Intelligence 53 (2015) 108–117
Contents lists available at ScienceDirect
Intelligence
Structure of phonological ability at age four Ulrika Wolff ⁎, Jan-Eric Gustafsson Department of Education and Special Education, University of Gothenburg, Sweden
a r t i c l e
i n f o
Article history: Received 20 July 2015 Received in revised form 20 September 2015 Accepted 22 September 2015 Available online 30 October 2015 Keywords: Phonological awareness Gf Cognitive ability Dimensions of phonology
a b s t r a c t Three research questions were investigated: 1) can phonological tasks be characterized in terms of two facets, a processing complexity facet and a linguistic complexity facet, 2) can performance on the phonological tasks be subsumed under one general phonological awareness factor, and 3) how are phonological abilities and other cognitive abilities related? A test battery of tasks with a complete crossing of the two phonological awareness facets was given to 364 children aged four, along with a test battery with tasks measuring cognitive abilities (Gf, Visual, Verbal). A “correlated trait-correlated method minus one” measurement model fit the data well, and supported the proposed model with two dimensions of phonological awareness. Further analyses demonstrated that the phonological tasks can be subsumed under a dominating general phonological awareness factor and orthogonal narrow factors representing the processing and linguistic complexity factors. A structural equation model in which the linguistic and processing complexity factors were regressed on Gf showed that Gf was highly related to the processing complexity factors but not to the linguistic factors. This suggests that the well-established impact of Gf on reading skills is mediated through phonological awareness. The Verbal factor was related to the linguistic complexity factors when Gf was partialed out but not to the processing complexity factors. The Visual factor did not relate to any of the phonological factors. These results support the view that the Verbal and Visual factors are modality specific, and do not represent general processing capacity. © 2015 Elsevier Inc. All rights reserved.
Phonological awareness can be defined as the ability to identify and manipulate speech sounds. Acquisition of phonological awareness implies that a child moves from implicit to explicit control of the sound structure of language, and this explicit control, or awareness, is critical when a child learns to understand and handle the alphabetic principle in both transparent and deep orthographies (e.g. Bradley & Bryant, 1985; Caravolas, Lervåg, Defior, Málková, & Hulme, 2013; Lundberg, Larsman, & Strid, 2010; Papadopoulos, Kendeou, & Spanoudis, 2012). However, even though considerable progress has been made in understanding the nature of phonological awareness and how to assess it among children at different age levels, several questions of a fundamental nature still remain unanswered. One of these is whether phonological awareness is unitary or whether it should be conceived of as a multidimensional phenomenon. A compromise position between these two positions is that phonological awareness is a single cognitive ability that is demonstrated in a variety of skills throughout development (Anthony & Francis, 2005). Thus, even though measures of phonological awareness typically are multidimensional, an underlying unitary construct may be assumed. However, our understanding both of the nature of the underlying construct and of its behavioral manifestations need to be further developed. Another fundamental question is how individual differences in phonological awareness relate to individual differences in other ⁎ Corresponding author at: University of Gothenburg, Department of Education and Special Education, Box 300, 405 30 Gothenburg, Sweden. E-mail address:
[email protected] (U. Wolff).
http://dx.doi.org/10.1016/j.intell.2015.09.003 0160-2896/© 2015 Elsevier Inc. All rights reserved.
cognitive abilities. Much of the research on the structure of phonological awareness has been restricted to deal with phonological tasks alone but there is reason to believe that phonological awareness is systematically related to other abilities to deal with complex information (Gustafsson & Wolff, 2015), so there is a need to investigate which overlap there is. The main aim of the current study was to approach these two fundamental questions, through an empirical study of individual differences in phonological awareness and other cognitive abilities among fouryear-old children. 1. The definition and dimensionality of phonological awareness There is no generally established consensus on how to measure phonological awareness. Tests of phonological awareness vary in terms of the size of the phonological units to be manipulated, the complexity of processing, and the degree of explicit awareness required. Examples of phonological tasks include judgments of rhyme, blending of phonological elements, deletion of phonological segments, phoneme and syllable counting, and judgments of shared phonemes in sequences of words (see, e.g., Alloway, Gathercole, Willis, & Adams, 2004). Anthony and Lonigan (2004) observed that some definitions of phonological awareness are highly inclusive of which different types of phonological skills are indicative of phonological awareness, while others are stricter. The strictest definition only includes tasks that involve manipulation of phonemes. A less strict definition includes identification or manipulation of all sub-syllabic skills, such as onsets,
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
rimes and phonemes. An even more inclusive definition was proposed by Stanovich (1992), who argued that the notion of conscious awareness should not be a definitional requirement. He instead proposed the term “phonological sensitivity,” described in terms of a continuum from a “shallow” sensitivity of large phonological units to a “deep” sensitivity of small phonological units (Stanovich, 1992, p. 317). This definition thus includes phonological skills involving any word unit, and it implies that phonological sensitivity can be seen as a developmental continuum, ranging from abilities to detect large phonological units such as words, and syllables, and ability to manipulate smaller units such as phonemes. This developmental conceptualization of phonological sensitivity implies that children's early developed phonological skills form the basis for more advanced phonological skills, while at the same time they reflect the same underlying ability (Anthony & Lonigan, 2004). The different conceptualizations of phonological awareness described above agree that there are multiple phonological skills that are distinguished by linguistic complexity and type of operation performed. The basic issue of disagreement is whether the different types of phonological skills belong to the same construct or whether they represent distinct abilities. Anthony and Lonigan (2004) applied confirmatory factor analysis in reanalyses of data from four independent studies of the dimensionality of phonological measures. The age of the children participating in the four studies varied between four and seven years. The main finding across the four studies was that models with a single phonological sensitivity variable fit the data for the younger children. For the older children a rhyme sensitivity factor was distinguishable, although highly correlated with a latent variable representing segmental awareness. Anthony and Lonigan (2004) interpreted these results as being consistent with the conceptualization of phonological sensitivity as “… a single ability that can be measured by a variety of tasks (e.g., detection, blending, and elision) that differ in linguistic complexity (e.g., syllables, rimes, onsets, and phonemes)” (p. 51). Papadopoulos et al. (2012) investigated the hypothesis that a single underlying dimension of phonological sensitivity can be identified also among Greek-speaking children. Greek is a language with transparent orthography, while English has a deep orthography. In transparent languages, the mapping of graphemes onto phonemes is relatively unambiguous, and in such languages phonological recoding operates at a smaller grain size (e.g., phoneme) than in nontransparent languages where phonological recoding operates at a larger grain size (e.g., syllable) (Ziegler & Goswami, 2005). Papadopoulos et al. (2012) investigated longitudinal data with three waves of data collection between Kindergarten and Grade 2. At each wave the children were tested on a set of ten phonological tasks. Six of the tasks measured phonological ability at the syllabic level, and four tasks tapped phonological ability at the phonemic level. Several theory-driven models were fitted to the data from each wave, and the favored model was a nested-factor model with a general phonological ability factor and two orthogonal residual factors, one representing supra-phonemic skills and one representing phonemic sensitivity skills. The general phonological ability factor accounted for the largest portion of variance for the vast majority of the subtests in all three waves. It also was demonstrated that only the general phonological factor contributed to prediction of a measure of word reading fluency. On the basis of these results, Papadopoulos et al. (2012) concluded that phonological ability is not to be seen as a set of discrete skills, but rather that it should be conceptualized as a unitary ability to deal with phonological tasks varying in linguistic complexity and the load of cognitive processing involved. They also observed that these findings obtained in a language with transparent orthography strengthen existing evidence for the universality of the phonological sensitivity construct as unitary in both transparent and nontransparent orthographies (see also Schatschneider, Francis, Foorman, Fletcher, & Mehta, 1999; Stahl & Murray, 1994; Wagner et al., 1997).
