Intelligence testing

Disciplines > Human Resources > Selection > Intelligence testing

Description | Development | Discussion | See also

Description

There are three schools of thought about intelligence and consequently how it may be tested.

Uni-factor models

Uni-factor models have a single dimension, defining 'intelligence' as a single thing that is fixed, unchangeable and can be measured. They usually have a socio-biological basis, where intelligence is defined as a combination of genetic and social factors.

Spearman (1904) did a factor analysis of ability, identifying ‘g’ as general ‘psychophysiological’ intelligence.

Unifactor models allow for racial and national differences in intelligence. In once sense, this may be seen as racism or xenophobia. On the other hand, it also calls into question the notion of 'intelligence' as a single thing -- usually the defining race/nationality/gender measures themselves as a standard, and any difference with others is very likely to show the others to be inferior.

Multi-factor models

Multi-factor models identify intelligence as a combination of distinct abilities. They tend to place emphasis on the role of the environment in learning and see intelligence as dynamic and situated., rather than a fixed ability in all circumstances.

Thurstone (1938) identified nine primary factors of intelligence:

Words	Spatial	Perceptual
Verbal	Numerical	Memory
Reasoning	Deduction	Induction

J. P. Guilford identified a cube (5 x 6 x 6) of abilities.

Multi-hierarchical models

Multi-hierarchical models measure the application of intelligence, not just cognitive ability. They try to provide more organization than multi-factor models that seem unwield (eg. Guilford’s 180 cubelets).

Horn and Cattell (1966) defined five second-order factors, notably including fluid and crystallized intelligence:

Fluid intelligence: non-verbal, general reasoning. [tacit?]
Crystallized intelligence: application of verbal or conceptual knowledge. [explicit?]
Visualisation
Retrieval
Cognitive speed

Vernon’s model (1950s) describes intelligence as an equation:

General intelligence
= Verbal ability (verbal + numerical)
+ Spatial/Mechanical ability (Spatial + Mechanical)

Carroll’s model (1993) is more complex:

General intelligence
= Fluid intelligence (inductive reasoning + sequential reasoning)
+ Crystallised (lexical knowledge + foreign language aptitude)
+ Visual perception (visual imagery + perceptual integration)

Gardner’s model (1983) of seven intelligences is favoured by educationalists:

Linguistic
Musical
Logical-mathematic
Spatial
Bodily-kinaesthetic
Intrapersonal
Interpersonal

Furnham (1992) defined five factors that predict occupational behaviour:

Ability (skill in the basic job)
Motivation
Personality traits
Demographic factors
Intelligence

Discussion

General intelligence tools are not used that often as preferences are usually to find more distinct abilities. The main areas of testing (related to Guilford) are verbal, numerical, spatial, dexterity, sensory. Cognitive ability tests are good predictors of initial job performance, but decline over time.

It is important to examine the test manual, to ensure you use the tool correctly (many do not do this well).

Testing people with disabilities is a problem as there is a lack of information available around this area.

There is a strong relationship between job performance and general intelligence. Tests are good at assessing this – interviews don’t add that much. Tests are also more objective.

Many tests seek maximum ability and are developed in a clinical environment. Ackerman et al (1989) noted that typical ability is more relevant in job environments. In a related early research, Terman (1934) identified chronic (long-term) vs. acute (short-term) intelligence.

The classic IQ test has been widely criticized as being biased towards white, middle-class Western adult males.

Culture-free testing is important for global organizations. A well known test is Raven’s ‘Progressive Matrices’ (1965), which measures three ability levels: children and people with learning disabilities up to fluid intelligence. Abstract reasoning is measured via pattern (‘matrix’) with a part missing. It is confusing to follow and the norm table difficult to use. There is little evidence of removing bias towards majority groups.

Some researchers say spatial ability is not as good as verbal ability. Differences may (or may not) be caused by geographic and organisational contexts. The Big Five test is claimed as culture-free (McCrea and Costa) but is also doubted.