What the 2025 Cloze-Test Meta-Analysis Means for IQ Scores

What did the 2025 synthesis actually pool?

The 2025 meta-analysis pooled relationships between many cloze tasks and cognitive ability measures; it did not pool one standardized cloze subtest or derive a universal conversion from cloze answers to IQ. A cloze task is a passage with deliberately omitted words that examinees supply. The included studies shared a broad format, but not necessarily identical passages, rules, scores, or purposes.

In *Cloze test performance and cognitive abilities: A comprehensive meta-analysis*, the synthesis covered 89 studies, 37,912 participants, and 634 effects across 110 years. The authors used a multilevel random-effects method to combine manifest correlations. Here, *r* is a correlation: it describes how scores vary together. It does not show that two scores agree, can be substituted for one another, or that one causes the other. The 634 effects are not 634 independent studies.

The reported average association was *r* = .54 with crystallized intelligence (485 effects; 95% confidence interval [.49, .59]), *r* = .48 with fluid intelligence (69 effects; [.42, .54]), and *r* = .61 with general intelligence (32 effects; [.46, .77]). Here, *k* labels the number of effects contributing to a result, not the number of studies. The general-intelligence estimate has fewer effects and a wider interval, so its higher average is less precise. Each interval describes uncertainty around a pooled association, not a range for an individual examinee.

These findings make it more plausible that cloze performance can carry cognitive information, particularly information related to crystallized ability. They change the useful question from whether cloze performance can relate to cognitive ability to what a particular cloze score justifies. Its long, century-wide evidence base does not establish a current norm group or validate a specific test’s scoring and interpretation.

Sources: Cloze test performance and cognitive abilities: A comprehensive meta-analysis

Why does the authors’ task-format distinction matter?

Cloze names a response format, not a single mental ability. In a cloze task, words or other units are omitted from a passage and an examinee supplies what is missing. The passage, omissions, response options and scoring rule shape what performance means. Tests sharing the label can call on different combinations of language knowledge, context use, inference or subject familiarity. “Cloze Test Performance and Cognitive Abilities: A Comprehensive Meta-Analysis” therefore informs a family of tasks, not one IQ score. The route from stimulus to score has several links. Passage content provides context, but it can also bring in knowledge the test maker did not intend to measure. An omission may ask the examinee to retrieve a word from vocabulary, use grammar to identify a fitting form, connect ideas across sentences, or infer a missing relation. These demands can overlap. A cloze component deliberately designed to sample cognitive ability may use passages and omissions for a different purpose from a reading-comprehension cloze, which is intended to assess understanding of text. The shared format does not make their targets identical. Response rules change the task further. If the examinee must produce an answer without prompts, retrieval and word production are involved; if the examinee selects from supplied alternatives, recognizing a fit among those choices may matter more. The scoring rule then decides whether only one exact answer counts, whether alternatives are accepted, or whether points are assigned in another way. These decisions shape the score and need to fit its intended interpretation. A score alone cannot show which mental route produced a correct answer. “Cloze Test Performance and Cognitive Abilities: A Comprehensive Meta-Analysis” treats cloze as a format, not a construct, and reports response type as the only significant moderator among those examined. A moderator analysis asks whether associations differ across coded study features. That makes response format relevant; it does not prove that passage base, deletion pattern or administration mode never matter. A non-significant result can reflect the evidence available, including how features were coded and how much information the studies supplied. It does not establish equivalence. This distinction also clarifies how to read the pooled relationships. An average association across varied cloze studies can support the proposition that cloze performance often carries information related to cognitive measures. It cannot identify which feature caused that association, establish that every cloze task measures the same thing, or show that an individual test’s score conversion is justified. That requires evidence linking a test’s items, response demands and scoring to its interpretation for the examinees. The Standards for Educational and Psychological Testing frame validity as evidence for specified score interpretations and uses, not as a property proved by a label. So when an IQ test includes a cloze component, “cloze” is only a starting description. Ask what the passage requires, how answers are elicited and scored, and why those performances support the stated construct. Language and prior knowledge may be part of the intended ability being sampled, or they may complicate a narrower claim; the label alone cannot decide. The 2025 synthesis strengthens the case for studying cloze as a potentially informative format. Its call for design-specific construct research is also the boundary: pooled relevance does not transfer automatically to a particular component, norm group or IQ interpretation.

Sources: Cloze test performance and cognitive abilities: A comprehensive meta-analysis; Standards for Educational and Psychological Testing

How can a cloze response combine different demands?