109
Anthony et al. (2011) examined the dimensionality of phonological awareness in the transparent orthography Spanish. They tested 1265 children, aged three to seven, on a set of phonological tasks on three linguistic complexity levels (word, syllable and phoneme) with two types of elision tasks and two types of blending tasks, each with two response formats (multiple-choice and free response). They found little impact of linguistic complexity on the difficulty of test items, whereas effects of task complexity were more evident, multiple-choice blending tasks being the easiest and free response elision tasks being the most difficult. The Spanish speaking children's phonological awareness performances were largely explained by a single, second-order phonological ability but there also were effects of task characteristics. However, several studies have found empirical support for the notion that phonological awareness should be conceived in terms of multiple distinct phonological abilities. For example, Høien, Lundberg, Stanovich, and Bjaalid (1995) examined one group of preschoolers and one group of first graders on phonological skills. They found distinct factors representing syllables, rhymes and phoneme awareness, where phoneme awareness was by far the most powerful predictor of word reading. Similar results have been reported by Muter, Hulme, Snowling, and Taylor (1997), and by Yopp (1988). Typically, studies that investigate different aspects of phonological sensitivity do so by examining tasks with different levels of linguistic complexity. However, phonological tasks also differ in terms of the complexity of the cognitive processing required and this aspect is often not explicitly recognized. For example, Wagner, Torgesen, and Rashotte (1994) identified five latent phonological variables. In addition to two naming variables and one phonological working memory variable, there were two phonological processing variables that differed in cognitive complexity, namely phonological analysis and phonological synthesis. These two variables included a mix of phoneme and rhyme awareness tasks. Another example is provided by Bryant, MacLean, Bradley, and Crossland (1990) who reported support for models according to which rhyme awareness leads to phoneme awareness. However, their rhyme tasks involved identification, whereas the phoneme tasks involved manipulation and segmentation. Given that it is more difficult to manipulate phonological units than to identify them (e.g. Høien et al., 1995; Lonigan et al., 2009), this measurement confounds linguistic complexity and task processing complexity. The fact that phonological tasks involve both a facet of linguistic complexity and a facet of processing complexity suggests that both these facets need to be explicitly recognized in the construction and interpretation of phonological tasks. These two facets are also recognized in theoretical models of development of phonological awareness that identify two patterns of development (Anthony & Francis, 2005; Anthony et al., 2011; Ziegler & Goswami, 2005). One of these is that children become increasingly sensitive to smaller and smaller linguistic units as they grow older, as is elaborated in the psycholinguistic grain size theory of the development of phonological awareness (Ziegler & Goswami, 2005). Thus, children detect words before they detect syllables, and they detect syllables before they detect phonemes. Development thus proceeds along a dimension of linguistic complexity which follows a hierarchical model of word structure (Goswami & Bryant, 1990; Ziegler & Goswami, 2005). The other pattern of development is that children first identify similar sounding word elements, then they blend and segment phonological information, and after that they manipulate phonological units by deleting, isolating, and reversing them. This pattern of development thus is along a dimension of processing complexity, and with a higher level of development children can perform increasingly complex operations and an increasing number of operations on phonological units (Anthony, Lonigan, Driscoll, Phillips, & Burgess, 2003). Development along the processing complexity dimension is likely to be highly related to development of working memory capacity (Anthony et al., 2003; Anthony et al., 2011).
110
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
We therefore propose a model where these two dimensions are taken into account simultaneously as two facets in the construction of phonological assessment tasks: the linguistic complexity facet (morphemes, syllables, phonemes), and the processing complexity facet (identification, segmentation/blending, manipulation). In the empirical study a test battery of phonological tasks with a complete crossing of these two dimensions (Wolff, 2013) was administered to a sample of children aged 4.
2. Phonological abilities and other cognitive abilities General fluid intelligence (Gf) is one of the central constructs in the field of intelligence. Gf is interpreted as the capacity to solve novel, complex problems, using operations such as inductive and deductive reasoning, concept formation, and classification. Cattell (1987) hypothesized the Gf-factor to be a causal factor influencing individual differences in development of knowledge and skills, and he also hypothesized Gf to be strongly related to g, which has obtained support (e.g.Gustafsson, 1984; Valentin Kvist & Gustafsson, 2008). Several studies also have shown that there are strong relations between Gf on the one hand and working memory and other executive abilities on the other hand (e.g., Blair, 2006). General cognitive ability has been shown to be the single most important individual difference variable predicting development of reading skills (Bowey, 2005), but little is known about the mechanisms behind this relationship. One hypothesis that would provide a partial explanation of the relationship between Gf and acquisition of knowledge and skills is that Gf is highly related to phonological awareness, which in turn is related to acquisition of reading skills, which in turn are related to acquisition of knowledge in many different fields. The hypothesis of a high relation between cognitive abilities and phonological awareness is supported by some previous research. McBride-Chang (1995) demonstrated among children in grades 3–4 that general cognitive ability, speech perception and verbal short term memory could account for 60% of the variance in a phonological awareness construct. In a structural model of working memory in a sample of children aged four to six, Alloway et al. (2004) found a strong relation of .88 between a latent variable representing central executive functions and a latent variable representing phonological awareness. De Jong and van der Leij (1999) found in a longitudinal study from kindergarten to second grade that both phonological abilities and Gf were related to early acquisition of reading skills. They also showed that the relation between phonological abilities and early reading skills decreased when Gf was controlled for and that the direct influence of Gf on reading skills diminished over time. Such an early, but successively diminishing influence of Gf on the acquisition of reading skills supports the hypothesis that the influence of Gf is mediated via the development of phonological awareness. A strong theoretical argument for why Gf would be related to phonological awareness is that Gf may be expected to be highly related to the processing complexity dimension of phonological development, which we identified above. We therefore hypothesize that among children at the age of four the relationship between Gf on the one hand and phonological awareness on the other hand is strong. However, even among children as young as four years of age there is differentiation of cognitive abilities. Gustafsson and Wolff (2015) identified three cognitive ability dimensions by fitting a bifactor model to a battery of 10 tests. One factor was a general dimension, interpreted as Gf, which was particularly highly related to complex working-memory tests, such as block tapping and comprehension of instructions. A second factor, nested under Gf, was labeled Visual, and it was related to problem solving tests with visual content, such as matrices and block design tasks. The third factor was labeled Verbal, and it was related to word and sentence span tasks. These three factors carry theoretical interest in relation to the hypothesized phonological ability dimensions.
