A core challenge for theories of word reading development is to explain how children transition from effortful, letter-by-letter decoding to fast and accurate identification of individual words. Many prominent models of reading development point to phonological awareness as central in establishing letter-sound correspondences, which in turn propel children’s learning to read individual words (e.g., Ehri, Reference Ehri2005). And yet phonological awareness is unlikely to be the only factor in this development. A word such as dishonest, for instance, would be mispronounced by reference to phonology alone, reflecting the fact that English is a morphophonemic orthography (e.g., Venezky, Reference Venezky1967). Morphological awareness, or the awareness of and ability to manipulate the smallest units of meaning in language (Carlisle, Reference Carlisle2000), has been gaining momentum as a second predictor (e.g., Castles et al., Reference Castles, Rastle and Nation2018; Deacon & Kirby, Reference Deacon and Kirby2004). Indeed, there is now over a decade of theoretical advocacy for a role of morphological awareness in supporting children in reading morphologically complex words such as dishonest (e.g., Kuo & Anderson, Reference Kuo and Anderson2006; Levesque et al., Reference Levesque, Breadmore and Deacon2021). These could be words with several base morphemes (i.e., compound words) or with a base morpheme and one or more affixes (i.e., prefixed and suffixed words). Morphologically complex words are incredibly prevalent in children’s texts (Anglin, Reference Anglin1993; White et al., Reference White, Power and White1989). And yet, most empirical studies have evaluated, and confirmed, a role for morphological awareness in children’s broader word reading skill, assessed by standardized measures of word reading that include both morphologically complex and simple words (e.g., Deacon & Kirby, Reference Deacon and Kirby2004). The mismatch between available evidence and theoretical predictions fact has led to repeated calls for research to clarify the role of morphological awareness in word reading development (Carlisle, Reference Carlisle2010; Nagy et al., Reference Nagy, Carlisle and Goodwin2014; Sénéchal & Kearnan, Reference Sénéchal, Kearnan and Kail2007). We report here on a longitudinal study designed to do so by contrasting the relative role of morphological awareness in developing skill in reading morphologically complex words and in broader word reading skill.
Theoretical context
Multiple theories (e.g., Perfetti & Stafura, Reference Perfetti and Stafura2014; see also Reichle & Perfetti, Reference Reichle and Perfetti2003; Schreuder & Baayan, Reference Schreuder, Baayen and Feldman1995) advocate a targeted role for morphological awareness in morphological decoding, or what has been defined as the use of morphemes in word reading (Deacon et al., Reference Deacon, Tong and Francis2017; Nagy et al., Reference Nagy, Berninger and Abbott2006; see also morphological decomposition, Verhoeven & Perfetti, Reference Verhoeven and Perfetti2011). Morphological decoding is a reading strategy in which children leverage individual morphemes to facilitate the reading of multimorphemic words (Carlisle & Stone, Reference Carlisle and Stone2005; Deacon et al., Reference Deacon, Tong and Francis2017; Goodwin et al., Reference Goodwin, Petscher, Carlisle and Mitchell2017). For instance, children could use morphological decoding to read the word dishonest, helping them to decide to pronounce sh by the morpheme boundary (dis+honest) rather than its more frequent pronunciation as a digraph (e.g., /ʃ/ in dishes). Accumulating research shows that English-speaking children use morphemes to read words (e.g., Carlisle & Stone, Reference Carlisle and Stone2005; Deacon et al., Reference Deacon, Whalen and Kirby2011; Goodwin et al., Reference Goodwin, Petscher, Carlisle and Mitchell2017; Goodwin, Gilbert, & Cho, Reference Goodwin, Gilbert and Cho2013). This finding resonates with results from priming paradigms (e.g., Rabin & Deacon, Reference Rabin and Deacon2008; Deacon, Campbell, Tamminga, & Kirby, Reference Deacon, Campbell, Tamminga and Kirby2010; McCutchen, Logan, and Biangardi-Orpe, Reference McCutchen, Logan and Biangardi-Orpe2009), several of which suggest a temporally late role for morphosemantic processing (e.g., Quémart, Casalis, & Colé, Reference Quémart, Casalis and Colé2011; Quémart, Gonnerman, Downing, & Deacon, Reference Quémart, Gonnerman, Downing and Deacon2018). Awareness of the morphological structure of words is argued to give developing readers insight into the morphemic structure of their writing system (Kuo & Anderson, Reference Kuo and Anderson2006) and make morpheme boundaries more salient when reading (Carlisle, Reference Carlisle2010). This enables the parsing of words into their constituent morphemes, which facilitates faster and more accurate reading of morphologically complex words. From this view, morphological awareness should have a stronger role in the development of children’s morphological decoding.
These ideas form the conceptual foundation for a recent model, Morphological Pathways Framework (Levesque et al., Reference Levesque, Breadmore and Deacon2021). In this framework, morphological awareness predominately influences reading processes by facilitating the recognition and use of morphemes in reading words. Accordingly, the Morphological Pathways Framework predicts that morphological awareness has a targeted effect on morphological decoding. This prediction adds mechanistic detail to prior models, such as the Reading Systems Framework (Perfetti & Stafura, Reference Perfetti and Stafura2014). In the Reading Systems Framework, as a part of the linguistic system (Perfetti et al., Reference Perfetti, Landi, Oakhill, Snowling and Hulme2005), morphology connects to both the orthographic system and the lexicon, and thereby to word reading (see also Reichle & Perfetti, Reference Reichle and Perfetti2003; Schreuder & Baayen, Reference Schreuder, Baayen and Feldman1995). The Morphological Pathways Framework builds on this deep foundation to make a specific prediction for the role of morphological awareness in morphological decoding.
