Introduction to the WAIS III
Richard Niolon, Ph.D.

This is the third edition of the WAIS, with the first edition coming out in 1955, and the revised (WAIS-R) coming out in 1981. Before the WAIS, however, there was the Wechsler Bellevue test (released in the 1930s and revised in the 1940s). He developed a children's version in 1949, and the popularity of it helped increase the popularity of the adult version, which he released again in 1955. By the 1960s, it was more popular than the Stanford Binet. This was in part because of the standard scores that allowed comparison across testings and ages (as we discussed when we talked about the Stanford Binet), as well as because of what it allowed psychologists to do. It provided a Full Scale IQ, or a one number estimate of someone's cognitive functioning. It also provided smaller numbers that were estimates of verbal functioning and visual-motor functioning, and differences between these numbers were helpful in detecting and diagnosing learning disabilities.

Basic Information
Basically it works like this. The test has 14 sections (called "subtests"). Each subtest begins with some extremely easy questions or tasks (collectively called "items"). You start with the hardest of the extremely easy ones which 95% of the population should be able to answer or do correctly (or "pass"). If the client fails either of the first two, however, you back up (the "reverse rule"), giving even easier items as a way to make sure they know how to do the subtest. You then return to items that get progressively harder, and you keep going until they miss a predetermined number (the "discontinue rule"). At the end of the test, you tally the scores for their answers on each subtest (your first set of numbers, called "raw scores"). You then convert these to a second set of numbers (called "scaled scores"), so we can compare performance on subtests. You then add these to come up with a third set of numbers (called "sums of scaled scores") which represent aggregate abilities. These are converted to a fourth set of numbers (called IQ or Index scores) that are standard scores (mean of 100 and standard deviation of 15) and so we can compare the abilities. These are the numbers that really tell you about a person's abilities and performance.

The WAIS-R was revised to the WAIS III for several reasons. Even if you don't know anything about the old test, you can still appreciate why they would update the test:
  1. it is important to update the norms, as people get higher scores every 10 years by three points on the Wechsler series of tests (Flynn effect)
  2. the age range of the norming sample needed to be expanded, since people live longer and get services at later ages, and some changes to the test itself were needed
  3. the items needed to be modified to make them more culturally up-to-date The Little King
  4. the artwork needed to be changed ("the little king" is gone because no one remembers the cartoon any more)
  5. the WAIS-R was criticized due to floor and ceiling effects, so this had to be fixed by adding more really easy items and really hard items
  6. the WAIS-R was criticized due to problems associated with the time-limit bonuses working against older clients, so scoring was changed on to minimize time bonuses
  7. the WAIS-R was criticized due to the lack of a measure for fluid reasoning, or on the spot reasoning processes
  8. the factor analysis needed to be updated
To assess cultural bias, WAIS III items were given to 200 African-American and Hispanic-American clients without discontinue rules; items which minorities seldom answered correctly were thrown out.

Computation of IQ scores is now based on comparison to age groups (thus, if you are 37, you get compared to other 35 - 44 year olds), rather than a standard age sample (like 20 - 34), like the WAIS-R. This changed because age may not matter for some tests (more the VCI and POI), and such comparisons can hide or pathologize the normal effects of aging (more for WMI and PSI). For example, the mean Digit Span score for the 70-74 year old range was 5.5 as opposed to 7 for the 20-34 year old group. In fact, Index scores (see below) start to diverge at age 30; by 45-54, POI and PSI have dropped by 12-14 points below WMI and VCI, and by 65-69 would be 10 points apart and span a 40 point range in the normal older adult if we still compared everyone to 20-34 year olds.

So, the sample includes 2450 adults, in 13 age groups ranging from 2 to 10 years wide (16-17… 35-44), which is a 50% increase in sample size since WAIS-R. At least 30 people were in each educational level (five of them) for each age group, and there were an equal number males and females.