One possible interpretation is that the two residual factors in the bifactor model represent capacity for memory representations in different modalities (e.g., Engle, Tuholski, Laughlin, & Conway, 1999), and another interpretation is that they represent processing capacity of different types of information (e.g., Colom, Rebollo, Abad, & Shih, 2006). Given that the problem-solving tests to which the Visual factor was related in much research have been taken to be the primary markers of Gf it may be asked to what extent the Visual factor is a general processing capacity factor or if it primarily is modality specific. If it represents more general processing capacity it may be expected to be related to performance on phonological tasks requiring complex processing, but if it is modality specific no such relations are expected. Similarly, the span tests loading on the Verbal factor have been taken both to measure general processing capacity and to represent verbal short-term memory capacity. If the Verbal factor represents general processing capacity it too may be expected to be related to performance on phonological tasks requiring complex processing, but if it is modality specific it may rather be expected to be related to the linguistic complexity dimension of phonological performance. The empirical study designed to investigate these questions involved the same 364 four-year-old children that took part in the study by Gustafsson and Wolff (2015). In addition to the battery of 10 cognitive tests the children were administered a comprehensive battery of tasks to measure phonological awareness. 3. Research questions Based on the review above, the study investigated three research questions: 1) Is it possible to describe phonological awareness tasks in terms of facets relating to linguistic complexity and processing complexity, and to measure phonological awareness in terms of factors representing these two facets? 2) Can the different phonological awareness factors be subsumed under a general phonological factor? 3) What are the relations between the three dimensions of the structural model of cognitive abilities (Gf, Visual and Verbal) and the phonological awareness factors at age four? 3.1. Method A group of typically developing preschool children was included in the study. The children were tested on a wide range of phonological and other cognitive tasks when they were four years old. 3.2. Participants The participating children (N = 364) were recruited from 58 different preschools in eight municipalities in Sweden. The children were between 3 years, 10 months old and 4 years, 4 months old with a mean age of 4 years, 1 month (SD = 2.3 months), 182 girls and 182 boys. Informed consent was provided from all parents before the testing was carried out. Children with Swedish as a second language who did not speak Swedish at the age of three were excluded from the study. As to our knowledge the children included in the study were typically performing children without any formal diagnosis of autism or specific language disorder. 3.3. Instruments 3.3.1. Cognitive tasks The cognitive measures were Colored progressive matrices (Raven, Raven, & Court, 2000); Wechsler preschool and primary scale of intelligence III, block design (Wechsler, 1991); Wechsler nonverbal scale of ability, matrices and recognition (Wechsler & Naglieri, 2006); Corsi block tapping (using the standardization of Farrell Pagulayan, Busch, Medina, Bartok & Krikorian, 2007); Verbal working memory (Wolff, 2013), Nepsy, comprehension of
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
instruction (Korkman, Kirk, & Kemp, 1998); Word span, backwards and forward (Thorell & Whålstedt, 2006); and Sentence memory, WIPPSI-R (Wechsler, 1991). The tests for cognitive abilities are not presented in detail here, as alternative measurement models for these tests were investigated in another study (for a detailed description see Gustafsson & Wolff, 2015). 3.3.2. Phonology tasks The tasks measuring phonological skills were designed to reflect two facets of phonological awareness (Wolff, 2013): The first facet is linguistic complexity, represented in increasing order of complexity by morphemes, syllables/rhyme, and phonemes. Compound words are frequently used in the Swedish orthography, and the morphemes used here are complete words that can be independently used but in this context they are parts of compound words (e.g. snow in snowball). The Swedish language is to some extent inflected, resulting in many long words, and there also are many consonant clusters, with a maximum of five consonants in one cluster. Therefore, it is more appropriate to work with syllables than onset rimes (cf. Anthony et al., 2011), which often is the case in studies with English speaking participants. Explicit awareness of phonemes is an important indicator of phonological awareness in Swedish (Lundberg, 2007; Lundberg et al., 2010), as in many other orthographies (Melby-Lervåg, Lyster, & Hulme, 2012). The second facet is the processing complexity facet, represented in increasing order of complexity by identification, blending/segmentation, and manipulation. Identification tasks require the child to identify sound units in words, blending/segmentation tasks include both blending and segmentation of sound units, and the manipulation tasks ask the child to manipulate speech sounds. The linguistic and processing facets are completely crossed, resulting in nine different types of phonological tasks, each comprising nine items. The test is organized into three blocks of tasks, starting with identification tasks. The tasks are described below. 3.3.2.1. Identification. The first task is on the morpheme level. The child is asked to point at the two pictures among three presented ones that represent compound words that begin with the same morpheme, in this case a complete word. An example in English would be pictures representing the words snowball and snowman with the distractor picture representing the word football. At the syllable level, the child is asked to judge whether two words rhyme or not. At the phoneme level, one task is to point at the picture among three presented ones that represents a word that begins with the same sound as sun (i.e./s/). There is one task (each task includes nine items) on the morpheme level, one task on the syllable level and three tasks on the phoneme level. In all three blocks, the children received two practice items and corrective feedback before each task of nine items. Corrective feedback was not provided on test items. When a child makes three subsequent errors the testing is interrupted and moves on to the next block of tasks, Blending/segmentation. 3.3.2.2. Blending/segmentation. At the morpheme level an example task is to tell which word you get if you have the word butter and you add the word fly, or vice versa, what you get if you have the word butterfly and separate it into two words. At the syllable level, one task is to listen to syllables presented in isolation, e.g. win-ter, and blend them together to a word, and another task is to indicate syllables in a word by using markers, e.g. summer. Similarly, at the phoneme level the task is to blend and segment phonemes. One example is to listen to d-o-g, and blend the phonemes together, another task is to segment the word boy into phonemes and indicate them by using markers. The first task in Blending/segmentation is on the morpheme level, the next two are on the syllable level and the last three are on the phoneme level. When the child makes three subsequent errors the testing is interrupted and moves on to the next block of tasks, which is Manipulation.