Empirical research to date
In contrast to these theoretical predictions of a targeted and/or stronger role in morphological decoding, most empirical studies have evaluated the role of morphological awareness in broader word reading skill. We use this term—broader word reading skill—here and throughout this manuscript to refer to measurement of word reading ability with standardized assessments. Standardized tests of word reading are designed to capture broad word reading skill in their strategic inclusion of a wide variety of words, including ones that with a single (e.g., forest) or multiple morphemes (e.g., courageous), that are regular and irregular, and also nonwords. Most studies to date on the effects of morphological awareness have included such standardized measures (e.g., Apel et al., Reference Apel, Diehm and Apel2013; Deacon & Kirby, Reference Deacon and Kirby2004; Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012). Indeed, a large set of cross-sectional studies have demonstrated a relation between morphological awareness and performance on these measures of broader word reading skill (e.g., Choi et al., Reference Choi, Tong, Law and Cain2017; Goodwin et al., Reference Goodwin, Huggins, Carlo, August and Calderon2013; Roman et al., Reference Roman, Kirby, Parrila, Wade-Woolley and Deacon2009). For instance, Kirby and colleagues (Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012) demonstrated significant relations between children’s morphological awareness and their performance on several standardized tests of word reading, including Word Identification (Woodcock, Reference Woodcock1998) and the Test of Word Reading Efficiency (TOWRE; Torgesen et al., Reference Torgesen, Wagner and Rashotte1999).
A smaller set of longitudinal studies have shown that morphological awareness makes a unique contribution to gains in broader word reading skill over time by including stringent autoregressive controls (Deacon et al., Reference Deacon, Benere and Pasquarella2013; Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012; Kruk & Bergman, Reference Kruk and Bergman2013). For instance, Deacon and colleagues (Reference Deacon, Benere and Pasquarella2013) showed that morphological awareness assessed in Grade 2 predicted children’s gains in broader word reading skill between Grades 2 and 3. Kruk and Bergman (Reference Kruk and Bergman2013) showed similar unique relations between Grades 1 and 3, including after controls for vocabulary and phonological awareness. Together, these studies provide evidence that morphological awareness supports the development of broader word reading skill.
This association with broader word reading skill fits with some views that morphological awareness supports word reading skill across a whole range of words. It has been argued to do so by facilitating the mappings between print and meaning (Perfetti, Reference Perfetti2007; Rastle, Reference Rastle2018) or by binding letters, sounds, and meaning of words in memory (Kirby & Bowers, Reference Kirby, Bowers, Cain, Compton and Parrila2017). Others have suggested morphological awareness reflects a broad form of linguistic awareness encompassing elements of meaning, syntax, and phonology, which together could have a widespread influence on word reading development (Carlisle, Reference Carlisle2003; Kuo & Anderson, Reference Kuo and Anderson2006). And indeed, just as phonological awareness initially serves to secure regular words in memory through robust letter-sound connections which then supports memory for both phonologically regular and irregular words (e.g., Ehri, Reference Ehri2014), the influence of morphological awareness could certainly extend beyond just the reading of morphologically complex words (see also Deacon & Kirby, Reference Deacon and Kirby2004).
And yet, the use of standardized measures of word reading in prior studies also means that the possibility that morphological awareness has a stronger influence on morphological decoding has not been directly tested. Roughly one-third of the words included on standardized assessments are morphologically complex (e.g., Deacon et al., Reference Deacon, Benere and Pasquarella2013; Goodwin et al., Reference Goodwin, Huggins, Carlo, August and Calderon2013), with most of these toward the end of the test. The use of stop rules means that readers reach these words to differing degrees, both within and across studies. Given this variability, it is not clear the extent to which performance on standardized tests captures, or is confounded by, morphological decoding. Indeed, it seems more likely that the effects of morphological awareness would be stronger for words with a morphological structure than for broader measures of word reading; effects on morphological decoding are in effect “near transfer,” while those for broader word reading constitute “far transfer.”Footnote 1
Testing the theoretically based prediction that effects of morphological awareness on word reading outcomes are stronger to morphological decoding requires more targeted measurement of morphologically complex words. A handful of studies have done so, providing evidence of relations between morphological awareness and morphological decoding (Carlisle, Reference Carlisle2000; Mann & Singson, Reference Mann, Singson, Assink and Sandra2003; Nagy et al., Reference Nagy, Berninger and Abbott2006). For instance, in a cross-sectional study with students in Grades 4 to 9, Nagy and colleagues (Reference Nagy, Berninger and Abbott2006) found that morphological awareness was uniquely related to multiple measures of morphological decoding, beyond phonological skills. These studies did not include measures of broader word reading skill though, limiting their ability to evaluate unique effects.
Three recent studies with mid- to upper elementary school aged children included measures of morphological decoding as well as of broader word reading skill (Goodwin et al., Reference Goodwin, Huggins, Carlo, August and Calderon2013; Kearns, Reference Kearns2015; Levesque et al., Reference Levesque, Kieffer and Deacon2017). This is an important step forward, as it enables a direct contrast of the relative size of the contributions of morphological awareness to reading of morphologically complex words and to broader word reading skill. With this approach, Goodwin et al. (Reference Goodwin, Huggins, Carlo, August and Calderon2013) and Kearns (Reference Kearns2015) showed that morphological awareness predicted morphological decoding above and beyond standardized assessments of word reading skills in their studies. Similar findings emerged with SEM modeling with children in Grade 3; Levesque and colleagues (2017) found that morphological awareness was related to morphological decoding, with no direct unique link to broader word reading skill; there were, however, indirect effects of morphological awareness on morphological decoding and in turn on broader word reading skill. Taken together, these studies support the idea that morphological awareness is more strongly related to morphological decoding than to broader word reading skill.
