Overview of the Woodcock-Johnson V Tests of Cognitive Abilities and Virtual Test Library

General Intelligence Clusters

There are three general intelligence clusters: General Intellectual Ability (GIA), Brief Intellectual Ability (BIA), and Fluid-Crystallized (Gf-Gc Composite). General intelligence clusters are estimates of psychometric g and are used for various purposes. For example, the GIA is used when evaluations require comprehensive assessments, such as for eligibility determinations. The GIA is also the best predictor of school achievement (LaForte et al., 2025, p. 331; Schrank et al., 2016). Table 5 illustrates that the composition of tests for the WJ V GIA differs from that of the WJ IV GIA in several respects. First, the WJ IV GIA consisted of seven tests, one for each of the seven CHC abilities, while the WJ V GIA consists of eight tests, also representing seven abilities. The eighth test, Verbal Analogies, measures aspects of Gf and Gc and was included to bolster the weight of Gf in the overall score ¹ (LaForte et al., 2025, p. 50). Second, the WJ IV GIA included a test of Auditory Processing (Ga) (i.e., Phonological Processing), which was replaced with a test of Gr (Semantic Word Retrieval). The WJ V GIA does not include a Ga test. Third, the Gf test included in the WJ IV GIA (Number Series) was replaced with a new test (Matrices). Fourth, the Gv test included in the WJ IV GIA (Visualization) was replaced with a new test (Spatial Relations), which was one of two subtests that comprised the WJ IV Visualization test. Fifth, the Letter-Pattern Matching test was replaced with the Number-Pattern Matching test as the indicator of Gs in the WJ V GIA. The comments column in Table 5 provides some additional information on the changes in the GIA from the WJ IV to WJ V.

OPEN IN VIEWER

Table 5.

A Comparison of the Composition of WJ IV and WJ V General Intellectual Ability (GIA)

CHC Ability Domain	WJ IV GIA Tests	WJ V GIA Tests	Comments
Gc	Oral Vocabulary	Oral Vocabulary b Verbal Analogies b	The WJ V GIA is more heavily weighted with reasoning, as Verbal Analogies is a mixed measure of Gc and Gf.
Gf	Number Series	Matrices	The Matrices test is a more robust measure of Gf than the Number Series test.
Gwm	Verbal Attention	Verbal Attention b	No change.
Glr a and Gl	Story Recall	Story Recall	No change.
Gv	Visualization	Spatial Relations	The WJ IV Visualization test consisted of two subtests: Spatial Relations and Block Rotation. Each was expanded to create two full-length, distinct tests.
Gs	Letter-Pattern Matching	Number-Pattern Matching	This change was prompted by WJ IV users who observed that performance on the Letter-Pattern Matching test can be influenced by orthographic processing weaknesses for individuals with reading difficulties. c
Ga	Phonological Processing	—	The WJ V GIA does not include a measure of Ga, suggesting it may be less susceptible to attenuation than the WJ IV GIA in individuals with phonologically based reading disabilities.
Gr	—	Semantic Word Retrieval	This test is on the WJ IV OL battery (i.e., the Retrieval Fluency test) and was renamed and included in the WJ V.

The changes regarding the WJ V GIA have implications for test users, two of which are noteworthy. First, practitioners should not expect an individual’s WJ IV and WJ V GIAs to be approximately the same. Individuals with above-average math skills may have a higher WJ IV GIA than the WJ V GIA due to an exceptionally high score on the Number Series test. Individuals with poor phonological processing may have a lower WJ IV GIA as compared to the WJ V GIA due to the attenuating effect of an exceptionally low score on the Phonological Processing test. Second, the WJ V GIA has slightly lower correlations with achievement than the WJ IV GIA. This is likely due to the reduction in shared content between cognitive and achievement tests (e.g., Number Series and math achievement tests, Phonological Processing and basic reading and writing skills tests). A well-designed and thoughtful series of simulation studies conducted by the WJ V authors supports this contention (see LaForte et al., 2025, Table 6–17, pp. 309–310).

The BIA, a three-test cluster, may be used for time-limited assessments, as well as for screening and reevaluation purposes. As a proxy for g, the BIA should be interpreted with caution, given its minimal breadth of component abilities (see Breit et al., this issue). The WJ IV and WJ V BIAs have two tests in common, Oral Vocabulary and Verbal Attention. The WJ IV Number Series test was replaced with Matrices as the third test comprising the WJ V BIA. The Gf-Gc Composite, which comprises tests with robust g-loadings (LaForte et al., 2025, Table 6.3, p. 226), is useful for identifying individuals who are intellectually gifted and intellectually disabled (Dumont et al., 2016; Floyd et al., 2016; Pfeiffer & Yarnell, 2016). The Gf-Gc Composite is also useful when the GIA does not provide the best description of intellectual ability for individuals suspected of having a specific learning disability due to domain-specific cognitive weaknesses, memory deficits, or difficulties processing information quickly and efficiently (Flanagan, 2025; Flanagan & Alfonso, 2017; McDonough & Flanagan, 2016; Schrank et al., 2015).