Percentages in WAIS III sample and population are below:
Age White (pop) White (sam) AA (pop) AA (sam) H (pop) H (sam)
16-17 64.84 69.5 15.76 15.00 11.76 11.00
25-29 69.05v 71.5 13.47 13.00 13.46 11.50
30-34 71.83 75.5 12.67 13.00 11.93 7.00

Index Structure
The IQ and Index scores mentioned earlier allow you to summarize and compare large ability areas reliably. There are actually three IQs and four Indexes we compute. The Full Scale Intelligence Quotient (FSIQ) score is the g number. The "old way" of analyzing the test included "dividing" the FSIQ into two factors, a Verbal IQ and a Performance IQ. However, updates to the test and additional subtests lead to a factor analysis resulting in four factors at the last update, and these factors form the four Indexes: Verbal Comprehension, Perceptual Organizational, Working Memory, and Processing Speed. We will focus mostly on the FSIQ an four Indexes, but will learn a little about the Verbal IQ and Performance IQ as well this semester. I'm thinking the Verbal IQ and Performance IQ will disappear from the next version of the test, but you should know the basics about it for now. Below, I discuss the Index, and the subtests combined to make them:

  Verbal Comprehension:

This is the first and most reliable Index. The Verbal Comprehension Index is a measure of general verbal skills, such as verbal fluency, ability to understand and use verbal reasoning, and verbal knowledge. It is based on both formal and informal educational opportunities, and requires understanding words, drawing conceptual similarities, and knowledge of general principles and social situations.


  Perceptual Organization:

This is the second most reliable Index. The Perceptual Organization Index is a measure of non-verbal and in-the-moment reasoning. It assesses ability to examine a problem, draw upon visual-motor and visual-spatial skills, organize thoughts, create solutions, and then test them. It can also tap preferences for visual information, comfort with novel and unexpected situations, or a preference to learn by doing.


  Working Memory:

The Working Memory Index (WMI) assesses ability to memorize new information, hold it in short-term memory, concentrate, and manipulate that information to produce some result or reasoning processes. It is important in higher-order thinking, learning, and achievement. It is important for cognitive flexibility and planning ability, as well as learning and ability to self-monitor. WM is similar to STM, but STM is traditionally seen as a passive process (sometimes called simple span of memory), whereas Working Memory is an active process (sometimes called complex span of memory).

Working Memory (WM) is a temporary storage and workspace in the brain, "the blackboard of the mind" according to Just and Carpenter, that allows for processing of moment-to-moment information, archived information, and a link between them both before storing new knowledge in long-term-memory. It is linked to:

  • arithmetic skill (think "carrying" a number in complex addition)
  • reading ability (think holding previous words in the sentence in memory while you sound out the next one)
  • verbal fluency (think about considering the impact of your word choice before or while you speak a sentence)
  • problem-solving and adapting (think monitoring progress at a problem, rating how close you are to a solution, redirecting your workflow to pursue a different course if you are not succeeding, and speeding up or slowing down your work speed to finish on-time)
  • possibly to ADHD (think the ability to integrate some past instructions with current stimuli deemed important, while ignoring other stimuli and information deemed unimportant)
  • possibly schizophrenia (think ability to take in new stimuli about the world, and compare and balance it with past experiences of logical and sequential events)
Working memory is thought to have three parts -
  • the "phonological loop" which serves an auditory processing and practicing function, which seems roughly tied to the speech centers of the brain and language development… Remember Miller's famous 7+2?
  • the "visuospatial sketchpad" or a visual analysis and processing part that holds 3 or 4 objects, and serves as an internal laboratory for visualizing problems and solutions… Remember Einstein's Theory of Relativity was largely developed in his head with complicated visualizations?
  • the "central executive" that controls these two parts, delegating work to them and receiving their analysis back in order to facilitate coordinated reasoning, comprehension, learning, and decision making, sometimes allowing rote habit to direct us while it works on other things
Students that excel in reading or math process letters and numbers differently. Thus, there might be another subdivision or different types of WM. WM appears to be related to the prefrontal cortex (the very front of the frontal lobes). It exist in monkeys too, but only after about two to four months of age. Monkeys are shown a toy or bit of food, which is covered and then the monkey is distracted. The monkey looks back and chooses the cover with the food or toy under it, much like Piaget did with children.

