Journal of School Psychology, Vol. 38, No. 1, pp.

51–61, 2000
Copyright  2000 Society for the Study of School Psychology
Pergamon Printed in the USA
0022-4405/00 $–see front matter

PII S0022-4405(99)00036-9

Speed of Information Processing:

Developmental Change and Links
to Intelligence
Robert Kail
University of Maryland

This article explores the nature and consequences of developmental change in

speed of information processing. The first part of the article summarizes evidence
indicating that age differences in processing speed reflect developmental change in
a global mechanism that limits processing speed on most tasks. The second part of
the article describes evidence that suggests a role for processing speed in the devel-
opment of intelligence. The article concludes with some implications of these find-
ings for theories and assessment of intelligence.  2000 Society for the Study of
School Psychology. Published by Elsevier Science Ltd

Keywords: Processing speed, Memory, Intelligence, Development.

For most of the history of psychology, the study of intelligence has been
closely associated with the psychometric tradition. In this approach, pat-
terns of performance on mental tests are used to identify the structure of
human intelligence. In 1904, for example, Charles Spearman reported
findings supporting the idea that a general factor for intelligence is respon-
sible for performance on all mental tests. Other psychometricians, how-
ever, argued that intelligence consists of distinct abilities. For example,
Thurstone and Thurstone (1941) used adults’ performance on a wide
range of mental tests to identify seven distinct elements of intelligence: per-
ceptual speed, word comprehension, word fluency, space, number, mem-
ory, and induction. Thurstone and Thurstone acknowledged the presence
of a general factor in intelligence, but claimed that the specific factors were
more useful in assessing and understanding intellectual ability.
These conflicting findings led many psychometric theorists to propose
hierarchical theories of intelligence that included both general and spe-
cific components. Carroll (1993), for example, proposed a hierarchical the-
ory with three levels. At the top of the hierarchy is general intelligence. The
next level includes eight broad categories of intellectual skill, ranging from
fluid intelligence to processing speed. Each of the abilities in the second
Received April 19, 1999; accepted October 6, 1999.
Address correspondence and reprint requests to Robert Kail, Department of Psychology,
University of Maryland, College Park, MD 20742-4411. Phone: (301) 405-8481; E-mail: rkail@
psyc.umd.edu

51
52 Journal of School Psychology

level is further divided into more specific skills in the third level. Fluid intel-
ligence, for example, includes sequential reasoning, induction, and quanti-
tative reasoning.
In the past 25 years, critics have argued that theories of intelligence de-
rived from the psychometric perspective are too narrow; theorists have pro-
posed a host of other perspectives that purport to provide a broader view
of intelligence (Gardner, 1983, 1993; Kail & Pellegrino, 1985; Sternberg,
1985). Although these new perspectives differ in many of their details, they
share a common emphasis on drawing upon other literatures in psychol-
ogy—cognitive, developmental, and neuropsychological—in their propos-
als for new perspectives on intelligence.
The research described in this article is rooted in these new perspectives
on intelligence. Specifically, research in cognitive psychology, develop-
mental psychology, and neuropsychology has revealed that speed of infor-
mation processing is a key element in people’s ability to think, reason, and
remember. In this article, I first provide an overview of research showing
how processing speed develops. This work is consistent with the hypothesis
that age differences in processing speed reflect developmental change in
a global mechanism that limits processing speed on most tasks. Next, I ex-
amine research that links speed of information processing to intelligence.

DEVELOPMENTAL CHANGE IN SPEED OF

INFORMATION PROCESSING
On a wide range of motor, perceptual, and cognitive tasks in which partici-
pants must respond rapidly, a common pattern of age differences emerges:
8- to 10-year-olds typically respond at a speed that is 5 to 6 standard devia-
tion units below the average speed for young adults; even 12- and 13-year-
olds respond at a speed more than a full standard deviation below the aver-
age for young adults (Kail, 1991). This consistent pattern of age differences
across many diverse tasks suggests that a global mechanism limits the speed
with which children and adolescents process information. The mechanism
is not specific to particular tasks or domains but is, instead, a fundamental
characteristic of the developing information-processing system.
In fact, considerable research conducted over the past 15 years provides
support for such a global mechanism. In this section, I describe evidence
supporting such a mechanism, then discuss possible characteristics of a
global mechanism.

