Developmental Psychology 1997, Vol.

33, No, 3, 549-554

Copyright l'J97 by the American Psychological Association, Inc. 0012-l649/97/$3.00

Developmental Changes in Speed of Information Processing in Sibling Children

Linda T. Miller and Philip A. Vernon
University of Western Ontario
This study investigated developmental increases in processing speed in young children, relative to adults, with only nonverbal stimuli. R. Kail's (1991) model of the rate of change in processing speed from childhood to adulthood was evaluated. Processing speed was measured in 34 children at 4 years, 37 at 5 years, and 38 at 6 years and in 43 adults, with a battery of 8 computer-administered tests. Results showed clear age-related increases in processing speed diat cannot be attributed to increased accuracy and error rate monitoring. Kail's model adequately accounted for the observed rate of developmental change in processing speed; however; the parameter estimates of R. Kail and Y. Park (1992) provided more accurate predictions than did the meta-analytically derived estimates of Kail (1991). Findings support the global developmental trend hypothesis and suggest that this trend extends beyond the range of verbal skills evaluated in previous research.

It is well established that adolescents and adults process information faster than young children; however, the precise reason for the increase in speed with increasing age is not entirely clear. Proposed reasons for the developmental increase in processing speed from early childhood to adolescence include increases in the speed of stimulus identification and encoding, reductions in decision-making time, and faster response selection. Brewer and Smith (1989) suggested that developmental increases in processing speed are due to age-related changes in the trade-off between speed and accuracy. Using a serial fourchoice reaction time (RT) task, they found younger children to be slower and less consistent in their speed of processing, and also less accurate in monitoring their errors. The most notable increase in processing speed observed by Brewer and Smith occurred between the ages of 5 and 9 years, concurrent with a marked improvement in error rate monitoring. In a meta-analytic review of research pertaining to developmental changes in RT during childhood and adolescence, Kail (1991) advocated that children's processing speed is systematically related to that of adults in a global developmental manner. Kail compiled the findings of 72 studies comparing the processing speed of adults and children and found that, in a wide variety of tasks and conditions, the RTs of children, although consistently slower than the RTs of adults, can be accurately expressed, for any given age, as a direct function of the RTs of adults. This expression has the form

where mage is the slowing coefficient for a given age, or the rate at which response latencies change with age (Kail, 1991). Kail interpreted this direct relationship between the processing speed of children and adults as an indication that a global developmental mechanism is responsible for the developmental change in RT. The slowing coefficient, or m value, can be computed for any given age. However, the m value will vary across ages from childhood to adulthood in a negatively decelerating curvilinear manner best described by an exponential function (Kail, 1991). To model the variability in mage, Kail used the equation = 1+ (2)

Linda T. Miller, Department of Occupational Therapy, Elborn College, University of Western Ontario, London, Ontario, Canada; Philip A. Vernon, Department of Psychology, University of Western Ontario, London, Ontario, Canada. Correspondence concerning this article should be addressed to Linda T. Miller, Department of Occupational Therapy, Elborn College, University of Western Ontario, London, Ontario, Canada N6G 1HI. Electronic mail may be sent via Internet to miller @audio.hhcru.uwo.ca. 549

containing two theoretical constants (b and c). In Equation 2, 1 + b defines the intercept (i.e., when age = 0, e~cXaBe = 1), e is the base of natural logarithms, and c is the decay parameter that determines the rate at which the value of m^ approaches unity with increasing age. Once the values of the constants b and c are known, the m value associated with any age can be estimated to determine the rate at which children, at a particular age, will perform slower than adults on a task for which adult performance is known. Applying nonlinear regression to the meta-analytic data, Kail accounted for 77% of the variability in slowing coefficients based on estimates of b and c as 5.16 and 0.21, respectively. These estimates of b and c yielded values of m ^ that declined with increasing age, with more than 90% of the variability in RTs of children across all ages accounted for by those of adults. In sum, Kail's findings indicate that processing speed in children is directly related to processing speed in adults on a broad range of tasks and under variable conditions, suggesting that some global factor contributes to the developmental change. Kail and Park (1992) used Kail's (1991) meta-analytic estimates (b = 5.16 and c = 0.21) to predict the slowing coefficients of 15 age groups ranging from 6 to 20 years. Although the Kail (1991) parameter estimates accounted for almost 93% of the variance in m values across the age groups, they underestimated the m values for young children (i.e., age 6 years) and overesti-

