You are on page 1of 27

Psychological Bulletin 1976, Vol. 83, No.

6, 1026-1052

Reflection-Impulsivity: A Review
Stanley B. Messer RutgersThe State University
Studies of refiection-impulsivity support the view that it is correlated with performance on a variety of similar tests and predicts the quality of problem solving in problems that contain response uncertainty. However, reflectionimpulsivity is typically not a predictor of performance in preschoolers who have not yet learned to delay and scrutinize alternative hypotheses. The response time component of reflection-impulsivity is relatively independent of IQ, whereas the error component is moderately related. Shortcomings of the major test used to measure reflection-impulsivity are described, and norms to guide the researcher are presented. Studies of scanning behavior show that reflectives gather more information more carefully and systematically than do impulsives. Attempts to relate this variable to personality and social variables have been only partially successful. Reflection-impulsivity is related to certain clinical syndromes including hyperactivity, brain damage, epilepsy, and mental retardation. It also affects school performance, as shown by the greater impulsivity of children with reading difficulties, learning disabilities, and school failure. Impulsivity has been found to be modifiable: The most consistently successful strategy in this regard is to teach impulsives improved scanning strategies by means of appropriate training materials while having them verbalize such strategies aloud.

Until the last decade, individual differences in the cognitive processes of perception, memory, and problem solving were viewed as reflecting differences in basic intelligence. More recently, the contribution of motivation, anxiety, conflict, and attitudes to these individual differences has been explored. A major example of this new research area is field dependence-independence, a construct originally introduced and studied by Witkin (Witkin, Dyk, Faterson, Goodenough, & Karp, 1962). This paper reviews research in a related area, reflection-impulsivity (Kagan, Rosman, Day, Albert, & Phillips, 1964). Reflection-impulsivity describes the tendency to reflect on the validity of problem solving under a very special condition, namely, when several possible alternatives are available and there is
The author acknowledges the support of the Rutgers Research Council and expresses his appreciation to Jerome Kagan and Peter E. Nathan for their valuable comments on earlier drafts of 'this paper. Helpful suggestions were also offered by John A. Carpenter, Cyril M. Franks, and Edith Neimark. Requests for reprints should be sent to Stanley B. Messer, Graduate School of Applied and Professional Psychology, Busch Campus, Rutgers University, New Brunswick, New Jersey 08903.

some uncertainty over which one is the most appropriate. Tasks used to measure reflectionimpulsivity generally present the subject with several highly plausible alternatives, only one of which is correct. Thus, experimental subjects (usually children) who respond quickly often err (impulsives), whereas those who pause to reflect on response alternatives are more often correct (reflectives). The most common operational definition of reflectionimpulsivity includes response time and errors; in cases in which only response time is used for this purpose, it is presumed that longer response times are associated with fewer errors. The instrument most often used to measure reflection-impulsivity is the Matching Familiar Figures Test (MFFT; Kagan et al., 1964). Different forms of the MFFT are available for preschoolers, school age children, and adults.1 The test format involves simultaneous presentation of a figure (e.g., a boat, a pair of scissors, a telephone) with four, six, or eight facsimiles differing in one or
1 Copies may be obtained by writing to Jerome Kagan, William James Hall, Harvard University, 33 Kirkland Street, Cambridge, Massachusetts 02138.

1026

RKFLECTION-IMPULSIVITY

1027

FIGURE 1. Sample item from the Matching Familiar Figures Test.

more details. On each of the test's 12 items, the subject is asked to select from the alternatives the one that exactly matches the standard (see Figure 1). Time to the first response and number of errors overall are then computed. All studies with the MFFT but one report a negative correlation between response time and errors, with a median r of about .48. (However, see below for the very different correlations for preschool versus school age children.) In any given sample of subjects, the child who is above the median on MFFT response time and below the median on errors is called reflective: the child

who is below the median on response time but above the median on errors is called impulsive. In the early stages of his research, Kagan used the two measures separately as indexes of reflection-impulsivity, but more recently a dual index of response time and errors has been used by him and other investigators to differentiate between those subjects whose fast response times are maladaptive because they are associated with errors and those subjects whose fast response times are accompanied by few errors and are, hence, adaptive. Thus, the median split on the negatively correlated measures defines a 2 X 2

1028

STANLEY B. MESSER

matrix in which reflectives and impulsives about two thirds of most samplesare those who fall in the two diagonal cells. The remaining one third of the subjects fall in the counterdiagonal cells labeled fast-accurate and slow-inaccurate, but these subjects have been studied much less often. While the MFFT response time - error index is a direct measure of the disposition to delay before responding, one may inquire whether it is also a sensitive index of the child's tendency to evaluate his hypotheses carefully. Several studies bear directly on this issue. Kagan, Pearson, and Welch (1966a) reported that MFFT response time was directly related to number of separate glances at the standard and variants. The longer the response time, the greater the number of distinct glances at the figures. This finding implies, albeit indirectly, that reflective subjects more actively evaluate the stimuli. Siegelman (1969) extended this line of research by devising an apparatus to study the glancing behavior of reflectives and impulsives. To view the standard or its alternatives, Siegelman's subjects had to press a button to bring the figure into focus. This procedure permitted the experimenter to know precisely what the subject wished to see. Reflectives looked more often and longer at all figures than did impulsives. Similarly, Drake (1970), employing the Mackworth eye camera to determine the loci of study, found that reflectives consistently gathered more information about the stimuli before offering an answer, as gauged by mean number of looks per figure, amount of eye coverage of the figure, and extent of comparison of homologous features of two or more figures. Other researchers have since confirmed these findings and described the specific scanning strategies employed by reflectives and impulsives (see below). Apparently, the reflective not only spends more time evaluating his hypotheses but also gathers more information on which to base his decisions, and he gathers it more systematically than the. impulsive. STATISTICAL ISSUES AND MFFT NORMS Citing Kerlinger and Pedhauzer (1973), Ault, Mitchell, and Hartmann (Note 1) have

pointed out that artificially dichotomizing the continuous variables of MFFT latency and errors by means of the median split squanders potentially valuable discriminating information, which results in a substantial loss of statistical power. Such dichotomization encourages the researcher to think of reflective, impulsive, fast-accurate, and slow-inaccurate subjects as separate groups rather than as arrayed on a continuum in which one subject classified as impulsive may differ by just a few errors from another classified as fastaccurate. Dichotomization has led, in turn, to the frequent use of analysis of variance, a statistic that is not ideally suited to trait variables like reflection-impulsivity. After performing the median split, researchers have typically dropped the fast-accurate and slow-inaccurate subjects from further consideration. This, too, is wasteful of data that could help to illuminate the relative contributions of latency and errors in predicting other variables. Recently, there has been a call for clearer evaluation of the degree of variance contributed separately to such predictions by MFFT errors, response time, and their interaction (Block, Block, & Harrington, 1974, 1975). One further drawback of the median-split procedure is that it leads to the use of sample-based divisions into reflectives, impulsives, and so on, rather than to a reliance on normative data, which would ensure that one researcher's reflectives are not another's impulsives. There are, however, solutions to these posed difficulties. In many instances, the most appropriate statistic to use in the analysis of reflection-impulsivity data is multiple regression. It is a more powerful method than analysis of variance because it allows latency and errors to be employed as continuous variables, thus avoiding the hazards of dichotomization as well as allowing for calculation of the amounts of variance contributed separately by latency, errors, and their interaction (Ault et al., Note 1). When researchers choose to use analysis of variance, they should consider using all four groups rather than reflectives and impulsives alone. How many and which of the groups they choose to include, however, should depend on the particular problem under investigation. If the object of a particular

REFLECTION-IMPULSIVITY

1029

study is to help impulsives delay to decrease errors, one might focus on this group alone (e.g., Meichenbaum & Goodman, 1971). If the object is to study the scanning strategies of reflectives and impulsives, the inclusion of fast-accurate and slow-inaccurate subjects could add important conceptual clarity (e.g., Ault, Crawford, & Jeffrey, 1972). The comparison of reflective and fast-accurate subjects in the latter study, for example, contributed to an understanding of how the fastaccurate subjects achieved as few errors as reflectives without the extra time spent in reflection. Because the lack of national norms for the MFFT has necessitated the use of the median split for the purpose of classification, MFFT response time and error data were culled from current reflection-impulsivity literature and are summarized in Table 1. Researchers may wish to consult these preliminary norms, especially when using discrete classification. RELIABILITY AND STABILITY Test-Retest Reliability The short-term test-retest reliability most suitable to reflection-impulsivity derives from equivalent forms of the MFFT, because immediate or short-term retesting with the same form of the test would permit responses influenced by memory of correct responses. In the sole attempt to create equivalent forms, Egeland (1974) divided the 12-item MFFT forms for children and adults into three eightitem forms. In a pilot study with 30 children (sample characteristics unspecified), correlations among the three tests ranged from .92 to .98, but whether these figures applied to response time, errors, or both was not stated. The construction of equivalent forms of the MFFT for use in repeated measures designs is obviously needed. There are four studies in which children 6-10 years of age were retested on the same version of the MFFT after periods of 1-8 weeks. Response time reliabilities were .58, .68, .73, and .96; corresponding error reliabilities were .39, .34, .43, and .80 (Adams, 1972; Duckworth, Ragland, Sommerfeld, & Wyne, 1974; Hall & Russell, 1974; Siegel-

man, 1969). However, in both the Adams and Siegelman studies, only reflective and impulsive, but not fast-accurate and slowinaccurate, subjects were included. In the Siegelman study, only one error per item was allowed in the retest, and in the Duckworth et al. study, the subjects were educable mental retardates, which may account for the much higher reliabilities they attained (.96 and .80). Because of sampling and procedural irregularities and the use of the same version (vs. equivalent versions) of the MFFT, these studies may not accurately represent true MFFT test-retest reliability. Response uncertainty for subjects taking the same version of the MFFT twice in close succession may decrease so much that the test becomes a less valid measure of reflection-impulsivity. Internal Consistency Reliability Block et al. (1974) and Ault et al. (Note 1) both reported an internal consistency reliability coefficient for MFFT response time of .89 but lower reliabilities of .62 and .58, respectively, for errors. The unpublished data of several researchers collated by Ault et al. yielded typical internal consistency reliabilities for MFFT errors in the .50s. Stability Among preschool children, MFFT errors are moderately stable over time, whereas response time is not stable (see Table 2). By contrast, among school age children, response time is moderately stable over time but errors are not. The rather low reliability of MFFT errors in school age childrenas measured by shortterm test-retest reliability, internal consistency reliability, or stability over longer periods of timeresults in several problems (summarized by Ault et al., Note 1). One of these problems has already been noted above: misclassification of subjects among the four conceptual tempo groups. As long as MFFT response time and errors are employed as continuous variables by the use of either correlation or multiple regression, this danger is avoided. However, when the same version of the MFFT is administered twice in a repeated measures design, the effect of low

1030

STANLEY B. MESSER TABLE 1

MEANS, MEDIANS, AND RANGES UF THE MATCHING FAMILIAR FIGURES TEST. (MFFT) RESPONSE TIME-' AHD ERRUKS Complete sample before median split
A/

After division into reflectives and impulsives

Mdn Errors on items Response item Errors on all 12 items Response ti me per item Reflectives Impulsives

M
Errors on all 12 items Reflectives Impulsives
-. ,
---

Response Grade Nursery item

M or Mdn
Range

28.6 5.9 4.3- 7.5 24.5-31.8

5.1
50 6.6 Ill

n
Kindergarten

174 9.5
8.0-15.0

174
23.9
22.2-30.2

31.0 50 22.0 Ill 14.7


11.0-16.0

M or Mdn
Range

13.4
11.9-20.2

6.1 5.4- 9.0

18.2
14.4-25.8

29.2
23.2-35.3

Grades 1-2

141
12.2
10.5-15.4

141

109

109 9.9
6.7-11.6

109 7.2 30 5.1


1.8- 9.1

109
24.1
21.0-25.8

M or Mdn
Range

13.5
10.9-17.7

11.3
10.2-13.8

20.8

n
Grades 3-5

411
16.9
13.7-22.6

411
10.4
7.4-14.1

443
13.6
8.0-22.0

443 7.7
5.8-12.0

30
30.0
21.1-38.6

87 9.1
8.1-17.0

87
11.4
8.4-18.0

M or Mdn
Range

427

427

423

423

110

330

110

330

Note. The figures in the left half of the table are from different samples than are those in the right half of the table. The ranges are based on means or medians of samples and not of individuals. (In the one study in which scoies on the adult version of the MFFT were presented for reflectives and impulsives, reflectives' (n = 15) mean response time \vus 60.1 sec and mean number of errors was 2.6. The comparable figures for impulsives (n ~ 15) \verc 15.7 sec and 7.8 eirors.) 11 All response times are given in seconds.