111
3.3.2.3. Manipulation. At the morpheme level, one example of a task is to delete the first part (morpheme) of the compound word doorstep, i.e. door, and indicate what word it makes by pointing to one of three pictures. The next morpheme task is similar, but a vocal response is required. At the syllable level, a word is presented orally, e.g. crocodile, and the task is to say which word is left if you take the syllable cro away. At the phoneme level an example of a task is to say the word mat without the sound/m/. The tasks increase in difficulty in the way that the correct responses are not real words. The two first tasks are on the morpheme level, the next two are on the syllable level and the last two are on the phoneme level. When a child makes three subsequent errors the testing is interrupted. 4. Test procedures The children were tested individually at their pre-schools. The total testing time was around four hours, with sessions spread over four days. Each session included a short break with a small snack. The test leaders were special needs education teachers at the child's preschool, or a special needs education teacher who was affiliated to the student health team in the school district. All test leaders received training from the research group before the testing. 5. Analytic procedures Hypotheses about the dimensionality of phonological awareness, and hypotheses about relations between cognitive and phonological abilities were examined with latent variable modeling techniques. The main analytic methods were confirmatory factor analysis (CFA) and structural equation modeling (SEM), conducted with the Mplus 7 program (Muthén & Muthén, 2012) under the STREAMS modeling environment (Gustafsson & Stahl, 2005). The two-facet structure of the phonological tasks, with one processing complexity facet with three levels and one linguistic complexity facet with three levels, was represented in a model inspired by the multitrait-multimethod (MTMM) approach (Brown, 2015). Each of the levels of the two facets was taken to correspond to a latent variable, and the latent variables under each facet were allowed to correlate. Each of the 17 phonological tasks was hypothesized to load on one of the processing factors and on one of the linguistic factors. However, given that MTMM models with correlated trait and correlated method factors typically suffer from empirical under-identification (Brown, 2015; Kenny & Kashy, 1992), we applied the correlated trait-correlated method minus one model (CT-C[M-1]; Eid, Lischetzke, Nussbeck, & Trierweiler, 2003; Eid et al., 2008) for estimation purposes. This model
Fig. 1. Hypothesized CT-CM[M-1] model for how the processing factors Identification (Id), Blending/segmentation (B/S) and Manipulation (Man), and the linguistic factors Syllables (syll) and Phonemes (phon) relate to the observed phonological tasks.
112
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
than .1 (Brown & Cudeck, 1993). WRMR should be less than 1.00, and CFI should be higher than .95 (Hu & Bentler, 1999). 5.1. Results First, descriptive data for the phonology tests are presented, followed by the measurement model for phonological awareness. The unity of phonological awareness was then investigated, and a series of higherorder CFA models with general phonological awareness factors was fitted. A structural model investigating relations between the cognitive and phonological abilities is finally presented. Table 1 presents descriptive statistics and reliability information for the 17 tasks in the phonology test. The means of the tasks support the hypothesized complexity levels. Given that the maximum score on each task was 9 it is obvious that many of the tasks, and particularly those involving manipulation and phonemes, were difficult for the four-year old children. However, by treating the observed variables as categorical, estimation problems were avoided. Fig. 2. Hypothesized CT-CM[M-1] model for how the processing factors Blending/ segmentation (B/S) and Manipulation (Man), and the linguistic factors Morphemes (Morph), Syllables (Syll) and Phonemes (Phon) relate to the observed phonological tasks.
implies that one of the hypothesized factors is not included in the model, but is instead taken to be a reference factor. The model thus hypothesized three correlated processing factors (Identification, Blending/segmentation, and Manipulation) and three correlated linguistic factors (Morphemes, Syllables, and Phonemes). However, following the CT-C[M-1] approach one factor was excluded and taken to be a reference factor for the other two factors. The least complex factor was excluded from each complexity facet in the two different models. The hypothesized model with the linguistic Morphemes factor excluded is presented in Fig. 1, and the hypothesized model with the processing Identification factor excluded is presented in Fig. 2. In order to take the cluster sampling into account the option ‘Complex’ in Mplus was used. The observed variables were taken to be categorical, and least squares estimation with mean- and varianceadjusted chi-square test statistics (WLSMV) was used. For tests of model fit χ2, Root Mean Square of Approximation (RMSEA) with 90% confidence intervals, Comparative Fit Index (CFI), and weighted root mean square residual (WRMR) are reported. To indicate good fit, the RMSEA estimate, and the upper range of its 90% confidence interval, should be lower than .07 (Steiger, 2007) or about .08 but not greater
Table 1 Means, standard deviations and Cronbach's alpha of the phonology tests. Test
Mean
SD
Alpha
Id_morph Id_syll Id_phon1 Id_phon2 Id_phon3 B/S_morph B/S_syll1 B/S_syll2 B/S_phon1 B/S_phon2 B/S_phon3 Man_morph1 Man_morph2 Man_syll1 Man_syll2 Man_phon1 Man_phon2
5.03 5.18 1.97 1.02 0.62 2.69 1.40 0.50 0.22 0.06 0.04 2.55 1.10 0.28 0.10 0.12 0.06
2.64 2.53 2.58 2.14 1.64 3.55 2.48 1.72 1.16 0.50 0.4 2.81 2.54 1.21 0.78 0.97 0.60
.82 .80 .87 .91 .89 .96 .92 .95 .94 .83 .89 .88 .96 .92 .93 .98 .93
Note. Test labels are constructed through combinations of category of processing task (identification, blending/segmentation, manipulation) and linguistic category (morpheme, syllable, phoneme).
6. Linguistic and processing complexity in phonological awareness tasks Before testing our hypothesis about two facets of phonology we examined if phonology may be a one-dimensional concept. Thus, all observed variables were related to one phonological factor. However, the model showed very poor fit to the data (χ2 = 835.17, df = 118, p b .00; RMSEA = .130, CI90 = .122–.138; CFI = .922; WRMR = 2.023). The hypothesized correlated trait-correlated method minus one model with the Morphemes factor excluded had acceptable fit (χ2 = 235.15, df = 102, p b .00; RMSEA = .060, CI90 = .050–.070; CFI = .986; WRMR = .895), and the factor loadings were generally high and statistically significant. In general, the loadings were lower on the linguistic factors than on the processing factors, showing that the processing factors accounted for a larger part of the performance differences than did the linguistic factors (see Table 2). The Identification factor correlated .55 with Blending/segmentation and .64 with Manipulation. The correlation between Blending/segmentation and Manipulation was .69. The highest correlation was between the two linguistic factors Syllables and Phonemes (.79). The hypothesized correlated trait-correlated method minus one model with the Identification factor excluded had an acceptable but slightly worse fit as compared with the model with the three processing factors (χ2 = 259.47, df = 103, p b .00; RMSEA = .065, CI90 = .055–.074; CFI = .983; WRMR = .971). The loadings were generally high and statistically significant (see Table 3). The Blending/segmentation factor correlated .53 with Manipulation, and Morphemes correlated .76 with Syllables and .72 with Phonemes. Syllables and Phonemes correlated .92. It also is interesting to compare the full model with both processing and linguistic factors, with models that only include one of the set of factors. The model that included the three processing factors but not the linguistic factors had a marginally acceptable fit (χ2 = 383.80, df = 116, p b .00; RMSEA = .080, CI90 = .071–.089; CFI = .971; WRMR = 1.31), but which clearly was poorer than the fit of the full model according to all the criteria. The model that included the three linguistic factors but not the processing factors had a poor fit (χ2 = 654.28, df = 116, p b .00; RMSEA = .113, CI90 = .105–.121; CFI = .942; WRMR = 1.76). These results thus show that both sets of factors were needed to account for the relations among the 17 phonological tasks, but that the processing factors were more important in accounting for the relations than were the linguistic factors. 7. Is there a general phonological awareness factor? The second research question concerned the existence of a general phonological factor and it was in a first step addressed by relating a
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
113
Table 2 Standardized factor loadings for the phonological tests in the measurement model with the morphemes factor excluded.