550

MILLER AND VERNON

mated the m values of the adolescent groups. Overestimation of m values for adolescents implies that the asymptote of m values is not accurately predicted by the meta-analysis parameter estimates. Applying nonlinear regression to their data, Kail and Park derived estimates of b and c of 11.87 and 0.30, respectively. Although these new parameter estimates increased the amount of explained variance in m values by only a trivial amount, they more accurately represented the m values of the adolescent groups. The utility of these estimates comes from their potential use in predicting the response time of children ages 3 years and older on the basis of adult performance; by inserting Equation 1 into Equation 2,

+ ^

)/?r a d u l l ,

(3)

and using the derived estimates of b and c, an accurate estimate of children's RTs on a wide variety of tasks could be obtained. Because of potential distortion in the meta-analytic parameter estimates due to fluctuations in the sample sizes, tasks, and conditions used in the studies compiled, the estimates of Kail and Park (1992) may be preferable. Kail (1993) compared predictions based on the parameter estimates (b = 11.87, c = 0.30) derived by Kail and Park to those based on Kail's (1991) meta-analytic parameter estimates (b = 5.16, c = 0.21). Results revealed that although both sets of parameters provided good prediction, the Kail and Park estimates provided slightly greater accuracy at the younger ages (i.e., ages 6 and 7 years). Differences in predictive ability between Kail's (1991) metaanalysis estimates and the estimates of Kail and Park (1992) are small. Although the Kail and Park estimates are slightly more accurate at younger ages, a recent review by Kail and Salthouse (1994) supports the use of the meta-analytically derived estimates. Prediction of m values appears to be difficult only at the younger ages, with predicted values being less than observed m values. The majority of studies evaluating processing speed in children have focused on children ages 6 years and older (cf. Kail, 1993; Kail & Park, 1992). Of the 72 studies included in Kail's (1991) meta-analysis, 13 studies included children younger than age 6 years, 5 included children younger than age 5 years, and only 1 study included children younger than age 4 years. It is possible that inaccuracy in predicting m values at younger ages is due, at least in part, to underrepresentation of young children in the study samples. It is worth noting that those studies that included children younger than 6 years of age relied largely on verbal stimuli and all but 2 of these studies used only one experimental task. Although past research has consistently shown developmental trends in speed of information processing and several models of the developmental trends have previously been proposed and tested, it is important to note that previous research in this area has focused, to a large extent, on school-aged children, with many of the research tasks requiring reading ability. It is assumed that the performance of young children can be accounted for by models estimated on older children. However, given the difficulty of previous estimates (i.e., Kail, 1991) to accurately predict the slowing coefficients of younger children in subsequent studies (i.e., Kail, 1993), it is necessary to consider the processing speed of children 6 years of age and younger to fully evaluate the predictive ability of a model such as that presented in Equation 3.

Many of the tasks used in previous studies of processing speed in children have been verbal in nature, relying on reading skills or, at the least, letter or numeral identification skills. To assess the response time of preschool- and kindergarten-age children, we found it necessary to use tasks that do not rely on reading or letter or numeral identification skills. The current study uses a battery of tasks that require only knowledge of several basic colors (red, green, yellow, and blue) and shapes (triangle, square, octagon ["stop sign"], and diamond). In fact, accurate identification of the colors and shapes is not necessitated by the tasks. Many past studies, such as those compiled by Kail's (1991) meta-analysis, have used only one RT task whereas other studies (cf. Kail & Park, 1992) have used four or five RT tasks and derived mean RJs for each participant under various conditions within particular tasks. We have developed eight different tasks, with each task yielding only one mean RT per participant. Because our study used a relatively large number of tasks that are nonverbal in content and evaluated children younger than those typically included in studies of processing speed, it offers a unique contribution to the research on children's processing speed. The objectives of this study are (a) to evaluate the global trend hypothesis in 4- to 6-year-old children by testing for differences in processing speed using a battery of nonverbal tasks; (b) to test, if speed differences are found, the predictive accuracy of Kail's (1991) meta-analytic parameter estimates and the Kail and Park (1992) parameter estimates relative to the observed m coefficients; and (c) to determine whether the global trend is related to improvements in accuracy, as suggested by Brewer and Smith (1989).