A correct completion does not reveal by itself why the examinee supplied it. Depending on the passage and the way blanks are chosen, an answer may draw on knowing a word, recognizing grammatical fit, using nearby context, or combining clues across a text. Those processes can overlap. A score records responses under the task’s rules; explaining what the score represents requires evidence about the task and the people who took it. Turner’s study, “The underlying factor structure of L2 cloze test performance in Francophone, university-level students,” makes this point in a defined setting. Eight cloze tests were given to 182 Francophone university-level students learning a second language. The study compared models of the relationships among language knowledge, content-domain knowledge, contextual constraints, and cloze-taking ability. Its abstract reports that a three-factor model best explained the data, with language factors and nonlinguistic specific knowledge contributing to performance. The finding challenges a simple equation between cloze performance and overall second-language proficiency in that sample. A model’s fit is evidence within that study, not proof that each individual response has separable causes. Because the study examined second-language tests in one university-level Francophone sample, it does not establish that the same factor structure applies to every cloze task, to first-language readers generally, or to an IQ test. Consider an illustration, not a test item or a finding from Turner’s participants. Imagine a short passage about a gardener who checks the soil before watering a plant. One blank could be completed with a familiar word that fits the sentence. Another might require noticing that the sentence calls for a verb rather than a noun. A later blank might depend on remembering that the soil was described as dry several lines earlier. If the passage uses an unfamiliar gardening term, prior topic knowledge could also help. A correct answer could therefore reflect several routes at once; the visible response does not identify which route was decisive. This matters when a test is intended to measure a particular ability. A task may deliberately use language or context because those demands belong to the ability being assessed. Alternatively, a demand may interfere with the intended interpretation if it affects scores for reasons outside that target. Whether an additional demand is relevant or problematic depends on the test’s stated purpose, its examinees, and evidence supporting the proposed score meaning. The mere presence of language does not automatically invalidate a cloze task, just as calling a task “reasoning” does not prove that language plays no part. The “Standards for Educational and Psychological Testing” frame validity as evidence and theory supporting a specified interpretation of scores for a proposed use. Applied here, that means asking what the cloze items require, how responses are scored, and why those results support the claimed interpretation for the intended examinees. If a test labels a cloze result as IQ, its developer needs evidence linking those actual response processes and scores to that interpretation. Turner’s study offers a reason to examine multiple contributors; it does not supply that test-specific link. Readers should therefore treat the score as performance on the cloze task unless evidence for that instrument, population, scoring, and use supports a broader conclusion.

Sources: The underlying factor structure of L2 cloze test performance in Francophone, university-level students; Standards for Educational and Psychological Testing

What does a bounded adult comparison add?

A bounded adult comparison helps answer a narrower question than “does cloze measure IQ?” The study titled “Explorations of crystallized intelligence: Completion tests, cloze tests, and knowledge” examined 167 adults ages 18 to 69, each with at least some college experience. It compared cloze and completion tests with measures of crystallized ability and four knowledge areas: biology, U.S. history, U.S. literature, and technology. The accessible publisher abstract reports that both formats accounted for variance in the crystallized-ability and knowledge measures. It also reports that completion tests accounted for all the variance in those criteria accounted for by the cloze tests, and added incremental predictive validity in both domains. In plain terms, in this study the completion measures contributed information about those outcomes beyond what the cloze measures contributed. That is a result about these tests, this sample, and these criteria; it is not a general ranking of formats. The distinction matters because “accounted for variance” describes statistical association in the study’s model. It does not mean that completing a word or passage caused crystallized ability or knowledge, nor does it show that a completion test is always better. The sample also sets a boundary: adults across a broad age span were included, but everyone had at least some college experience. The result therefore cannot, on its own, establish how the same tasks perform for adults with different educational backgrounds, for younger examinees, or for every language community. The 2025 “Cloze test performance and cognitive abilities: A comprehensive meta-analysis” answers a complementary question at a broader scale. It synthesized associations from 89 studies, 37,912 participants, and 634 effects accumulated across 110 years of research. Its average correlation was .54 with crystallized intelligence, based on 485 effects; .48 with fluid intelligence, based on 69 effects; and .61 with general intelligence, based on 32 effects. A correlation describes how scores vary together, not whether two scores are interchangeable. The synthesis estimates recurring relationships across varied tasks, whereas the adult comparison shows that two related response formats can contribute differently when tested against named ability and knowledge criteria. These findings are not contradictory. The pooled analysis makes it more plausible that cloze performance can carry cognitive information across research settings. The adult comparison shows why that broad association cannot decide what one cloze component means: task design and the selected comparison measures matter. Nor does the older study validate a present-day IQ test, provide a score conversion, or establish norms for a cloze result. Taken together, the studies support a practical inference: use pooled associations to motivate closer evaluation, then inspect evidence for the exact subtest, scoring, population, and intended interpretation. A broad average can identify a promising relationship; it cannot substitute for evidence about the instrument in front of the reader.