Broad Ability Clusters

Table 3 shows that 14 of the 20 WJ V cognitive tests contribute to seven distinct CHC cognitive clusters (Gc, Gf, Gwm, Gs, Gr, Gl, and Gv). The CHC broad ability clusters have many purposes. They are helpful for in-depth diagnostic assessments designed to identify neurodevelopmental disorders, such as specific learning disabilities (SLD) and attention-deficit/hyperactivity disorder (ADHD) (e.g., Decker et al., 2017; Flanagan, 2025; Harvey, 2012; Hoelzle et al., 2023; Miller et al., 2016). They are also helpful in examining the cognitive consequences of brain damage and brain disease (Sweet et al., 2019). Furthermore, research shows that performance on tests of Gl, Gf, Gs, and Gwm differentiates depression and dementia in older adults (Mazur-Mosiewicz et al., 2011). Similarly, adolescents with depression score lower on tests of Gl, Gwm, and Gs (Basnet et al., 2015; see also Kriesche et al., 2023). The broad ability clusters are also used to identify an individual’s unique pattern of strengths and weaknesses, aiding in diagnostic decision making, instructional planning, and the selection of targeted interventions, accommodations, and support strategies (e.g., Flanagan et al., 2024; Mather & Wendling, 2024; Schrank & Wendling, 2015; Woodcock et al., 2014).

A mounting body of research on the relations between CHC broad and narrow abilities and specific academic skills shows that some cognitive abilities (e.g., Gwm and Gs) predict performance across multiple academic domains, while others are more domain-specific, further supporting their utility in understanding learning difficulties and disabilities (Hajovsky et al., 2025a, 2025b; Niileksela et al., 2025; Niileksela et al., this issue). A comparison of the relationships between cognitive abilities and academic skills in the WJ V standardization data (Niileksela & Hajovsky, 2025) and clinical samples of individuals with learning disabilities in the literature (Flanagan et al., 2024; McDonough et al., 2017) reveals remarkable similarities. See Table 6 for an example of the relations between cognitive abilities and reading subskills. Moreover, neuroimaging studies have identified specific brain networks associated with different cognitive abilities and shown how these networks relate to academic skill development (Hawes et al., 2019; Owens et al., 2020; Peters & De Smedt, 2017; Tablante et al., 2023; Xu et al., 2017).

OPEN IN VIEWER

Table 6.

A Comparison of Relations Between Cognitive Abilities and Reading Subskills in the WJ V Standardization Sample and Independent Samples of Individuals with Reading Disabilities and Dyslexia

WJ V Cognitive Correlates a	Clinical Samples of RD and Dyslexia b
Word Reading Accuracy
Ga is the strongest predictor at all ages	Phonological Awareness (Ga:PC)
Gc, Gwm, and Gr are moderate predictors at some ages	Phonological Memory (Ga:UM; Gwm:Wa)
	Rapid Automatized Naming (Gr:NA)
Reading Rate and Fluency
Gs is the strongest predictor at all ages	Processing Speed (Gs:RS)
Gc is a moderate predictor; it decreases in size with age	Rapid Automatized Naming (Gr:NA)
Gr is a small predictor; it increases in size with age	Orthographic Processing/Mapping Skills (Gc)
Reading Comprehension
Basic Reading is a strong predictor; it decreases in size with age	Oral Language (Gc: VL, MY, CM) and Listening Comprehension (Gc: LS)
Gc is the strongest predictor; it increases in size with age	Short-Term Auditory Storage and Working Memory Capacity (Gwm:Wa, Wc)
Gf is a moderate predictor at all ages	Reasoning (Gf:I, RG)
	Executive Functions (shifting/inhibition/AC)

Table 7 provides a comparison of the tests that comprise the CHC broad ability clusters on the WJ IV and WJ V cognitive batteries. Except for Gwm and Gs, the composition of all the broad ability clusters changed. The most salient changes are listed below.

• The Gf cluster was made more robust by introducing the Matrices test, a measure of the narrow Gf ability of Induction (I). Matrices replaced Concept Formation, also a test of I, but with a lower receptive language demand. Analysis-Synthesis is a measure of the narrow Gf ability of General Sequential Reasoning (RG). Interestingly, the Gf tests comprising the Gf cluster have medium psychometric g-loadings. A large body of research predicts that these tests would have high g-loadings (Gustafsson, 1984). However, LaForte et al. (2025) reported that, similar to their findings, more recent analyses (e.g., Gignac, 2015) demonstrate that Gf tests like the Raven’s Progressive Matrices are not among the highest g-loaded tests. See Laforte and colleagues (2025, p. 228) for details.

OPEN IN VIEWER

Table 7.

A Comparison of the CHC Broad Ability Clusters of WJ IV and WJ V

WJ IV Broad Cluster Tests	WJ V Broad Cluster Tests
Gf Number Series Concept Formation	GfMatricesAnalysis-Synthesis
Gc Oral Vocabulary General Information	GcOral VocabularyVerbal Analogies
GwmVerbal AttentionNumbers Reversed	GwmVerbal AttentionNumbers Reversed
Glr Story Recall Visual Auditory Learning	GlStory RecallStory Comprehension
—	GrSemantic Word RetrievalPhonemic Word Retrieval
GvVisualizationPicture Recognition	GvSpatial RelationsBlock Rotation
GaPhonological ProcessingNonword Repetition	—
GsLetter-Pattern MatchingPair Cancellation	GsNumber-Pattern MatchingLetter-Pattern Matching

Note. Tests in bold contribute to the General Intellectual Ability cluster in the respective batteries, WJ IV and WJ V. The four WJ IV tests printed in italics are part of the WJ V COG Extended battery and do not contribute to any clusters on the WJ V.