Identification iId_morph Id_syll Id_phon1 Id_phon2 Id_phon3 Blending/segmentation B/S_morph B/S_syll1 B/S_syll2 B/S_phon1 B/S_phon2 B/S_phon3 Manipulation Man_morph1 Man_morph2 Man_syll1 Man_syll2 Man_phon1 Man_phon2
r
SE
t-value
0.75 0.60 0.75 0.79 0.77
0.039 0.047 0.038 0.043 0.055
19.29 12.74 19.92 18.17 13.96
0.96 0.90 0.80 0.70 0.78 0.53
0.88 1.00 0.84 0.71 0.68 0.96
0.032 0.029 0.041 0.068 0.100 0.055
0.013 0.013 0.033 0.045 0.035 0.048
3.05 31.08 19.46 10.25 7.82 9.68
67.92 79.61 25.27 15.89 19.52 20.00
r
SE
t-value
Syllables Id_syll S/B_syll1 S/B_syll2 Man_syll1 Man_syll2
0.25 0.34 0.35 0.54 0.74
0.064 0.065 0.081 0.040 0.052
3.81 5.22 4.26 13.49 14.24
Phonemes Id_phon1 Id_phon2 Id_phon3 B/S_phon1 B/S_phon2 B/S_phon3 Man_phon1 Man_phon2
0.41 0.54 0.53 0.66 0.59 0.42 0.69 0.45
0.055 0.064 0.068 0.071 0.108 0.077 0.069 0.077
7.40 8.45 7.76 9.39 5.46 5.36 9.77 5.90
Note. Test labels are constructed through combinations of category of processing task (identification, blending/segmentation, manipulation) and linguistic category (morpheme, syllable, phoneme).
second-order phonological latent variable to the three processing factors. The higher-order model part of this model is just-identified, so the model had the same fit as the model shown in Fig. 1. However, the estimated loadings strongly supported the hypothesis of a general factor (Identification .71, Blending/Segmentation .77, and Manipulation .90). It may be noted, however, that this second-order factor was defined by first-order processing factors alone (first model in Fig. 3), and attempts to relate also the first-order linguistic factors to the general factor failed because of model non-convergence. By imposing equality constraints on factor loadings, convergence could be achieved, but loadings of the linguistic factors on the general factor were low and insignificant. It may thus be concluded that it is possible to define a second-order factor that represents the processing complexity facet. In a second step we wanted to examine if it was possible to estimate a general phonological latent variable related to the three linguistic factors. To do this the linguistic factor Morphemes was included in the model, and the processing factor Identification was excluded from the
model. In this model too, the higher-order part is just identified, and the model had the same fit as the model shown in Fig. 2. The factor loadings were .73 for Morphemes, .98 for Syllables, and .93 for Phonemes, which provides support for a general linguistic factor. Here too it proved impossible to extend this second-order factor with processing factors, so it may be concluded that it is possible to define another secondorder factor which represents the linguistic complexity facet. The two models are depicted in Fig. 3. However, with the modeling approach used here it was not possible to estimate a general phonological factor that encompassed both the processing and the linguistic complexity facets. In the third step we therefore fitted a bifactor model that included a general phonological awareness factor on which all phonological tasks had loadings, along with four residual factors which represented two each of the processing and linguistic factors (Fig. 4). The reason for not including all of the factors was to avoid under-identification of the bifactor model. All factors were taken to be orthogonal. The fit of the model was almost equally
Table 3 Standardized factor loadings for the phonological tests in the measurement model with the identification factor excluded. r
Blending/segmentation B/S_morph B/S_syll1 B/S_syll2 B/S_phon1 B/S_phon2 B/S_phon3 Manipulation Man_morph1 Man_morph2 Man_syll1 Man_syll2 Man_phon1 Man_phon2
SE
t-value
0.75 0.79 0.66 0.57 0.65 0.54
0.048 0.042 0.054 0.056 0.094 0.059
15.46 18.76 12.93 10.17 6.95 9.26
0.66 0.73 0.61 0.61 0.62 0.59
0.047 0.053 0.039 0.042 0.030 0.058
13.97 13.70 15.66 14.31 20.60 10.25
Morphemes Id_morph B/S_morph Man_morph1 Man_morph2 Syllables Id_syll B/S_syll1 B/S_syll2 Man_syll1 Man_syll2 Phonemes Id_phon1 Id_phon2 Id_phon3 B/S_phon1 B/S_phon2 B/S_phon3 Man_phon1 Man_phon2
r
SE
t-value
0.77 0.57 0.58 0.74
0.043 0.051 0.051 0.056
17.91 11.35 11.26 13.11
0.66 0.59 0.56 0.77 0.71
0.040 0.053 0.059 0.051 0.049
16.73 11.14 9.50 14.96 14.44
0.87 0.96 0.93 0.71 0.71 0.40 0.65 0.88
0.018 0.015 0.019 0.066 0.075 0.054 0.033 0.060
48.60 64.34 49.53 10.73 9.48 7.34 19.76 14.54
Note. Test labels are constructed through combinations of category of processing task (identification, blending/segmentation, Manipulation) and linguistic category (morpheme, syllable, phoneme).
114
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
Fig. 3. Two hypothesized higher-order models with a general phonological factor (GenPhon). In the first model the processing factors Identification (Id) Blending/segmentation (B/S) and Manipulation (Man) relate to the general phonological factor, and in the second model the linguistic factors Morphemes (Morph), Syllables (Syll) and Phonemes (Phon) relate to the general phonological factor.
good as the model with correlated processing and linguistic factors shown in Fig. 1 (χ2 = 231.55, df = 95, p b .00; RMSEA = .063, CI90 = .053–.073; CFI = .985; WRMR = .875). The mean of the standardized factor loadings on the general phonological awareness factor was .74, while the mean loadings were lower for the residual factors (Blending/ segmentation .35, Manipulation .43, Morphemes .37 and Phonemes .12). These results thus demonstrate that there was a dominating general phonological awareness factor, along with a set of minor factors representing the processing and linguistic complexity facets. 8. Relations among cognitive abilities and phonological awareness The third research question concerned the relations between Gf, Verbal and Visual on the one hand and the phonological factors on the other hand. However, the bifactor model estimated by Gustafsson and Wolff (2015) did not prove useful for investigating this question, because inclusion of the phonological tests caused the general factor to tilt over into a general verbal factor. The fact that general factors tend to be flavored by the composition of the test battery is well known (see, e.g., Horn & Noll, 1997), so for the purposes of this analysis the bifactor model was transformed into an oblique three-factor model (see Fig. 5). In the bifactor model (Gustafsson & Wolff, 2015), the general factor was related to all tests, four tests (Colored progressive matrices, Block design, Wechsler matrices, Wechsler recognition) were related to Visual, and three tests to Verbal (Word span forward, Word span backward, Sentence memory). In the current oblique three-factor model
Fig. 4. A bifactor model with a general phonological factor related to all phonological awareness tasks. Four residual factors represent the processing factors: Blending/ segmentation (B/S) and Manipulation (Man), and the linguistic factors Morphemes (morph) and Phonemes (phon). Test labels of the manifest variables are constructed through combinations of category of processing task (identification, blending/segmentation, manipulation) and linguistic category (morphemes, syllables, phonemes).
