The g factor, or general intelligence, is the common ability that explains why a person who performs well on one type of cognitive test tends to perform well on others as well. It is estimated using factor analysis of reasoning tests, and in personnel selection it is one of the most studied predictors of performance, although recent research has reduced its predictive power.
This article explains the origin of the concept, how it is measured, what meta-analyses say about its relationship to job performance, and how to use it judiciously, along with other evidence. For the general framework of measurement in selection, see the science behind the tests.
Table of Contents
What is the g factor of intelligence?
In 1904, British psychologist Charles Spearman published in the American Journal of Psychology the article «»General Intelligence,» Objectively Determined and Measured». He observed that scores on very different tasks were positively correlated with each other and proposed that performance on each task depends on a general factor, common to all, and on factors specific to each task. He called this common factor g.
This pattern of positive correlations between cognitive tests is known as manifold positive and has been replicated with many batteries and populations. It is important to understand clearly what is and what is not:
- Is a statistical construct: the variance shared by cognitive tests.
- It is related with reasoning, understanding complex ideas, solving new problems, and learning quickly.
- It's not a measure of a person's worth, nor of their personality, motivation or values.
- It's not a fixed and immutable figure: a specific score depends on the test, the scale and the conditions in which it was applied.
From Spearman's model to hierarchical models (CHC)
Spearman spoke of a general factor and specific factors. Later, other authors proposed broader capacities, such as Cattell and Horn's distinction between fluid and crystallized intelligence. Today, the most widely used reference is the Cattell-Horn-Carroll (CHC) model, which integrates these ideas into a three-level hierarchy, as described by [the author]. McGrew (2009) in the journal Intelligence:
| Level | What it contains | examples |
|---|---|---|
| Stratum III | General capacity (g) | General intelligence |
| Stratum II | Extensive capabilities | Fluid reasoning, comprehension-knowledge, visual processing, working memory, processing speed |
| Stratum I | Specific capabilities | Inductive reasoning, vocabulary, perceptual speed |
The hierarchical model has a practical consequence: a numerical reasoning test primarily measures g, and also a more specific ability. Therefore, two "intelligence" tests are not interchangeable, and it's advisable to know what broad skills the position requires before choosing one.
The role of g remains a subject of theoretical debate. Some authors see it as a real capability; others, as the statistical result of many overlapping processes. For selection practice, this discussion is less important than the question of whether the measure predicts anything useful.
How is the g-factor measured?
It is not measured with a single question or a single type of task. It is estimated from several cognitive tests:
- A battery is applied with diverse tasks: verbal, numerical and abstract reasoning (e.g., matrices), working memory or processing speed.
- The correlations are analyzed Among tests with factor analysis; the general factor is the one that captures what they all share.
- A composite score is calculated. which is interpreted in relation to a scale: the position of the person with respect to a reference population.
A full clinical battery is rarely used in selection. Usually, they are psychometric reasoning tests Shorter versions, which serve as an approximation to g. Their quality depends on the usual factors: reliability, evidence of validity for the intended use, and norms appropriate for the population being assessed. Here's a guide for choosing valid candidate assessment instruments.
Does the g factor predict job performance?
Yes, but less than was believed a few years ago. There are two key references.
Schmidt and Hunter (1998). Their meta-analysis in Psychological Bulletin reviewed 19 selection procedures and estimated a validity of 0.51 for general mental ability tests as a predictor of performance, the highest among methods applicable to people without prior experience in the position (Schmidt and Hunter, 1998). For two decades, this was the main argument for putting cognitive tests at the center of selection.