It is worth noting that numerous studies have argued that general intelligence test scores should be the primary or exclusive focus when interpreting cognitive assessments (e.g., Canivez, 2017; Dombrowski et al., 2018; McGill et al., 2018). This conclusion is based primarily on bifactor and Schmid-Leiman (BF/SL) statistical approaches, which have been criticized for methodological problems and a lack of a theoretical foundation (Decker et al., 2020; Murray & Johnson, 2013; Schmank et al., 2021). Drawing from their BF/SL research, the Canivez/Dombrowski research team supports psychometric g-theory and believes interpretation should focus on general intelligence (g) rather than broad cognitive abilities. However, the level of interpretation that is most appropriate varies depending on the statistical methods employed, and this “g-focused” research group frequently disregards studies that support interpreting CHC broad abilities (see McGrew et al., 2023, for a detailed discussion).

The Canivez/Dombrowski research team demonstrates bias through heavy self-citation, while disregarding contrary evidence from groups such as the Keith-Reynolds researchers, who endorse multifactorial intelligence theories (e.g., Caemmerer et al., 2020; Keith & Reynolds, 2018; Reynolds et al., 2013). This one-sided citation approach allows them to advocate for a g-only interpretation without sufficient empirical support (McGrew et al., 2023). As Meehl (1992) observed, statistical methods do not inherently produce truth. The BF/SL models lack grounding in intelligence and cognitive theory. Ironically, while this group acknowledges that their bifactor models have computational issues, they continue to use these problematic models to argue against the interpretation of CHC broad ability clusters.

Extensive validation evidence supports distinct CHC abilities beyond statistical factor analysis alone, encompassing developmental trajectories, connections to academic performance, brain-based findings, and genetic studies (e.g., Haier & Jung, 2018; Horn & Knoll, 1997; LaForte et al., 2025). Understanding g and broad abilities necessitates knowledge from disciplines outside psychometrics and school psychology, especially neuroscience and cognitive science (McGrew et al., 2023).

The general intelligence factor (g) arises from positive correlations between cognitive test scores, representing positive manifold or psychometric g (Van der Maas et al., 2006). Although g shows robust predictive power, what it represents remains uncertain. It is plausible that g might result from the test measures themselves, making it a formative rather than a causative factor. From this viewpoint, the abilities beneath g (broad and narrow abilities) become more crucial for understanding cognition (see Kovacs & Conway, 2016).

Contemporary brain network theories provide more convincing explanations for g than BF/SL research. Process Overlap Theory (POT; Kovacs & Conway, 2016) proposes that general executive processes in frontal and parietal brain regions generate positive manifold by overlapping with specialized cognitive processes. Differences in these executive processes can limit cognitive performance. Parietal Frontal Integration Theory (P-FIT; Haier & Jung, 2018) provides an alternative brain-centered explanation. These theories account for psychometric g while rejecting the idea of a single general factor of intelligence, or psychological g.

The WJ V technical manual provides ample empirical support for the structure of the WJ V through various methodologies, yielding similar findings. According to the WJ V authors, “The application of multiple structural analysis methods on the same data limits the risk of a single statistical method (viz., factor analysis) obscuring conclusions, missing nuances in the data, or limiting the discovery of new findings” (LaForte et al., 2025, p. 257). The network of validity evidence for the CHC broad abilities supports the interpretation of the WJ V CHC broad ability clusters. The bold claims of the Canivez/Dombrowski research team that broad ability score interpretation lacks empirical support need to be tempered. The emergence of a strong general factor is dependent on the methods used to analyze the data (see Decker et al., 2020; Horn & Blankson, 2012).

Narrow Ability and Clinical Clusters

Table 3 shows there are six CHC narrow ability and clinical clusters. The Cognitive Efficiency (CE) cluster is comprised of one Gwm and one Gs test, which are the cognitive domains that represent the “parameters of cognitive information processing efficiency” (LaForte et al., 2025, p. 54; see also McGrew, 2023; Schneider & McGrew, 2018). This cluster reflects an individual’s ability to hold and manipulate information, control and sustain attention, and execute tasks quickly and accurately, which is important for understanding and diagnosing attention problems, learning disabilities, and other neurological impairments (Mather & Wendling, 2024; Schrank et al., 2014). The Phonemic Retrieval Fluency cluster is comprised of the Phonemic Word Retrieval test from the COG battery, a measure of the narrow Word Fluency (FW) ability, and Rapid Phoneme Naming, a new test to the WJ V, housed in the VTL. This test requires the examinee to pronounce phonemes for single letters quickly, a task that best fits within the broad ability domain of Gr. The authors provide preliminary empirical support for the Phonemic Retrieval Fluency cluster in Chapter 6 of the technical manual (LaForte et al., 2025). Additionally, because the Rapid Phoneme Naming test does not align well with any existing narrow Gr abilities, the authors provide support for a potential new narrow ability, called Phoneme Retrieval Fluency (FP; Laforte et al., 2025, p. 250). Additional validation evidence by independent researchers may pave the way for FP to join other Gr narrow abilities in the next iteration of the CHC taxonomy.