The next key step in this line of research is to assess whether morphological awareness supports gains in morphological decoding over time, an important move beyond demonstrating associations at a single point in time. This can be done by controlling for prior performance on word reading tests (i.e., the autoregressor) when evaluating the contributions of earlier levels of morphological awareness on later word reading performance. In this way, autoregressive models identify factors that determine change over time (Raudenbush, Reference Raudenbush2001; Selig & Little, Reference Selig, Little, Laursen, Little and Card2012). This approach has been used to demonstrate that morphological awareness is related to children’s gains in broader word reading skill (e.g., Deacon et al., Reference Deacon, Benere and Pasquarella2013; Kruk & Bergman, Reference Kruk and Bergman2013). We apply this approach here to test the theoretically driven prediction that morphological awareness has a stronger role in the development of morphological decoding skill by contrasting its effects on the development of broader word reading skill.
The current study
In this study, we evaluate whether morphological awareness has a stronger role in supporting gains in morphological decoding or in broader word reading skill. We did so in a longitudinal study of children from Grade 3 to 4. At Grade 3, we assessed morphological awareness with two orally presented and widely used tasks: the Word Analogy Task (Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012) and the Test of Morphological Structure (Carlisle, Reference Carlisle2000). We also included vocabulary and phonological awareness, measured with standardized tests, as controls in our models given their well-established connections to reading-related skills (e.g., Deacon & Kirby, Reference Deacon and Kirby2004). At both Grades 3 and 4, we included multiple measures of morphological decoding (Levesque et al., Reference Levesque, Kieffer and Deacon2017; Nunes et al., Reference Nunes, Bryant and Barros2012) and standardized tests of word reading skill at both grades (Torgesen et al., Reference Torgesen, Wagner and Rashotte1999). The standardized measures include morphologically complex words (e.g., factories and straighten) and so they likely capture morphological decoding to some extent. This means that models that contrast effects on morphological decoding with those in gains in broader word reading skill offer a particularly stringent test of the prediction that morphological awareness is more strongly related to gains in morphological decoding.
The mid-elementary grades are a highly relevant time period to evaluate these effects. In line with developmental phases of word reading, progress in word reading efficiency during this period is driven by the use of salient, recurring letter patterns such as morphemes (Ehri, Reference Ehri2014). Moreover, texts during these grades are replete with morphologically complex words (Anglin, Reference Anglin1993; White et al., Reference White, Power and White1989), and children are also rapidly developing knowledge of derivational morphology (Carlisle, Reference Carlisle2003; Foorman et al., Reference Foorman, Petscher and Bishop2012). Within this context, we tested whether children’s morphological awareness in Grade 3 has stronger unique effects on gains in morphological decoding or in broader word reading skill between Grades 3 and 4.
Answering this question requires distinguishing between morphological decoding and broader word reading skill. And yet, skill in morphological decoding is likely to be related to broader word reading skill; reading single- and multimorphemic words and nonwords engages the same lexical architecture for word identification (Perfetti & Stafura, Reference Perfetti and Stafura2014). In this context, we begin our analyses by contrasting a two-factor model separating measures of morphological decoding from standardized measures of word reading with a unidimensional model in which all tasks converge as a single “word reading” factor. Establishing separability of morphological decoding from broader word reading skill is a necessary first step toward answering our primary research question. In doing so, we build on the small set of prior work showing the empirical separability of morphological decoding from broader word reading skill (Levesque et al., Reference Levesque, Kieffer and Deacon2017; Nunes et al., Reference Nunes, Bryant and Barros2012).
Based on the Morphological Pathways Framework (Levesque et al., Reference Levesque, Breadmore and Deacon2021), we predicted that children’s morphological awareness would predict unique variance in gains in their morphological decoding between Grades 3 and 4, beyond its effects on broader word reading. This finding would also be in keeping with the limited cross-sectional research testing these questions (e.g., Goodwin et al., Reference Goodwin, Huggins, Carlo, August and Calderon2013; Kearns, Reference Kearns2015). It offers a critical extension to these findings by testing effects on change in skill level over time.
Methods
Participants
Participants were 197 children (106 girls) who were followed from Grade 3 to 4 as a part of a larger study. Data for these children at Grade 3 were reported in Levesque et al. (Reference Levesque, Kieffer and Deacon2017), and we report here on data from both Grades 3 and 4. Participants were on average 8 years and 10 months old (SD age = 3.91 months) in Grade 3 (Grade 4: M age = 9y;10m, SD age = 3.84 months). Ninety-three percent of children spoke English as their first language, and most of the participants (85%) were English monolingual speakers. Participants were recruited from 13 elementary schools. The children’s instruction was entirely in English at Grade 3 and in the schools in which recruitment occurred, literacy instruction generally integrates both whole-language and phonics instruction. Children were in the average range for their age on standardized assessments of phonological awareness, vocabulary, and word reading (see Table 1), suggesting that the sample was representative of typically developing children. On average, households in the catchment areas for the schools are in the middle to upper-middle SES class (Statistics Canada, 2017).
Descriptive statistics for Grade 3 and Grade 4 measures
Note. M = mean. SD = standard deviation.
a Age-based standard score with a mean of 100 (SD = 15).
b Reliability from manual.
c Cronbach’s alpha reliability.
d Age-based standard score with a mean of 10 (SD = 3).
Measures
Standardized tasks were administered and scored as per manual instructions. Task reliability estimates (Cronbach’s alpha) are provided in Table 1.
Morphological decoding
Morphological decoding was measured with three tasks evaluating children’s accuracy in reading aloud morphologically complex words and pseudowords. These tasks were designed to engage morphological processes in reading. Together, these tasks gauge participants’ ability to use the structure and morpheme boundaries within written multimorphemic words to produce their correct pronunciation.