Fig. 5. An oblique three-factor model of cognitive skills with the latent variables Visual, Gf and Verbal. CPM = Colored progressive matrices, Blockd = block design, WNV_MA = Wechsler matrices, WNV_RG = Wechsler recognition, CORSI = Corsi block tapping, WMW = verbal working memory, COMPINST = comprehension of instructions, STMWB = word span backward, STMWF = word span forward, SENTMEM = sentence memory.
three tests loaded on Gf (Corsi block tapping .58, Comprehension of instructions .66, and Verbal working memory .54), four tests loaded on Visual (Colored progressive matrices .50, Block design .54, Wechsler matrices .54, Wechsler recognition .53) and three tests loaded on Verbal (Word span forward .61, Word span backward .65 and Sentence memory .61). Thus, the same relations as in the bifactor model were kept for Visual and Verbal, while the three tests that only were related to the general factor in the bifactor model loaded on Gf. It may be noted that while Word span backward and Verbal working memory both can be considered to measure verbal working memory, Word span backward loaded on Verbal, and Verbal working memory loaded on Gf. One explanation may be the high correlations among the latent variables at age four. Another explanation may be that the manipulation element is more salient in the Verbal working memory task whereas the verbal element is more salient in the Word span backward task. In the present model Gf correlated .82 with both Visual and Verbal, and Visual and Verbal correlated
Fig. 6. A structural equation model of how the cognitive abilities Visual, Verbal and Gf relate to the processing factors Identification (Id) Blending/segmentation (B/S) and Manipulation (Man), and to the linguistic factors Syllables (Syll) and Phonemes (Phon). Only significant correlations between Visual and Verbal, and the phonological variables are depicted in the figure. Note: *** = p b .001.
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
.56. The model fit was good (χ2 = 55.54, df = 32, p b .006; RMSEA = .045, CI90 = .024–.064; CFI = .963; WRMR = .678) and only marginally worse than the fit of the bifactor model (Gustafsson & Wolff, 2015). The phonological factors in the model with three processing factors and two linguistic factors were first regressed on Gf. Highly significant regression coefficients were found for the processing factors (see Fig. 6), the relations being highest with Identification and somewhat lower with Blending/segmentation and Manipulation. Given that Gf represents the ability to manipulate complex relations, Manipulation might have been expected to have the highest relation with Gf and Identification the lowest. However, at age four the Blending/segmentation and Manipulation abilities are generally not developed, while there are substantial individual differences in the Identification ability. In the next step of analysis relations between the phonological factors and the Visual and Verbal factors were introduced as well. This was done by correlating the residuals in the phonological variables, after Gf was introduced, with the two ability variables. Alternative models, in which the phonological factors were regressed on Verbal and Visual, also were tried but they failed to converge. Nevertheless, the pattern of correlations with the residuals was clear-cut, there being highly significant correlations between the two linguistic factors and the Verbal cognitive factor, and no correlations at all with Visual (see Fig. 6). The fit of this structural model was good (χ2 = 521.47, df = 289, p b .00; RMSEA = .047, CI90 = .040–.053; CFI = .967; WRMR = 1.085). When the alternative model with two processing factors and three linguistic factors was estimated instead, the correlation between Morphemes and Verbal was found to be .13 (t = 1.96). The results of this analysis thus strongly support the hypothesis that the Verbal and Visual factors are modality factors rather than processing factors. The results also showed that the Verbal factor related to the linguistic phonological factors, while the Visual factor did not relate to any of the phonological factors. 8.1. Discussion and conclusions Three research questions were investigated in this study. The first question was if phonological tasks can be characterized in terms of two facets, the processing complexity facet and the linguistic complexity facet. The second question was if the phonological factors can be subsumed under one general phonological awareness factor, and the third question concerned relations among cognitive abilities and phonological abilities. To investigate the first question a large set of phonological tasks were systematically constructed to represent three levels of the processing complexity facet (Identification, Blending/segmentation, and Manipulation) and three levels of the linguistic complexity facet (Morphemes, Syllables and Phonemes). A measurement model was defined as a fully crossed “correlated trait-correlated method minus one model” with three correlated processing factors and two correlated linguistic factors. The model fit was good, and the fact that all phonological tasks were significantly related to their expected processing and linguistic factors must be interpreted as strong support for the validity of the model. We also demonstrated that a model with three linguistic factors and two processing factors had the same good fit. Thus there clearly are two facets of phonological awareness tasks, and both these facets have to be taken into account in designs of assessments and in theoretical accounts of phonology. However, the results also indicated the processing factors to be more important in accounting for the relations among the phonological tasks than were the linguistic factors. Previous research on phonological awareness has particularly recognized the linguistic facet. According to Ziegler and Goswami (2005), children acquire syllable awareness by age three to four, but they argue that they do not acquire phoneme awareness skills until they are taught to read, no matter which age. Other researchers claim that phoneme awareness is more important to take into account in assessments than syllable awareness, both in transparent (Høien et al.,
115
1995; Lundberg, 2010), and deep orthographies (Muter et al., 1997). The present results show that it is not sufficient to choose either syllables or phonemes as the only level of the linguistic complexity facet. However, the children were younger in this study than in the studies referred to, and the variance due to syllables may disappear when the children get older. But it may also be that the fully crossed two-facet approach made it possible to separate otherwise confounded effects, and thereby to identify the factors representing morphemes, syllables and phonemes. In a study with young Spanish-speaking children, Anthony et al. (2011) suggested that the tasks along the linguistic dimension could be kept constant, and that any level of word structure could be used, whereas the processing complexity should vary in assessments. Anthony et al. pointed out that for English-speaking children the opposite is true, and that one level of processing difficulty can be used as long as the word structures vary. On a continuum, Spanish is a transparent orthography, whereas English is a very opaque orthography. The language used in this study, Swedish, is much less opaque than English, but