Four narrow ability and clinical clusters are derived from tests in the VTL. The Phonological Manipulation cluster comprised of the Sound Deletion and Sound Substitution tests, measures the narrow phonetic coding (PC) ability. These two tests were subtests of the WJ IV Sound Awareness test and were extended to full-length tests on the WJ V. This cluster, along with the Phonemic Retrieval Fluency cluster, is likely to play a prominent role in evaluating individuals suspected of having a reading disability or dyslexia. Another cluster that will be important in dyslexia evaluations is the RAN-Reading cluster, comprised of the Rapid Picture Naming, Rapid Letter Naming, and Rapid Phoneme Naming tests. Rapid automatized naming (RAN) tests like these have long been used in dyslexia evaluations because they have been found to differentiate dyslexic children from typical readers (see, for example, Denckla & Rudel, 1976). Furthermore, phonological and RAN processes explain unique variance in different types of reading (Wolf & Bowers, 1999). A meta-analysis of 60 years of research found that knowledge of letters and sounds, along with phonological awareness and rapid naming in kindergarten, best predicted reading skills in 1^st and 2^nd grade (Schatschneider et al., 2004). Two additional rapid naming tests from the VTL, Rapid Number Naming and Rapid Quantity Naming, comprise the RAN-Math cluster. Research shows that number-specific RAN predicted arithmetic fluency best (e.g., Kirby et al., 2003).

The WJ V authors’ separation of RAN tests into reading and math clusters suggests that a general RAN factor may obscure specific cognitive difficulties. Research shows that, in general, alphanumeric RAN (letters and numbers) correlates more strongly with reading performance, while nonalphanumeric RAN (colors and objects) better predicts math fluency (e.g., Norton & Wolf, 2012). Typically, colors and objects are not processed serially, making their naming less automatic and less similar to reading. Thus, even though a general RAN factor exists, research indicates that targeted RAN tasks differentially predict early reading and math performance, helping to identify children at risk for learning difficulties (Hornung et al., 2017). As such, separating RAN tasks into separate clusters, as the WJ V authors did, is helpful in clinical contexts. Empirical support for separating the RAN tests into separate RAN-reading and RAN-math clusters is found in the WJ V technical manual. For example, developmental growth curves for each cluster show that they “diverge from one another in early rate of growth and approximate age of plateau” (LaForte et al., 2025, p. 233). Concurrent validity studies show patterns of convergent and discriminant validity that also support the contention that the RAN-reading and RAN-math clusters are not redundant (Laforte et al., 2025, pp. 306–307).

Norming Sample

The WJ V norming study targeted 6,000 participants across 24 age groups, from ages 3 to 80 years and above, with 250 participants per group. The sampling aligned with the 2017 U.S. Census Bureau’s national population projections (U.S. Census Bureau, 2017), taking into account ethnicity, race, and education. Data collection between February 2022 and August 2023 involved 5,837 individuals, comprising 562 children aged 3 to 5 years who were not enrolled in K–12 education, 3,106 students aged 4 to 19 years enrolled in K–12, and 2,169 individuals aged 17 years and older who were no longer in high school. Although some age groups fell slightly short of the target of 250 examinees, the average number of cases per age group, ranging from 3 to 79 years, was 245. The lower number of participants (190) in the 80+ age group was anticipated due to challenges in recruiting older adults, particularly amid health concerns post-pandemic. To ensure the WJ V norms accurately represented the demographic distributions in the U.S. population, weights were applied to adjust for underrepresented groups, ensuring representativeness (LaForte et al., 2025, pp. 129, 157–170).

Calibration/Alternate Form/Test-Retest (CAR) Study

The Calibration and Alternate-Form Reliability (CAR) studies conducted for the WJ V aimed to ensure the equivalence of alternate test forms, involving 1,430 participants, to assess the reliability of second forms across the Cognitive, Achievement, and Virtual Test Library batteries (LaForte et al., 2025, pp. 132–133). These studies evaluated 27 tests: 5 in the Cognitive battery, 17 in the Achievement battery, and 5 in the Virtual Test Library. High alternate-form reliability coefficients were observed, typically above 0.8 for ages 4 to 19, indicating consistent rank-ordering of scores between forms (LaForte et al., 2025, p. 202). For adults, most coefficients remained above 0.8, with exceptions like Symbol Inhibition, Word Reading Fluency, and Oral Reading showing coefficients in the 0.7's, and Passage Comprehension, Written Language Samples, and Paragraph Reading Comprehension in the 0.6's (LaForte et al., 2025, p. 202).

Test-retest reliability was similarly high for the WJ V Semantic Word Retrieval, Phonemic Word Retrieval, and Letter Writing Fluency tests, with coefficients above 0.9 for younger participants and above 0.8 for adults (LaForte et al., 2025, p. 202). Content equivalence was ensured by calibrating items to the standard W scale, with curriculum experts consulting on tests like Letter-Word Identification, Calculation, and Spelling to verify comparable item types.

The CAR studies also examined the impact of prior exposure to specific WJ V tests, noting minimal effects for most tests. However, the Rapid Automated Naming tests, such as Rapid Letter Naming and Rapid Quantity Naming, showed small to moderate practice effects, particularly in adults. Rapid Phoneme Naming exhibited moderate effects for children and adults, suggesting slight advantages from recent exposure (LaForte et al., 2025, p. 202). Overall, the CAR studies validate the reliability and equivalence of WJ V’s alternate forms, with caution advised for interpreting scores from speeded tests due to potential practice effects.