There are four clear things we can say about WM:
  1. WM is a real construct:
    • While some have said WM is g, this is not so. WM is correlated with general intelligence, ranging from .30 to .80 depending on the control for content overlap, but the controlled studies average .48, or 23% of the variation in general intelligence
    • WM seems to be an ability that can be strengthen, at least with some kinds of tasks
    • The WM subtests have good specificity, meaning they are more s than g

  2. WM develops over time like other cognitive abilities:
    • Executive functions like "talking through something" in children initially are conducted aloud, but by age 6 years begin to be internalized to subvocal and finally silent modes by age 9-10
    • Recall for single units of spatial information (e.g., where on a screen a single dot appears) develops at age 11-12
    • Recall for multiple units of spatial information (e.g., a sequence of dots tapped by the examiner) develops around 13-15
    • Self-organized strategies (finding hidden dots with an efficient strategy) develops around 16-17
    • About 50% of the growth in fluid intelligence from age 7-19 is associated with WM, and 75% of that is associated with PS development.

  3. WM seems to have a neurological underpinning:
    • WM seems based in the pre-frontal cortex, and improvements in WM are associated with improved GMR there as well. Human and animals studies can also show similarities in immature and damaged cortexes, as well as repair to performance with injections of Dopamine into the area.

  4. WM is a part of "intelligence":
    • PS in part determines how long it takes information to be called up from long-term memory. On the one hand, slowed PS can mean information already in WM deteriorates while it waits for information from long-term memory to arrives. Thus, WM has to delegate rehearsal of the information to one of the two subsystems in order to continue its work. On the other hand, faster processing (quickly matching shapes, recalling memorized information like your times tables, letter and number recognition, and sequencing skills) can mean more economical processing, and that WM (and the skills that draw upon it like reading and math) is more effective.
    • Developments in PS support developments in WM, which support developments in Gf, although this is not a simple step-wise process.


  Processing Speed:

The last Index, The Processing Speed Index (PSI), assesses skills focusing attention and quickly scanning, discriminating between, and sequentially ordering visual information. It requires persistence and planning ability, but is sensitive to motivation, difficulty working under a time pressure, and motor coordination too. It is related to reading, mathematical, and memory skills as well. Cultural factors seem to have little impact on processing speed.

Processing Speed (PS) refers to the speed at which cognitive processes can be carried out. Faster is more efficient because it improves the power of the WM in two ways. First, WM can recall information and do more with it before it deteriorates (think about a multiplication problem where you have to recall both 5x9 and 6x8 from your stored multiplication tables). Second, WM can integrate more complex and complete concepts before storing them in long-term memory. In fact, differences in PS account for much of the differences in mathematical ability (Bull and Johnston, 1997).

You might assume that a PET or MRI scan of the brain should show lots of activity for very smart people. Not so. For simple reaction time tasks (e.g., press a button when a bird appears on the screen), high and low IQ people perform the same. However, when you add a decision to the task (e.g., press a button when a yellow bird appears on the screen, but not when any other color bird appears), higher IQ people have high activity which diminishes with practice much faster than with low IQ people. This works on the WAIS too - higher g loaded subtests (the more demanding ones) correlate .79 with the level of glucose metabolized in the brain.

There are four clear statements we can make about Processing Speed (PS):

  1. PS is a real construct:
    • People with faster PS scores perform the tasks at faster rates (in other words, there is good external validity)
    • Reaction time to a novel stimulus at age 3.5 months correlates .51 with the same at 4 years (in other words, PS is reliable and stable over time)
    • PSI subtests had good specificity in the WAIS III and WISC III and the WISC IV probably increases on this

  2. PS develops over time like other cognitive abilities:
    • Developmental changes in Gf correlate .53 with changes in PS (in other words, as other cognitive abilities develop, so does PS). Age differences in reasoning, spatial memory, and episodic memory share 70% of their variance with PS, and age differences in recall of factual information shares 46% of its variance with PS (more good news about validity).