Evidence That a Global Mechanism Limits Developmental Change in

Processing Speed
Three lines of evidence provide support for the hypothesized global mech-
anism.
 Kail 53

Studies of developmental functions. If speeds of different processes are

limited by a common global mechanism, then the same pattern of develop-
mental change in processing speed is expected for these processes. If, in-
stead, the speed of each process reflects experiences or practice specific to
that process, then patterns of developmental change should vary across
processes. In reality, speeds for processes such as mental addition, mental
rotation, memory search, and simple motor skills all change at a common
rate that is well described by an exponential function (Kail, 1991).

Studies using the method of statistical control. Another method is based

on calculating partial correlations to eliminate common variance. Suppose
speeds of two processes, X and Y, both change substantially during child-
hood and adolescence, which would produce large correlations between
age and speeds of each process. If age-related changes in X and Y both re-
flect age-related change in a global mechanism, then statistical control of
the variance in one measure should greatly reduce the age-related variance
in the other. That is, the correlation between age and speed of X should
be greatly reduced with the speed of Y partialled out:
rAX · Y ¿ rAX
where A is age, X is the speed of process X, and Y is the speed of process Y.
If, instead, two measures of processing speed reflect independent age-
related changes, then statistical control of the variance in one measure
should have minimal influence on age-related variance in the other. The
partial correlation between age and speed of X should be nearly as large
as the simple correlation between age and the speed of X:
rAX · Y ⬇ rAX.
In fact, the typical result is for age-related differences in processing speed
to be attenuated substantially when the variance with a second speeded pro-
cess is eliminated. For example, Kail and Salthouse (1994) reported that a
correlation of .54 between age and speed of responding on analogical rea-
soning problems was reduced to .20 when performance on a paper-and-
pencil measure of processing speed was partialled out. This pattern sug-
gests a substantial general component to the development of processing
speed.

Studies using the method of systematic relations. Suppose adults’ re-

sponse time (RTa) on a task involves several processes (Equation 1)
RTa ⫽ f ⫹ g ⫹ h ⫹ . . . (1)
where f is the time to execute process F, g is the time for process G, and so
on. If children at age j execute each process more slowly than adults, by the
same factor, mj, then the corresponding equation for children would be
54 Journal of School Psychology

RTj ⫽ mjf ⫹ mjg ⫹ mjh . . . ⫽ mj(f ⫹ g ⫹ h . . .). (2)

Rewriting Equation 2 to express children’s response times (RTs) as a func-
tion of adults’ RTs yields
RTj ⫽ mjRTa. (3)
Based on Equation 3, RTs for children at age j should increase linearly
as a function of adults’ RTs from the same experimental conditions, and
the slope of this function, which estimates mj, should decrease with increas-
ing age.
Both predictions have been supported repeatedly. Some of the initial evi-
dence came from a meta-analysis of studies of speeded performance that
included nearly 2,000 pairs of RTs for children, adolescents, and adults
(Kail, 1991). As shown in Figure 1, at each age, children’s RTs increased
linearly as a function of adults’ RTs; across the 11 ages studied, the correla-
tion between children’s and adults’ RTs was never less than .95. Further-
more, the slope of the linear function—which estimates m, the factor by
which children respond more slowly than adults—changed at an exponen-
tial rate across childhood and adolescence (Kail, 1991). These predictions
hold when tested with new data rather than archival data (Hale, 1990;
Kail & Park, 1992) and when they are based on individuals’ data rather than
on group means (Kail, 1993).1
Equation 3 also accounts for other important phenomena associated
with speeded processing, such as the impact of practice, the speed-accuracy
tradeoff, and variability in RTs (Kail, 1995). To illustrate, consider the
trade-off between speed and accuracy. Children and adults can regulate re-
sponse speed, emphasizing accuracy at the cost of speed or speed at the
cost of accuracy. The impact of these different emphases can be incorpo-
rated into Equations 1–3 by introducing additional constants to both chil-
dren’s and adults’ RTs. Emphasizing accuracy might increase RTs by a fac-
tor of 1.2 over RTs in a neutral condition, whereas emphasizing speed
might decrease RTs by a factor of .8. Assuming these constants are the same
for children and adults, the result is that across various speed and accuracy
conditions, children’s RTs should be greater than adults’ RTs by m, and this
result has been obtained (Kail, 1995).
1
One common interpretation of studies of developmental change in processing speed is
that the results reflect, in large part, developmental change in the motor acts required to re-
spond (e.g., pushing a button). These movements are part of total RT, of course, and they do
become faster with age. However, the common pattern of developmental change reflects
more than the time to implement motor responses. This can be shown by examining the logic
of Equations 1–3 in more detail. Suppose that f in these equations represents the time to
initiate a motor response and mfj is the factor by which children at age j activate motor re-
sponses more slowly than adults. Further, assume that g, h, . . . are nonmotoric processes in
different tasks, that mgj, mhj, . . . are unique slowing coefficients, and that mfj ⬆ mgj ⬆ mhj. Because
the various m values differ and because tasks use different nonmotor processes (i.e., unique
combinations of f, g, h, . . .), RTj should not be accurately predicted by RTa. Yet Figure 1 shows
that RTj is very well predicted, an outcome most compatible with the assumption that values
of mj are the same for motoric and nonmotoric processes.
 Kail

