causal link from phonological awareness to success in learning to read? Cognition, 91, 77–111.
doi:10.1016/S0010-0277(03)00164-1
Christensen, C. A., & Bowey, J. A. (2005). The efficacy of orthographic
rime, grapheme-phoneme correspondence, and implicit phonics approaches to teaching decoding skills. Scientific Studies of Reading, 9,
327–349. doi:10.1207/s1532799xssr0904_1
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993). Models of
reading aloud: Dual route and parallel distributed processes. Psychological Review, 100, 589 – 608. doi:10.1037/0033-295X.100.4.589
Comaskey, E. M., Savage, R. S., & Abrami, P. (2009). A randomized
efficacy study of a web-based literacy intervention among disadvantaged
urban kindergarten children. Journal of Research in Reading, 32, 92–
108. doi:10.1111/j.1467-9817.2008.01383.x
Deavers, R., Solity, J., & Kerfoot, S. (2000). The effect of instruction on
early nonword reading strategies. Journal of Research in Reading, 23,
267–286. doi:10.1111/1467-9817.00122
Di Stasio, M., Savage, R. S., & Abrami, P. (2010). A follow up study of the
ABRACADABRA web-literacy intervention in Grade 1. Journal of
Research in Reading. Advance online publication. doi:10.1111/j.14679817.2010.01469.x
Duncan, L. G., Seymour, P. H. K., & Hill, S. (1997). How important are
rhyme and analogy in beginning reading? Cognition, 63, 171–208.
doi:10.1016/S0010-0277(97)00001-2
204
SAVAGE, DEAULT, DAKI, AND AOUAD
Duncan, L. G., Seymour, P. H. K., & Hill, S. (2000). A small to large unit
progression in metaphonological awareness and reading? Quarterly
Journal of Experimental Psychology, 53A, 1081–1104.
Dunn, L. M., & Dunn, L. M. (1996). Peabody Picture Vocabulary Test (3rd
ed.). Circle Pines, MN: American Guidance Service.
Ehri, L. C. (1992). Reconceptualizing the development of sight word
reading. In P. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 107–143). Hillsdale, NJ: Erlbaum.
Ehri, L. C. (2005). Learning to read words: Theory, findings, and
issues. Scientific Studies of Reading, 9, 167–188. doi:10.1207/
s1532799xssr0902_4
Ehri, L. C., & Robbins, C. (1992). Beginners need some decoding skill to
read words by analogy. Reading Research Quarterly, 27, 12–26. doi:
10.2307/747831
Farrington-Flint, L., Coyne, E., Stiller, J., & Heath, E. (2008). Variability
in children’s early reading strategies. Educational Psychology, 28, 643–
661. doi:10.1080/01443410802140958
Farrington-Flint, L., & Wood, C. (2007). The role of lexical analogies in
beginning reading: Insights from children’s self-reports. Journal of
Educational Psychology, 99, 326 –338. doi:10.1037/0022-0663.99.2.326
Farrington-Flint, L., Wood, C., Canobi, K. H., & Faulkner, D. (2004).
Pattern of analogical reasoning among beginning readers. Journal of
Research in Reading, 27, 226 –247. doi:10.1111/j.1467-9817.2004
.00229.x
Geudens, A., Sandra, D., & Martensen, H. (2005). Rhyming words and
onset-rime constituents: An inquiry into structural breading points and
emergent boundaries in the syllable. Journal of Experimental Child
Psychology, 92, 366 –387.
Glushko, R. J. (1979). The organisation and activation of orthographic
knowledge in reading aloud. Journal of Experimental Psychology: Human Perception and Performance, 5, 674 – 691. doi:10.1037/00961523.5.4.674
Goswami, U. C. (1986). Children’s use of analogy in learning to read: A
developmental study. Journal of Experimental Child Psychology, 42,
73– 83. doi:10.1016/0022-0965(86)90016-0
Goswami, U. C. (1988). Orthographic analogies and reading development.