Morphologically Derived Word Reading
For this morphological decoding task (taken from Levesque et al., Reference Levesque, Kieffer and Deacon2017), participants read aloud 40 derived words (questionable). These multimorphemic words consisted of a base morpheme (question) and a derivational suffix among the 20 most common English suffixes (-able, -al, -ly, -ment, -ful, -less, -ness, and -ous; Blevins, Reference Blevins2001). All derived words had a low whole-word frequency (<5 occurrences per million words in text, U, based on Zeno et al., Reference Zeno, Ivens, Millard and Duvvuri1995; M surface U = 1.60). Critically, derived words had a high-frequency base morpheme (U ≥ 48; M base U = 148.18). The high-frequency base provides an opportunity for morphological decoding to occur while children are reading these infrequent morphologically complex words (see e.g., Hay, Reference Hay2001; McCutchen & Logan, Reference McCutchen and Logan2011). About half of the derived words were phonologically and orthographically transparent with their base morpheme (e.g., reasonless); the remaining half had either a phonological change (e.g., publicity), an orthographic change (e.g., heaviness), or both a phonological-orthographical change (e.g., studious) between the base and the derive word (see Appendix).
Use of morphemes in reading real words
Taken from Nunes and colleagues (2012), the second morphological decoding task assessed children’s use of morphemes in reading individual words (19 items). In this task, the correct pronunciation of the words depends on their segmentation into morphemes (Nunes et al., Reference Nunes, Bryant and Barros2012). For instance, the words unusual and uniform, both beginning with un, are pronounced differently based on their morphological structure. Segmenting the words at their correct morpheme boundary (un + usual vs. uni + form) enables readers to arrive at the accurate pronunciation. In contrast, identifying the wrong morpheme boundary to segment the word leads to incorrect responses (e.g., pronouncing the prefix un in uniform).
Use of morphemes in reading pseudowords
Also taken from Nunes and colleagues (2012), the third morphological decoding task assessed children’s use of morphemes in reading pseudowords (15 items). Pseudowords were items with real morphemes put together to form nonwords (e.g., prefix uni + base match to form the pseudoword unimatch). As with the previous task with real words, the correct pronunciation of pseudowords words depends on their segmentation at the correct morpheme boundary (Nunes et al., Reference Nunes, Bryant and Barros2012). For instance, both mishope and unishaped include sh; correct pronunciation of these pseudowords requires distinguishing that the digraph sh is treated differently based on whether it aligns with a morpheme boundary (mis + hope vs. uni + shaped). In this sense, this pseudoword task strongly elicits morphological decoding processes in reading, because the whole-word forms do not exist in spoken or written language.
DirectRT software (Jarvis, Reference Jarvis2008) displayed the stimuli for the three morphological decoding tasks. The three tasks were administered separately, with items within each task appearing in a random order. Single words were displayed in the center of the screen in black 40-point Arial font. Words were preceded by a 1-s central fixation and appeared for a maximum of 5 s. Participants read the words aloud, and the experimenter recorded the accuracy of responses.
Broader word reading skill
Sight Word Efficiency
The Sight Word Efficiency subtest of the Test of Word Reading Efficiency (Torgesen et al., Reference Torgesen, Wagner and Rashotte1999) includes a list of real words printed in order of increasing difficulty (e.g., bat, plates, and forest). The subtest measures children’s ability to recognize words as whole units quickly and accurately (i.e., sight word reading).
Phonemic Decoding Efficiency
The Phonemic Decoding Efficiency subtest presents a list of pronounceable nonwords in order of increasing difficulty (e.g., daf, shloo, straler; Torgesen et al., Reference Torgesen, Wagner and Rashotte1999). This subtest evaluates children’s ability to decode, or sound out, printed nonwords quickly and accurately.
In keeping with prior literature, we chose the Sight Word Efficiency and Phonemic Decoding Efficiency subtests in an effort to capture children’s broader word reading skill. Together, these two standardized subtests provide a reliable index of word reading accuracy and fluency. Roughly, 30% of words on the Sight Word Efficiency subtest are morphologically complex (e.g., factories and straighten), and some nonwords have prefix- and suffix-resembling letter patterns (e.g., de- in depate; -er in debmer). In keeping with best practices, the children took a different form of the TOWRE in Grade 3 than in Grade 4.
Morphological awareness
We assessed morphological awareness with two widely used tasks: the Word Analogy Task (Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012) and the Test of Morphological Structure (Carlisle, Reference Carlisle2000). Both were administered orally, in keeping with standard practice (e.g., Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012).
Word Analogy Task
As in Kirby et al. (Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012), the Word Analogy Task presented pairs of morphologically related words in an analogy format (A:B::C:D). Participants heard a pair of words (A:B), followed by the first word of a second pair (C). Participants were then asked to complete the pattern by providing the missing word of the second pair (D) (e.g., help: helped:: say: [said]). Participants completed five practice and 20 test items. Of the 20 test items, half were inflected words and half were derived. Both inflected and derived items included phonologically transparent (walk-walked) and phonological opaque transformations (stood-stand).
Test of Morphological Structure
For the Test of Morphological Structure (Carlisle, Reference Carlisle2000), a target item was presented followed by an incomplete sentence. Participants were asked to change the target word to complete the sentence (e.g., Farm. My uncle is a ___ [farmer]). Feedback was provided for three practice items. The task included 28 target items. Half involved the production of a derived word from a base word (e.g., protect to protection) and half the decomposition of a derived form to its base form (e.g., growth to grow). Further, half of the items were phonologically transparent transformations (e.g., accept–acceptance) and half were phonologically opaque (e.g., revise–revision).