Figure 1. Each panel depicts children’s RTs as a function of adults’ RTs in corresponding conditions. Also shown is the linear equation corresponding
to Equation 3. From Kail (1991). Copyright  1991 by the American Psychological Association. Reprinted with permission.
55
56 Journal of School Psychology

To summarize the three lines of evidence, speeds of many different cog-

nitive, perceptual, and motor processes develop at a common exponential
rate, age-related variance in speeded measures is effectively eliminated
through statistical control of other measures of processing speed, and chil-
dren’s RTs are often a simple multiple of adults’ RTs across a wide range
of tasks and conditions. These findings appear to converge on the conclu-
sion that some sort of global mechanism limits the speed with which chil-
dren process information. They are not readily explained in terms of expe-
riences and practice that are specific to particular processes or tasks.

Nature of the Hypothesized Global Mechanism

Although the evidence consistently implicates a global mechanism in devel-
opmental change in processing speed, the characteristics of the hypothe-
sized mechanism are relatively poorly understood. One approach has been
to look to cognitive psychology for basic parameters of mental functioning
that might correspond to such a global mechanism. For example, cognitive
theories sometimes depict the mind as a network of many nodes with inter-
connecting links. Networks are thought to be analogous to brain processes
and neural networks, but at a higher level of abstraction that is much less
complex than the actual human nervous system. Typically, a limited num-
ber of nodes in the network can be active at any time, and such activation is
thought to spread via links to adjacent nodes. In this framework, processing
speed would correspond to the rate of propagation from one node to an-
other. Indeed, changing this rate parameter yields systematic differences
across a range of tasks that resemble those described earlier (Salthouse,
1988).
Taking this approach a step further would suggest that developmental
change in speed of processing might reflect age-related change in rate of
neural communication (Myerson, Hale, Wagstaff, Poon, & Smith, 1990). In
fact, developmental change in processing speed has been linked to neural
development and function. During childhood and adolescence, periods
when processing speed changes dramatically, important neural changes oc-
cur. There are age-related changes in the number of transient connections
in the central nervous system (e.g., Huttenlocher, 1979) as well as age-
related increases in myelinization (e.g., Yakovlev & Lecours, 1967). Fur-
thermore, individuals with impairments to the nervous system often have
slower processing rates. For example, in multiple sclerosis, neurons be-
come demyelinated, a change that results in slower information processing
(e.g., Kail, 1997, 1998). Thus, although the evidence is often indirect and
sometimes speculative, I believe it is reasonable to view processing speed as
a basic parameter of cognitive functioning that changes with development,
due, at least in part, to underlying biological factors.
 Kail 57

In the next part of this article, I describe research that links this develop-
mental change in processing speed to intelligence.