Test and Cluster Norms

A Reference W (REF W) is a normative score calculated for each test according to the normative-median score for each age/grade interval. These scores allow for the creation of developmental growth curves for the respective age and grade groupings. They are foundational for determining age and grade equivalent scores, the RPI, and instructional range features in the WJ V. Furthermore, consideration of the standard deviations (SDs) of the REF W scores provides the basis for calculating other norm-referenced metrics, such as standard scores and percentile ranks. LaForte and colleagues (2025) detail the steps for establishing age- and grade-based norms for tests, clusters, and comparison base rate procedures in Chapter 4 of the technical manual (pp. 175–188).

Cluster Analysis

CA links the most highly correlated WJ V tests into initial groupings, which then combine with other groups or individual tests to form progressively larger clusters. This hierarchical structure is visualized through tree dendogram diagrams (Jacoby & Ciuk, 2018), culminating in a single all-encompassing group. The Ward’s cluster analysis of the initial target age group (10 to 14 years) revealed meaningful groupings that align with broad CHC abilities, including Ga, Gl, Gr, Gv, Gq, Gf, Gc, and Gs. Additionally, the analysis identified potential narrow CHC subgroups within Gwm and Grw. Within the Gwm domain, the analysis distinguished between tests requiring complex attentional control (Gwm-Wc) and those requiring simpler memory span or short-term auditory storage (Gwm-Wa). For Grw abilities, a meaningful separation emerged between tests measuring basic reading skills (Grw-skills) and those requiring application through connected discourse (Grw-app).

Multidimensional Scaling (MDS)

Like CA, MDS is an exploratory analysis that aims to identify the underlying structure in datasets by examining relationships between variables. MDS is inherently more qualitative than factor analysis, as it can visually depict test relationships, providing an essential layer of evidence for the validity of the WJ V by clarifying test content and processing domains. The WJ V authors applied Guttman’s Radex two-dimensional MDS procedure to correlation data from all 60 WJ V tests across six age groups (4–5, 6–9, 10–14, 15–19, 20–49, and 50–80+). LaForte et al. (2025) used the detailed results from the 10- to 14-year-old age group as their primary model. Their results revealed six distinct groupings or content facets: (1) the Verbal (V) facet includes mainly Gc and Gl tests, which require access to semantic knowledge; (2) the Auditory (A) facet includes Ga tests and tests of the Gwm narrow ability of Auditory Short-Term Storage (Wa); (3) the Figural-Visual (FV) facet includes Gv, Gf, and the Gwm narrow ability of Visual Short-Term Storage (Wv); (4) the Quantitate-Numeric (QN) facet includes tests of the Gq narrow ability of Math Achievement (A3) and the Gf narrow ability of Quantitative Reasoning (RQ); (5) the Reading-Writing (RW) facet includes nine reading and writing tests from the WJ V ACH battery; and (6) the Speed-Fluency (SF) facet consists of Processing Speed (Gs) tests from the WJ V COG and ACH batteries, Retrieval Fluency (Gr) tests from WJ V COG and VTL, and Gs/Gr tests that are classified as Naming Facility (NA)—a narrow Gr ability. ⁷

In general, the CA and MDS methods validated each other’s findings. That is, when CA identified certain groupings of cognitive abilities, MDS provided spatial confirmation by showing those same abilities positioned close together in dimensional space. The findings from these analyses provide empirical support for interpreting tests as measuring similar content domains, moving beyond professional judgment based on task characteristics and task demands. Empirical CHC and content classifications of tests provide a dual lens for understanding test performance.

Exploratory Principal Axis Factor Analysis (PAF)

The WJ V authors utilized the open-source JASP (Version 0.17.2.1) exploratory factor analysis module with oblique oblimin rotation (JASP Team, 2022) to examine both Set A and Set B test collections. Consistent with cluster analysis and multidimensional scaling results, the Set A analyses identified nine broad CHC factors, including Gc, Gs, Gs/Gr-NA, Gwm-Wa, Gwm-Wc, Wa, Gv, Gr, Gl, and Gf. However, the Ga factor did not emerge as distinct, being defined by only two Ga tests: Sound Reversal (.63) and Segmentation (.34), with Number Series (.45) loading interpreted as a chance finding. An interesting discovery was the Gr-FP factor, characterized by the pairing of Rapid Phoneme Naming (fluency in retrieving and pronouncing phonemes from graphemes) and Sound Blending (hearing and blending phonemes). Both tests require efficient phoneme processing and retrieval, representing a combination not currently recognized in the CHC taxonomy (Schneider & McGrew, 2018).

Three tests demonstrated notably high uniqueness values, indicating substantial variance not explained by identified factors. Symbol Inhibition showed the least shared variance (.61) with other WJ V tests, followed by Rapid Picture Naming (.59) and Oral Language Samples (.55). These patterns were consistent across analytical methods, with Symbol Inhibition being the last to join the Gs grouping in cluster analysis, Rapid Phoneme Naming showing spatial distance in multidimensional scaling results, and Oral Language Samples not fitting clearly into any multidimensional scaling grouping (LaForte et al., 2025).

The Set B factor analysis identified four clear broad factors: Grw, Gc, Gq, and Gs. Tests with the highest uniqueness included Oral Language Samples (.62), Reading Recall (.57), and Sentence Writing Fluency (.57). The uniqueness of Reading Recall likely stems from its shared measurement methodology and Meaningful Memory (Gl:MM) construct variance with Story Recall. The uniqueness of Sentence Writing Fluency was evident in its spatial separation from other speed tests in multidimensional scaling and its late grouping with Rapid Picture Naming in cluster analysis results (LaForte et al., 2025).