Controls
We included vocabulary and phonological awareness as controls given their well-established connections to reading-related skills (e.g., Deacon & Kirby, Reference Deacon and Kirby2004).
Vocabulary
Using the Peabody Picture Vocabulary Test—Fourth Edition (Dunn & Dunn, Reference Dunn and Dunn2007), participants heard individual words of increasingly difficulty spoken by the experimenter. They were shown four pictures and asked to indicate the picture that best represented the word.
Phonological awareness
We used the Elision subtest of the Comprehensive Test of Phonological Processing (Wagner et al., Reference Wagner, Torgesen and Rashotte1999). Participants heard up to 20 words presented one at a time by the experimenter. Children repeated each target word verbatim and again without a specific sound (e.g., “Say stale. Now say stale without saying /t/”; [sale]).
Procedure
This study was approved by the institutional and school board ethics committees. Written consent was obtained from parents, and oral assent was obtained from each child at every testing session. In each of Grades 3 and 4, testing took place from February to May with approximately 12 months separating the two testing points for any given participant.
The measures were administered within a battery of tasks as part of a larger longitudinal study, in which measures of morphological decoding and broader word reading skill were administered in Grades 3 and 4. Measures of morphological awareness, vocabulary, and phonological awareness were administered in Grade 3. All tasks were administered individually in a fixed order by a research assistant in a quiet location in the child’s school across two or more sessions based on the child’s interest and the school’s schedule.
Data analysis
We used structural equation modeling to test the contribution of morphological awareness to morphological decoding and to broader word reading skill from Grade 3 to 4. Analyses were conducted with participants’ raw scores using Mplus 8.0 (Muthén & Muthén, 2017). We used full-information maximum likelihood robust (MLR) to account for missing data (Enders, Reference Enders2013) and guard against bias due to non-normality and non-independence of observations (Finney & DiStephano, Reference Finney, DiStefano, Hancock and Mueller2013). MLR necessitates the Satorra–Bentler scaled chi-square difference test when comparing the goodness of fit across competing theoretical models (Bryant & Satorra, Reference Bryant and Satorra2012). Model fit was evaluated across several indices, such as the chi-square statistical test, comparative fit index, Tucker–Lewis index, root mean square error of approximation, and standardized root mean residual, as per Kline, Reference Kline2016 (see also Hooper et al., Reference Hooper, Coughlan and Mullen2008; Schreiber, Reference Schreiber2017). Across all analyses, our modeling decisions considered fit statistics and guidance from theory and prior research.
Results
Preliminary analyses
Table 1 presents descriptive statistics. Inspection of the data (e.g., outliers and skewness) showed no indication of non-normality or other concerns (Field, Reference Field2009). The amount of missing data was small, averaging less than 2% across measures (M = 1.97%; SD = 2.36%). MLR enabled the use of data from the full sample (Brown, Reference Brown2006; Enders, Reference Enders2013). Correlations are presented in Table 2. Brief inspection of this table reveals that morphological awareness at Grade 3 has a stronger relation to vocabulary than to phonological awareness at the same grade (see e.g., Sparks & Deacon, Reference Sparks and Deacon2015). Critically, there were significant relations of morphological awareness at Grade 3 to both morphological decoding and broader word reading at Grade 4, with stronger relations to morphological decoding than to broader word reading. Stable correlations between each of morphological decoding and broader word reading between Grades 3 and 4 support the use of autoregressor controls to examine these relations.
Correlations between measures
Note. N = 197. Gr3 = Grade 3. Gr4 = Grade 4. All correlations are significant, p < .01.
Building an autoregressive SEM model
As a foundation for a theoretically justified autoregressive model (Agresti & Finlay, Reference Agresti, Finlay, Agresti and Finlay2009), we assessed the theoretical structure among latent variables by fitting a two-factor model consisting of a morphological decoding factor and a broader word reading skill factor at both grades. The indicators for these factors and their loadings are listed in Table 1. At each grade, we contrasted this two-factor model with a unidimensional word reading factor (i.e., all tasks converging as a single construct). Model fit estimates for the two-factor model and unidimensional model for Grade 3 and 4 are presented in Table 3. Contrasting the two-factor and single-factor models tests whether morphological decoding is separable from broader word reading skill.
Model fit statistics for the two-factor and unidimensional word reading models in Grade 3 and 4
Note. The Satorra–Bentler scaled χ2 difference test, which uses a scaled correction factor, compared the fit of the 2-factor and 1-factor models at each grade. CFI = comparative fit index. TLI = Tucker–Lewis index. RMSEA = root mean square error of approximation. SRMR = standardized root mean square residual. BIC = Bayesian information criterion.
The two-factor model was a better fit than the unidimensional model; this difference was significant in Grade 3, Satorra–Bentler Δχ2 = 7.36, Δdf = 1, p < .01, and a near-significant trend in Grade 4, Satorra–Bentler Δχ2 = 3.34, Δdf = 1, p = .06. These findings suggest that morphological decoding is separable from broader word reading skill. The two-factor model was retained in both grades given these results as well as the a priori theoretically driven goal to evaluate contributions of morphological awareness to morphological decoding and broader word reading skill simultaneously. We were further motivated to keep the two-factor model in both grades to evaluate measurement invariance between constructs over time, as this is key to longitudinal autoregressive modeling.
