INTELLIGENCE 6, 265-274 (1982)

Reaction Time, Movement Time, and Intelligence: A Replication and Extension* JERRY S . C A R L S O N AND C . M A R K J E N S E N

University of California, Riverside

The purpose of the investigation was to replicate and extend a study by Jensen and Munro which found reaction time (RT) and movement time (MT) parameters to correlate negatively and moderately with Raven matrices performance. A sample of 20 ninth-grade girls was used. Relationships between RT and MT and Raven scores were found to be negative and moderate to high, thus replicating the Jensen and Munro study. In addition, moderate to strong negative correlations were found between RT and MT parameters and reading comprehension and performance on the California Test of Basic Skills. Weaker relationships were found for mathematics and English grades although the direction was consistently negative.

The development of mental chronometry began in Germany when in 1850, Helmholz was able to carry out measurements of the conduction time of impulses in nerves. Prior to this work, it was assumed that conduction velocity was infinitely fast and therefore nonmeasurable. In the 1860s, F. C. Donders of Holland raised the question if thought really has the infinite speed usually associated with it. He asked " i f it wouldn't be possible to determine the time required for shaping a concept or expressing one's will" (Donders, 1969). Inherent in Donders' question is the assumption that mental processes are related to time. Subsequent empirical work has validated this assumption and leads to the following definition of mental chronometry: " T h e study of the time course of information processing in the human nervous system" (Posner, 1978, p. 7). Although several approaches can be used which fall under the general rubric of mental chronometry, the most common is the study of reaction time. This, in turn, can be viewed in several ways. The common approach is to consider reaction time as a dependent variable which aids inferences concerning the time it takes to process information. Most often, research has been done using either simple or choice reaction time. Simple reaction time measures a uniform response to a uniform "This research was partially supported by a research grant from the University of California, Riverside. A modified form of this paper was presented at the Annual Meeting of the American Educational Research Association, Los Angeles, 1981. The authors wish to express their gratitude to Mr. William Conlon, Principal at La Sierra High School for his cooperation and assistance. Reprint requests should be addressed to Jerry S. Carlson, School of Education, University of California, Riverside, CA 92521-0128.

265

266

CARLSON AND JENSEN

stimulus; choice reaction time assesses latency to one of a number of stimuli or where different responses are made to one or a number of stimuli presented. Within the last two decades considerable research effort has been given to studying reaction times. Most of the work has focused on methodological issues and information processing systems and models. Recently the relationship of reaction time indices to standard intelligence test scores and other psychometric abilities has been addressed, although the actual number of studies is quite small. From the li~terature as well as his own work, Jensen (1980-a, 1980-b) concludes that reaction time correlates consistently and negatively with psychometric measures purported to assess g. The results of a study carried out by Fairweather and Hutt (1980) challenge the generality of this conclusion, however, as they found nonsignificant correlations between WlSC IQ and reaction time. The purpose of this investigation is to replicate and extend a recent study of Jensen and Munro (1979) which bears directly on this issue; the relationship of reaction time to psychometric measures of g. In their study, 39 ninth-grade gi~'ls were tested with a reaction time (RT)--movement time (MT) apparatus. The results obtained were correlated with peformance on Raven's Standard Progressive Matrices. Significant negative correlations between both RT and MT and Raven matrices performance were reported. They are given in detail in the results section of this paper, allowing for comparisons with the present findings. In this study the Jensen and Munro research design is extended to the analysis of relationships between RT and MT and school achievement measures such as reading comprehension, performance on the California Test of Basic Skills (CTBS), and grades earned in mathematics and English.

METHOD

Subjects The subjects were 20 ninth-grade girls randomly selected from a high school serving a middle-class neighborhood in the Riverside area. The mean age of the sample was 14.4 years, the standard deviation was .62.

Procedures All subjects were tested individually with a reaction time--movement time apparatus. The design of the equipment conforms precisely to the apparatus used by Jensen (see Jensen, 1980-a, p. 689 for details). It consists of a panel which has a " h o m e " button and eight light-button combinations. The subject is asked to place her index finger of the preferred hand on the home button. (A warning buzzer is followed approximately 3 seconds later by the onset of the lights.) When

REACTIONTIME, MOVEMENTTIME, AND INTELLIGENCE

267

one of the lights goes on the subject is to remove her finger from the home button as quickly as possible and move to press a button directly in front of the light. The elapsed time from when the light goes on to when the finger is removed from the home button is the reaction time. The time taken to move from home to press the button which turns off the light is the movement time. RT and MT are assessed independently from one another with electronic timers. Simple reaction time is assessed when all lights but one are masked, i.e., the subject is presented with zero bits of information. Choice reaction time is assessed when one, two, or three bits of information are presented, corresponding to 2, 4, or 8 lights respectively. Although in the Jensen and Munro study 30 trials at each bit of information were involved, the present study had 20 trials at each bit. The Raven Standard Progressive Matrices (Raven, 1947) was administered to all subjects in small groups. Everyone completed the test within a 50-minute period. Reading comprehension was assessed by the Rate of Comprehension Scales (Van Wagener, 1953). This test assesses rate of reading. The score is affected by familiarity with vocabulary, language structure and content of material read. Scores from the California Test of Basic Skills (CTBS) were available for 11 of the subjects; mathematics for 18 and English grades for all 20. The CTBS scores were recent, resulting from testing within six months of when the reaction time data were gathered. The grades for mathematics and English were from the semester immediately prior to when the study was carried out.