Psychometric Network Analysis (PNA), Exploratory Graph Analysis (EGA)

The WJ V authors included PNA-EGA as a special supplementary approach within their comprehensive exploratory structural analyses (Laforte et al., 2025, p. 240). Recently, PNA methods and network-based cognitive ability models, including process overlap theory (POT) and dynamic mutualism, have emerged as alternatives to traditional factor-analytic approaches for identifying dimensional structures in cognitive and achievement variables (Laforte et al., p. 251; see McGrew, 2023; McGrew et al., 2023). The WJ V Technical Manual (Laforte et al.) appears to be the first to include PNA-EGA.

In PNA, individual tests are represented as nodes within a multidimensional visual graphic network (see Figure 1 for an illustration). In Figure 1, the nodes (circles) are labeled with narrow ability codes, representing what each test measures. Connections between nodes, depicted as lines in Figure 1, are called edges and represent statistical estimates (partial correlation coefficients) of the strength of nondirectional relationships between test pairs. Thicker edges indicate stronger associations. For example, in Figure 1, the lines connecting the RG, RQ, and I nodes are thicker than the lines connecting these nodes to all other nodes, indicating a strong association between these Gf nodes. Significant pairwise partial correlations estimate node relations conditionally, with variance from all other tests statistically removed to ensure edge weights remain independent of other test relationships (Hevey, 2018).OPEN IN VIEWER

Network topography can be evaluated using network science tools (Borsboom et al., 2021; Bulut et al., 2021; Jones et al., 2018; Neal & Neal, 2021; Robinaugh et al., 2016; cf. Laforte et al., 2025, p. 252), including various centrality metrics such as closeness, betweenness, and strength—metrics that are fundamental to understanding the functional importance and clinical relevance of nodes within psychometric networks (Borgatti, 2005; Hevey, 2018). For example, Figure 1 shows the PNA results of a hypothetical battery of cognitive tests. The largest nodes in PNA (such as K0, RG, and Wc in Figure 1) represent core cognitive abilities that are most central or influential in the overall network structure. PNA methods have expanded to include exploratory graph analysis (EGA), which is used to determine the number of underlying factors or dimensions present in the data. For example, the illustration in Figure 2 shows that an EGA analysis of the hypothetical battery of cognitive tests identified four distinct dimensions with high modularity (or well-defined groups of nodes), suggesting a robust four-factor structure underlying the hypothetical cognitive battery. While PNA and EGA produce visually similar network plots with nodes, edges, and color-coded groupings, they serve fundamentally different analytical purposes. The PNA visualization (Figure 1) emphasizes individual node centrality and connectivity patterns to identify the most influential variables in the network. In contrast, the EGA visualization (Figure 2) emphasizes community detection to determine the dimensional structure of an instrument. Although the visual representations appear similar, the former addresses questions about variable importance, while the latter addresses questions about factor structure (Golino & Epskamp, 2017).OPEN IN VIEWER

The hierarchical EGA results for the WJ V identified first-order CHC community dimensions, including Gf, Gq, Gv, Gs, Gr, Gwm-Wa, Gwm-Wc, Wa, Grw-application, and Gs/Gr-NA. These results are generally consistent with findings from traditional exploratory structural analysis. Notably, two differences emerged. Grw-skills merged with Ga tests to form a community, while Gc and Gl tests formed a general, semantically meaningful verbal abilities dimension (LaForte et al., 2025). EGA analysis also identified a speed of reasoning (Gs-RE) community that consistently appeared across four primary age groups, prompting its inclusion in the CHC taxonomy. Speed of reasoning is defined as the ability to perform reasoning tasks within time limits (LaForte et al., 2025, p. 257), as exemplified by tests that require fluent semantic processing and reasoning-based decision-making (i.e., Sentence Reading Fluency, Symbol Inhibition, and Word Reading Fluency).

The higher-order community structure suggested three dimensions: cognitive processes, cognitive speed/fluency, and cognitive working memory. A robust differentiation between cognitive level and speed abilities appeared across all age-differentiated samples, supporting Schneider and McGrew’s (2018) suggestion that cognitive processing speed may represent a higher-order taxonomy equivalent to psychometric g. Edge weight patterns between speed-related dimensions indicated core elements of this g-speed dimension, suggesting the need to explore alternative higher-order dimensional structures beyond traditional single psychometric g models (LaForte et al., 2025, p. 257).

The exploratory analyses resulted in strong evidence for 10 broad cognitive abilities (Gc, Gf, Gv, Ga, Gl, Gr, Gwm, Gs, Grw, and Gq). They also revealed narrow ability substructures within broader domains. For example, within the Gwm domain, two narrow substructures were identified: Gwm-Wa (Auditory Short-Term Storage) and Gwm-Wc (Working Memory Capacity with Attentional Control). Another example occurred within the Gs domain, where three substructures were identified: Gs-Cognitive (speed on cognitive tasks, such as Number-Pattern Matching), Gs-Achievement (speed on academic tasks, such as Math Facts Fluency), and Gs/Gr-NA (Naming Facility; defined by rapid automatized naming tasks). These substructures are clinically important because they help pinpoint a problem more precisely, leading to more targeted interventions. One of the most significant findings of the Stage 2 analyses was the consistent patterns observed across age groups and analytic methods (Laforte et al., 2025).