still more opaque than Spanish (Seymour, Aro, & Erskine, 2003). If the results in the Anthony et al. study can be generalized to other orthographies, an idea is that in a “semi-transparent” language both linguistic and processing complexity are important task facets, at least in ages as young as four. The second research question concerned whether phonological awareness can be subsumed under a general phonological factor. Two models were estimated with a general phonological factor related to the two sets of factors, one at a time. The models gave strong support for the assumption of broad phonological abilities subsuming either processing or linguistic abilities. The factor representing Syllables was almost identical to the broad linguistic phonological ability, while Manipulation related very strongly to the broad phonological processing factor. Given previous Swedish research (e.g. Lundberg et al., 2010), one may expect Manipulation to remain being the most powerful indicator of the broad processing factor, whereas there will be a shift towards Phonemes being the most powerful linguistic factor as children get older. Since the two-facet modeling approach did not allow identification of a single general phonological awareness factor, a bifactor model with a general factor and four narrow processing and linguistic factors was also fitted. The model fit was good, and it demonstrated the general phonological awareness factor to be the dominant source of individual differences in performance. This model thus supports the conclusion about phonological awareness as an underlying unitary construct, even though it also is necessary to recognize sources of variance related to the content and processing requirements of the phonological tasks. The third question concerned the relations between the cognitive abilities Gf, Visual and Verbal on the one hand and phonological abilities on the other hand. Gf represents capacity to identify patterns, understand relationships and solve novel problems. Phonological awareness partly involves such activities, and one hypothesis is that Gf works as a trigger for development of phonological awareness skills. This hypothesis was supported by the fact that Gf was significantly related to the three processing factors, which is in line with the Alloway et al. (2004) and de Jong and van der Leij's (1999) studies. Adding further support to our hypothesis, Gf had no relation to the linguistic factors. Given the well-established relationship between phonological awareness and development of reading skills, the strong relationship between Gf and the processing factors also provides an explanation for the relationship between Gf and development of reading skills, even though Gf may also influence development of reading skills via other mechanisms. If the assumption is correct that Gf mainly is a trigger for development of phonological awareness, the impact of Gf should diminish over time. Some research speaks against this, as high correlations between cognitive abilities and phonological awareness among older children have been found (e.g. McBride-Chang, 1995). However, in the McBride-Chang study speech perception was part of the cognitive construct, and there was no control for correlations between these
116
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117
abilities at a younger age. There is a need for studies that follow children's phonological awareness abilities and other cognitive abilities over time, and investigate the development and the relationship between these abilities. The Verbal factor correlated with the residuals of the linguistic factors, after Gf was partialed out, but not with the residuals of the processing factors. The Visual factor did not relate to any of the phonological variables. These results support the hypothesis that Visual and Verbal are modality specific factors, and that they do not represent general processing capacity. The interpretation put forward by Gustafsson & Wolff (2015) that the Visual component in non-verbal problemsolving tests is a source of construct-irrelevant variance when the aim is to measure Gf is also supported by these findings. Also, the fact that Gf, Visual and Verbal showed different patterns of relations to the two categories of phonological awareness factors, provides in itself further support for the need to distinguish between the two facets of phonological awareness. The results in the current study thus support the hypothesis that it is possible to separately identify both a facet of linguistic complexity and a facet of processing complexity in phonological tasks, which is of great theoretical importance. The results are also important to consider in the development of instruments to measure phonological awareness. The multitrait-multimethod approach used here was appropriate for the purpose to map the structure of phonological abilities at one point in time. However, this approach rapidly becomes too complex to be useful in structural models, in which development over time is investigated. A better solution could be to use more parsimonious models that represent either the factors of the processing dimension of phonological abilities, or the factors of the linguistic dimension. The current study is limited by the fact that it is a cross-sectional study that primarily focuses on individual differences among children in their early phases of phonological development. Many of the phonological tasks therefore were too difficult for them, and only a restricted range of phonological abilities could be investigated. The crosssectional design also makes it impossible to evaluate hypotheses about directions of influence between the cognitive abilities on the one hand and the phonological abilities on the other. While we here assume cognitive abilities to influence development of phonological abilities, alternative theoretical frameworks (e.g. Demetriou & Kazi, 2001) may imply other hypotheses. However, given that the data analyzed here only represent the first wave of measurement in a longitudinal study we will in future research investigate the development of phonological abilities over time, and how this relates to other cognitive abilities and to acquisition of reading skills. Acknowledgements The research reported here has been supported by grants from the Swedish Research Council and from the Lennart Israelsson Foundation for Research on Individual and Society. References Alloway, T. P., Gathercole, S. E., Willis, C., & Adams, A. -M. (2004). A structural analysis of working memory and related cognitive skills in young children. Journal of Experimental Child Psychology, 87, 85–106. Anthony, J. L., & Francis, D. J. (2005). Development of phonological awareness. Current Directions in Psychological Science, 14, 255–259. http://dx.doi.org/10.1111/j.09637214.2005.00376.x. 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. http://dx.doi.org/10.1037/0022-0663.96.1.43. Anthony, J., Williams, J., Durán, L., Gillam, S., Liang, L., et al. (2011). Spanish phonological awareness: dimensionality and sequence of development during the preschool and kindergarten years. Journal of Educational Psychology, 103(4), 857–876. Anthony, J. L., Lonigan, C. J., Driscoll, K., Phillips, B. M., & Burgess, S. R. (2003). Phonological sensitivity: A quasi-parallel progression of word structure units and cognitive operations. Reading Research Quarterly, 38, 470–487.