RESULTS As can be seen from inspection of Figure 1, mean reaction time increases as a function of bits. This monotonic increase conforms to Hick's Law (Hick, 1952) and the results of most reaction time studies. Mean movement time remains relatively constant over bits. The reliabilities of the measures tended to be moderate. Split-half reliabilities were calculated at each bit and corrected with the Spearman-Brown formula to give an estimate for all 20 trials at each bit. The average reliability across all bits for RT was .68. The average MT reliability was .72. The reliability of the Raven matrices was .68. In order to assess the relationship between RT and MT and Raven matrices performance, correlational analysis and analysis of variance procedures were employed. For the ANOVA, subjects were evenly divided into those who scored above and below the mean on the Raven test. The mean RTs and MTs for these groups are shown in Figure 1. For RT, all the F ratios at each bit were significant (p < .05). For MT, significant F values (p < .05) were calculated at zero and two bits.

268

CARLSON AND JENSEN 6 7 0 r-

640

RT

~-

610 1:3

Z

580

--

550

0 I..) LIJ O3 _J 1 1

v I--

520

rr 0

490

I-rr

SIGNIFICANT ( p < . 0 5 ) NOT S I G N I F I C A N T

MT 460

Z W

t I

430

'

400

I I

I

H

370

I

I

I

I

0

I

2

3

INFORMATION

(BITS)

FIG. 1. Mean reaction time and movement time as a function of bits of information, plotted separately for high (above mean) and low (below mean) scorers on Raven's PM (N = 20)

Table 1 presents the correlations, uncorrected for attenuation, between Raven matrices performance and RT and MT at each level of information. The Jensen and Munro (1979) results are presented for comparison. The correlations corrected for attenuation are also shown inTable I. As can be seen from inspection of this table, all the correlations are negative and range from moderate to high. Correlations for total RT and MT on Raven matrices scores are presented in Table 2. Inspection of this table again reveals moderate to high relationships between both RT and MT and Raven performance: Especially noteworthy are the high correlations between the standard deviations of RT + SD RT and MT + SD MT. Figure 2 shows more clearly the relationship between average intraindividual variability in RT and Raven performance at each of the three bits of information.

REACTION TIME, MOVEMENT TIME, AND INTELLIGENCE

269

TABLE 1. Correlations between Raven PM Scores and RT and MT as a Function of Bits of Information. Correlations corrected for attenuation are in parentheses. Jensen & Munro

Number of Bits

RT

Carlson & Jensen MT

RT

0

-.26

(-.30)

-.38"

1

-.33'

(-.36*)

-.43'" ( - . 4 9 " )

-.42

-.41"* -.49*" -.35" -.37"

(-.45") (-.54"') (-.39") (-.41"*)

-.36* -.40"" -.36" -.39"

-.60"* (-.89"') -.58"" (-.89") -.52" (-.78"*)

2 2.58 3 MEAN

(-.42") (-.45"*) (-.48") (-.41") (-.46')

.46" (-.66"*) (-.69*')

MT -.45"

(-.64"*)

-.40

(-.60")

-.48" -.31 -.41

(-.70**) (-.49") (-.61"')

•p < .05

•"p < .01

Separating the sample into l o w and high scorers on the R a v e n matrices allows for c o m p a r i s o n o f differences in intraindividual variability for the two groups at each bit. In each instance, significant F values (p < 0.05) were calculated. This indicates that those subjects w h o have less variability or greater consistency in RT, score higher on the R a v e n than those w h o tend to be less consistent. The relationships b e t w e e n reading c o m p r e h e n s i o n and R T and M T as a function o f bits o f information are presented in Table 3. M o d e r a t e n e g a t i v e correlations w e r e found at e a c h bit for both R T and M T . All the R T correlations were significant with the e x c e p t i o n o f 1 bit of information. Analysis including total R T and M T parameters on reading c o m p r e h e n s i o n is s h o w n in T a b l e 4. The results are consistent with those found for the relationships o f the standard deviations o f R T and M T and R a v e n matrices p e r f o r m a n c e as these s a m e parameters correlate significantly with reading c o m p r e h e n s i o n . C o m b i n a -

TABLE 2. Correlations for Total RT and MT Parameters and Raven PM Scores

Raven x Total RT: r Raven x SD RT: r Total RT x SD RT: r Total RT + SD RT x Raven: R Raven x Total MT: r Raven x SD MT: r Total MT x SD MT: r Total RT × Total MT: r Total RT + Total MT x Raven: R Slope (of RT on Bits) x Raven "p < .05 "'p < .Ol

Jensen & Munro

Carlson & Jensen

-.39" -.31" .48"" .42"" -.43"* .07 .00 .37' .50"" -.30

-.54" -.71"* .83"" .72"" -.43 -.64" .81"" .61'" .55" -.20

270

CARLSON AND JENSEN I10-104 SIGNIFICANT ( p < . 0 5 )

>..

I--

98

w

92

b

.._/

'~ n,-

>~

86

J ~ D

80

Z

1::30

74

~-~

68

z

m

..J

L

62 z

56

LIJ (_9

5O

r~ Ld >