We evaluated measurement invariance of morphological decoding and broader word reading skill from Grade 3 to 4. Establishing measurement invariance is important in longitudinal modeling with latent variables to ensure that the same constructs are being measured over time (MacCallum & Austin, Reference MacCallum and Austin2000). As is typical, residuals of corresponding indicators were correlated across time points. Next, progressively constrained models were tested such that factor loadings, followed by indicator intercepts, and finally indicator residual variances were fixed to be equal over time (as per Brown, Reference Brown2006; Little, Reference Little2013; Putnick & Bornstein, Reference Putnick and Bornstein2016). Partial strict invariance was attained following this process, χ2(33) = 70.44, p = < .01, comparative fit index (CFI) = .98, Tucker–Lewis index (TLI) = .98, root mean square error of approximation (RMSEA) = .08, SRMR = .05, supporting the evaluation of change in the latent variable means over time (Little, Reference Little2013). In our case, “partial” refers to a single intercept, that of Morphologically Derived Word Reading in Grade 4, which was permitted to vary.
Testing the contribution of morphological awareness to gains over time
Our research question lies in whether morphological awareness contributes more strongly to unique gains in morphological decoding versus in broader word reading skill between Grades 3 and 4. To examine gains in abilities over time, we included autoregressive factors of both morphological decoding and broader word reading skill in Grade 3, which accounts for prior levels of abilities in these constructs. Vocabulary and phonological awareness were added as control variables in the model (see Figure 1). A latent factor of morphological awareness was created from the Test of Morphological Structure and Word Analogy Task (factor loadings in Table 1). Two key predictive paths were added: one between Grade 3 morphological awareness and Grade 4 morphological decoding and another between Grade 3 morphological awareness and Grade 4 broader word reading skill. The paths between Grade 3 morphological awareness and Grade 3 factors of morphological decoding and broader word reading skill were fixed to zero so that the influence of morphological awareness was not subsumed within the autoregressive paths; without doing so in this study, the contributions of morphological awareness on gains in morphological decoding and broader word reading skill are conflated across direct and indirect effects. Using this targeted autoregressive modeling approach enabled us to more precisely test and contrast the cross-lagged effects between morphological awareness in Grade 3 and each of morphological decoding and broader word reading skill in Grade 4. Recall that for a specific influence of morphological awareness, we expected an effect on morphological decoding only.
Structural longitudinal model testing the contribution of morphological awareness to gains in morphological decoding and broader word reading skill from Grade 3 to 4.
Note. Standardized path coefficients are shown in the figure. *p < .05. Latent variable indicators are as follows: 1 = Word Analogy Task. 2 = Test of Morphological Structure. 3 = Morphologically Derived Word Reading. 4 = Use of morphemes in reading real words. 5 = Use of morphemes in reading pseudowords. 6 = Sight Word Efficiency. 7 = Phonemic Decoding Efficiency.
The resulting model (Figure 1) showed good fit to the data, χ2 (69) = 168.63, p = < .05, CFI = .96, TLI = .95, RMSEA = .08, SRMR = .07. Standardized path coefficients (β) are included in the figure. As reflected in the large autoregression effects, morphological decoding (β = .93, 95% CI [.75, .89]) and broader word reading skill (β = .97, 95% CI [.85, 1.02]) were highly stable from Grade 3 to 4. And yet, beyond the autoregressive effect and controls for vocabulary and phonological awareness, morphological awareness had a small significant effect on Grade 4 morphological decoding (β = .08, 95% CI [.01, .18]). In contrast, morphological awareness had no statistically detectable unique effect on word reading efficiency in Grade 4, beyond the other variables in the model (β = .03, 95% CI [−.06, .14]. Together, these findings suggest that morphological awareness contributed uniquely to gains across Grades 3 to 4 in children’s morphological decoding skills with no such unique effects on their broader word reading skill.
Discussion
This study was designed to provide much-needed empirical clarity (Rastle, Reference Rastle2018) as to how morphological awareness contributes to word reading development (e.g., Kuo & Anderson, Reference Kuo and Anderson2006; Nagy et al., Reference Nagy, Carlisle and Goodwin2014). We evaluated the relative size of the unique contribution of morphological awareness to morphological decoding with that to broader word reading skill. We tested this in a developmental framework, with highly conservative latent autoregressive modeling of longitudinal data for children followed from Grade 3 to 4 and with controls for phonological awareness and vocabulary levels. This is a time period during which children’s texts are rife with morphologically complex words (e.g., White et al., Reference White, Power and White1989), and their morphological awareness skill is increasing rapidly (e.g., Carlisle, Reference Carlisle2003). We found that morphological awareness in Grade 3 was a significant unique predictor of gains from Grade 3 to 4 in morphological decoding, but not of broader word reading skill. These findings point to a stronger effect of morphological awareness on the development of morphological decoding than for broader word reading skill for English-speaking children in the mid-elementary grades.
Building on other cross-sectional studies with mid-elementary school children (e.g., Goodwin et al., Reference Goodwin, Huggins, Carlo, August and Calderon2013; Kearns, Reference Kearns2015; Levesque et al., Reference Levesque, Kieffer and Deacon2017), our findings suggest that morphological awareness contributes to unique gains in morphological decoding when contrasted against broader word reading skill. Accordingly, we extend this prior work by showing that morphological awareness contributes to the development of morphological decoding between Grades 3 and 4; we did not detect unique effects on gains in broader word reading skill across this year. These findings provide empirical support for morphological awareness as a foundational metalinguistic skill that provides mid-elementary school aged readers with critical insight into the morphological structure of written words, which in turn facilitates the reading of words with a complex morphological structure (Carlisle, Reference Carlisle2003, Reference Carlisle2010; Kuo & Anderson, Reference Kuo and Anderson2006).