Quarterly Journal of Experimental Psychology, 40A, 239 –268.
Goswami, U. C. (1990a). Phonological priming and orthographic analogies
in reading. Journal of Experimental Child Psychology, 49, 323–340.
doi:10.1016/0022-0965(90)90062-D
Goswami, U. C. (1990b). A special link between rhyming skill and the use
of orthographic analogies by beginning readers. Journal of Child Psychology and Psychiatry and Allied Disciplines, 31, 301–311. doi:
10.1111/j.1469-7610.1990.tb01568.x
Goswami, U. C. (1993). Towards an interactive analogy model of reading
development: Decoding vowel graphemes in beginning reading. Journal
of Experimental Child Psychology, 56, 443– 475. doi:10.1006/
jecp.1993.1044
Goswami, U. C. (1999). Causal connections in beginning reading: The
importance of rhyme. Journal of Research in Reading, 22, 217–240.
doi:10.1111/1467-9817.00087
Goswami, U. C., & Bryant, P. E. (1990). Phonological skills and learning
to read. London, England: Erlbaum.
Goswami, U. C., & Mead, F. (1992). Onset and rime awareness and
analogies in reading. Reading Research Quarterly, 27, 152–162. doi:
10.2307/747684
Goswami, U., Ziegler, J. C., & Richardson, U. (2005). The effects of
spelling consistency on phonological awareness: A comparison of English and German. Journal of Experimental Child Psychology, 92, 345–
365.
Holopainen, L., Ahonen, T., & Lyttinnen, L. (2002). The role of reading by
analogy in first grade Finnish readers. Scandinavian Journal of Educational Research, 46, 83–98. doi:10.1080/00313830120115624
Howell, D. C. (1997). Statistical methods for psychology (4th ed.). Belmont, CA: Duxbury Press.
Hulme, C., Hatcher, P., Nation, K., Brown, A., Adams, J., & Stuart, G.
(2002). Phoneme awareness is a better predictor of early reading skill
than onset-rime awareness. Journal of Experimental Child Psychology,
82, 2–28. doi:10.1006/jecp.2002.2670
Johnston, F. (1999). The timing and teaching of word families. The
Reading Teacher, 53, 65–75.
Kay, J., & Marcel, A. J. (1981). One process, not two in reading aloud:
Lexical analogies do the work of non-lexical rules. Quarterly Journal of
Experimental Psychology, 33A, 397– 414.
MacMillan, B. (2002). Rhyme and reading; A critical review of the
research methodology. Journal of Research in Reading, 25, 4 – 42.
doi:10.1111/1467-9817.00156
Masterson, K., Stuart, M., Dixon, M., Lovejoy, D., & Lovejoy, S. (2003).
The Children’s Printed Word Database. Retrieved from http://
www.essex.ac.uk/psychology/cpwd/
Masterson, K., Stuart, M., Dixon, M., & Lovejoy, S. (2010). Children’s
printed word database: Continuities and changes over time in children’s
early reading vocabulary. British Journal of Psychology, 101, 221–242.
doi:10.1348/000712608X371744
McBride-Chang, C., Jeung-Ryeul, C., Liu, H., Wagner, R., Shu, H., Zhou,
A., . . . Muse, A. (2005). Changing models across cultures: Associations
of phonological awareness and morphological structure awareness with
vocabulary and word recognition in second graders from Beijing, Hong
Kong, Korea, and the United States. Journal of Experimental Child
Psychology, 92, 140 –160. doi:10.1016/j.jecp.2005.03.009
Muter, V., Snowling, M. J., & Taylor, S. (1994). Orthographic analogies
and phonological awareness: Their role and significance in early reading
development. Journal of Child Psychology and Psychiatry and Allied
Disciplines, 35, 293–310. doi:10.1111/j.1469-7610.1994.tb01163.x
Nation, K., Allen, R., & Hulme, C. (2001). The limitations of orthographic
analogy in early reading development: Performance on the clue word
task depends on phonological priming and elementary decoding skill,
not use of orthographic analogy. Journal of Experimental Child Psychology, 80, 75–94. doi:10.1006/jecp.2000.2614
Papadopoulos, T. C., Spanoudis, G., & Kendeou, P. (2009). The dimensionality of phonological abilities in Greek. Reading Research Quarterly, 44, 127–143. doi:10.1598/RRQ.44.2.2
Perfetti, C. (2003). The universal grammar of reading. Scientific Studies of
Reading, 7, 3–24. doi:10.1207/S1532799XSSR0701_02
Perfetti, C. (2007). Reading ability: Lexical quality to comprehension.