Cross-Validation Confirmatory Analysis

The most plausible and best-fitting CHC models emerged from Stage 2 exploratory analyses. These models were called Carroll hierarchical g broad CHC model (a model with psychometric g at the apex, most closely resembling Carroll’s [1993] three-stratum theory), Carroll hierarchical g+narrow CHC model (includes narrow abilities under broad factors), and Horn no-g broad CHC model (similar to the Carroll hierarchical g broad CHC model, but there is no psychometric g factor). The Carroll hierarchical g broad CHC model and the Horn no-g broad CHC model have the same measurement models but different structural models. The WJ V authors did not evaluate a bifactor g broad CHC model, indicating that it is not consistent with their position or Woodcock’s legacy (LaForte et al., 2025, p. 271).

Confirmatory analysis was conducted at Stage 3 for cross-validation using Sample B. The WJ V authors concluded that all three models were plausible, with the Horn no-g broad CHC model considered the most parsimonious. However, the Carroll hierarchical g+narrow CHC model “offers potentially important insights regarding the structure of the WJ V battery, possibly clinically relevant interpretations, and potential new insights into CHC theories” (LaForte et al., p. 277). The confirmatory structural model cross-validation provided evidence for the “stability and generalizability of the structural validity of the WJ V measurement model operationalized as three types of CHC models” (p. 277). The WJ V effectively measures broad and narrow CHC abilities, as well as general cognitive functioning. Overall, the combination of results from the advanced and strategic methods and analyses conducted by LaForte and colleagues provides strong evidence of the validity of the WJ V.

W Score, Reference W Score, and W Difference Score

A W score is the foundational metric of the WJ V that represents an individual’s ability level on a test. It is derived from raw scores using an equal-interval scale centered on 500, which represents the average performance of 10-year-olds in the norm sample. Every test on the WJ V has a Reference W score, representing the point along the W scale at which 50% of individuals of the same age or grade succeed and 50% do not, providing a baseline for determining if someone is above, at, or below typical performance. Reference W scores are established separately for each age and grade level in the normative sample. At age 10, the Reference W score is 500. The Reference W is the comparison point against which an individual’s W score is measured, so it is akin to a “zero point” for calculating difference scores. An individual’s W score minus their same age/grade peer’s Reference W score is the W Difference Score, which can be zero, positive, or negative. See Figure 3 for an example of the application of the Reference W and W Difference scores.OPEN IN VIEWER

It is important to note that the Reference W score represents an optimal level of difficulty for instruction. As such, a W Difference score of zero is the point where the level of difficulty is optimal for promoting learning (referred to as an individual’s instructional level). When an individual’s W Difference is positive (e.g., +10 or higher, like Student A in Figure 3), they will find items at the 50% difficulty level for same-age peers easier than those peers do (referred to as an individual’s independent level). When an individual’s W Difference is negative (e.g., −10 or lower, like Student B in Figure 3), they will find items at the 50% difficulty level for same-age peers harder than those peers do (referred to as an individual’s frustration level).

Relative Proficiency Index

The median W score for each age/grade group in the norming sample serves as a helpful reference point (Reference W score) because it represents the difficulty level at which 50% of that age/grade group succeeds. However, 50% is a reference point typical in a traditional norm-referenced approach (e.g., it is helpful for rank ordering), but it is not considered proficient in educational settings. Instead, educators typically consider 90% to be proficient (Jaffe, 2009). To create a score that shows an individual’s level of proficiency, the point at which 90% of same-age peers are successful is a more meaningful reference than 50%. As such, the reference point was set at 20 W units below the Reference W, where items are easier than those associated with the median W, representing the level of difficulty where 90% of same-age/grade peers succeed. With this criterion adjustment from 50% to 90%, a W Difference score of zero indicates that an individual will perform with 90% proficiency on tasks where their median same-age/grade peers also perform with 90% proficiency, expressed as an RPI of 90/90. The RPI denominator is always 90. The numerator is the individual’s likelihood of a correct response on items that peers perform with 90% success. The conversion of a W Difference to the numerator of the RPI is based on mathematical probabilities derived from the Rasch model (LaForte, personal communication, July 6, 2025). Riverside Score® generates RPIs automatically. Refer to Table 10 for practical examples showing various RPIs and their real-world implications for educational planning.

OPEN IN VIEWER

Table 10.

Real-World Examples of Relative Proficiency Indexes and Their Educational Implications

W DIFF	RPI	Proficiency Level	Task Difficulty	Example Scenario	Educational Recommendation
+35	100/90	Very Advanced	Extremely Easy	Student completes 6th-grade math problems while peers struggle with 4th-grade concepts	Acceleration, gifted programming, and independent study projects
+20	99/90	Advanced	Very Easy	Reads chapter books fluently while classmates read picture books	Above-grade curriculum, enrichment activities, peer tutoring opportunities
+10	96/90	Average to Advanced	Easy	Completes grade-level writing assignments quickly with high quality	Grade-level work with extension activities and leadership roles
+3	90/90	Average	Manageable	Performs similarly to typical peers on grade-level tasks	Standard curriculum with occasional support for challenging concepts
−3	82/90	Average	Manageable	Needs extra time but masters most grade-level material	Regular curriculum with some accommodations and progress monitoring
−10	73/90	Limited to average	Difficult	Struggles with multi-step math problems that peers solve easily	Modified assignments, additional instruction time, and visual supports
−20	48/90	Limited	Very difficult	Reads at a 2nd-grade level while in a 5th-grade classroom	Intensive intervention, specialized instruction, and assistive technology
−35	15/90	Very limited	Extremely difficult	Cannot decode basic sight words while peers read fluently	Alternative curriculum, functional academics, one-on-one support
−50	2/90	Negligible	Impossible	Requires significant support for basic academic concepts	Life skills curriculum, adaptive materials, comprehensive support