Method Participants
A sample of thirty-four 4-year-olds (M = 4.52 years, SD = 0.32; 18 boys and 16 girls), thirty-seven 5-year-olds (M = 5.58 years, SD = 0.32; 20 boys and 17 girls), and thirty-eight 6-year-olds (M = 6.44 years, SD = 0.34; 22 boys and 16 girls) participated in the study. The adult sample consisted of 43 undergraduate students (M = 23.15 years, SD = 2.97; 12 men and 31 women).

Measures
A battery of eight RT tests was administered using a Commodore 128 computer with a color monitor (Commodore). Participants made their responses using a three-button response console (two response keys and a home key). The left response key was red and the right response key was yellow; the home key, centered below the response keys, was white. Different colored keys were used so that test instructions could refer to color rather than to position of the response keys. The response console was rectangular, 10 X 8.5 in. (25.4 X 21.6 cm) in dimension. The response keys were I in. in diameter and were raised slightly above the surface of the response console to facilitate responding. All RT tasks were nonverbal in content, requiring no reading ability. The test stimuli consisted of geometrical figures of varying color and size and the tasks required participants to distinguish between shapes, sizes, and colors. Computer programs for all the RT tasks were run using a software timer with RT measured as the interval between the onset of the display of the stimulus to the release of the home button (Lee & Miller, 1995). The timer ensured that all measured RTs are accurate to within 1 ms. The tests were designed such that the presentation of the display is

DEVELOPMENTAL CHANGES IN PROCESSING SPEED contingent on pressing the home button; as such, progression through each test was self-paced, with participants controlling the initiation of each trial. Display times are accurate to 17 ms. All eight RT tests comprised 24 test items as well as six practice trials that were not included in the calculation of a participant's mean RE1 During the practice trials, a tone was used to signal an incorrect response; no feedback was given during the actual test trials. A tune was used to signal the end of the practice trials, and a different tune was used to signal the end of the test trials. For all tests, the red response key was used to indicate "same" or "yes," and the yellow response key was used to indicate ' 'different'' o r ' 'no.'' All the tests were balanced such that there were an equal number of ''same" or " y e s " responses and ' 'different' * or ' 'no'' responses. For all tests except the Arrows Test, the shape set used in the stimuli included triangles, squares, octagons, and diamonds presented in blue, green, red, and yellow. The occurrence of each shape and color was balanced within each test. The time between pressing the home button and the appearance of a test stimulus or item varied randomly between 500 to 1,500 ms. Shape test. For this test, two shapes (triangle, square, octagon, and diamond, in red or blue) are presented in the center of the screen side by side. The participant's task is to respond "same" if the two shapes are the same or "different" if the shapes differ. During any given trial, color is held constant. Therefore, trials consist of two red shapes or two blue shapes. Color test. The participant's task in this test is to indicate whether two shapes appearing side by side in the center of the screen are the same color or different colors. This test uses four colors (red, green, blue, and yellow) and two shapes (triangle and square), with shape always held constant within trials. Like the shape test, this test includes 24 trials (12 "same" trials and 12 "different" trials). Size test. In this test, a participant's task is to indicate whether two shapes appearing side by side in the center of the screen are the same size or different sizes. Three different shapes (triangle, square, and octagon) and two different colors (red and yellow) are used in this test. Color is varied across trials but is held constant within trials. Number test. A participant's task in this test is to indicate whether two series of vertically presented shapes are the same or different in quantity. This test uses only squares and triangles presented in either red, yellow, blue, or green. Again, color is varied only across trials. Each trial consists of one vertical series of squares on the left side of the screen and one vertical series of triangles on the right side of the screen. The maximum number of squares or triangles in any series is three and the minimum number is one. Shape string test This test involves the presentation of a target shape, followed by the presentation of a string of shapes, all of the same color. Participants must respond "yes" if the target shape occurs in the string, or " n o " if the target does not occur in the string. The target shape appears on the screen for 1,000 ms and is followed by the shape string with an interstimulus interval of 250 ms. The string remains on the screen until participants make their response. All four shapes are used in this test (triangle, diamond, square, and octagon) and are presented in red, green, blue, and yellow. Color string test. In the same manner as the shape string test, this test involves the presentation of a target color, followed by the presentation of a string of colors, all of the same shape. Participants must respond ' 'yes'' if the target color occurs in the string or " n o " if the target color does not occur in the string. All four colors are used in this test (red, green, blue, and yellow); however, shape is held constant across all trials, with only squares used in this test. Arrows test. In this test, participants are required to respond by pressing the red (left) response key when a left arrow appears on the screen, or by pressing the yellow (right) response key when a right arrow appears on the screen. On all trials, the arrows are presented in