LINKS BETWEEN PROCESSING SPEED AND INTELLIGENCE

There are many bases for expecting that global developmental change in
processing speed should be related to performance on traditional mea-
sures of intelligence. One is that, if change in processing speed is truly sys-
temic, then it should be manifest on most tasks including those used to as-
sess intelligence. Another is that, except for theories that emphasize
general intelligence, most psychometric theories (e.g., Carroll, 1993) in-
clude a factor called perceptual speed or processing speed.
In fact, three bodies of research show that speed of information pro-
cessing is an important element in intelligence.

Studies Comparing Populations With Intellectual Impairment

If processing speed is an important element of intelligence, then groups of
children known to differ in intelligence should also differ in their speed of
processing. This outcome has been obtained many times in studies of indi-
viduals with mental retardation: On most tasks that required rapid re-
sponding, individuals with mental retardation respond more slowly than in-
dividuals without mental retardation. Furthermore, Kail (1992) found that
RTs for individuals with and without mental retardation were related in
much the same manner as RTs for children and adults. That is, Equation
3 can be rewritten so that RTj denotes a mean RT for a group of persons
with mental retardation and RTa denotes the mean RT for the same condi-
tion for a group of chronologically age-matched persons without mental re-
tardation. Kail (1992) found that RTs for individuals with mental retarda-
tion increased as a function of RTs for individuals without mental
retardation and that m (the value by which individuals with mental retarda-
tion respond more slowly) decreased as a function of the IQ level of the in-
dividuals with mental retardation.

Studies of Developmental Cascade

Processing speed, memory, and reasoning all improve with age, raising the
possibility that these changes are not independent but related. Fry and
Hale (1996) proposed a strong form of interdependence, which they called
developmental cascade, in which “processing speed becomes faster, leading
to improvements in working memory, and improved working memory, in
turn leads to increases in [reasoning and problem solving]” (p. 237). Fry
and Hale evaluated this hypothesis by administering measures of pro-
cessing speed, working memory, and fluid intelligence to 7- to 19-year-olds.
58 Journal of School Psychology

Structural equation modeling provided evidence that age-related increases

in speed were associated with increases in working memory capacity, which,
in turn, were associated with higher scores on the test of fluid intelligence.
Thus, as children process information more rapidly, they can use working
memory more effectively, which allows them to solve problems like those
on the test of fluid intelligence more successfully.

Studies Predicting Childhood Intelligence From Infancy

For testing intelligence in infants, many psychologists use the Bayley Scales
of Infant Development (Bayley, 1970, 1993). Designed for 2- to 42-month-
olds, the Bayley Scales consist of mental and motor scales. Although scores
on the Bayley Scales can be used to determine whether development is
progressing normally, scores obtained during infancy usually are not re-
lated to IQ scores obtained later in childhood, adolescence, or adulthood
(McCall, 1989). Apparently, children must be at least 18 to 24 months old
before scores from the Bayley (or similar scales) predict IQ scores during
childhood (Kopp & McCall, 1982).
One reason that scores on infant intelligence tests may predict childhood
IQ so poorly is that infant tests measure different abilities than tests admin-
istered to children and adolescents: Infant tests place more emphasis on
sensorimotor skills and less on tasks involving cognitive processes such as
language, thinking, and problem solving. According to this argument, a
measure of infant cognitive processing might yield more accurate predic-
tions of later IQ. In fact, several investigators have found that measures of
infant processing speed do predict later IQ more effectively than scores
from the Bayley.
Fagan (1992) pioneered work on infants’ information processing by tak-
ing advantage of the fact that infants usually prefer novel stimuli to familiar
stimuli. Infants initially pay much attention to a novel stimulus, but gradu-
ally pay less attention as the stimulus is shown repeatedly, a phenomenon
known as habituation. Some infants habituate to stimuli more rapidly than
other infants, and these individual differences in rate of habituation are re-
lated to intelligence in childhood. The average correlation between rate of
habituation and later IQ is approximately .5 (Bornstein, 1998; Fagan,
1992). That is, 1- to 6-month-olds who habituate to visual stimuli more rap-
idly tend to have higher IQs as children.
Processing speed may contribute to individual differences in rate of ha-
bituation, but the two are not synonymous. A more direct test of links be-
tween processing speed in infancy and intelligence in childhood comes
from work by Dougherty and Haith (1997). These investigators recorded 3
1/2-month-olds’ eye movements while a series of pictures was shown rap-
idly to them. For each infant, they calculated a mean RT, defined as the
amount of time between presentation of a picture and the start of the eye’s
 Kail 59