MFFT reliability is regression toward the mean, the severity of which is inversely proportional to the degree of MFFT reliability. It is especially important, therefore, that when a pretest-posttest design is employed, a testretest/only control group be included. Low reliability of MFFT errors increases the variance of such scores, so that a larger

sample is required to detect treatment effects. In addition, as pointed out by Ault et al. (Note 1), when the reliability of a variable is low, its correlation with another variable will underestimate the true value of the correlation. Two examples offered by these authors illustrate the potential importance of this problem. "With an average internal con-

TABLli2 STABILITY OF THE MATCHING FAMILIAR FIGURES TEST (MFFT) RESPONSE TIME AND ERRORS Age or grade at first testing Study Wright (Note 3) Block, Block, & Harrington (1974) Ward (Note 5)b
a

Time span to second testing Years Months

MFKT Response time Boys Girls Errors Hoys Girls


.42

Years Months

0 7 3 3 11
Grade 1 Grade 1 Grade 1

-.16
.19

.67 .13 .24 .17 .70 .50

.09 ,34 .47 .40 .42 .21 .25 .33

Yando & Kagan (1968) Kagan (1965b) Messer (1970b)

.21 .14 .21 .13 .48 .31

.49 .34 .51 .23 .51

Note. All correlations are Pearson product-moment correlations (r) for test-retest. tt Kansas Reflection-Impulsivity Scale for Preschoolers was used, rather than the MFFT, b Longitudinal sample tested at 4 years 3 months, 4 years 11 months, and 5 years 11 months. With correction for attenuation, correlations are higher. A more difficult version of the MFFT was used at 5 years 11 months.

REFLECTION-IMPULSIVITY

1031

sisteiicy reliability for errors of .52 and for latency of .89, the correction for attenuation gives a true correlation between latency and errors of .80" (Ault et al., Note 1, p. 4). Using the figure of .80, latency accounts for about 64% of the variance in errors (vs. about 31% for the average attenuated correlation of .56). Similarly, an MFFT errorIQ correlation of .36, corrected for attenuation, becomes a correlation of .50. Ault et al. concluded that calculating estimates of true correlations by means of the correction for attenuation formula will yield a true picture of the relation of MFFT errors to other variables. Although such problems with the MFFT have solutions as stated, a different strategy might be to refine the test in such a way as to increase its reliability. This is a project for the future, however, and in the interim the reader might keep the above factors in mind in evaluating the studies reviewed here. Simpler matching familiar figures tests those containing fewer variantshave recently been developed for use with younger children (Banta, 1970; Lewis, Rausch, & Goldberg, 1968; Wright, Note 2). Of these, the Kansas Reflection-Impulsivity Scale for Preschoolers (KRISP; Wright, Note 2, Note 3) has been developed and studied most systematically. Norms, reliability data, and age and sex differences for this instrument, along with a preliminary investigation of scanning strategies of impulsives and reflectives, were reported by McCluskey and Wright (Note 4). The smaller negative correlation between response time and errors (about .11) and the lack of stability of conceptual tempo over 1 year in young boys suggest that either the KRISP is not a measure of reflection-impulsivity or that reflection-impulsivity is typically not a potent phenomenon in younger age groups (see section on development.) CONVERGENT VALIDITY If conceptual tempo is a construct with some generality, reflectives and impulsives should remain reflective or impulsive on tests similar to the MFFT, as well as in other test situations that contain response uncertainty. Yando and Kagan (1970) constructed 10 different matching familiar figures tests, each

with a different number of variants (ranging from 2 to 12). The tests were administered to 7-year-olds, one test each week for 10 weeks. In spite of differences in task complexity, which were accompanied by increases in response time and errors, most children retained their relative rank on both response time and errors. The median correlation over 10 weeks was .73 for response time and .68 for errors. Similarly, Ward (1968) administered two different matching tests to kindergarten childrenone with three response alternatives and one with five alternatives made up of geometrical or meaningful figures. Response times on the two tests intercorrelated .64 for males and .57 for females. Two other tests used to explore the generality of reflection-impulsivity are the Design Recall Test (DRT) and the Haptic Visual Matching Test (HVMT; Kagan et al., 1964). In the DRT, the standard and alternatives are geometrical forms rather than familiar objects, and the subject must choose the correct alternative from his memory of the standard, which is viewed first and then removed. The standard in the HVMT is either a three-dimensional geometrical form or a familiar object that the subject feels but does not see; the correct alternative is also chosen from memory. Response times to the MFFT, DRT, and HVMT are moderately intercorrelated, with intercorrelations ranging from .33 to .52 (Kagan, 1965a; Kagan et al., 1966a; Kagan et al., 1964). On Raven's Coloured Progressive Matrices, in which the subject is asked to select the correct variant from an array of six variants that will complete a matrix, MFFT latency correlated .54 with mean latency to choice of variant (Hall & Russell, 1974). Conceptual tempo can also extend to situations in which the subject has to answer on the basis of self-generated alternatives or questions rather than from alternatives presented to him by the examiner. On a "20questions" type of task in which the subject could ask preliminary questions to help determine which of several items the experimenter had in mind, the correlation over many items between response time on the MFFT and response time to the first ques-

1032

STANLEY B. MESSER

tion asked was .45 (Denney, 1973). Kagan (196Sc) asked children questions such as, "What is your favorite subject in school?" Their response times to 20 such questions significantly correlated with MFFT response times (.30 for boys and .38 for girls). Asked to evaluate the adequacy of their performance, children with longer MFFT response time took longer to respond (Kagan et al., 1966a). When subjects had to describe a tachistoscopically presented incongruous stimulus (Kagan, 196Sa), response time on this task was related to MFFT response time; when asked to decide how far to stand from a target (Mann, 1973), reflectives took longer than impulsives. Similarly, Ward (1968) gave kindergarten children MFFT-like tests as well as other tests offering several response possibilities but not containing the matching-to-sample feature of the MFFT. Testing conditions were deliberately varied, with some tests given in an evaluative context and others in a nonevaluative atmosphere. Significant moderate correlations (rs were typically in the .30s) between response time and errors were obtained across tests and across conditions. In a concept identification task requiring consideration of several alternative solution modes, the time taken to complete a problem was longer for reflectives than impulsives (Nuessle, 1972). The importance of the presence of response uncertainty in studies of generality of reflection-impulsivity is illustrated by two studies in which toy choices were offered to children and their decision times coded. When the toys offered had previously been rated comparably attractive or unattractive by the subjects (and, as a result, response uncertainty was present), reflectives took more time than did impulsives to choose the toy (Mann, 1973); when no preference rating was first obtained, no generality of response time was found (Eska& Black, 1971). The relevance of task involvement to generality of reflection-impulsivity was illustrated in another choice situation. When asked to tell stories to pictures, generality of response time was found for girls but not for boys, leading the investigators first to examine the picture content and then to con-

clude that the pictures were not "involving" for the boys (Eska & Black, 1971). Similarly, when subjects had to choose between two groups of words to spell, with penalties for misspelling (thereby enhancing task involvement), reflectives took longer than impulsives to make their choice (Mann, 1973). In sum, the reflection-impulsivity construct remains moderately robust over changes in the MFFT; it also extends to tests containing different requirements and content. Correlations between performance on the MFFT and such tests are not always high, however, suggesting the importance of the specific problem context in obtaining generality. RELATION OF REFLECTIONIMPULSIVITY TO PERFORMANCE ON OTHER TESTS Other motor and cognitive processes have been related to reflection-impulsivity, including motor inhibition, conceptual style, and field dependence-independence. Motor Inhibition The tendency to delay offering a solution in problem situations apparently extends to the ability to delay motor response in tasks in which there is no pressure to find a correct answer. Thus, on the Motor-Inhibition Test (Maccoby, Dowley, Hagen, & Degerman, 196S), which requires children to execute motor acts slowly, reflectives 5-J years old (Ward, Note 5) and 4 years old (Harrison & Nadelman, 1972; Hess, Shipman, Brophy, & Baer, Note 6) were better able to inhibit their action than were impulsives. In studies with younger preschoolers (Nadeau, 1968; Ward, Note S; Shipman, Note 7) no such relationship was found. However, the typical negative relation between MFFT response time and errors was absent among these preschoolers. Analytic Versus Relational Conceptual Style When asked to say which of several pictures of objects "go together" and why, some children prefer to group objects on the basis of similarity of certain objective physical attributes of the stimuli (e.g., "the watch and the ruler both have numbers": analytic), whereas others do so on the basis of a functional relationship ("the watch and the ruler

R EFLECTION-IMPULSIVITY

1033

both measure something": relational [Kagan, Moss, & Sigel, 1963]). Children who have a preference for analytic concepts were found to take longer before offering their responses than were children who prefer relational concepts (Kagan et al., 1963, 1964). This finding was subsequently extended by having children answer quickly or slowly when asked to group objects and then noting consequent effects on the production of analytic responses (Kagan et al., 1964). Under instructions to respond slowly, children produced more analytic responses than they did when instructed to answer quickly, leading Kagan and his colleagues to conclude that conceptual tempo is one prerequisite of analytic responding. Ostfeld and Neimark (1967) and Zelniker, Cochavi, and Yered (1974) confirmed this finding for children who were initially nonanalytic. In the case of children classified previously as analytic, however, forcing them to respond quickly did not diminish their tendency to respond analytically. It may be that once an analytic style is well established it is maintained under conditions of forced fast responding. Several investigators, however, have failed to replicate the relationship between analytic responding and reflectivity (Block et al., 1974; Denney, 1971, 1972; Wyne, Coop, & Brookhouse, 1970). In addition, Denney found that in spite of the demonstrable effect of a model's conceptual tempo on an ob-

server's conceptual tempo, there was no corresponding effect of a model's conceptual tempo on the observer's conceptual style. Conversely, a model's conceptual style did not influence a child's conceptual tempo, from which Denney concluded that conceptual style and conceptual tempo are unrelated. Alternately, one might conclude from these data that the modeling of conceptual style is simply an ineffective means of varying conceptual tempo, and vice versa. Field Dependence-Independence In the studies reported in Table 3, the test used most frequently to measure field dependence-independence is some variant of the Embedded Figures Test (EFT). In this test, subjects are asked to discover a simple figure embedded in complex designs on each of about a dozen items. The number of correct solutions is recorded, as is response time to first response in the studies listed in Table 3. There is a consistent, moderate overlap between MFFT and EFT, especially of error scores. Reflectives are significantly more field independent than impulsives (Campbell & Douglas, 1972; Keogh & Donlon, 1972; Massari, 197S; Massari & Massari, 1973; Mumbauer & Miller, 1970; Neimark, 1975; Schleifer & Douglas, 1973). Further, on the Children's Embedded Figures Test (CEFT), as on the MFFT, a negative correlation between response time and errors (.42) typically ap-

TABLE 3 CORRELATIONS OP REFLECTION-IMPULSIVITY AND FIELD-INDEPENDENCE


MFFT response time
ARC

MFFT errors
EFT

EFT

Study

Test

Years Months

time
.35 .46

score

KFT

time

score

Massari & Massari (1973) Schleifer & Douglas (1973) Murabauer & Miller (1970) Campbell & Douglas (1972) Keogh & Donlon (1972) Neimark (1975)

EC-EFT KC-EI-T

CEl'T CEFT
RET EFT

3 3 4 4 5
6 8 10 8-14 9-13 9-13

9 9 3 10 6 0 0 0

.26 .20 .18 .26 .38"

-.09

-.45

-.17

.43


.19

-.46 -.62 -.62 -.56 -.35 -.46 -.45" -.52 -.38 -.46

Note. Abbreviations: MFFT = Matching Familiar Figures Test; EC-EFT = Early Childhood Embedded Figures Test; CEFT = Children's Embedded Figures Test; RFT = rod-and-frame test. Correlations for all three samples combined, with age partialed out, were .25 and .45.