551

white with only the direction of the arrow (left vs. right) varied. The arrows remain on the screen until participants make their response. Matching test. The participant's task in this test is to indicate which side of the screen (left or right) the previously presented match to a target object appeared. Two different shapes (triangle, square, octagon, or diamond) of the same color (red, green, blue, or yellow) appear on the screen simultaneously for 1,000 ms, one shape on the left side of the screen and one shape on the right side of the screen. After an interstimulus interval of 250 ms, a target shape appears centered on the screen below where the two shapes were previously presented. Participants indicate which of the previously presented shapes the target matches by pressing the left (red) response key if the target matches the shape on the left side of the screen, or by pressing the right (yellow) response key if the target matches the shape presented on the right side of the screen.

Procedure
All testing was completed in one session, lasting approximately 30 to 45 min with the tests administered in the same order for all participants: arrows, size, shape, color, shape string, color string, matching, and number tests. All testing was conducted in a laboratory with only the participant and the tester present. For all tasks, the participants were instructed to use their preferred hand to press the buttons on the response console and to make their responses as quickly as possible while maintaining accuracy.

Results Differences in Processing Speed Across the Four Age Groups

Mean RTs for each of the tasks were calculated for all participants using only those items responded to correctly. As is apparent in Table 1, there are notable differences between the three child groups and the adult group in terms of mean KIs; a clear trend of decreasing KIs with increasing age is apparent. To assess the significance of these differences across age groups, we performed a single-factor multivariate analysis of variance (MANOVA) using KTs on the eight tasks as multiple dependent variables. A significant multivariate effect of age was found, Pillai's F(24,429) = 5.59,/? < .001, indicating that significant differences in processing speed across the four age groups exist. Univariate tests revealed that all eight tasks demonstrated significant age effects. This finding is not surprising, given that the adult RTs were considerably faster than the RTs of the three child groups. To determine whether significant differences in processing speed exist among the three child groups, we performed Tukey post hoc comparisons on all eight tasks. Results of the post hoc comparisons revealed that, for all tests, the mean KIs for adults differed significantly from those of all three child groups. In addition, 4-year-olds were found to differ significantly from 6-year-olds on all eight tasks; five tasks (shape, size,

1 Mean RTs were computed without any prior truncating of outliers. Recent research (Ulrich & Miller, 1994) has revealed that truncation of outliers may introduce bias by removing extreme but valid responses. Further, Ulrich and Miller reported that elimination of extreme scores may adversely affect the estimation of linear relations between reaction time and other variables. For this reason, extreme scores were not eliminated from analyses.

552
Table 1 Mean Response Time (in Milliseconds) for the Four Age Groups
Test Shape M SD Color M SD Size M SD Number M SD Shape string M SD Color string M SD Arrows M SD Matching M SD 4-year-old 5-year-old 6-year-old

MILLER AND VERNON

Comparisons

of Observed mage to Predicted mage

Adult

1,346 417 1,377 642 1,384 408 2,074

1,124 382 1,137 462 1,181 371 1,725 552 1,656 622 1,394 457 818 275 1,226 441

1,014 384 1,010 349 1,032 325 1,561 506 1,406 370 1,222 397 659 182 1,048 384

471
80 439 67 504 83 608 82 636 107 516 91 359 45 459 68

638
1,946 914 1,614 703 963 304 1,506 626

Slowing coefficients were calculated for each of the three child age groups by regressing the mean KTs of the eight tasks for each child group on adult RTs (e.g., the mean RT of 4-yearolds on each of the eight tasks was regressed on the mean RT of adults on the eight tasks). Adult RTs accounted for 90%, 93%, and 90% of the variability of RTs of 4-, 5-, and 6-yearolds, respectively. In addition, the m values were found to decrease with age, dropping from 3.74 for 4-year-olds to 3.20 for 5~year-olds, and to 2.92 for 6-year-olds. The relationship of adult RT to that of each of the three child groups is illustrated in Figure 1, where the mean RT for each of the child groups on each task is plotted against that of adults on the corresponding task. Using the mean age for each child group, we calculated two sets of predicted m values for the three child groups, one set using Kail's (1991) meta-analytic parameter estimates (b = 5.16, c 0.21) and a second set using Kail and Park's (1992) parameter estimates (b = 11.88. c = 0.30). As can be seen in Table 3, Kail's (1991) meta-analytic parameters lead to underestimated m values, whereas predicted values based on the Kail and Park (1992) estimates appear to be more consistent with the observed values. In fact, the Kail (1991) estimates lead to predicted values that are significantly lower, f(2) = 11.20, p < .01, than the observed values. Values predicted using the Kail and Park estimates do not differ significantly from the observed estimates.

number, arrows, and matching) revealed significant differences between 4-year-olds and 5-year-olds; and only one task (arrows) revealed a significant difference between 5-year-olds and 6-yearolds.