movement toward the picture. At approximately 4 years of age, children

were tested on the Wechsler Preschool and Primary Scale of Intelligence–
Revised. The correlation between infant RT and childhood IQ was –.44: In-
fants with faster eye movements tended to have higher IQ scores as 4-year-
olds.
Of course, the correlations between infants’ processing speed and child-
hood IQ are far from perfect. Some of this can be attributed to lack of relia-
bility in the measures. Some undoubtedly reflects the fact that processing
speed is but one of several factors contributing to childhood intelligence
(Fagan, 1992).
To summarize research on links between processing speed and intelli-
gence, processing speed is (a) slower in persons with mental retardation,
(b) associated concurrently with more efficient memory and reasoning,
and (c) associated longitudinally with traditional IQ scores. Collectively,
these results implicate a role for processing speed in intelligence.

CONCLUSIONS
Processing speed is an element of intelligence. Why is this conclusion pro-
vocative or newsworthy? After all, Thurstone and Thurstone (1941) made
the same claim nearly 60 years ago, and processing speed has typically been
an element of multifactor theories of intelligence ever since. The key differ-
ence is that, in contrast to psychometric research, modern research por-
trays a dynamic assessment of processing speed. That is, processing speed
is not simply one of many different independent factors that contribute to
intelligence; instead, processing speed is thought to be linked causally to
other elements of intelligence. As children develop, they process informa-
tion more rapidly, apparently reflecting age-related change in the central
nervous system. This more rapid processing results in more effective use of
working memory, which enhances performance on many reasoning tasks
like those used to assess intelligence. Thus, in contrast to a prototypic
multifactor theory of intelligence (e.g., Carroll, 1993) that might include
processing speed, memory, and inductive reasoning as separate and inde-
pendent factors, the current view links these constructs. More rapid pro-
cessing enhances memory, which, in turn, enhances reasoning.
In addition, in the dynamic view, processing speed can influence perfor-
mance on intelligence tests both directly and indirectly. An indirect effect
is illustrated by the impact of processing speed on memory. By allowing
working memory to be used more efficiently, increased processing speed
enhances performance on intelligence tests indirectly. However, process-
ing speed may also affect performance directly by, for example, speeding
retrieval of task-relevant information from long-term memory.
This dynamic view of different elements of intelligence also has implica-
tions for the measurement of intelligence and interpretation of test scores.
60 Journal of School Psychology

Obviously, tests like the Stanford-Binet that only assess general intelligence
are seen as much less useful than those that assess different elements of in-
telligence. However, in a typical test that assesses multiple skills, the results
are cast as a profile of an individual’s independent strengths and weak-
nesses. The dynamic view reminds us that many of these skills may be inter-
related. In the psychometric tradition, a child with below average scores on
processing speed, memory, and reasoning is subpar in three areas. In the
dynamic view, the child may well be limited primarily in one area (pro-
cessing speed); the other subpar performances are largely byproducts of
deficits in processing speed. Consequently, if this child’s processing speed
were improved, memory and reasoning would be expected to follow suit.
In conclusion, more rapid processing speed is an important aspect of
greater intellectual power. As additional research continues to clarify the
nature of links between intelligence and processing speed, theories of intel-
ligence will be enhanced and test information more valuable.

ACKNOWLEDGMENTS
The research described in this article was supported by grants from the Na-
tional Institute of Child Health and Human Development and the National
Science Foundation.