Sources: Explorations of crystallized intelligence: Completion tests, cloze tests, and knowledge; Cloze test performance and cognitive abilities: A comprehensive meta-analysis

When can cloze design make a task-specific interpretation stronger?

A cloze measure can support a focused interpretation when its gaps and scoring are designed for a defined ability and the resulting scores are evaluated against evidence suited to that ability. A close relationship with a reading-comprehension measure, for example, can support a reading-comprehension interpretation. It does not by itself establish that the score measures IQ. The 2013 study “Cloze Tests May be Quick, But Are They Dirty? Development and Preliminary Validation of a Cloze Test of Reading Comprehension” shows how design can target a purpose. Its authors argued that gaps can be selected so that filling them requires understanding ideas across a text, not only predicting a nearby word. They report a preliminary study of 204 Danish adults attending reading courses or general education. Participants took a 10-minute comprehension-focused cloze measure and a 30-minute question-answering comprehension test. The scores correlated at r = .84. The abstract says decoding ability and vocabulary accounted for only a small part of the correlation. This supports a bounded conclusion: in this sample, the cloze measure tracked a longer measure intended to assess reading comprehension. It does not show that every cloze test measures comprehension equally well, or that a cloze score is an intelligence score. Correlation describes co-variation, not interchangeability or causation. Here, the comparison criterion was another reading-comprehension test, not a measure of general intelligence. Calling the result evidence of IQ validity would extend the finding beyond the comparison the study made. The authors also note a qualification: their cloze measure was somewhat more sensitive to decoding ability than the question-answering test. This qualification matters when deciding what the score represents. A measure may be useful for its intended purpose while also reflecting another skill. If the score is meant to represent comprehension, a decoding contribution merits examination. If it is meant to represent a broader ability, the test needs evidence and a rationale for that broader interpretation. The “Standards for Educational and Psychological Testing” give the general rule: validity concerns how far evidence and theory support a specific score interpretation for a given use. They say validation begins by specifying the intended interpretation, construct and proposed use. “Performance on a comprehension-focused cloze test” and “IQ” are distinct interpretations. Evidence linking the first score to another comprehension measure bears on the first claim; the second requires its own supporting argument, suited to the instrument and examinee population. For an IQ claim, that argument would need to address the actual test’s design and scoring, the cognitive construct claimed, and evidence connecting its scores to relevant intelligence measures. It would also need support for any conversion to a broader score meaning. The Standards do not prescribe one identical checklist for every test; relevant evidence depends on the proposed purpose. This study supports neither blanket dismissal nor automatic acceptance; it shows the value of purpose-specific design, while an IQ interpretation still needs separate evidence.

Sources: Cloze Tests May be Quick, But Are They Dirty? Development and Preliminary Validation of a Cloze Test of Reading Comprehension; Standards for Educational and Psychological Testing

What does the meta-analysis add—and what does it leave open?

The 2025 synthesis shows cloze performance covaries with measured cognitive abilities. It leaves open how any particular cloze test generates its score and what that score can mean. A correlation describes an association between measures; it does not show that scores are interchangeable, that one caused the other, or that a result converts to an IQ scale. The meta-analysis pooled 89 studies, 37,912 participants and 634 effects over 110 years. Average correlations were .54 with crystallized intelligence, .48 with fluid intelligence and .61 with general intelligence. The latter drew on 32 effects and had a broad 95% confidence interval, .46 to .77. These relationships are not norms for a current test. (“Cloze test performance and cognitive abilities: A comprehensive meta-analysis.”) A focused comparison differs. “Explorations of crystallized intelligence: Completion tests, cloze tests, and knowledge” examined 167 adults aged 18 to 69, all with at least some college experience. Particular completion and cloze tests were compared with crystallized-ability measures and knowledge in biology, U.S. history, U.S. literature and technology. In that sample, completion tests accounted for all the variance in these criteria accounted for by cloze tests and added incremental predictive validity. This is evidence about those test sets and outcomes, not a general verdict that completion always outperforms cloze or that cloze lacks cognitive relevance. Its sample and criteria provide detail the meta-analysis cannot, while also limiting how far the result travels. A second study helps show that cloze performance can reflect mixed demands. “The Underlying Factor Structure of L2 Cloze Test Performance in Francophone, University-Level Students” analyzed eight second-language cloze tests completed by 182 Francophone university students. Its three-factor model linked performance to language factors and nonlinguistic, domain-specific knowledge associated with cloze-taking ability. This concerns second-language performance in that university sample, not IQ validity. It illustrates why an answer pattern does not identify its own cause: language and other knowledge may contribute, depending on the task. These sources are complementary. The meta-analysis asks whether cloze scores tend to relate to cognitive measures across research; the adult comparison examines defined formats and criteria in one sample; the language study analyzes contributors to performance on particular second-language tasks. Each informs a different part of interpretation. None substitutes for evidence about another instrument. The meta-analysis spans a century of research, and its database predominantly comes from language research, where cloze has often measured language proficiency. Its authors propose that the format could be tailored to assess specific cognitive abilities, but call for systematic studies of how design affects construct validity. That qualification matters. The findings do not show that cloze cannot contribute to intelligence assessment, and they do not establish that every cloze score measures IQ. They make cognitive relevance plausible and sharpen the next question: does evidence for the exact task, score and examinee group support the intended interpretation? Historical pooled associations are not contemporary norms. The update is promising background evidence, still insufficient for a particular IQ score claim.