Estimates of b and c Based on Current Data

Using Equation 3, we applied nonlinear regression to the current data to solve for b and c. We computed the nonlinear

Differences in Accuracy Rates Across the Four Age Groups

As can be seen in Table 2, accuracy rates for all four age groups on the eight tasks were quite high, with the accuracy rates for adults being somewhat higher than those of the child groups on all eight tasks. A single-factor MANOY\ was performed on the four age groups using percentage correct on the eight tasks as multiple dependent variables. A significant multivariate effect of age was found, Pillai'sF(24,429) = 3.58, p < .001, indicating that significant differences in accuracy across the four age groups were apparent. Univariate tests revealed significant effects of age on accuracy for all eight tasks. Tukey post hoc comparisons revealed that most of the significant comparisons were between the adult group and the child groups. Adults differed from 4-year-olds on all tasks except size and differed from 5-year-olds on all tasks except color; adults differed from 6-year-olds on only four tasks (size, shape string, color string, and matching). Significant comparisons were found between child groups on only two tests. Four-year-olds were found to differ significantly from 6-year-olds on the color string test, and 4-year olds differed from both 5-year-oLds and 6-yearolds on matching. It is worth noting that the matching test, one of only two tests to reveal accuracy differences between the child groups, had the lowest accuracy rates of all the RT tests for the youngest children.

Table 2 Mean Percentage Correct for the Four Age Groups

Test Shape M SD Color 4-year-old 5-year-old 6-year-old Adult

96

5.7 94
9.6 96 5.7 92 11.4 85 11.9 88 10.7 95 6.2 75 16.1

96 4.8 96 4.0 94 6.9 96 4.9 90 9.1 91 8.5 95 5.4 96 12.6

97 3.6 96 5.3 95 6.1 94 6.3 90 7.6 93 6.6 97 5.0 89 14.8

99 1.8 99 2.3 98

M
SD Size M SD Number M SD Shape string M SD Color string M SD Arrows M SD Matching M SD

2.6 99
2.1 96 4.4 98 3.5 99 1.6 99 1.5

DEVELOPMENTAL CHANGES IN PROCESSING SPEED

553

4-year-okls
5-year-o*ds

.__-

6-year-oWs

t3 a.

200

400

600

800

1000 1200 1400 1600 1800 2000 2200

Reaction T i m e of Adults

Figure 1. Linear relationship between adult reaction time and child reaction time for each child group (each slope is fitted to eight data points, representing the eight tests).

regression twice, once using the Kail (1991) estimates as start values for b and c, and once using the Kail and Park (1992) estimates as start values. These analyses rendered identical results. Equation 3 was found to fit the data well, accounting for 91% of the variance in the KIs of the children. However, resulting estimates of b 6.81 and c = 0.26 should be interpreted cautiously because of the small number of age groups used in their estimation. Discussion The tendency for young children to process information slower than adults is clearly demonstrated. In addition, the current data indicate that, even between the ages of 4 and 6 years, substantial gains in processing speed are apparent. The predominance of significant comparisons between 4- and 6-year-olds and between 4- and 5-year-olds would suggest that the rate of change in processing speed is faster in younger children and that, even between the ages of 4 and 6 years, declines in the rate of change in processing speed are detectable. Comparisons of the accuracy rates of the four age groups revealed significant differences in accuracy, largely attributable to differences between the adult group as compared with each of the three child groups; very few differences were found among the child groups. In fact, it would seem that age affected accuracy only when the task was particularly demanding or difficult across all three child groups (e.g., matching). Age-related differences in the accuracy of children were not compelling, and age-related changes in processing speed were not paralleled by changes in accuracy rates. On the basis of the current data, Brewer and Smith's ( 1989) hypothesis that slower processing speed in younger children is a function of less accurate response monitoring is not strongly supported. Although increased processing speed may relate to increased accuracy when comparing adults and young children, developmental in-