Sackett, Zhang, Berry, and Lievens (2022). In an article from Journal of Applied Psychology They reviewed how the meta-analyses had been corrected for range restriction and concluded that these corrections inflated the validity estimates. With their method, the estimated operational validity of cognitive ability drops to 0.31, and the structured interview takes first place, with 0.42, according to the summary published by the SIOP.
| Study | Estimated validity of cognitive ability | Practical reading |
|---|---|---|
| Schmidt and Hunter (1998) | 0,51 | Lead preacher in many positions |
| Sackett et al. (2022) | 0,31 | A useful predictor, among others of similar value |
A validity of 0.31 is still useful. What changes is the order of priorities: cognitive ability ceases to be the criterion that decides almost everything and becomes just one more piece of evidence, which gains value when the position requires learning quickly or solving new problems, and when it is combined with other measures.
Adverse impact and equity: what to consider
Studies conducted primarily in the United States show average differences between ethnic groups in cognitive tests that can translate into adverse impacts: one group being selected at a significantly lower rate than another. This is known as the validity-diversity dilemma.
A subsequent work by the same team, Berry, Lievens, Zhang and Sackett (2024), He recalculated this dilemma with the new estimates. His conclusion is that excluding general mental ability tests generally has little or no effect on the validity of a test battery, but substantially reduces the adverse impact. It does not eliminate the dilemma: according to the same study, achieving an adverse impact ratio of 0.80 still requires some loss of validity, although less than previously thought.
In Spain, the Law 15/2022, comprehensive law for equal treatment and non-discrimination, The law considers any criteria or systems for accessing employment that produce indirect discrimination to be discriminatory (Article 9.2). A test that disproportionately excludes a group, without a clear connection to the position, is a legal risk as well as a matter of fairness. Consult your legal advisor to see how this applies to your case.
How to use the g-factor prudently in selection
If you decide to include a reasoning test, These guidelines reduce risks:
- Justify your relationship with the position. Use it when job analysis shows that reasoning, learning, or solving new problems is central, not out of habit.
- Do not use it as the sole or eliminatory filter. A high cut-off point at the beginning of the process concentrates the adverse impact and excludes people who other evidence would have valued.
- Combine it with other measures. The structured interview, the knowledge tests, and the competency tests They provide information that cognitive ability cannot capture. More ideas on how supplement interviews with skills tests.
- Pay attention to the application conditions. Clear instructions, practice examples, and reasonable accommodations for people with disabilities.
- Review the results by group. If you have enough data, check if the process generates selection differences and correct the design.
- Leave the decision to the people. The result guides the interview and the team's deliberation; it does not replace it.
At Hirint, we apply that criterion to every assessment battery: several job-related tests, comparable results across candidates, and a decision always made by the team. selection.
Frequently Asked Questions
Is the g factor the same as the intelligence quotient (IQ)?
Not exactly. IQ is a specific score on an intelligence test, expressed relative to a norm. The g factor is the construct that these tests attempt to capture: the common ability across cognitive tasks. A good overall IQ is usually heavily weighted in g, but each test measures it with its own margin of error and its own mix of specific abilities.
Can the g-factor be improved with training?
Practicing a particular type of test can improve scores on that test due to familiarity with the format; whether this actually changes overall ability is much more debatable. Therefore, it's advisable to offer practice examples to all candidates equally: this way, the score reflects less prior experience with the format and more the ability being measured.
What weight should a reasoning test have in the decision?
It depends on the position. In roles with high cognitive complexity or requiring constant learning, it may carry more weight; in others, experience, technical knowledge, or soft skills tend to be more informative. With current estimates, it makes sense to treat it as one piece of evidence among several of similar value, and not as the deciding factor.
Is it legal to use intelligence tests in selection processes in Spain?
Psychometric tests are commonly used in recruitment processes, but their use must respect data protection, equal treatment, and the requirement to inform candidates, in accordance with applicable regulations. The test must be relevant to the position and not result in indirect discrimination. For specific questions, consult your legal advisor.
Is the g-factor useful for evaluating potential in inexperienced candidates?
It can provide useful information because it relates to learning speed. Even so, it should be combined with motivation tests, soft skills assessments, and a structured interview. This article on how Identifying the potential of junior talent It gives more ideas for those processes.
If you want to design a balanced assessment battery for your positions, you can see how Hirint works in a meeting with the team.