Sources: Cloze test performance and cognitive abilities: A comprehensive meta-analysis; The underlying factor structure of L2 cloze test performance in Francophone, university-level students; Explorations of crystallized intelligence: Completion tests, cloze tests, and knowledge

What evidence bridges association to an IQ interpretation?

The bridge is an evidence-backed argument for the specific score meaning and intended use, bounded to the examinees and conditions it covers. The Standards for Educational and Psychological Testing define validity in terms of how well evidence and theory support a proposed score interpretation for a particular use. The question is whether this score, produced in this way, supports this conclusion for these examinees and this decision. Trace the argument from task to use. The test developer should state the intended ability and population. “Cloze” identifies a format in which missing words are supplied; it does not specify whether the target is reading comprehension, vocabulary, reasoning, or broader cognitive ability. The items and response demands need to fit. If language knowledge is intended, its role should be clear; if incidental, evidence should address whether it weakens the interpretation. Next comes scoring. A reader needs to know how responses become a raw result and whether the instrument reports a cloze subscore, combines it with other tasks, or interprets a composite as IQ. Each step adds a claim. Evidence that a cloze score relates to cognitive measures may support task relevance, but does not justify the weights or aggregation used in a composite. Evidence for a composite also does not establish that its cloze component is a standalone IQ measure. The 2025 meta-analysis, “Cloze test performance and cognitive abilities: A comprehensive meta-analysis,” contributes evidence at the relationship stage: it reports average associations between cloze performance and crystallized, fluid, and general intelligence measures across varied studies. This makes cloze worth investigating. But the synthesis combines tasks and measures; it does not establish the scoring model or intended meaning of one test. A pooled relationship is not a conversion table and does not show that scores are interchangeable. For a norm-referenced IQ interpretation, evidence must support the comparison and precision claimed. A norm is a defined reference group used to interpret performance relative to others; that group and the score conversion must fit the intended examinees. A pooled correlation supplies neither the reference group nor an individual score’s uncertainty. The meta-analysis confidence interval describes uncertainty around a pooled association, not a range for one examinee’s IQ. Administration and population matter. Timing, instructions, response mode, language demands, and testing conditions may affect the performance being interpreted. Evidence should support the interpretation for the relevant population and conditions, rather than assume findings from another language group or task version transfer unchanged. The Standards treat these as considerations, not a universal checklist. Support should fit the claim and its consequences. Finally, state the proposed use. Describing performance on a low-stakes task requires less than a decision about diagnosis, placement, or eligibility. The meta-analysis strengthens one part of that case by showing associations between cloze performance and cognitive measures. It cannot alone establish how a particular test constructs its score, whether its norms fit a reader, how precise its estimate is, or whether it supports a consequential decision. Without those links for the exact instrument and population, describe task performance only.

Sources: Standards for Educational and Psychological Testing; Cloze test performance and cognitive abilities: A comprehensive meta-analysis

Which links in the interpretation chain remain unproven?