creases in the processing speed of 4- to 6-year-olds on the tasks used in this study cannot be attributed to increased accuracy and error rate monitoring. The declining rate of change in processing speed with increasing age in childhood is apparent when considering the slowing coefficients calculated for the 4-, 5-, and 6-year-olds. The magnitude of the slowing coefficient decreases with increased age. Further, the slopes illustrated in Figure 1 reveal that, when plotting RTs on the eight tests for each of the child groups as a function of adult RTs, the slope is steepest for the youngest children and becomes progressively less steep with increasing age. These findings indicate that, as age increases through adolescence, the slowing coefficient would be expected to decrease to a value of 1.0. When comparing the slowing coefficients calculated for the current data with those predicted using parameter estimates derived in previous research (Kail, 1991; Kail & Park, 1992), the current m values are best approximated by the values predicted using the estimates of Kail and Park. This finding concurs with speculations that the meta-analytically derived estimates may be affected by fluctuations in the sample sizes, tasks, and ages included in the studies compiled in the meta-analysis. The comparability of the m values predicted using the Kail and Park estimates to the observed m values validates the use of these estimates in predicting the KIs of young children; the Kail and Park estimates appear to be reasonably effective in predicting the RTs of children as young as age 4. Although the Kail and Park (1992) estimates were more accurate than the meta-analytic estimates in predicting slowing coefficients consistent with the observed slowing coefficients for 4to 6-year-olds, the predicted values using the Kail and Park parameter estimates somewhat overestimated the m value for the 4-year-olds; the poorest prediction using the Kail and Park parameter estimates was found for the youngest children. The estimates of b and c derived from the current data for 4- to 6year-olds were 6.81 and 0.26, respectively. In view of the fact that the meta-analytic estimates lead to underestimated slowing coefficients for young children, whereas the Kail and Park estimates lead to an overestimate of the slowing coefficient for the youngest children, it is interesting to note that the decay parameter for the current data (c = .26) falls between the meta-analytically derived decay parameter (c = .21) and the decay parameter (c = .30) of Kail and Park. In sum, the current study provides further support for the global developmental trend hypothesis. The strong relationship

Table 3 Observed and Predicted Slowing Coefficients for 4-r 5- and 6-Year-Olds
Regression of child RT on ad^ilt RT Age (years) 4 5 6 Note. R1 .90 .93 .90 Intercept -338.30 -313.75 -339.10 m (slope) 3.74 3.20 2.92 Predicted values Kail (1991) 3.00 2.64 2.33 Kail and Park (1992) 4.06 3.30 2.72

KT = reaction time.

554

MILLER AND VERNON Kail, R. (1991). Developmental change in speed of processing during childhood and adolescence. Psychological Bulletin, 109, 490-501. Kail, R. (1993). Processing time decreases globally at an exponential rate during childhood and adolescence. Journal of Experimental Child Psychology, 56, 254-265. Kail, R., & Park, Y. (1992). Global developmental change in processing time. Merrill-Palmer Quarterly, 38, 525-541. Kail, R., & Salthouse, T. A. (1994). Processing speed as a mental capacity. Ada Psychologica, 86, 199-225. Lee, C.J., & Miller, L. T. (1995). Measuring reaction time without measuring movement time: A modification of the Hormann & Allen millisecond timer for the Commodore 128. Behavior Research Methods, Instruments, e & Computers, 27, 83-87. Ullrich, R., & Miller, J. (1994). Effects of truncation on reaction time analysis. Journal of Experimental Psychology: General, 123, 34-80. Received Rsbruary 15, 1995 Revision received May 22, 1996 Accepted May 24, 1996

found between the KIk of adults and young children and the adequacy of Equation 3 in accounting for the RTs of 4- to 6year-olds implies that the decline in RT from childhood to adulthood is strongly influenced by a global maturational factor. Previous research has focused on somewhat older children, using tasks that are often verbal in content. The findings of the current study, focusing on younger children and using a battery of nonverbal tasks, concur with those of previous research. This suggests that the global developmental trend is more than a verbal developmental trend and encompasses processes used in relatively simple, nonverbal tasks as well as those used in verbal tasks.

References Brewer, N., & Smith, G. A. (1989). Developmental changes in processing speed: Influence of speed-accuracy regulation. Journal of Experimental Psychology: General, 118, 298-310.