As we consider our positive findings, we also need to discuss what we did not find; there was no unique effect of morphological awareness on gains in broader word reading skill. At first glance, this null finding is inconsistent with the small set of longitudinal studies showing that morphological awareness supports the development of word reading skill when assessed across a wide range of words for readers early in the elementary years and up (Deacon et al., Reference Deacon, Benere and Pasquarella2013; Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012; Kruk & Bergman, Reference Kruk and Bergman2013). In these studies, word reading was measured with standardized tests (much like our study). Critically, these prior longitudinal studies did not include measures of morphological decoding. When both metrics are included, our results specify that, at least for English-speaking children at the mid-elementary level, the contribution of morphological awareness is stronger to the development of morphological decoding than to broader word reading skill. This finding converges with the concept of larger (and more easily detectable) effects on the “near-transfer” task of morphological decoding than on the “far-transfer” task of broader word reading.
Theoretical and educational implications
Our results specify that morphological awareness plays a critical role in word reading development, one that is strongest for the reading of morphologically complex words in the mid-elementary school years. We think that this stronger role emerged because readers’ sensitivity to the structure of words in the oral language enables them to segment morphologically complex words into smaller morpheme constituents (Carlisle, Reference Carlisle2003; Nagy et al., Reference Nagy, Carlisle and Goodwin2014; Sénéchal & Kearnan, Reference Sénéchal, Kearnan and Kail2007), which then facilitates their ability to use written morphemes as units to support their reading (Carlisle, Reference Carlisle2010; Kuo & Anderson, Reference Kuo and Anderson2006; Rastle, Reference Rastle2018). These interpretations are consistent with the recent Morphological Pathways Framework (Levesque et al., Reference Levesque, Breadmore and Deacon2021), which explicitly articulates a role for morphological awareness in reading multimorphemic words. These ideas also likely reflect in part the morphophonemic nature of the English writing system (e.g., Venezsky, Reference Venezky1967); both these components need to be fully integrated into reading theory and instruction. Certainly, it is possible that these effects spill over to broader word reading, particularly over time; to this point, this possibility was tested in Levesque et al.’s (Reference Levesque, Kieffer and Deacon2017) cross-sectional study but truly requires three time point longitudinal modeling. That said, between Grades 3 and 4, the unique effects of morphological awareness are stronger for morphological decoding than on word reading processes broadly defined (Perfetti & Stafura, Reference Perfetti and Stafura2014).
Our finding that morphological decoding is distinguishable from other aspects of word reading has a further implication for theory. We found that a two-factor model—one that distinguished morphological decoding from broader word reading skill—fit the data better than a model with a unidimensional word reading model, a difference that was significant at Grade 3 and marginal at Grade 4. These results point to morphological decoding as empirically separable from measures of word reading efficiency, as has emerged in a few prior studies (see also Nunes et al., Reference Nunes, Bryant and Barros2012). These findings run counter to the predictions of phase theory (Ehri, Reference Ehri2014) and other conceptualizations (e.g., Grainger & Ziegler, Reference Grainger and Ziegler2011) which make no distinction in the utility of morphemes units (e.g., dis- and -able) from other recurring grapho-syllabic units (e.g., -ight and -ump) in supporting word reading. This prediction has long been surprising to us and others (e.g., Kirby & Bowers, Reference Kirby, Bowers, Cain, Compton and Parrila2017) given that morphemes are semantically and syntactically rich, features absent in other letter patterns. Morphemes in print are meaningful orthographic units that bring a considerable degree of consistency to opaque orthographies such as English (Rastle, Reference Rastle2018). Our findings and those of others (e.g., Nunes et al., Reference Nunes, Bryant and Barros2012) suggest that the use of morphemes in reading is a distinct dimension in children’s word reading. The priority of morphemes over other letter patterns merits inclusion in reading theory, and it also has strong implications for the development of reading instruction.
Our finding of unique effects of morphological awareness on gains in morphological decoding leads us to speculate that explicit instruction in morphological awareness is likely to have its strongest effects on students’ skill in morphological decoding in the mid-elementary school years. This idea has some empirical support from existing intervention studies (Bowers et al., Reference Bowers, Kirby and Deacon2010; Goodwin & Ahn, Reference Goodwin and Ahn2013). Such intervention studies have included activities whereby students create banks of word families that share the same morphemes (roots and affixes) or break apart larger complex words into their component morphemes. Other “construction” activities could be created that teach children how to combine roots and affixes together into morphological complex words (e.g., Casalis & Colé, Reference Casalis and Colé2009). Perhaps more speculatively, we think that targeted instruction in morphological decoding would be appropriate given its separability from children’s broader word reading skill, at least during the mid-elementary grades. Such instruction in morphological decoding has been integrated into a few reading programs. The PHAST program (Lovett et al., Reference Lovett, Lacerenza and Borden2000), for instance, includes targeted instruction on identifying common, productive affixes such as un- and -less and “peeling them off” (p. 464) words, thereby supporting reading of complex words through these component parts (see also Gaskins et al., Reference Gaskins, Downer, Anderson, Cunningham and Gaskins1988; Gaskins, Downer, & Gaskins, Reference Gaskins, Downer and Gaskins1986). Building on these ideas, children could also be taught that the pronunciation of some letter patterns, such as -ive, is based on their status as morphemes (e.g., detective and arrive, respectively). Determining the relative emphasis on rime versus morpheme patterns will be critical, though there is some precedent for including both in instruction (e.g., Lovett et al., Reference Lovett, Lacerenza and Borden2000). All these suggestions await further empirical testing, including the extent to which such instruction needs to include other aspects of literacy instruction (Bowers et al., Reference Bowers, Kirby and Deacon2010).