A correlation with cognitive measures does not show how a raw cloze score maps to an IQ scale, which comparison group supplies that conversion, how precisely it estimates performance, or whether the interpretation fits a population. The 2025 meta-analysis, “Cloze test performance and cognitive abilities: A comprehensive meta-analysis,” reports pooled correlations between varied cloze tasks and cognitive measures. Correlation describes co-variation: higher scores on one measure tended to accompany higher scores on another. It does not tell a reader what IQ value to assign to one person’s number of correct completions. That requires evidence about the particular test and scoring. A raw score is the count or summary produced by applying a test’s scoring rules to a person’s responses. A norm-referenced score interprets performance relative to a defined comparison group through a supported conversion. A raw cloze score describes performance on those passages under those rules; alone, it cannot say where that performance falls among adults, students, or another population. A norm comparison depends on who supplied the reference, whether that group is relevant to the examinee, and whether the conversion belongs to the exact test version. The meta-analysis’s pooled relationship supplies no individual score conversion. Precision is a separate link. A score can be calculated consistently yet fail to support its interpretation. Reliability concerns consistency or measurement precision under specified conditions; validity concerns whether evidence and theory support a proposed interpretation and use. Neither follows automatically from a pooled correlation. A stable cloze score would not become an IQ score merely because cloze measures correlate with intelligence measures. A cloze component could contribute information inside a broader battery if that test’s own evidence supports the role. Association alone does not establish the component’s weight, how scores are combined, or whether a stand-alone interpretation is warranted. The confidence interval in the meta-analysis answers a different question from an individual score interval. It describes uncertainty around a pooled association estimated from included studies. It does not give an examinee’s likely score range or show the precision of a specific cloze test. Treating the intervals as interchangeable would move group-level evidence to an individual without test-specific information. The Standards for Educational and Psychological Testing frame validity around evidence and theory supporting interpretations of scores for proposed uses. This makes the missing links visible: what score was produced, how it was derived, which examinees the interpretation covers, and what decision it is meant to inform. Before treating a number or band as a population placement, ask for the test version, scoring basis, reference population, and evidence for the intended interpretation. If those are not supplied, describe the result as performance on that cloze task. The meta-analysis makes cognitive relevance plausible; it does not complete the path from task performance to an IQ comparison.

Sources: Standards for Educational and Psychological Testing; Cloze test performance and cognitive abilities: A comprehensive meta-analysis

How should a reader use a cloze-based result now?

For curiosity, treat a cloze result as performance on the task you were given. Review the passage, omissions, response rules and your answers. This can show where reasoning succeeded or where context, vocabulary or grammar made an item difficult. It does not, by itself, say what your IQ is. The 2025 meta-analysis makes cloze relevant to cognitive research, but pooled associations do not convert an individual task score into an IQ score. If a test reports an IQ interpretation, ask for its exact version and score, the population used for comparison, and evidence supporting that interpretation for its intended use. The Standards for Educational and Psychological Testing frame validity as support for a specified interpretation and use, not a permanent property of a test label. Evidence about a cloze subscore also does not automatically establish the meaning of a larger composite. For diagnosis, educational placement, accommodations or another consequential decision, consult a qualified professional who can select an appropriate assessment and consider multiple sources of evidence. A browser quiz cannot substitute for that process. Test IQ Free offers a separate low-stakes practice option: its adaptive 50-question test covers pattern, number and spatial reasoning, not cloze. Its provisional estimated IQ range comes from an unvalidated conversion with no validated adult reference population; it is neither a normed comparison nor a confidence interval. Use the free test as a challenge and review its worked answers, while treating the range as provisional performance feedback. The optional $9 report adds explanation and practice structure, not accuracy or validation. Carefully designed cloze tasks may be relevant evidence, but an unsupported result is not a statement about identity, diagnosis or fixed ability.

Sources: Standards for Educational and Psychological Testing

What is the proportionate next step?

Treat a cloze score as observed performance unless evidence for that instrument supports an IQ interpretation for the relevant people and use. The 2025 meta-analysis, “Cloze test performance and cognitive abilities: A comprehensive meta-analysis,” makes cloze performance’s cognitive relevance plausible. But a task-format association cannot supply a test’s scoring conversion, reference group, precision, or justification for a decision. The Standards for Educational and Psychological Testing frame validity around evidence and theory supporting a specified score interpretation and proposed use. That is the bridge an IQ claim needs.

For curiosity, inspect the task and review its answers; keep the result within the evidence. Test IQ Free offers a free adaptive 50-question challenge across pattern, number, and spatial reasoning. Its estimated IQ range comes from an unvalidated conversion with no validated adult reference population, so it is neither a normed comparison nor a confidence interval. The optional $9 report adds explanation and practice structure, not accuracy or validity. For placement, diagnosis, accommodations, or another consequential decision, consult a qualified professional who can select and interpret a suitable assessment. Take the meta-analysis as a reason to examine cloze tests carefully, not as permission to treat every cloze score as IQ.

Sources: Standards for Educational and Psychological Testing

Questions readers ask