Scientific Studies of Reading, 11, 357–384.
Roberts, L., & McDougall, S. (2003). What do children do in the rime
analogy task? An examination of the skills and strategies used by
readers. Journal of Experimental Child Psychology, 84, 310 –337. doi:
10.1016/S0022-0965(03)00029-8
Savage, R. (1997). Do children need concurrent prompts in order to
perform lexical analogies? Journal of Child Psychology and Psychiatry
and Allied Disciplines, 38, 235–246. doi:10.1111/j.1469-7610.1997
.tb01857.x
Savage, R. (2001). A re-evaluation of the evidence for orthographic analogies: A reply to Goswami (1999). Journal of Research in Reading, 24,
1–18. doi:10.1111/1467-9817.00129
Savage, R. S., Abrami, P., Hipps, G., & Deault, L. (2009). A randomised
control trial study of the ABRACADABRA web-based reading intervention program in grade 1. Journal of Educational Psychology, 101,
590 – 604. doi:10.1037/a0014700
Savage, R. S., Blair, R., & Rvachew, S. (2006). Rimes are not necessarily
favored by prereaders: Evidence from meta- and epilinguistic phonological tasks. Journal of Experimental Child Psychology, 94, 183–205.
doi:10.1016/j.jecp.2006.03.005
Savage, R. S., & Carless, S. (2005). Phoneme manipulation but not
onset-rime manipulation is unique predictor of early reading. Journal of
ANALOGIES FROM MULTIPLE CLUE WORDS
Child Psychology and Psychiatry and Allied Disciplines, 46, 1297–
1308. doi:10.1111/j.1469-7610.2005.01436.x
Savage, R. S., Carless, S., & Stuart, M. (2003). The effects of rime- and
phoneme-based teaching delivered by learning support assistants. Journal of Research in Reading, 26, 211–233. doi:10.1111/1467-9817.00199
Savage, R. S., & Rvachew, S. (2006, June 18 –20). Epi-linguistic and
meta-linguistic phonological knowledge of preschoolers. Poster presented at the annual Canadian Language and Literacy Research Network
Conference, Charlottetown, Prince Edward Island, Canada.
Savage, R., & Stuart, M. (1998). Sub-lexical inferences in beginning
reading: Medial vowels as functional units of transfer. Journal of Experimental Child Psychology, 69, 85–108. doi:10.1006/jecp.1998.2439
Savage, R., & Stuart, M. (2001). Orthographic analogies and early reading:
Explorations of performance and variation in two transfer tasks. Reading
and Writing: An Interdisciplinary Journal, 3, 1–27.
Savage, R. S., & Stuart, M. (2006). A developmental model of reading
acquisition based upon early scaffolding errors and subsequent vowel
inferences. Educational Psychology, 26, 33–53. doi:10.1080/
01443410500340983
Seymour, P. H. K., Duncan, L. C., & Bolik, F. M. (1999). Rhymes and
phonemes in the common unit task: Replications and implications for
beginning reading. Journal of Research in Reading, 22, 113–130. doi:
10.1111/1467-9817.00077
Share, D. L. (2008). On the anglocentricities of current reading research
and practice: The perils of overreliance on an “outlier” orthography.