REFERENCES
Bayley, N. (1993). Bayley scales of infant development: Birth to two years (2nd ed.). San Anto-
nio, TX: Psychological Corporation.
Bayley, N. (1970). Development of mental abilities. In P. H. Mussen, (Ed.), Carmichael’s
manual of child psychology (pp. 1136–1209). New York: Wiley.
Bornstein, M. H. (1998). Stability in mental development from early life: Methods, mea-
sures, models, meanings, and myths. In G. E. Butterworth & F. Simion (Eds.), The de-
velopment of sensory, motor, and cognitive capacities in early infancy: From sensation to cogni-
tion (pp. 301–332). Hove, England: Psychology Press.
Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analytic studies. New York:
Cambridge University Press.
Dougherty, T. M., & Haith, M. M. (1997). Infant expectations and reaction time as pre-
dictors of childhood speed of processing and IQ. Developmental Psychology, 33, 146–
155.
Fagan, J. F. (1992). Intelligence: A theoretical viewpoint. Current Directions in Psychological
Science, 1, 82–86.
Fry, A. F., & Hale, S. (1996). Processing speed, working memory, and fluid intelligence:
Evidence for a developmental cascade. Psychological Science, 7, 237–241.
Gardner, H. (1983). Frames of mind: The theory of multiple intelligences. New York: Basic
Books.
Gardner, H. (1993). Multiple intelligences: The theory in practice. New York: Basic Books.
Hale, S. (1990). A global developmental trend in cognitive processing speed in children.
Child Development, 61, 653–663.
Huttenlocher, P. R. (1979). Synaptic density in human frontal cortex—Developmental
changes and effects of aging. Brain Research, 163, 195–205.
 Kail 61

Kail, R. (1991). Developmental change in speed of processing during childhood and ad-
olescence. Psychological Bulletin, 109, 490–501.
Kail, R. (1992). General slowing of information processing by persons with mental retar-
dation. American Journal of Mental Retardation, 97, 333–341.
Kail, R. (1993). Processing time decreases globally at an exponential rate during child-
hood and adolescence. Journal of Experimental Child Psychology, 56, 254–265.
Kail, R. (1995). Processing speed, memory, and cognition. In W. Schneider & F. E. Wein-
ert (Eds.), Memory development: State of the art and future directions (pp. 71–88). Hillsdale,
NJ: Erlbaum.
Kail, R. (1997). The neural noise hypothesis: Evidence from processing speed in adults
with multiple sclerosis. Aging, Neuropsychology, and Cognition, 4, 157–165.
Kail, R. (1998). Speed of information processing in patients with multiple sclerosis. Jour-
nal of Clinical and Experimental Neuropsychology, 20, 1–9.
Kail, R., & Park, Y. (1992). Global developmental change in processing time. Merrill–
Palmer Quarterly, 38, 525–541.
Kail, R., & Pellegrino, J. W. (1985). Human intelligence: Perspectives and prospects. New York:
Freeman.
Kail, R., & Salthouse, T. A. (1994). Processing speed as a mental capacity. Acta Psycholog-
ica, 86, 199–225.
Kopp, C. B., & McCall, R. B. (1982). Predicting later mental performance for normal,
at-risk, and handicapped infants. In P. B. Baltes & O. G. Brim (Eds.), Life-span develop-
ment and behavior: Vol. 4 (pp. 33–61). New York: Academic Press.
McCall, R. B. (1989). Commentary. Human Development, 32, 177–186.
Myerson, J., Hale, S., Wagstaff, D., Poon, L. W., & Smith, G. W. (1990). The information
loss model: A mathematical theory of age-related cognitive slowing. Psychological Re-
view, 97, 475–487.
Salthouse, T. A. (1988). Initiating the formalization of theories of cognitive aging. Psy-
chology and Aging, 3, 1–16.
Spearman, C. (1904). “General intelligence” objectively determined and measured.
American Journal of Psychology, 15, 201–293.
Sternberg, R. J. (1985). Beyond IQ: A triarchic theory of human intelligence. Cambridge, En-
gland: Cambridge University Press.
Thurstone, L. L., & Thurstone, T. G. (1941). Factorial studies of intelligence. Psychometric
Monograph, No. 2.
Yakovlev, P. I., & Lecours, A. R. (1967). The myelogenetic cycles of regional maturation of
the brain. In A. Minkowski (Ed.), Regional development of the brain in early life (pp. 1–70).
Oxford, UK: Blackwell.