1034

STANLEY B. MESSER

pears (Mumbauer & Miller, 1970), although this finding did not hold in very young children who took the Early Childhood Embedded Figures Test (EC-EFT; Massari & Massari, 1973). The moderate association between reflection-impulsivity and field dependence-independence may rest, in part, on the similarity of the requirements of the MFFT and the EFT. Both tests contain response uncertainty and require scanning and analysis of a visual field. When a different kind of test was used to measure field dependence (the rod-andframe test; Witkin et al., 1962), scores were still moderately associated with MFFT errors (Keogh & Donlon, 1972). Perhaps response uncertainty is the common process involved in all three tests. Another study indicates that both of these two individual difference variables contribute independently to success at problem solving. In this study, in which 913-year-old children had to identify a concealed pattern from among eight possibilities by opening as few clue-containing shutters as possible, reflectives (vs. impulsives) and fieldindependent (vs. field-dependent) subjects solved the problems most efficiently (Neimark, 1975). DEVELOPMENT OF REFLECTION-IMPULSIVITY Children typically become more reflective with age. They increase response times and decrease errors on the MFFT, a finding confirmed both by cross-sectional and longitudinal studies (Ault, 1973; Campbell & Douglas, 1972; Fancher, 1969; Kagan, 196Sc; Ward, Note 5) and by the age norms shown in Table 1. In addition, the negative correlation between response time and errors tends to become larger with age. This relationship is considerably less robust in preschool age than in school age children. The median correlation for all available samples of subjects 5 years or younger is .28, as seen in Table 4 (Block et al., 1974; Fancher, 1969; Nadeau, 1968; Wright, Note 3; Hess et al., Note 6; Klein, Note 8; Shipman, Note 9), whereas for school age samples it is .S6.2 Of 26 correlations from studies of preschool children, only 5 rose above the .30s, whereas of 23 correlations from studies of school age children, only 4 were as low as the .30s. In the single longitudinal study of preschool chil-

dren available, a significant negative correlation between response time and errors was absent at 4 years 3 months but present at S years 11 months in boys. It appeared earlier in girls (4 years 11 months) and in children from higher socioeconomic status homes (Ward, Note S). Wright (Note 3), who administered the KRISP to a cross-sectional sample of 3-, 4.J-, and 6-year-old middle-class children, found that the relationship between response time and errors increased with age for boys but decreased with age for girls. This relationship appeared at an even earlier age for girls in Wright's study than in Ward's (age 3 vs. age S). To summarize, delay in response time comes to be associated with fewer errors only after school entrance. Perhaps the young child does not realize that delay is useful in solving problems and does not activate the strategies of delay. Ward (Note S) speculated that the preschooler is more oriented to the tester than to the task. (It has been observed during MFFT administration that nursery school children are frequently engaged in activities not germane to the task, such as talking to the examiner and daydreaming. In fact, some preschoolers responded to the MFFT without even looking at the stimuli | Nadeau, 1968]). Implied in Ward's explanation is the view that the young child does not yet possess a motive for performing well and hence does not bother reflecting on his response. This is consistent with Messer's (1970a) data suggesting that concern about intellectual performance is one antecedent of a reflective disposition. RELATION TO INTELLIGENCE It has been suggested that results obtained with the MFFT may be due largely to the relation of reflection-impulsivity to IQ (e.g., Block et al., 1974). For that reason, studies were scanned for reports of correlations between MFFT response time, errors, and IQ (see Table 4). They revealed that there is a median correlation between MFFT response time and IQ (for both sexes) of .165 (.14 for
2 The studies listed in Table 4 are only those in which MFFT-IQ correlations were reported. There is no reason to expect that they are biased in regard to MFFT response time-error correlations.

REFLECTION-IMPULSIVITY TABLE 4 CORRELATIONS OF MATCHING FAMILIAR FIGURES TEST (MFFT) RESPONSE TIME, ERKOKS,
AND IQ FOR BOYS AND GlRLS MFFT response time-IQ IQ test
S-B S-B PSI PS I

1035

Age

MFFT errors - IQ Male Female


.67

MFFT response time -errors Both


-.50 -.63

Study Lewis, Rausch, Goldberg, & Dodd (1968) Massari & Massari (1973) Ward (Note 5) Harrison & Nadelman (1972) Mumbauer & Miller (1970) Meichenbaum & Goodman (1969) Ward (1968) Ward (Note 5) Campbell & Douglas (1972) Kagan (196Sc) Kagan, Pearson, & Welch (1966a) Kagan (1965c) Yando & Kagan (1970) Denney (1973) Kagan (1965a) Campbell & Douglas (1972) Kagan, Rosman, Day, Albert, & Phillips (1964) Kagan (1965c) Eska & Black (1971) Kagan et al. (1964) Mollick & Messer (in press) Campbell & Douglas (1972) Meinster (1972)

Years Months

Male Female Both


.30 .26 -.06 -.04 .46 . -
.23 .15 .19 .00 .27 .18 .00 .12 .13 .05 .36 .36 .12 .33 .23

Male Female Both

3
3 4 4 4 5

.45 .22 .03 -.03 .00

-.03 -.02 .26 .52


.39

-.40 -.29 -.47 -.66 -.30

-.54 -.69 .09 -.01 -.25


-.02 -.39 -.30

-.37 -.50 .01 -.20 -.35


-.39 -,26 -.43

.01 -.20

3 11

-.35 -.52 -.57 -.34

8 4 8
9

PPVT
S-B PMA PPVT PPVT PSI

-.29 -.53

-.36 -.48 -.64 -.36


.
_ -.39 . . . .

5
5

5 5 6 6 6 7 7
7 8 8

9 11 4 6 6 6 6 6
0

.39 .29 .35 .10 .46 .25


.10

.23 _ .07 .

-.09 -.27 -.53 -.06 -.26

-.33 -.22 -.51


-.32

-.62 -.53

WISC-FS WISC-V

-.30 -.50 -.56 -.65


-.61 -.63 -.65

wise v

WISC-V

K-A L-T WISC-V

.14"

-.21 -.25 -.18


.05

-.22 -.20 -.33


-.14

3 6 6
6 6 6 6 0 4

WISC-FS CTMM WISC-V WISC-V WISC-V 0-L WISC-V WISC-FS WISC-FS WISC-V

.15 .22 .19 -.16 .45 -.03


.08

-.39 -.57 -.66 -.75 -.46 -.69 -.65

8
8 8 8 8 9 10 10 10

-.13

-.47 -.53 -.36 -.38 -.47 .10


-.08 -.295

-.21 -.28 -.30 -.02 -.35 -.40 -.36


-.335

-.51 -.47 -.47 -.62 -.56 -.60 -.47

Mdn

-.16 .04 .14

.22

-.38
-.48

-.47

Note. Abbreviations: S-B = Stanford-Binet; PSI = Preschool Inventory; PPVT = Peabody Picture Vocabulary Test; PMA = Primary Mental Abilities; WISC-FS = Wechsler Intelligence Scale for Children-Full Scale; WISC-V = Wechsler Intelligence Scale for Children-Verbal; K-A = Kuhlmann-Anderson; I.-T = Lorgo-Thorndike; CTMM = California Test of Mental Maturity; O-L = Otis-Lennon. " Estimated median conelatinn based on 11 coefficients.

boys and .22 for girls). The correlation between MFFT errors and IQ is -.295 for boys and .335 for girls. Thus, conceptual tempo is moderately related to IQ when IQ falls in the normal range, and the relationship is higher for errors than for response time and slightly higher for girls than for boys. Other studies that compared the IQ of impulsives and reflectives using t tests bear out these general findings (Ridberg, Parke, & Hetherington, 1971; Siegelman, 1969; Nadelman & Wallace, Note 10). However, order of presentation of the MFFT and Wechsler Intelligence Scale for Children (WISC) IQ may affect MFFT scores and size of reflectionimpulsivity - IQ correlations (Plomin & Buss, 1973). When the MFFT was administered after the WISC, second-grade subjects had more reflective scores on the MFFT than did

those who took the MFFT first, and correlations between IQ (Verbal, Performance, and Full Scale) and MFFT (response time and errors) were somewhat lower. The MFFT-IQ correlations may be partly a function of age (see Table 4). Thus, when subjects of these studies were categorized as preschool (5 years or below) or school age (6 years or above), and as above or below the median for their sex on response timeIQ correlation, the resulting 2 X 2 contingency table revealed a significant association between the two variables (x2 = 4.61, p < .05, two-tailed), suggesting that a higher correlation exists between MFFT response time and IQ for preschool than for school age children. It is important to note that the intelligence tests employed in these studies differed in

1036

STAiNLEY B. MESSER

two ways: question format and verbal content. Question format was either the standard question-and-answer variety, with the subject generating an answer (as in WISC Verbal scales), or the multiple-choice variety in which the subject is presented several similar alternatives pictorially (as in the Otis-Lennon Test). It is clear that question-and-answer tests place a premium on productive verbal skills, whereas multiple-choice tests require more evaluation of perceptual attributes, as on the MFFT. It is not surprising that MFFT scores relate more highly to IQ when both the format and the task requirements are similar to the MFFTas they are on group-administered intelligence scales. To test this latter hypothesis, four experienced clinical psychologists were asked first to take the MFFT and then to inspect the 10 intelligence tests listed in Table 4 in order to predict whether, on the basis of format or content, or both, they expected the correlation between MFFT response time and IQ to be above or below the overall median correlation. On 7 of the 10 tests, all four of the raters agreed on their predictions (Low: WISC, WISC Verbal IQ, Preschool Inventory; High: Lorge-Thorndike Intelligence Tests, Kuhlmann-Anderson Intelligence Tests, Primary Mental Abilities Test, Otis-Lennon Mental Abilities Test). These predictions were then tested against the actual findings. A 2 X 2 contingency table revealed a significant association (x2 = 3.62, p < .05, onetailed) between type of test (i.e., like MFFT or unlike MFFT) and response time- IQ correlation (above median or below median) in the expected direction. Thus, when the content of an IQ test is primarily nonverbal and the format requires decision about alternatives (multiple choice), correlations of MFFT response time to IQ are higher than when the test calls for verbal responses, especially to items with minimal response uncertainty. This was confirmed in studies that found higher correlations between both MFFT variables and nonverbal (vs. verbal) sections of the Otis-Lennon Mental Abilities Test (Eska & Black, 1971), and higher correlations between MFFT errors and WISC Performance (vs. Verbal) IQ (Plomin & Buss, 1973). One way to view these data

is to conclude that reflection-impulsivity overlaps with some components of intelligence tests, particularly those that are multiple choice and nonverbal and, hence, that IQ may account for many reported relationships between reflection-impulsivity and other variables such as problem solving (discussed below). Zigler (1963) presented a similar argument in his critique of the field dependence-independence research. Alternately, one can conclude that performance on nonverbal parts of intelligence tests is influenced by conceptual tempo as reflected in the higher response time - IQ relationship for such tests. In either case, a prudent course to follow for future studies would be to control at least for verbal IQ, especially when MFFT errors are included in combination with response time as a predictor. SCANNING STRATEGIES OF REFLECTIVES AND IMPULSIVES Several investigators have attempted molecular analyses of the way reflectives and impulsives of different ages actually deploy their attention when responding on MFFTlike tasks (Ault et al., 1972; Drake, 1970; Nelson, 1969; Siegelman, 1969; Zelniker, Jeffrey, Ault, & Parsons, 1972; McCluskey & Wright, Note 4; Wagner & Cimiotti, Note 11). The subject's eye movements to each of the stimuli were recorded in all studies but two, in which the subject had to press a button to bring the MFFT standard or variant into focus (Nelson, 1969; Siegelman, 1969). Reflectives typically examined more variants and made more eye fixations per stimulus than did impulsives. Adult reflectives, unlike adult impulsives, scanned all the alternatives before choosing one (Drake, 1970). Reflectives also spent a higher percentage of their total viewing time comparing pairs of stimuli that included the standard and an alternative as well as scanning a larger number of different pairs. Among preschoolers, reflectives viewed a higher percentage than impulsives of pairs made up of two alternatives (Wright, Note 2). Reflectives were more thorough in comparing the pairs, in that they looked back and forth between one and the other more often than impulsives.