Limitations and future directions
An important limitation of our work lies in measurement. It is remarkably challenging to identify comparable metrics of morphological decoding and broader word reading skill. In our work, we used some of the few available measures of morphological decoding (Levesque et al., Reference Levesque, Kieffer and Deacon2017; Nunes et al., Reference Nunes, Bryant and Barros2012), along with, in keeping with prior studies (e.g., Kirby et al., Reference Kirby, Deacon, Bowers, Izenberg, Wade-Woolley and Parrila2012), an established standardized measure (TOWRE; Torgesen et al., Reference Torgesen, Wagner and Rashotte1999) of broader word reading skill. These measures were reasonably comparable; both were age-appropriate, had good reliability, and included real words and nonwords. This relative similarity means that we need to take seriously findings of unique effects of morphological awareness on morphological decoding and not on broader word reading. That said, other differences need to be acknowledged. One is that our measures of morphological decoding were based on accuracy, whereas the TOWRE evaluates both accuracy and speed. Slight differences in the processes involved in timed word reading tests might explain, at least in part, why morphological awareness accounted for unique gains in morphological decoding but not broader word reading skill. And yet, we think that this explanation is unlikely because the design was in fact weighted toward finding effects for broader word reading skill; TOWRE has higher reliability and includes some morphologically complex words and captures morphological decoding to some extent. That said, our measures of morphological decoding focused predominantly on complex words with accessible morphological structures (e.g., high-frequency base words, common suffixes and prefixes). This may have fortified children’s ability to exploit their morphological awareness in decoding these multimorphemic words. Further still, in the absence of an available parallel form, we used the same tests of morphological decoding in Grades 3 and 4 (see e.g., Bowers et al., Reference Bowers, Kirby and Deacon2010) and alternate forms for TOWRE in Grades 3 and 4. Any of these differences might account, at least in part, for different results between morphological decoding and broader word reading development. Clearly, there are several ways in which task development remains important for future research.
As with all studies, our findings are specific to the age range and reading level in our study. The link between morphological awareness and morphological decoding might be particularly strong in Grades 3 and 4, when children’s reading strategies become increasingly attuned to consolidated units like morphemes (Ehri, Reference Ehri2014) and when there is a steep increase in the morphological complexity of texts (Anglin, Reference Anglin1993; Nagy & Anderson, Reference Nagy and Anderson1984; White et al., Reference White, Power and White1989). Earlier in development, effects of morphological awareness might be detected on children’s broader word reading skill because children encounter fewer morphologically complex words, both in texts and in standardized tests. Certainly, any influence of morphological awareness to word reading development prior to Grade 3 is likely subsumed in our findings, and thus, it is impossible to untangle prior developmental effects in this study of children in Grades 3 to 4. Beyond developmental differences, there might be different findings across reading levels. Recent findings point to the possibility that reading ability may moderate the impacts of morphological awareness on literacy skills (e.g., see Gilbert et al., Reference Gilbert, Goodwin, Compton and Kearns2013; Goodwin et al., Reference Goodwin, Huggins, Carlo, August and Calderon2013; Kearns, Reference Kearns2015). Kearns and colleagues (Reference Kearns, Steacy, Compton, Gilbert, Goodwin, Cho, Lindstrom and Collins2016), for instance, reported a stronger effect of morphological awareness on morphological decoding for children with early- rather than late-emerging reading difficulties (see also Beyersmann et al., Reference Beyersmann, Grainger, Casalis and Ziegler2015). This finding converges with earlier suggestions of stronger effects of morphological interventions for poorer readers (e.g., Bowers et al., Reference Bowers, Kirby and Deacon2010). Clearly exploring moderation by reading and developmental level are important steps forward.
Further still, we interpret our findings specifically within the language studied here: English. It is not clear that similar effects would emerge in more phonologically transparent orthographies. Evidence of relations between morphological awareness and broader word reading have emerged in studies of children learning to read in languages represented with more phonologically transparent orthographies such as Portuguese and Greek (de Freitas et al., Reference de Freitas, Mota and Deacon2018; Rothou & Padeliadu, Reference Rothou and Padeliadu2015). These studies have not, to our knowledge, tested links of morphological awareness to morphological decoding, though children learning to read in more transparent languages are clearly capable of morphological decoding (e.g., Angelelli et al., Reference Angelelli, Marinelli and Burani2014). Future research needs to test the contribution of morphological awareness to gains in morphological decoding for children learning to read in different languages.
In conclusion, our findings clarify the role of morphological awareness in children’s reading development, unique effects that are strongest in supporting gains in children’s morphological decoding over time. Given the prevalence of morphologically complex words in the texts that elementary children read, we think that such effects are educationally important and theoretically relevant. We hope that these findings inspire detailed inclusion of morphology in models of word reading development as well as in comprehensive programs of literacy instruction.
Availability of data and material
Available upon request.
Declarations
Financial Support
This research was supported by the Social Sciences and Humanities Research Council of Canada (grant 435- 2012- 0630) to S. Hélène Deacon and a Joseph-Armand Bombardier Canada Graduate Scholarship to Kyle C. Levesque.
Conflict of interest
The authors report no conflicts of interest.
Ethical approval (include appropriate approvals or waivers)
This research was approved by the Social Sciences & Humanities Research Ethics Board at Dalhousie University, REB #: 2012-2861.
Appendix
Morphological decoding task: Morphologically Derived Word Reading
Note. U = frequency of occurrence per million words in text; U frequency values obtained from Zeno et al. (Reference Zeno, Ivens, Millard and Duvvuri1995). MorphoLexSegm = morphological segmentation of the lexical item, where each morpheme is represented by its canonical form (Sánchez-Gutiérrez et al., Reference Sánchez-Gutiérrez, Mailhot, Deacon and Wilson2017). Nmorph = number of morphemes contained in the morphological segmentation of the lexical item (Sánchez-Gutiérrez et al., Reference Sánchez-Gutiérrez, Mailhot, Deacon and Wilson2017). Transp = transparency between base morpheme and derived word: T = transparent. P = phonological change. O = orthographic change.
Open access