Psychological Bulletin, 134, 584 – 615. doi:10.1037/0033-2909.134
.4.584
Snowling, M. J., & Hulme, C. (2005). The science of reading: A handbook.
Oxford, England: Blackwell.
Stuart, M., & Coltheart, M. (1988). Does reading develop in a sequence of
stages? Cognition, 30, 139 –181. doi:10.1016/0010-0277(88)90038-8
Stuart, M., Dixon, M., Masterson, J., & Gray, R. (2003). Children’s early
reading vocabulary: Description and word frequency lists. British Journal of Educational Psychology, 73, 585–598. doi:10.1348/
000709903322591253
Stuart, M., Masterson, J., Dixon, M., & Quinlan, P. (1999). Inferring
sub-lexical correspondences from sight vocabulary: Evidence from 6and 7-year olds. Quarterly Journal of Experimental Psychology, 52,
353–366. doi:10.1080/027249899391106
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics.
(5th ed.) New York, NY: Allyn & Bacon.
Tanenhaus, M. K., Flanigan, H. P., & Seidenberg, M. (1980). Orthographic
and phonological activation in auditory and visual word identification.
Memory & Cognition, 8, 513–520.
Torgerson, C., Brooks, G., & Hall, J. (2006). A systematic review of the
research literature on the use of phonics in the teaching of reading and
205
spelling. Department for Education and Skills research report no:
RR711.
Treiman, R., Mullennix, J., Bijeljac-Babic, R., & Richmond-Welty, E. D.
(1995). The special role of rimes in the description, use and acquisition
of English orthography. Journal of Experimental Psychology: General,
124, 107–136. doi:10.1037/0096-3445.124.2.107
van Daal, V., Reitsma, P., & van der Leij, A. (1994). Processing units in
word reading by disabled readers. Journal of Experimental Child Psychology, 57, 180 –210. doi:10.1006/jecp.1994.1009
Walton, P. D. (1995). Rhyming ability, phoneme identity, letter-sound
knowledge, and the use of orthographic analogy by pre-readers. Journal
of Educational Psychology, 87, 587–597. doi:10.1037/00220663.87.4.587
Wilkinson, G. S. (1993). Wide Range Achievement Test administration
manual. Wilmington, DE: Wide Range.
Wood, C. (2000). Rhyme awareness, orthographic analogy use and phonemic awareness: An examination of relationships. Educational Psychology, 20, 5–15. doi:10.1080/014434100110344
Wood, C. (2002). Orthographic analogies and phonological priming effects. Journal of Research in Reading, 25, 144 –159. doi:10.1111/14679817.00165
Wood, C., & Farrington-Flint, L. (2001). Orthographic analogy use and
phonological priming effects in non-word reading. Cognitive Development, 16, 951–963. doi:10.1016/S0885-2014(02)00071-0
Wright, D. M., & Ehri, L. C. (2007). Beginners remember orthography
when they learn to read words: The case of doubled letters. Applied
Psycholinguistics, 28, 115–133. doi:10.1017/S0142716406070068
Wyse, D., & Goswami, U. (2008). Synthetic phonics and the teaching of
reading. British Educational Research Journal, 34, 691–710. doi:
10.1080/01411920802268912
Ziegler, J. C., & Goswami, U. C. (2005). Reading acquisition, developmental dyslexia, and skilled reading across languages: A psycholinguistic grain size theory. Psychological Bulletin, 131, 3–29. doi:10.1037/
0033-2909.131.1.3
Zorzi, M., Houghton, G., & Butterworth, B. (1998a). The development of
spelling-sound relationships in a model of phonological reading. Language and Cognitive Processes, 13, 337–371.
Zorzi, M., Houghton, G., & Butterworth, B. (1998b). Two routes or one in
reading aloud? A connectionist dual process model. Journal of Experimental Psychology, Human Perception and Performance, 24, 1131–
1161. doi:10.1037/0096-1523.24.4.1131
Received January 6, 2010
Revision received September 7, 2010
Accepted September 13, 2010 䡲