REFLECTION-IMPULSIVITY

1037

These studies suggest that the search of reflectives involves greater concentration on homologous parts of the variants. By contrast, the viewing behavior of impulsives is less systematic and more global. In particular, they do not scan the field for distinctive features as systematically as reflectives. That the performance of reflectives (vs. impulsives) is more clearly influenced by their attention to feature differences is also suggested by a study in which the type of information being processed by these conceptual tempo groups was assessed by a transfer of training procedure (Odom, Mclntyre, & Neale, 1971). Surprisingly, the average duration of a fixation is the same for impulsives and reflectives, and neither group employs the strategy of checking distinctive features one at a time across all stimuli. When reflectives and impulsives were studied cross-sectionally at two different ageschildren versus adults (Drake, 1970) and 3- versus S-year-olds (McCluskey & Wright, Note 4)the scanning differences between younger and older subjects resembled the difference discovered between impulsives and reflectives. The major area of disagreement among these studies of scanning strategy is in percentage of eye fixations devoted to the standard. Siegelman (1969) and Nelson (1969) reported that impulsive children devoted more time to the standard, Drake (1970) and Wright (Note 2) found that it was the reflectives who devoted more time to the standard, whereas Ault et al. (1972), Wagner and Cimiotti (Note 11), and Zelniker et al. (1972) reported no difference. Subjects who were either fast and accurate or slow and inaccurate on the MFFT were included in one important comparative study (Ault et al., 1972), which revealed that fastaccurate subjects looked at significantly fewer different variants and spent a greater percentage of their time comparing just a few pair of stimuli than did impulsives, reflectives, and slow-inaccurate subjects. The fast-accurate children achieved both speed and accuracy by attempting to match the standard in as few comparisons as possible. This contrasts with the more cautious strategy of eliminating all the incorrect variants before

responding, which is characteristic of reflectives. PROBLEM SOLVING AND SPEECH OF REFLECTIVES AND IMPULSIVES On a variety of perceptual, conceptual, and perceptuomotor problem-solving tasks involving response uncertainty (i.e., where the answer is not immediately obvious), reflectives consistently perform better than impulsives, behaving like more mature children. On a color-form matching test, which permitted subjects to match the standard on the basis either of color or form, reflectives gave more form responses (the more mature answer) than did impulsives (Katz, 1971). Reflectives revealed better short-term auditory memory on a serial learning task than did impulsives (Kagan, 1966a) and superior short-term visual memory on a visual recognition task in which the child had to recall which of two similar pictures had been presented previously (Siegel, Kirasic, & Kilburg, 1973). They were also less reliant on memory supports during a concept attainment task (McKinney & Banerjee, Note 12). Reflectives solved Porteus mazes more successfully than impulsives (Shipe, 1971; Weintraub, 1973); they were even better card players at the game of Hearts (Jeffers, 1974). In an analogical reasoning task in which the subject was asked to complete analogies such as "Five is to number as black is to ," impulsives were more likely to answer with high-probability, but incorrect, associations such as "white," whereas reflectives more typically used analogical (and correct) reasoning as a basis for response (Achenbach, 1969). In one of two conceptually related inductive reasoning tasks, geometrical forms were presented in sequence, and the subject was required to continue the sequence correctly with one of several presented alternatives. In a second task resembling the WISC Picture Arrangement subtest, the subject was shown a sequence of pictures and asked to complete the picture story with one of several pictures. In both tasks, reflectives performed better than impulsives (Kagan et al., 1966a). On a two-choice discrimination learning task, reflectives made more correct responses than did impulsives (Massari &

1038

STANLEY B. MESSER

Schack, 1972). On a three-choice probability low affective salienceinvolved the presence learning task, impulsives showed more imma- or absence in the cartoon of interpersonal agture problem-solving strategies than reflec- gression, hostility, or imminent danger. Comtives in one study (Adams, 1972) but not in pared with impulsive and slow-inaccurate subjects, reflectives at 6 and 8 years of age, another (Fancher, 1969). In four studies that employed different but not at 10, revealed greater understandvariations of the 20-question game (Mosher ing of the visually incongruous, high-affective& Hornsby, 1966), reflectives were more salience cartoons and of the conceptually likely than impulsives to ask questions that incongruous, low-affective-salience cartoons. quickly eliminated a great number of pos- The fast-accurate subjects fell between resible answers and led them to the correct an- flectives and the other two groups but were swers most efficiently (Ault, 1973; Denney, not statistically different from them. In ad1973; Finch & Montgomery, 1973; Mc- dition, the reflectives were better able than Kinney, 1973). In a related concept identi- the other three groups to maintain their comfication task, the subjects were presented with prehension of those cartoons that were high stimuli that varied along four dimensions in affective salience. However, it was the im(e.g., size, position) with two levels of each pulsives who displayed the greater mirth to (e.g., large and small; right and left). The the high-affective-salience cartoons, especially subject had to decide which stimuli would ap- compared with the fast-accurate subjects. In their use of speech, as in problem solvpear in which positions by trying out a hypothesis and then getting feedback as to its ing, reflectives are more mature than impulcorrectness. Reflectives made more efficient sives. Meichenbaum (Note 13) studied the use of feedback information than did impul- spontaneous verbal behavior of a small group of 4^-year-old impulsives and reflectives. sives in solving such tasks (Nuessle, 1972). In two studies using college students as sub- (Response time-error correlations were not jects, reflection-impulsivity did not relate presented in this report.) Impulsives verbalto short-term memory, habituation of the ized less than reflectives, though they disorienting response, introversion-extraversion played twice as much egocentric speech. (Young, 1973), or to abstract versus con- Reflectives, by contrast, displayed more selfcrete conceptual structure (Wolfe, Egelston, guiding private speech, especially in response & Powers, 1972). The authors of both studies to specific tasks, as well as more outer-dispeculated that their negative findings might rected private speech (e.g., animal noises). be due to a skew in their college samples to- Overall, reflective preschoolers, compared with ward reflectivity and, therefore, ambiguous impulsives, possessed a more mature level of separation of subjects into reflectives and im- speech, which they used in a self-guiding pulsives. Unfortunately, neither study pre- fashion. sented the MFFT mean scores or response In a similar vein, Meichenbaum and Goodtime - error correlations by which their specu- man (1969) examined the ability of reflective lations could be confirmed. Except for these and impulsive kindergarten children to use two studies, the previous data support the their own speech (either overt or covert) to conclusion that reflective children employ control their own motor behavior. In one better problem-solving strategies in tasks with task, each child was required to press a finger response uncertainty. tapper while saying the word faster or slower In a unique study of cartoon humor, Brod- and to tap accordingly. In a second task, a zinsky (1975) compared all four conceptual blue light on a panel indicated that the subtempo groups on their comprehension and jects should say the word push aloud or to appreciation of cartoons that varied on two themselves and push a foot lever, whereas a dimensions. Some cartoons contained obvious yellow light indicated that they should say visual incongruity, whereas the incongruity don't push and refrain from pushing the foot in others was more conceptual and abstract, lever. Reflectives manifested more verbal conapparently requiring concrete operational trol of their motor behavior. The difference thought, The second dimensionhigh versus between reflectives and impulsives was espe-

REFLECTION-IMPULSIVITY

1039

dally apparent when covert speech was required. Impulsives used the words as a metronome, tapping each time they uttered a word, but reflectives used the words as cues, tapping several times for each utterance. PERSONALITY TRAITS AND SOCIAL BEHAVIOR OF REFLECTIVES AND IMPULSIVES Several personality and social variables have been studied in relation to reflectionimpulsivity in an effort to bridge the gap between cognitive and emotional dispositions two aspects of behavior that are often treated separately. Anxiety Over Error Following the arousal of anxiety over error after failure or perceived failure in an anagrams task, both reflective and impulsive boys showed increases in MFFT response time. Impulsives showed a decrease in errors, but errors of reflectives were constrained by a floor effect. A success experience, by contrast, led to faster response times for all subjects and increased errors for reflectives, leading to the conclusion that anxiety over error is one antecedent of a reflective disposition (Messer, 1970a). Freebody (Note 14) attributed Messer's results to an accuracy set engendered by the experimenter's feedback on the correctness or incorrectness of subjects' answers. Perhaps what occurs is that feedback on subjects' performance arouses concern or anxiety over performance, which, in turn, induces a set for accuracy resulting in greater cautiousness and evaluation.3 Consistent with Messer's results, both reflectives and impulsives increased MFFT response time following failure on MFFT items or items on an anagrams task (Reali & Hall, 1970; Ward, 1968; Weiner & Adams, 1974). Two studies have shown the greater impact on reflectives (vs. impulsives) of failure or potential failure on intellective tasks, supporting the hypothesis that reflectives are more anxious than impulsives, specifically about the quality of their intellectual performance. After receiving feedback as to the correctness of their hypotheses in a concept identification task, reflectives had significantly longer latencies following incorrect versus correct hypotheses than did impulsives.

Such longer latencies were also associated with more effective problem solving (Nuessle, 1972). Elsewhere, reflectives showed more disruption of memory on a serial learning task following a communication arousing anxiety over possible failure than did impulsives (Kagan, 1966a). These studies also suggest that impulsives or reflectives can improve their performance on such tasks only when anxiety over response accuracy can be readily coped with by increased reflectiveness. Anxiety Over Competence Based on teacher ratings of the personality characteristics of reflective, impulsive, fastaccurate, and slow-inaccurate children, Block et al. (1974) concluded that impulsive 4-yearolds are the most fearful, inhibited, and highest in anxious self-concern of the four groups. The latter terms refer to a broad personality trait associated with weak ego functioning. Tt may be that reflectives possess anxiety associated with uncertainty over making mistakes on intellectual tasks (Messer, 1970a), whereas impulsives are anxious over a basic inability to perform with competence on any task (Block et al., 1974). Rather than limit the significance of their findings to impulsives and reflectives, Block and his colleagues suggested that the personality correlates of accuracy or inaccuracy per se are far greater than the correlates of response time alone or in combination with errors, an assumption that casts doubt on the utility of the reflection-impulsivity variable It must be noted however, that their findings may be a function of the age of their subjects (4 years old), for whom the correlation of response time and errors was only .33. Under such circumstances, response time is not a measure of the tendency to evaluate hypotheses, and hence it is not surprising that the claims made for MFFT response time based on many studies of school age children, where the correlation of response time and errors is higher ( .56), did not hold (see
3 Anxiety, as measured by scales such as the Test Anxiety Scale for Children (TASC) or the StateTrait Anxiety Index (STAI), does not relate to reflection-impulsivity (Messer, 1970a; Freebody, Notq 14; Ferguson, Note 15).

1040

STANLEY B. MESSER

Kagan & Messer, 1975, for their reply to Block et al., 1974, and the subsequent comment by Block et al., 1975). A ttentiveness Reflective 4- and 8-year-old children sustain attention in play longer than impulsives (Campbell, 1973). Impulsive preschoolers are likely to start and stop their activities and to chat or roam between activities, but reflectives sustain attention even while chatting (Welch, Note 16). Impulsive school children are rated by teachers as equal to reflective, slow-inaccurate, and fast-accurate children in motivation to learn, as less attentive than reflective but equally attentive to fast-accurate and slow-inaccurate children, and as more hyperactive than reflective and fastaccurate children but similar in this respect to slow-inaccurate children (Ault et al., 1972). The ability of reflectives and impulsives to sustain attention has also been explored by testing their reaction times with a variable interval between the signal to "get ready" and the presentation of the stimulus to which they are to respond (Zelniker et al., 1972). As predicted, no difference between impulsives and reflectives was observed on the short preparatory intervals, but on the longer ones, impulsives took longer to respond, suggesting their greater reluctance to sustain attention in a laboratory task. Aggressiveness On a modification of Buss's aggression machine, 7-^-year-old impulsive boys displayed more aggression than did reflectives (Thomas, 1971). Perhaps impulsives exercise not only less cognitive control than do reflectives but less behavioral control as well. Locus of Control Individuals who delay in order to evaluate may do so because they see themselves, not others, as responsible for their successes and failures. Using a single index of conceptual tempo (errors/latency),Shipe (1971) found a significant but moderate correlation between reflectivity and internal locus of control in a sample of vocational school boys but not in a similar sample of institutionalized boys.