  3. PS seems to have a neurological underpinning:
    • Decreases in PS are associated with deterioration of white matter and decreased GMR
    • Processing speed is lower in pre-term children compared to full-term children, and this difference is detected as early as 5 months of age, and maintained as late as late as 11 years (Rose and Feldman, 1996). For 11 year olds, this amounts to about 6 points in PS on the WISC III. The graphs for the cognitive development of normal and premature births are parallel, meaning they develop at the same rate starting at where they were at birth. The difference doesn't show up in simple reaction time (as noted above), but premature children never catch up to full-term children when it comes to complex decisions and seem to need more time to process the stimuli if they are to reach equal performance. There seems to be three plausible explanations for this (Rose, Feldman, Jankowski, 2002):
    • the loss of oxygen in the brain (especially the hippocampus, thalamus, and basal ganglia) due to respiratory distress and sometimes mild hemorrhages
    • the loss of special chemicals like docosahexaenic acid which builds up during the last trimester of birth
    • the impact of the environment on their immature sensory system
    • PS is the best predictor on the WISC for severity of childhood brain damage, with a 70% hit rate, compared to 65% PO, 55% VC, and 45% WM
    • Processing Speed is lower in patients with multiple sclerosis, which results from the immune system breaking down and attacking the myelinated sheath around nerve axons (Demaree, DeLuca, Gaudino, and Diamond, 1999)
    • PS is lower in elderly drivers with accidents. Processing speed tasks actually are better predictors of accidents in elderly drivers than are visual acuity tests. Intervention programs that provide training to improve processing speed have been shown to decrease accidents in elderly drivers
    • PS is lower in children with LD, reflecting problems in attention, writing, rate of learning, and fatigue, often with Cd falling below SS because of the motor problems
    • Children with ADD score lower than children with ADHD, who also score below normal
    • Children with depression also obtain lower PS scores

  4. PS is a part of "intelligence":
    • Reaction time to a novel stimulus at age 3.5 months correlates .37-.56 with FSIQ at age 4 years - the idea is that infants who quickly process the stimulus also quickly habituate, and thus do not mistake the stimulus as novel the next time they see it
    • PS is related to verbal fluency, or the speed and ease with which words are generated
    • Children with reading problems perform at a 10-15% slower rate than good readers, and PS is separate from phonological awareness (the idea is that PS relates to automated recognition of letters and words, and quickly making a decision based on their meaning)
    • FSIQ (and thus PS) account for 32% of the variance in reading comprehension and 8% of the variance in word recognition, whereas phonological awareness accounts for 20% and 39% of reading comprehension and word recognition respectively (the idea is that PS helps with the "rapid recognition of letters and words and the access to and the simultaneous integration of semantics, syntactics, and text-level information")


Reliability and Validity

The IQs and Indexes are the most reliable numbers we generate from the WAIS III:
Test-Retest Reliability
Full Scale
Verbal IQ
Performance IQ
Verbal Comprehension
Perceptual Organization
Working Memory
Processing Speed

As you can see above, the reliabilities are very good (1.00 would be considered perfect), and the Standard Error of Measurement (SEM) is generally small. Most subtests have test-retest reliabilities between .81 to .94, with a few falling lower. As to practice effects over a one to three month time, Verbal IQ increased about 3 points, Performance IQ about 6 or 7 points, and Full Scale IQ about 4 or 5 points. Inter-rater reliability is also rather good.

Content Validity was established by expert judges who reviewed the items. Criterion Validity was established by correlating WAIS-R and WAIS III. The numbers are good, and Full Scale IQ is about three points higher on the WAIS-R, as expected by the Flynn effect. The WAIS III is also correlated with the SB4, the WISC-III, the WIAT, etc…

Construct Validity was established using a factor analysis. Studies found that g was supported, and that verbal subtests correlated better with each other than performance subtests. The same was true for performance subtests verses verbal, but not as strongly.

Other efforts to establish Construct Validity come from correlating the WAIS III with other intelligence tests. Groth-Marnat tells us the WAIS III Performance IQ and Perceptual Organization Index correlated with the Ravens Matrices .79 and .65 respectively. While these numbers may not seem spectacular, recall the the test-retest reliability of the Ravens Matrices for a one to three month period is about .80, so given the reliability of the test, the WAIS III correlates almost perfectly.

Studies of cognitive disorders revealed:

Working Memory is 8.3 points lower than Verbal Comprehension, and IQ is usually average
Verbal Comprehension is 7 points higher than Working Memory for a reading LD and 13 points for a math LD. Perceptual Organization was 7 points higher than Processing Speed in both LDs. 24% show a partial ACID profile (lower Arithmetic, Coding, Information, and Digit Span
Typically, a relatively flat and rather low profile is seen with no real spikes or significant elevations