Cognitive and Academic Language Proficiency (CALP)

The WJ V provides a method for evaluating a student’s capacity to comprehend and utilize the sophisticated language demands inherent in academic settings, a concept first promoted by Cummins (1984). This scoring system operates on a six-point scale that correlates directly with a student’s ability to manage instructional content delivered in English. Students achieving the highest levels of proficiency (CALP Levels 5–6) demonstrate advanced to very advanced cognitive and academic language skills, with academic instruction presenting minimal challenge. These individuals typically find coursework extremely easy to very easy, indicating strong readiness for grade-level academic demands. Students at Level 4 exhibit fluent language proficiency, where instructional content remains manageable and accessible within their current developmental capacity (Mather et al., 2025c).

The intermediate range (CALP Levels 3–4) represents students transitioning from limited to fluent proficiency. Students at Level 3 are likely to encounter significant difficulties with standard instructional approaches, while students at Level 4 may find academic instruction challenging but not insurmountable. These students often require additional support and modified instructional strategies to access curriculum content effectively. The lower proficiency levels (CALP Levels 1–2) indicate significant challenges with academic language demands. Students at Level 2 demonstrate very limited proficiency, making standard instruction extremely difficult to navigate. Students at Level 1 face the most substantial barriers, with conventional academic instruction being nearly impossible without intensive intervention and support (Mather et al., 2025c).

The CALP levels are valuable for educational planning and can be derived from various cognitive and academic clusters, including measures of oral language, reading comprehension, and written expression. By understanding where students fall along this continuum, educators can tailor instructional approaches more effectively, select appropriate instructional materials, and provide interventions that closely align with students’ current language proficiency level. This targeted approach ultimately supports more effective learning outcomes by ensuring that academic demands align with students’ cognitive and academic language capabilities (Mather et al., 2025c).

CALP in a Broader Context and a Word of Caution

Based on Cummins’ (1979, 1980, 1984) research, he distinguished between conversational language, which he referred to as “Basic Interpersonal Communicative Skills” (BICS), and the advanced language that develops when someone attends school, which he termed “Cognitive-Academic Language Proficiency” (CALP). He noted that it takes approximately 5–7 years for CALP level proficiency to develop and that formal education is required to achieve this level of proficiency. Thus, CALP level proficiency begins to emerge at around 4th to 5th grade (ages 9–10) and continues to develop throughout one’s schooling. His research helped illustrate the fact that language, for educational purposes, does not reach a simple threshold upon the attainment of BICS, which occurs at age 3 when a child is capable of engaging in general conversation and expressing their thoughts, needs, and feelings. By age 5, children entering Kindergarten are at the advanced BICS level and ready for formal education, given that teachers can now communicate with them well enough to provide formal learning experiences, many of which are designed expressly to expand language abilities directly (e.g., reading and writing).

A potential problem with the reporting of CALP Levels on the WJ V, is that these levels are given to students for whom language development has not yet reached the point where CALP exists (Rhodes et al., 2005). If the average, monolingual English-speaking student reaches early or beginning CALP by age 9, then CALP levels do not exist for students younger than 9 (Rhodes et al.). Moreover, CALP is not a binary concept any more than BICS, which means that neither appears in a student’s development suddenly nor fully formed. Instead, both are dynamic and grow from rudimentary levels to more advanced levels, much like any other form of development. Thus, CALP does not abruptly emerge at age 9; instead, it begins to emerge at that point and then continues into an intermediate level before progressing on to advanced levels as dictated by age or grade and education. CALP levels that are reported on the WJ V are used to describe, for example, higher performance in younger children who, by definition, are not expected to have CALP. Thus, these descriptive levels have been described as misleading, albeit unintentionally (see Ortiz, 2019; Ortiz & Cehelyk, 2024, for more information).

Instructional and Developmental Zones

Instructional Zones and Developmental Zones serve different assessment purposes and are reported for different test types. Instructional Zones are explicitly associated with achievement tests and clusters, providing information about the type of academic tasks that are appropriate for instruction. These zones help educators determine what level of academic material a student can successfully engage with, given appropriate instructional support. Developmental Zones are reported for cognitive ability tests and clusters, relating to the student’s cognitive developmental level and potential for growth in cognitive abilities. These zones provide insight into the student’s cognitive readiness for learning and the types of cognitive demands that would be within their developmental capacity. The Developmental Zone framework acknowledges that cognitive abilities develop over time and that students may have different zones of readiness for various cognitive and academic tasks (Mather et al., 2025c).

Both types of zones encompass a range from 10 W units below an individual’s W ability (RPI 75/90) to 10 W units above an individual’s ability (RPI 96/90), with the lower and higher limits represented by age or grade equivalents for planning purposes. Tasks at the lower end of the range will be easy for the individual, while tasks at the higher end will be quite difficult. This technical framework provides the foundation for understanding a student’s zone of optimal challenge across different domains. These zones essentially provide a psychometric way to approximate what Vygotsky described theoretically as zones of proximal development (Vygotsky, 1978), offering concrete grade/age levels for instruction. Vygotsky’s framework guides the provision of effective support within these zones (Chaiklin, 2003).