However, Massari (1975) reported only a trend toward internality among reflective first- and third-grade black children, Berzonsky (1974) found no difference in first and second graders, and Finch, Nelson, Montgomery, and Stein (1974) discovered no difference in locus of control between reflectives and impulsives among emotionally disturbed children whose mental age was 10.7 years. In addition, Messer (Note 17) found no relationship in a group of fourth-grade middle-class boys and girls. In this latter paper, Messer proposed that the superior performance in academic-intellectual tasks of children with an internal locus of control (cf. Messer, 1972) may be due to greater reflectivity on the part of internals. Because the internals were not more reflective than the externals, the substantial relation of locus of control to grades and achievement test scores was not affected by conceptual tempo. In a related study, Campbell and Douglas (1972) observed that reflective boys were more likely than impulsives to tell stories on a projective test that involved expectation of success in meeting a threat of frustration. By contrast, the impulsive subjects passively accepted the inevitability of such events. Moral Behavior Moral judgments were elicited by Schleifer and Douglas (1973) from 6^-year-old children to pairs of stories such as those used by Piaget. One described a well-intentioned act resulting in considerable damage; the other described a selfishly or maliciously motivated act with minor consequences. Longer MFFT latencies and fewer MFFT errors were associated with the tendency to label the actor in the latter story as naughtier. That is, reflectives made moral judgments on the basis of intentions rather than consequences, reflecting a more advanced stage of moral judgment. Risk Taking No difference was found between impulsives and reflectives on two quite different risk-taking tasks, one involving easy or difficult goal setting (Mann, 1973) and another involving a strong gambling element (Kopfstein, 1973). Based on the scanning study by

REFLECTION-IMPULSIVITY

1041

Ault et al. (1972), which revealed that fastaccurate subjects scanned the fewest alternatives before choosing, one might predict that it would be these subjectswho take the greatest error risk on the MFFTrather than impulsives or reflectives who would exhibit greatest risk taking in goal setting and gambling ventures. Delay of Gratification Mischel (1958) has devised a task in which children are offered the options of a small but immediate reward (e.g., one piece of candy today) or a larger delayed reward (two pieces tomorrow). Each of three studies using this technique to examine relations between conceptual tempo and delay of gratification has reported a different finding. Thus, Mann (1973) reported that reflective first graders were more likely to choose delayed rewards than were impulsives. By contrast, Shipe (1971), using a single index of reflection-impulsivity (errors/latency), found that institutionalized mentally retarded impulsive adolescents delayed gratification more than reflectives, but no such relationship adhered for a like group of mildly retarded vocational school boys. Finally, Ward (Note 5) reported no relations at all for impulsive and reflective preschoolers. It is possible that the relation of reflection-impulsivity to delay of gratification depends on the age or composition of the sample. SOCIAL CLASS AND SEX DIFFERENCES IN REFLECTIVES AND IMPULSIVES

times and made more grammatical errors than did middle-class boys (Heider, 1971; Schwebel, 1966). Finally, when asked to describe events of the day or to group objects on any basis, lower-class boys once again had shorter response times than did middle-class boys (Schwebel, 1966). Sex Differences Ward (Note S) reported no consistent sex difference in MFFT response time for children tested at 4, S, and 6 years of age, though girls displayed fewer errors than did boys at each of the three ages. Harrison and Nadelman (1970) found 4^-year-old girls to be more reflective than boys on both response time and errors, although no such sex difference was found in samples of 3J-year-old middle-class whites (Lewis et al., 1968), 5year-old lower-class blacks and middle-class whites (Zucker & Strieker, 1968), or middleclass 6- and 8-year-olds (Adams, 1972). In only one instance did girls have faster response times (Meichenbaum & Goodman, 1969). Small but consistent sex differences in the direction of fewer errors were reported in samples of 6-, 7-, and 8-year-old girls (Kagan, 196Sc). Overall, then, these data lack consistency, with some studies reporting sex differences and others not. When sex differences do appear, the girls are slightly more reflective. Sex differences are also reflected in relations between MFFT response time and errors. Lewis et al. (1968), who reviewed four relevant studies, concluded that boys show a larger correlation between MFFT response time and errors than do girls and that girls show a higher correlation between MFFT errors and IQ than do boys. Boismier (1971) criticized these conclusions on the basis of faulty statistical comparisons, arguing that Lewis et al. contrasted each correlation coefficient with the null hypothesis of a zeroorder coefficient rather than comparing the correlation coefficients of each male and female sample. In a rejoinder, Lewis (1971) agreed with Boismier but attempted to support his earlier findings by the addition of four subsequent studies. Table 4 contains 18 relevant comparisons for boys and girls ranging in age from 3 to 10 years. There is no

Social Class
Lower-class children from S to 12 years of age are consistently more impulsive on the MFFT, as measured by both response time and errors, than are comparable samples of middle-class children (Heider, 1971; Mumbauer & Miller, 1970; Schwebel, 1966; Weintraub, 1973; Zucker & Strieker, 1968). However, in a sample of S-year-olds in which MFFT response time and errors were highly correlated with IQ, the relation of social class to reflection-impulsivity disappeared when IQ was controlled (Mumbauer & Miller, 1970). In another study, lower-class boys given several words and asked to incorporate them into a sentence showed shorter response

1042

STANLEY B. MESSER

overall sex difference in either response time error or error-IQ relations, thus disputing Lewis's earlier findings. Massari and Massari's (1973) claim that girls develop an MFFT error-IQ relation earlier than boys is also not supported by these data. Instead, they show a small sex difference in the response time - IQ correlations, with girls attaining a slightly higher median coefficient (r = .22) than boys (r .13) and, as noted above, girls developing the response time error relation earlier. BODY BUILD, PHYSIOLOGICAL VARIABLES, AND BIRTH ORDER Kagan (1966b) explored the relation of body build in boys and girls to reflectionimpulsivity. In two samples of third- and fifth-grade children, boys who were short and had a broad chest girth were more likely to be impulsive than boys who were tall and of narrow chest girth (who were more often reflective). An attempt to replicate these results with first-grade boys was unsuccessful, however. No consistent relation between body build and conceptual tempo in girls was observed in any of these three samples of subjects. In the one sample in which the relation of perception of body size to conceptual tempo was studied, impulsive third-grade boys tended to perceive themselves as shorter than reflective boys of similar stature. In studies employing samples of first(Hemry, 1973) and third-grade boys (Campbell, 1973a), no difference in birth order between reflectives and impulsives was found. Using a slide version of the MFFT, Schwartz and Tursky (1969) took continuous physiological measures on 20 male adults who were taking the MFFT. Resting cardiac variability showed a moderate positive correlation (r = .45) with impulsivity. At slide onset, as well as during discrete eye movements, cardiac acceleration was obtained for impulsives and cardiac deceleration for reflectives. Since cardiac deceleration is a physiological state varying with high concentration, the findings are consistent with the greater attentiveness shown by reflectives (as discussed above). EDUCATIONAL AND CLINICAL PROBLEMS Several groups of children, some of them deficient in school-related skills and others

with clinical diagnoses, have been studied using the MFFT. A group of 8- to 14-yearold boys with severe learning disabilities and social and emotional problems was more impulsive than a group of boys with moderate to mild learning disabilities (Keogh & DonIon, 1972). In a retrospective study of 65 first-grade boys who were first administered the MFFT in Grade 1, 7 were found to have failed a grade 2 years later. Of these seven, five were impulsive, one was reflective, and one was slow and inaccurate in Grade 1. Two years later the same seven children, though comparable in verbal intelligence to those children who had not failed, were still significantly more impulsive, as gauged by MFFT response time and errors, than was the sample as a whole (Messer, 1970b). Similarly, impulsives were found to be placed two grades below reflectives in a group of 12-year-old, emotionally disturbed children in short-term residential treatment, in spite of their comparability on achievement tests (Finch, Pezzuti, Montgomery, & Kemp, 1974). Several studies have explored the relation of a more specific learning skill, reading proficiency, to conceptual tempo. In one such study, children in a reading readiness class were found to be significantly more impulsive than children in a regular first-grade class (Nadelman & Wallace, Note 10). However, among older middle-class children in Grades 2-5, MFFT latency and errors did not discriminate between poor and average readers on the Gilmore Oral Reading Test, or on four subtests of the Gates-McKillop Reading Diagnostic Tests (Denney, 1974). In more detailed studies of reading proficiency in first-grade boys and girls (Kagan, 1965b) and third-grade boys (Hall & Russell, 1974), in which the children were asked to select the correct word from five alternatives (e.g., moon: noon, boom, etc.), word recognition errors were negatively related to MFFT response time and positively to errors. For low-verbal boys in the Kagan study, letter recognition errors were also related to MFFT response time. One year later, those children who were impulsive in first grade tended to have the highest reading error scores when asked to read a paragraph aloud. The MFFT errors, however, were a better

REFLECTION-IMPULSIVITY

1043

predictor of reading performance among girls, whereas response time was a better predictor among boys. Kagan concluded from these results that the relation between conceptual tempo and reading depends on the presence of sufficient response uncertainty in the reading materials for the particular sample of children tested. If this conclusion is valid, it may be that for high-verbal boys the letter recognition task may not have evoked response uncertainty (as it did for low-verbal boys). Further support for the importance of a reflective attitude in reading proficiency comes from a study by Egeland (1974), who found improved reading comprehension among second-grade impulsives from inner-city schools 5 months after employing a training procedure that successfully increased their reflectivity. These data suggest that an impulsive tempo is one factor contributing to poor school performance whether it is labeled as general learning disability, school failure, or a reading problem. Children diagnosed as hyperactive in an outpatient psychiatric clinic were found to be more impulsive than normals. When the drug methylphenidate was given, their impulsivity decreased (Campbell, Douglas, & Morgenstern, 1971). Elsewhere, Campbell (1973b) reported that the MFFT scores of diagnosed hyperactive children were more impulsive than those of normal reflectives and comparable to the scores of normal impulsives. Juliano (1974) has also reported a greater ratio of impulsives to reflectives among hyperactive as compared with normal children. However, the low MFFT latency-error correlation ( .37), along with a pattern of correlations favoring MFFT errors and not response time as a useful predictor, suggests that in Juliano's study reflection-impulsivity was not the operative discriminating variable. Children with diagnoses of brain damage, epilepsy, and mental retardation have also been compared with normals on the MFFT. Children in residential treatment who were diagnosed as brain damaged had shorter latencies and more errors than did normal elementary school children of comparable mental age (Ollendick & Finch, 1973). Similarly, 5- to 11-year-old children, with either generalized or focal epilepsy, had shorter MFFT latencies and more errors than did matched

normal controls (Lowry & Campbell, 1973). Finally, a study by Gozali (1969) revealed that a group of 9-year-old educable retarded children had extremely impulsive scores on the MFFT and characteristically adopted a position set in responding. Employing a symptom checklist, subsamples of 11- to 13-year-old male clinical cases were classified either as externalizers, whose symptoms were action oriented (e.g., fire setting), or as internalizers, whose symptoms were thought or somatically oriented (e.g., worrying, apathy, etc. [Weintraub, 1973]). Externalizers had much shorter MFFT latencies and many more errors than did internalizers or normal controls. Internalizers were comparable to normals in response latency but made more errors. In line with these findings, Montgomery and Finch (1975) reported that among emotionally disturbed children, impulsives were typically rated by their teachers as externalizers and reflectives as internalizers, according to criteria comparable to those employed by Weintraub. Kenney (Note 19), who tested a small sample of adolescents for reflection-impulsivity in a psychiatric hospital, obtained results that also distinguished between two types of psychiatric patients. Of four reflectives, all were diagnosed as schizophrenic or schizoid with paranoid features, and of five impulsives, none was so diagnosed. Two of the impulsives were diagnosed as hysterics, two as acting out character disorders, and one was not diagnosed. The above results suggest a link between psychopathology and conceptual tempo. Those patients (children or adolescents) who internalize their conflicts, manifested in excessive fears or worry, somatic difficulties, thought disorder, and the like, are reflective, and those most prone to act out their conflicts by lying, stealing, rebelliousness, and so on, are impulsive (cf. Shapiro, 1965). MODIFYING CONCEPTUAL TEMPO Attempts have been made to alter childrens' conceptual tempo, typically in the interest of inducing improved cognitive performance. To this end, children have been instructed to delay their responses and have been rewarded for so doing. In addition, they have been exposed to others who modeled re-