Blair, C. (2006). How similar are fluid cognition and general intelligence? A developmental neuroscience perspective on fluid cognition as an aspect of human cognitive ability. Behavioral and Brain Sciences, 29, 109–125. Bowey, J. A. (2005). Predicting individual differences in learning to read. In C. Hulme, & M. J. Snowling (Eds.), The science of reading: A handbook.(pp. 155–172) (pp. 155–172). Blackwell. Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading and spelling. Ann Arbor, MI: University of Michigan Press. Brown, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. Bollen, & J. Long (Eds.), Testing structured equation models (pp. 136–162). Newbury Park, CA: Sage. Brown, T. A. (2015). Confirmatory factor analysis for applied research (2nd edition ). New York: Guilford Press. Bryant, P. E., MacLean, M., Bradley, L., & Crossland, J. (1990). Rhyme and alliteration, phoneme detection and learning to read. Developmental Psychology, 26, 429–438. Caravolas, M., Lervåg, A., Defior, S., Málková, G. S., & Hulme, C. (2013). Different patterns, but equivalent predictors, of growth in reading in consistent and inconsistent orthographies. Psychological Science, 24, 1398–1407. http://dx.doi.org/10.1177/0956797612473122. Cattell, R. B. (1987). Intelligence: Its structure, growth, and action. New York: Elsevier Science. Colom, R., Rebollo, I., Abad, F. J., & Shih, P. C. (2006). Complex span tasks, simple span tasks, and cognitive abilities: A reanalysis of key studies. Memory & Cognition, 34(1), 158–171. De Jong, P. F., & Van der Leij, A. (1999). Specific contributions of phonological abilities to early reading acquisition: Results from a Dutch latent variable longitudinal study. Journal of Educational Psychology, 91, 450–476. Demetriou, A., & Kazi, S. (2001). Unity and modularity in the mind and self: Studies on the relationships between self-awareness, personality, and intellectual development from childhood to adolescence. London: Routledge. Eid, M., Lischetzke, T., Nussbeck, F. W., & Trierweiler, L. I. (2003). Separating trait effects from trait-specific method effects in multitrait-multimethod models: A multipleindicator CT-C(M-1) model. Psychological Methods, 8, 38–60. http://dx.doi.org/10. 1037/1082-989X.8.1.38. Eid, M., Nussbeck, F. W., Geiser, C., Cole, D. A., Gollwitzer, M., & Lischetzke, T. (2008). Structural equation modeling of multitrait-multimethod data: Dfferent models for different types of methods. Psychological Methods, 13(3), 230–253. Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999). Working memory, short-term memory, and general fluid intelligence: A latent variable approach. Journal of Experimental Psychology: General, 125, 309–331. Farrell Pagulayan, K., Busch, R., Medina, K., Bartok, J., & Krikorian, R. (2007). Developmental normative data for the corsi block-tapping task. Journal of Clinical and Experimental Neuropsychology, 28, 1043–1052. Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. Hove, East Sussex, England: Psychology Press. Gustafsson, J. -E., & Stahl, P. A. (2005). STREAMS 3.0 user's guide. Mölndal, Sweden: MultivariateWare. Gustafsson, J. -E. (1984). A unifying model for the structure of cognitive abilities. Intelligence, 8, 179–203. Gustafsson, J. -E., & Wolff, U. (2015). Measuring fluid intelligence at age four. Intelligence, 50, 175–185. Høien, T., Lundberg, I., Stanovich, K. E., & Bjaalid, I. -K. (1995). Components of phonological awareness. Reading and Writing, 7, 171–188. Horn, J. L., & Noll, J. (1997). Human cognitive capabilities: Gf–Gc theory. In D. P. Flanagan, J. L. Genshaft, & P. L. Harrsion (Eds.), Contemporary intellectual assessment (pp. 53–91). New York, NY: Guilford Press. Hu, L., & Bentler, P. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. Kenny, D. A., & Kashy, D. A. (1992). Analysis of the multitrait-multimethod matrix by confirmatory factor analysis. Psychological Bulletin, 112(1), 165–172. Korkman, M., Kirk, U., & Kemp, S. (1998). NESPY: A developmental neuropsychological assessment. San Antonio, TX: The Psychological Corporation. Lonigan, C. J., Anthony, J. L., Phillips, B. M., Purpura, D. J., Wilson, S. B., & McQueen, J. D. (2009). The nature of preschool phonological processing abilities and their relations to vocabulary, general cognitive abilities, and print knowledge. Journal of Educational Psychology, 101, 345–358. Lundberg, I. (2007). Bornholmsmodellen. Vägen till läsning. Språklekar i förskoleklass [Routes to reading. Phonological games in Kindergarten]. Stockholm: Natur & Kultur. Lundberg, I. (2010). Läsningens psykologi och pedagogik [The psychological and educational view of reading]. Stockholm: Natur & Kultur. Lundberg, I., Larsman, P., & Strid, A. (2010). Development of phonological awareness during the preschool year: The influence of gender and socio-economic status. Reading and Writing: An Interdisciplinary Journal, 25, 305–320. McBride-Chang, C. (1995). What is phonological awareness? Journal of Educational Psychology, 87, 179–192. Melby-Lervåg, M., Lyster, S. -A. H., & Hulme, C. (2012). Phonological skills and their role in learning to read: A meta-analytic review. Psychological Bulletin, 138, 322–352. Muter, V., Hulme, C., Snowling, M. J., & Taylor, S. (1997). Segmentation, not rhyming, predicts early progress in learning to read. Journal of Experimental Child Psychology, 65, 370–396. Muthén, L. K., & Muthén, B. O. (2012). Mplus User 's Guid. Statistical Analysis with Latent Variables. Version 7. Los Angeles, CA: Muthén & Muthén. Papadopoulos, T. C., Kendeou, P., & Spanoudis, G. (2012). Investigating the factor structure and measurement invariance of phonological abilities in a sufficiently transparent language. Journal of Educational Psychology, 104, 321–336. http://dx.doi.org/10. 1037/a0026446.
U. Wolff, J.-E. Gustafsson / Intelligence 53 (2015) 108–117 Raven, J., Raven, J. C., & Court, J. H. (2000). Standard progressive matrices. Including the parallel and plus versions. Oxford: Oxford Psychologist Press. Schatschneider, C., Francis, D. J., Foorman, B. R., Fletcher, J. M., & Mehta, P. (1999). The dimensionality of phonological awareness: An application of item response theory. Journal of Educational Psychology, 91, 439–449. Seymour, P. H. K., Aro, M., & Erskine, J. M. (2003). Foundation literacy acquisition in European orthographies. British Journal of Psychology, 94, 143–174. Stahl, S. A., & Murray, B. A. (1994). Defining phonological awareness and its relationship to early reading. Journal of Educational Psychology, 86, 221–234. Stanovich, K. E. (1992). Speculations on the causes and consequences of individual differences in early reading acquisition. In P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. 307–342). Hillsdale, NJ: Erlbaum. Steiger, J. (2007). Understanding the limitations of global fit assessment in structural equation modeling. Personality and Individual Differences, 42, 893–898. Thorell, L. B., & Whålstedt, C. (2006). Executive functioning deficits in relation to symptoms of ADHD and/or ODD in preschool children. Infant and Child Development, 15, 503–518. http://dx.doi.org/10.1002/icd.475. Valentin Kvist, A., & Gustafsson, J. -E. (2008). The relation between fluid intelligence and the general factor as a function of cultural background: A test of cattell's investment theory. Intelligence, 36, 422–436.
117
Wagner, R. K., Torgesen, J. K., & Rashotte, C. A. (1994). Development of reading-related phonological processing abilities: New evidence of bidirectional causality from a latent variable longitudinal study. Developmental Psychology, 30, 73–87. Wagner, R. K., Torgesen, J. K., Rashotte, C. A., Hecht, S. A., Barker, T. A., Burgess, S. R., & Garon, T. (1997). Changing relations between phonological processing abilities and word level reading as children develop from beginning to skilled readers: A 5-year longitudinal study. Developmental Psychology, 33, 468–479. Wechsler, D. (1991). Manual WPPSI-R. Wechsler preschool and primary scale of intelligencerevised. (Psykologiförlaget AB). Wechsler, D., & Naglieri, J. A. (2006). Wechsler nonverbal scale of ability (WNV). San Antonio, TX: Harcourt Assessment. Wolff, U. (2013). MiniDUVAN. Kartläggning av fonologisk förmåga hos barn mellan 4 och 6 år [Assessment of phonological skills in children 4 – 6 years old]. Stockholm: Hogrefe Psykologiförlaget. Yopp, H. K. (1988). The validity and reliability of phonemic awareness tests. Reading Research Quarterly, 23, 159–177. Ziegler, J. C., & Goswami, U. (2005). Reading acquisition, developmental dyslexia, and skilled reading across languages: a psycholinguistic grain size theory. Psychological Bulletin, 131, 3–29.