1044

STANLEY B. MESSER

flective or impulsive behavior or who taught them more efficient scanning strategies, and they have been presented with tasks that compelled them to attend to the stimuli more carefully. Forced Delay Schwebel (1966) first asked middle- and lower-class boys to describe a picture presented to them and then to construct a sentence from three words under conditions of free and forced delay. Middle-class boys showed no difference in performance under the two delay conditions, whereas lower-class children improved their performance on both tasks under forced delay. In a subsequent partial replication, forced delay again improved performance of lower- but not middleclass boys on the sentence task (Heider, 1971), though it did not reduce MFFT errors. In related studies, the subjects were not only requested to delay their responses but were also given some minimal instructions about what to do during the delay. In one such study, first-grade children were instructed to think about and check their answers before responding after the enforced delay to avoid making a mistake (Kagan, Pearson, & Welch, 1966b). In a second condition, the similarity of trainer to trainee was emphasized, with the trainer describing himself to the trainee as reflective. Both conditions produced lengthened MFFT response time, but the errors were unaffected. Similarly, Gaines (Note 19) required subjects to withhold response on the Haptic Visual Matching Test and to explore the object until given the signal to respond. The subjects were pretested and posttested on the MFFT. Compared with a control group, there was a tendency for all subjects, especially boys, to increase in response time and decrease in errors, but these results did not achieve significance. Albert (1969) told second- and third-grade children to look carefully at all the MFFT designs during an enforced delay and to think about their answers before responding. Although response time and errors did not differ significantly from those of a control group, the change in these scores was in the direction of increased reflectivity.

Reinforcement jor Increasing Response Time or Decreasing Errors Reflective and impulsive fourth-grade boys were reinforced by means of colored lights for showing either increased or decreased latency from their previous responses (Briggs, 1968). Reinforcement for increasing latencies produced both longer latencies and fewer errors, whereas reinforcement for decreasing latencies led to shorter latencies and more errors. These effects generalized to response time on the WISC Picture Arrangement subtest but not to a motor inhibition test or a time estimation task. In a similar study, Weinberg (1969) extended the range of reinforcers for long latencies to include social approval, tangible rewards, and achievement mastery. During training with these contingencies in force, latencies increased and errors were not affected. However, there was no significant change in either latencies or errors from pretest to posttest on the MFFT. Instead of reinforcing subjects for increasing latencies, Scher (1971) reinforced them for decreasing errors. Because of large changes in the control group, no significant difference between experimental and control groups was achieved. However, trends in the data indicated that reward and withdrawal of reinforcement had affected latencies but not errors. In a third condition, MFFT items were broken down and both the standard and alternatives were presented in steps comprised of their comparable (homologous) areas. In addition, subjects were rewarded for correct matching. The most substantial increase in latency and decrease in errors occurred in this condition. In another study emphasizing responsereinforcement relations, the subjects were told that if they made few enough mistakes on the MFFT and on a sentence construction task, they would be rewarded by being allowed to play with some attractive toys at the end of testing (Heider, 1971). For lower- but not for middle-class boys, this promise of reward increased MFFT latencies only, leaving errors on both tasks largely unaffected. In a related study (Errickson, Wyne, & Routh, 1973), 14-year-old children whose mean IQ was 71 and who were in a special class, were

REFLECTION-IMPULSIVITY penalized for MFFT errors by having to give up tokens exchangeable for candy for each error they made. Compared with standard MFFT administration, this punishment condition led to increased response times and decreased errors. Modeling Yando and Kagan (1968) set out to determine whether differences in the conceptual tempo of teachers, whom they regarded as natural models in the classroom, would affect the conceptual tempo of their students in the time period of 1 academic year. Both boys and girls in classrooms of experienced reflective teachers showed sizable increases in response time over the year, though their error scores were not altered appreciably. Other studies have attempted to alter cognitive tempo with experimental models rather than under natural conditions (Debus, 1970; Denney, 1972; Meichenbaum & Goodman, 1971). These studies report changes in response time as a consequence of children observing models who respond quickly or make explicit verbal statements about their conceptual strategy (e.g., "I take my time"; "I check my answers"; "I look back and forth from one to the other"), but none of the models produced any overall significant change in errors. In an extension of his earlier modeling study, Debus (Note 20) had an adult experimenter enhance the salience of the reflective sixth-grade model by commenting on how well he was doing, linking the positive reinforcements to the model's reflective verbal statements. In addition, the experimenter repeated aloud the effective strategy the model was using. Modeling effects, apparent in increased latency as well as decreased errors, were sustained in a 3-week follow-up. In a modeling study with fourth-grade boys, Ridberg, Parke, and Hetherington (1971) reported a significant increase in MFFT response time and decrease in errors, which they attributed to the influence of a reflective model. However, a similar shift toward reflectivity in their control condition casts serious doubt on this interpretation of their data. It is more likely that the increased latency and decreased errors were due to

1045

arousal of anxiety over errors in all conditions by instructions given immediately before the posttest, which emphasized poor pretest performance. This would also explain the paradoxical increase in latency following observation of an impulsive model because there, too, instructions preceding the posttest might have influenced the subjects to respond reflectively (cf. Messer, 1970a). Contrasting with the above studies, in which the models were not specifically matched to the subjects (e.g., sixth-grade children were used as models for third-grade subjects [Debus, 1970, Note 20] or a videotape of a female adult model was employed [Denney, 1972]), Cohen and Przybycien (1974) paired each subject with an age-appropriate peer model who had been rated by the subject on a sociometric scale as his choice on popularity and as a work partner. The models were trained in reflectivity and to express their strategy (e.g., "I'll look at this one and then carefully at each of these"). There was a significant increase in response times and a decrease in errors for the impulsive observers from pretest to posttest and between the experimental and no-model control groups. The interpretation of the results as due to the specific relationship between the model and observer would be more convincing had the authors included, in addition to a no-model control group, a group consisting of models who were selected randomly. Rather than requiring subjects to delay responses, or reinforcing or modeling reflective responding only, several studies presented subjects with MFFT-like tasks that forced them to concentrate on the distinctive features of a stimulus array (Duckworth et al., 1974; Egeland, 1974; Zelniker et al., 1972; Zelniker & Oppenheimer, 1973). In Egeland's study, the subjects were instructed (a) to correct each of several alternatives that deviated from a standard on a match-to-sample task by drawing in the corrections and verbalizing the change, and (b) to describe a geometric design so that others could tell which among several it was. In the studies by Zelniker and her co-workers, the subjects were required to find transformation? of a standard figure from a set that also included

1046

STANLEY B. MESSER

figures identical to the standard. As a result, the subjects had to make themselves aware of details distinguishing among the stimuli. Errors on the MFFT and other tasks were found to decrease following such training in all four studies. Response time also increased in the Duckworth et al. and Egeland studies. It remained the same in the study by Zelniker et al. (1972), who reported that impulsives improved by changing their scanning strategy to include more comparisons among variants rather than comparisons just between the standard and the variants. Egeland also concluded that forced strategy change in combination with forced delay was sufficient for improved performance without explicit verbalization of scanning rules. Teaching Scanning Strategies Directly Based on the early studies of Siegelman (1969) and Drake (1970), several investigators have attempted to instruct impulsive children to employ the scanning strategy of reflectives (Albert, 1969; Egeland, 1974; Nelson, 1969; Gaines, Note 19; Patterson & Debus, Note 21). Egeland, whose subjects were trained on the special materials described above, was most explicit in teaching these rules and basic strategies, which included looking at the standard and all the alternatives, breaking down the alternatives into component parts, and checking the standard to determine its correct form. All six studies in which scanning strategies were taught directly were unanimous in their findings: increased MFFT latencies and decreased errors. In addition, teaching proper task strategy was more effective in increasing response time and decreasing errors than either enforced delay of response or increasing motivation (Heider, 1971). These findings must be qualified in several ways. Heider's (1971) results held for lower- but not for middle-class children (who, however, had averaged only two MFFT errors to begin with and, therefore, were constrained by a floor effect). Generalization of effects following training did not extend to performance on the WISC Picture Arrangement sub test (Nelson, 1969), the Word Discrimination subtest of the Monroe Reading

Diagnostic Test, the Visual Decoding and Visual Motor Association sub tests of the Illinois Test of Psycholinguistic Abilities (Albert, 1969), or to the Map Reading Test (Patterson & Debus, Note 21), though it did extend to the Comprehension subtest of the Gates-MacGinitie Reading Test (Egeland, 1974). Where the original training procedure had been brief (Nelson, 1969), its effects were no longer present on the long-term 6month follow-up, but where the training procedure had been more elaborate or the followup period briefer (1-4 months), the MFFT errors remained low (Albert, 1969; Egeland, 1974; Patterson & Debus, Note 21). More than one session of training may be necessary to achieve reliably durable and generalizable effects. Meichenbaum and Goodman (1971) not only taught an efficient scanning strategy to two groups but went a step further with one of them to ensure that the impulsive subjects were both attending to and incorporating the scanning instructions. The subjects in this group were instructed to perform the task while speaking the instructions aloud to themselves, much as the model they observed had just done (e.g., "I have to look carefully at this one, then this one"). Both groups increased response times, but only 'the selfinstructed group improved error scores relative to its own performance and to that of other groups. Purdue (Note 22) employed this self-instruction procedure along with fading from a more overt to a more covert mode of expression for two groups of impulsive fifth graders. One group received self-reinforcement and the other external reinforcement. The self-reinforcement group manifested greater changes in MFFT response time and errors in the direction of reflectivity on both immediate and delayed posttests. Each attempt to change conceptual tempo has enjoyed some success, especially in altering response time. Nevertheless, certain proceduresincluding forced delay of response, response reinforcement, and the modeling of partial strategieshave typically not been sufficient to help the impulsive child become reliably more reflective. (However, an attrac-

REFLECTION-IMPULSIVITY

1047

live model who is the subject's well-liked friend or adored teacher may improve modeling effectiveness.) The most consistently positive results, in terms of both response time and errors, have been obtained either by forcing a strategy change through exposure to particular training materials or by teaching the scanning strategy employed by reflectives. Such strategies were derived directly from careful studies on the attention deployment of impulsive and reflective children. Although not initially intended as such, the effects of anxiety over error on conceptual tempo suggests another potentially effective way to increase reflectivity (Messer, 1970a). If concern over intellectual performance is an important factor contributing to reflectivity, a longer term strategy would be to help those impulsive children who are not particularly concerned about failure to become more concerned. This may have special relevance to lower-class children whose impulsivity may be more strongly related to the absence of such a motive. SUMMARY AND CONCLUSIONS 1. Among school age children, reflectivity and impulsivity are moderately stable, generalizable across similar tasks, and predictive of quality of performance in problem solving on tasks with response uncertainty. Scanning studies imply that reflectives evaluate alternative hypotheses more carefully than impulsives before they offer a response. 2. Long MFFT response time has a different meaning in preschoolers and in school age children. The lower correlation between response time and errors among preschoolers on the MFFT, as well as on similar but simpler tests such as the Kansas ReflectionImpulsivity Scale for Preschoolers, suggests that response time is not as reliable an index of reflection in this group as in older children. Perhaps very young children do not realize that delay is adaptive or do not yet possess a motive for performing well. When such children delay, it is usually for reasons other than to check their answers (such as being unsure of the task requirements). Investigators of preschool children would do well to inspect the response time-error correlation

before proceeding to label subjects as reflective or impulsive. 3. The correlation of MFFT response time with IQ is small (.16); the correlation of MFFT errors with IQ is moderate (-.31). At the same time, there is considerable variability in the size of these relationships, stemming in part from the subject's age and the nature of the intelligence test employed. In children 5 years old and younger and on intelligence tests with format and content similar to the MFFT, the relation of response time to IQ is higher. In children over 5 and on verbally oriented IQ tests, the relation is lower. Hence, some correlations observed between reflection-impulsivity and other variables may be accounted for, at least in part, by IQ. In future studies a check should be made of the reflection-impulsivity - IQ relation and it should be controlled where necessary. 4. Another factor that may underlie reflection-impulsivity is anxiety over error on an intellectual task. The reflective child seems to be more concerned than the impulsive child about the quality of his cognitive product. Research into other personality and social variables indicates that the reflective child, as compared with the impulsive, is better able to sustain attention, is less aggressive, makes more advanced moral judgments, and is less pessimistic about overcoming obstacles. 5. Impulsivity is related to certain educational deficits and clinical syndromes. Impulsive children succeed less well in school than do reflectives and appear to have deficient reading skills. They are found in higher proportion than are reflectives among children diagnosed as hyperactive, brain damaged, epileptic, and mentally retarded. In addition, impulsives tend to externalize emotional conflicts through action-oriented, deviant behaviors, whereas reflectives internalize their conflicts via thought processes and psychophysiological disorders. 6. Impulsivity is modifiable, at least to a degree. The most potent way to make impulsives more reflective seems to be to teach them improved scanning strategies while having them verbalize what they are doing'and to use appropriate training materials that require such scanning. Such teaching is en-

1048

STANLEY B. MESSER port Series; Technical Report 11.) Washington, D.C.: Project Head Start, Office of Child Development, U.S. Department of Health, Education, and Welfare, 1972. 10. Nadelman, L., & Wallace, E. The relationship of self-concept, conceptual tempo, and intelligence to reading achievement. Unpublished manuscript, 1973. (Available from L. Nadelman, Department of Psychology, University of Michigan, Ann Arbor, Michigan 48104.) 11. Wagner, I., & Cimiotti, E. Eye movements of cognitively impulsive and reflective children in a visual problem solving task. Paper presented at the Conference on Psychology of Human Learning and Problem Solving, Prague, 1973. 12. McKinney, J. D., & Banerjee, C. Concept attainment by reflective and impulsive children as a function of memory support. Unpublished manuscript, 1974. (Available from J. McKinney, Frank Porter Graham Child Development Center, Highway 54 Bypass, Chapel Hill, North Carolina 27514.) 13. Meichenbaum, D. H. The nature and modification of impulsive children: Training impulsive children to talk to themselves. (Research Report No. 23). Unpublished manuscript, 1971. (Available from D. H. Meichenbaum, Department of Psychology, University of Waterloo, Waterloo, Ontario, Canada.) 14. Freebody, P. R. Coping styles in conceptual tempo and problem solving. Unpublished manuscript, 1973. (Available from Ray L. Debus, Department of Education, University of Sydney, Sydney, New South Wales 2006, Australia.) 15. Ferguson, G. The relationship of anxiety to reflection and impulsivity. Unpublished manuscript, 1973. (Available from R. L. Debus, Department of Education, University of Sydney, Sydney, New South Wales, 2006, Australia.) 16. Welch, L. R. A naturalistic study of the free play behavior of reflective and impulsive four year olds. Paper presented at the meeting of the Society for Research in Child Development, Philadelphia, April 1973. 17. Messer, S. B. Internal-external control, reflection-impulsivity and academic performance. Unpublished manuscript, 1971. (Available from S. B. Messer, Psychology Building, Busch Campus, Rutgers University, New Brunswick, New Jersey 08903.) 18. Kenney, H. J. Improving the education of emotionally disturbed adolescents through educational programming in terms of curriculum and methods (Final report of Project No. 6-2315, Grant No. OEGI-6-062315-1S86). Washington, D.C.: Office of Education, Bureau of Research, U.S. Department of Health, Education, and Welfare, June 1967. 19. Gaines, P. D. The modification of attentional strategies in children (Report No. 1, Developmental Program, Department of Psychology).

hanced if done by an adult or peer whom the learner likes. Parents or teachers who help their children develop a concern about performing accurately are thereby also fostering reflectivity. For clinically hyperactive children, medication appears useful in decreasing impulsivity. 7. Because a child's tendency to be reflective or impulsive extends to many tasks and influences the quality of his performance on these tasks, it is suggested that studies of problem-solving in children control for this variable, especially when the test situations contain response uncertainty.
REFERENCE NOTES 1. Ault, R., Mitchell, C., & Hartmann, D. P. Some methodological problems in reflection-impulsivity research. Paper presented at the meeting of the Society for Research in Child Development, Denver, April 1975. 2. Wright, J. C. Reflection-impulsivity and associated observing behaviors in preschool children. Paper presented at the meeting of the Society for Research in Child Development, Minneapolis, April 1971. 3. Wright, J. C. The KRISP: A technical report. Unpublished manuscript, 1973. (Available from J. C. Wright, Kansas Center for Research in Early Childhood Education, University of Kansas, Lawrence, Kansas.) 4. McCluskey, K. A., & Wright, D. C. Age and reflection-impulsivity as determinants of selective and relevant observing behavior. Paper presented at the meeting of the Society for Research in Child Development, Philadelphia, March 1973, 5. Ward, W. C. Disadvantaged children and their first school experiences: Development oj seljregulatory behaviors (PR-73-18). Princeton, N.J.: Educational Testing Service, 1973. 6. Hess, R. D., Shipman, V. C., Brophy, J. E., & Bear, R. M. The cognitive environments of urban preschool children; Follow-up phase. Chicago: University of Chicago Press, 1969. (Available from R. D. Hess, Graduate School of Education, Stanford University, Stanford, California.) 7. Shipman, V. C. Disadvantaged children and their first school experiences: Structure and development of cognitive competencies and styles prior to school entry (PR-71-19). Princeton, N.J.: Educational Testing Service, 1971. 8. Klein, R. E. Malnutrition and mental development. Unpublished report, 1974. (Available from R. E. Klein, Institute of Nutrition of Central America and Panama, Guatemala City, Guatemala.) 9. Shipman, V. C. (Ed.). Disadvantaged children and their first school experience. (Technical Re-

REFLECTION-IMPULSIVITY Unpublished manuscript, March 1971. (Available from L. Nadelman, Department of Psychology, University of Michigan, Ann Arbor, Michigan 48104.) 20. Debus, R. L. Modification of an impulsive tempo through modeling and training procedures. Unpublished manuscript, 1974. (Available from R. L. Debus, Department of Education, University of Sydney, Sydney, New South Wales, 2006, Australia.) 21. Patterson, K., & Debus, R. L. Strategy training in modification of conceptual tempo in impulsive children. Unpublished manuscript, 1974. (Available from R. L. Debus, Department of Education, University of Sydney, Sydney, New South Wales, 2006, Australia.) 22. Purdue, W. A. Self-instruction in the modification of impulsivity. Paper presented at the Annual Conference of the Australian Association for Research in Education, Sydney, Australia, November 1973. REFERENCES Achenbach, T. M. Cue learning, associative responding, and school performance in children. Developmental Psychology, 1969,1, 717-72S. Adams, W. V. Strategy differences between reflective and impulsive children. Child Development, 1972, 43, 1076-1080. Albert, J. Modification of the impulsive conceptual style. Unpublished doctoral dissertation, University of Illinois, 1969. Ault, R. L. Problem-solving strategies of reflective, impulsive, fast-accurate, and slow-inaccurate children. Child Development, 1973, 44, 259-266. Ault, R, L., Crawford, D. E., & Jeffrey, W. E. Visual scanning strategies of reflective, impulsive, fast-accurate, and slow-inaccurate children on the Matching Familiar Figures test. Child Development, 1972, 43,1412-1417. Banta, T. J. Tests for the evaluation of early childhood education: The Cincinnati Autonomy Test Battery (CATB). In J. Hellmuth (Ed.), Cognitive studies (Vol. 1). New York: Brunncr/Mazel, 1970. Berzonsky, M. Reflectivity, internality, and animistic thinking. Child Development, 1974, 45, 785789. Block, J., Block, J. H., & Harrington, D. M. Some misgivings about the Matching Familiar Figures test as a measure of reflection-impulsivity. Developmental Psychology, 1974,10, 611-632. Block, J., Block, J. H., & Harrington, D. M. Comment on the Kagan-Messer reply. Developmental Psychology, 197S, 11, 249-252. Boismier, J. D. Comments on "Sex differences in cognitive style." Perceptual and Motor Skills, 1971, 33, 766. Briggs, C. H. An experimental study of reflectionimpulsivity in children (Doctoral dissertation, University of Minnesota, 1966). Dissertation Abstracts, 1968, 28, 3891-B. (University Microfilms No. 681610)

1049

Brodzinsky, D. M. The role of conceptual tempo and stimulus characteristics in children's humor development. Developmental Psychology, 1975, 11, 843850. Campbell, S. B. Mother-child interaction in reflective, impulsive, and hyperactive children. Developmental Psychology, 1973, 8, 341-349. (a) Campbell, S. B. Cognitive styles in reflective, impulsive and hyperactive boys and their mothers. Perceptual and Motor Skills, 1973, 36, 747-752. (b) Campbell, S. B., & Douglas, V. I. Cognitive styles and responses to the threat of frustration. Canadian Journal of Behavioural Science, 1972, 4, 30-42. Campbell, S. B., Douglas, V. I., & Morgenstern, G. Cognitive styles in hyperactive children and the effect of methylphenidate. Journal of Child Psychology and Psychiatry, 1971,12, 55-67. Cohen, S., & Przybycien, C. A. Some effects of sociometricaily selected peer models on the cognitive styles of impulsive children. Journal of Genetic Psychology, 1974,124, 213-220. Debus, R. L. Effects of brief observation of model behavior on conceptual tempo of impulsive children. Developmental Psychology, 1970, 2, 22-32. Denney, D. R. The assessment of differences in conceptual style. Child Study Journal, 1971, 1, 142155. Denney, D. R. Modeling effects upon conceptual style and cognitive tempo. Child Development, 1972, 43, 105-119. Denney, D. R. Reflection and impulsivity as determinants of conceptual strategy. Child Development, 1973, 44, 614-623. Denney, D. R. Relationship of three cognitive style dimensions to elementary reading abilities. Journal of Educational Psychology, 1974, 66, 702-709. Drake, D. M. Perceptual correlates of impulsive and reflective behavior. Developmental Psychology, 1970, 2, 202-214. Duckworth, S., Ragland, G. G., Sommerfeld, R. E., & Wyne, M. D. Modification of conceptual impulsivity in retarded children. American Journal of Mental Deficiency, 1974, 79, 59-63. Egeland, B. Training impulsive children in the use of more efficient scanning techniques. Child Development, 1974,45, 165-171. Errickson, E. A., Wyne, M. D., & Routh, D. K. A response-cost procedure for reduction of impulsive behavior of academically handicapped children. Journal oj Abnormal Child Psychology, 1973, 1, 350-357. Eska, B., & Black, K. N. Conceptual tempo in young grade-school children. Child Development, 1971, 42, 505-516. Fancher, P. E. M. Problem solving strategies of children as a function of conceptual tempo. Unpublished doctoral dissertation, Vanderbilt University, 1969. Finch, A. J., Jr., & Montgomery, L. E. Reflectionimpulsivity and information seeking in emotionally disturbed children. Journal of Abnormal Child Psychology, 1973, 1, 358-362.

1050

STANLEY B. MESSER Kagan, J., Pearson, L., & Welch, L. Modifiability of an impulsive tempo. Journal of Educational Psychology, 1966, 57, 359-36S. (b) Kagan, J., Rosman, B. L., Day, D., Albert, J., & Phillips W. Information processing in the child: Significance of analytic and reflective attitudes. Psychological Monographs, 1964, 78, (1, Whole No. 578). Katz, J. M. Reflection-impulsivity and color-form sorting. Child Development, 1971, 42, 745-754. Keogh, B. K., & Donlon, G. Field dependence, impulsivity, and learning disabilities. Journal oj Learning Disabilities, 1972, 5, 331-336. Kcrlinger, F. N., & Pedhauzer, E. J. Multiple regression in behavioral research. New York: Holt, Rinehart & Winston, 1973. Kopfstein, D. Risk taking and cognitive style. Child Development, 1973, 44, 190-192. Lewis, M., Rausch, M., Goldberg, S., & Dodd, C. Error, response time and IQ: Sex differences in cognitive style of preschool children. Perceptual and Motor Skills, 1968, 26, 563-568. Lewis, M. Sex differences in cognitive style: A rejoinder. Perceptual and Motor Skills, 1971, 33, 1006. Lowry, C. B., & Campbell, S. B. Some psychological correlates of generalized and focal epilepsy in children. Journal oj Abnormal Child Psychology, 1973, 1, 267-279. Maccoby, E. E., Dowley, E. M., Hagen, J. W., & Degerman, R. Activity level and intellectual functioning in normal preschool children. Child Development, 1965, 36, 761-770. Mann, L. Differences between reflective and impulsive children in tempo and quality of decision making. Child Development, 1973, 44, 274-279. Massari, D. J. The relation of reflection-impulsivity to field dependence-independence and internalexternal control in children. Journal of Genetic Psychology, 1975, 126, 61-67. Massari, D. J., & Massari, J. A. Sex differences in the relationship of cognitive style and intellectual functioning in disadvantaged preschool children. Journal of Genetic Psychology, 1973, 122, 175-181. Massari, D. J., & Schack, M. L. Discrimination learning by reflective and impulsive children as a function of reinforcement schedule. Developmental Psychology, 1972, 6, 183. McKlnney, J. D. Problem solving strategies in impulsive and reflective second graders. Developmental Psychology, 1973, 8, 145. Meichenbaum, D. H., & Goodman, J. Reflectionimpulsivity and verbal control of motor behavior. Child Development, 1969, 40, 785-797. Meichenbaum, D. H., & Goodman, J. Training impulsive children to talk to themselves: A means of developing self-control. Journal of Abnormal Psychology, 1971, 77, 115-126. Meinster, M. O. Individual differences in communication processes in children. Unpublished master's thesis, Rutgers University, 1972. Messer, S. B. The effect of anxiety over intellectual

Finch, A. J,, Jr., Nelson, W. M. Ill, Montgomery, L. E., & Stein, A. B. Reflection-impulsivity and locus of control in emotionally disturbed children. Journal of Genetic Psychology, 1974, 125, 273-275. Finch, A. J., Jr., Pezzuti, K. A., Montgomery, L. E., & Kemp, S. R. Reflection-impulsivity and academic attainment in emotionally disturbed children. Journal of Abnormal Child Psychology, 1974, 2, 71-74. Gozali, J. Impulsivity-reflcctivity as problem-solving styles among educable mentally retarded children. American Journal of Mental Deficiency, 1969, 73, 864-867. Hall, V., & Russell, W. Multitrait-multimethod analysis of conceptual tempo. Journal oj Educational Psychology, 1974, 66, 932-939. Harrison, A., & Nadelman, L. Conceptual tempo and inhibition of movement in black preschool children. Child Development, 1972, 43, 657-668. Heider, E. R. Information processing and the modification of an "impulsive conceptual tempo." Child Development, 1971, 42, 1276-1281. Hemry, F. Effect of reinforcement conditions on a discrimination learning task for impulsive versus reflective children. Child Development, 1973, 44, 657-660. Jeffers, V. W. Card playing as a means of studying cognition. Unpublished master's thesis, Rutgers University, 1974. Juliano, D. B. Conceptual tempo, activity, and concept learning in hyperactive and normal children. Journal of Abnormal Psychology, 1974, 83, 629634. Kagan, J. Individual differences in the resolution of response uncertainty. Journal oj Personality and Social Psychology, 1965, 2, 154-160. (a) Kagan, J. Reflection-impulsivity and reading ability in primary grade children. Child Development, 1965, 36, 609-628. (b) Kagan, J. Impulsive and reflective children: Significance of conceptual tempo. In J. D. Krumboltz (Ed.), Learning and the educational process. Chicago: Rand McNally, 1965. (c) Kagan, J. Reflection-impulsivity: The generality and dynamics of conceptual tempo. Journal of Abnormal Psychology, 1966, 71, 17-24. (a) Kagan, J. Body build and conceptual impulsivity in children. Journal of Personality, 1966, 34, 118128. (b) Kagan, J., & Messer, S. B. A reply to "Some misgivings about the Matching Familiar Figures test as a measure of reflection-impulsivity." Developmental Psychology, 1975, 11, 244-248. Kagan, J., Moss, H. A., & Sigel, I. E. Psychological significance of style of conceptualization. In J. C. Wright & J. Kagan (Eds.), Basic cognitive processes in children. Monographs of the Society for Research in Child Development, 1963, 28, (2, Serial No. 86). Kagan, J., Pearson, L., & Welch, L. Conceptual impulsivity and inductive reasoning. Child Development, 1966, 37, 583-594. (a)

REFLECTION-IMPULSIVITY performance on reflection-impulsivity in children. Child Development, 1970,41, 723-735. (a) Messer, S. B. Reflection-impulsivity: Stability and school failure. Journal of Educational Psychology, 1970, 61, 487-490. (b) Messer, S. B. The relation of internal-external control to academic performance. Child Development, 1972, 43, 1456-1462. Mischel, W. Preference for delayed reinforcement: An experimental study of a cultural observation. Journal of Abnormal and Social Psychology, 1958, 66, 57-61. Mollick, L. R., & Messer, S. B. The relation of reflection-impulsivity to intelligence. Journal oj Genetic Psychology, in press. Montgomery, L. E., & Finch, A. J., Jr. Reflectionimpulsivity and locus of conflict in emotionally disturbed children. Journal of Genetic Psychology, 1975,126, 89-92. Mosher, F. A., & Hornsby, J. R. On asking questions. In J. S. Bruner, R. R. Olver, & P. M. Greenfield (Eds.), Studies in cognitive growth. New York: Wiley, 1966. Mumbauer, C. C., & Miller, J. O. Socioeconomic background and cognitive functioning in preschool children. Child Development, 1972, 41, 471-480. Nadeau, G. Cognitive style in preschool children: A factor analytic study. Unpublished doctoral dissertation, University of Minnesota, 1968. Neimark, E. D. Longitudinal development of formal operations thought. Genetic Psychology Monographs, 1975, 91, 171-225. Nelson, T. F. The effects of training in attention deployment on observing behavior in reflective and impulsive children (Doctoral dissertation, University of Minnesota, 1969). Dissertation Abstracts, 1969, 29, 2659 B. (University Microfilms No. 68-17, 703) Nuessle, W. Reflectivity as an influence on focusing behavior of children. Journal of Experimental Child Psychology, 1972,14, 265-276. Odom, R. D., Mclntyre, C. W., & Neale, G. The influence of cognitive style on perceptual learning. Child Development, 1971, 42, 883-891. Ollendick, T. H., & Finch, A. J. Jr. Reflection-impulsivity in brain damaged and normal children. Perceptual and Motor Skills, 1973, 36, 654. Ostfeld, B. M., & Neimark, E. D. Effect of response time restriction upon cognitive style scores. Proceedings of the 75th Annual Convention of the American Psychological Association, 1967, 2, 169170. (Summary) Plomin, R., & Buss, A. Reflection-impulsivity and intelligence. Psychological Reports, 1973, 33, 726. Rcali, N., & Hall, V. Effect of success and failure on the reflective and impulsive child. Developmental Psychology, 1970, 3, 392-402. Ridberg, E. H., Parke, R. D., & Hetheringlon, E. M. Modification of impulsive and reflective cognitive styles through observation of film mediated models. Developmental Psychology, 1971, 5, 369-377. Scher, S. The effects of fading, reinforcement, or withdrawal of reinforcement on impulsive respond-

1051

ing. Unpublished master's thesis, C. W. Post College of Long Island University, 1971. Schleifer, M., & Douglas, V. I. Moral judgments, behaviour and cognitive style in young children. Canadian Journal of Behavioural Science, 1973, 5, 133-144. Schwartz, G. E., & Tursky, B. Some autonomic correlates of conceptual impulsivity. Psychophysiology, 1969, 5, 589. (Abstract) Schwebel, A. Effects of impulsivity on performance of verbal tasks in middle and lower-class children. American Journal of Orthopsychiatry, 1966, 36, 12-21. Shapiro, D. Neurotic styles. New York: Basic Books, 1965. Shipe, D. Impulsivity and locus of control as predictors of achievement and adjustment in mildly retarded and borderline youth. American Journal of Mental Deficiency, 1971, 76, 12-22. Siegel, A. W., Kirasic, K. C., & Kilburg, R. R. Recognition memory in reflective and impulsive preschool children. Child Development, 1973, 44, 651-656. Siegelman, E. Reflective and impulsive observing behavior. Child Development, 1969, 40, 1213-1222. Thomas, S. A. W. The role of cognitive style variables in mediating the influence of aggressive television upon elementary school children. Unpublished doctoral dissertation, University of California, Los Angeles, 1971. Ward, W. C. Reflection-impulsivity in kindergarten children. Child Development, 1968, 39, 867-874. Weinberg, R. A. The effects of different types of reinforcement in training a reflective conceptual tempo (Doctoral dissertation, University of Minnesota, 1969). Dissertation Abstracts, 1969, 29, 2578-A. (University Microfilms No. 69-1560) Weiner, A. S., & Adams, W. V. The effect of failure and frustration on reflective and impulsive children. Journal oj Experimental Child Psychology, 1974, 17, 353-359. Weintraub, S. A. Self-control as a correlate of an internalizing-externalizing symptom dimension. Journal oj Abnormal Child Psychology, 1973, 1, 292-307. Witkin, H. A., Dyk, R. B., Faterson, H. F., Goodenough, D. R., & Karp, S. A. Psychological difJerentation. New York: Wiley, 1962. Wolfe, R., Egelston, R., & Powers, J. Conceptual structure and conceptual tempo. Perceptual and Motor Skills, 1972, 35, 331-337. Wyne, M. D., Coop, R. H., & Brookhouse, D. B. Information processing in young mildly retarded children. American Journal of Mental Deficiency. 1970, 75, 371-375. Yando, R. M., & Kagan, J. The effect of teacher tempo on the child. Child Development, 1968, 39, 27-34. Yando, R., & Kagan, J. The effect of task complexity on reflection-impulsivity. Cognitive Psychology, 1970,1, 192-200. Young, J. Some correlates of reflection-impulsivity in adults. Unpublished master's thesis, Rutgers University, 1973.

1052

STANLEY B. MESSER Zelniker, T., & Oppenheimer, L. Modification of information processing of impulsive children. Child Development, 1973, 44, 445-450. Zigler, E. A measure in search of a theory. Contemporary Psychology, 1963, 8, 133-135. Zucker, J., & Strieker, G. Impulsivity-reflectivity in preschool head-start and middle class children. Journal of Learning Disabilities, 1968,1, 24-30. (Received May 27, 1975)

Zelniker, T., Cochavi, D., & Yered, J. The relationship between speed of performance and conceptual style: The effect of imposed modification of response latency. Child Development, 1974, 45, 779784. Zelniker, T., Jeffrey, W. E., Ault, R., & Parsons, J. Analysis and modification of search strategies of impulsive and reflective children on the Matching Familiar Figures test. Child Development, 1972, 43, 321-33S.

You might also like