The middle childhood temperament questionnaire: Factor structure in a German sample

The middle childhood temperament questionnaire: Factor structure in a German sample

Person. indiuid. 013 Vol. 13, No. 2, pp. 205-210, Printed in Great Britain. All rights reserved 1992 Copyright 0191-8869/92 $5.00 + 0.00 0 1991 Perg...

519KB Sizes 9 Downloads 235 Views

Person. indiuid. 013 Vol. 13, No. 2, pp. 205-210, Printed in Great Britain. All rights reserved

1992 Copyright

0191-8869/92 $5.00 + 0.00 0 1991 Pergamon Press plc

THE MIDDLE CHILDHOOD TEMPERAMENT QUESTIONNAIRE: FACTOR STRUCTURE IN A GERMAN SAMPLE* TATIANA CZESCHLIK Department of Psychology, Philipps-University, GutenbergstraBe Fed. Rep. Germany

18, D-3550 Marburg,

(Received 14 February 1991)

Summary-This study analysed the factor structure and psychometric properties of the German translation of the “Middle Childhood Temperament Questionnaire MCTQ” designed to measure the nine temperament categories proposed by Thomas and Chess (Temperumem and development. New York: Brunner/Mazel, 1977). The parents of 307 German elementary-school children (fourth-graders) completed the questionnaire (99 items). Principal component analyses yielded an eight-dimensional solution, supporting only partially (with shorter scales composed only partly by the corresponding original items) six of the nine a priori scales proposed by Hegvik et al. (Developmental and Behavioral Pediatrics 3, 197-200, 1982). Three categories were not present at all, and two new factors emerged. The eight components (Mood, Approach/Withdrawal, Persistence, Activity, Distractibility, Intensity, Organisation, and Threshold) and the problems of internal consistency and unidimensionality are discussed. INTRODUCTION

The recently growing temperament research activity is characterized by conceptualization and measurement problems and, at present, these two seem to be interrelated (Bates, 1989). While different temperament measurement methods are available (observations, interviews, questionnaires, assessment of physiological and biochemical variables) questionnaires are still the most common method. In the centre of an ongoing controversy on conceptualization and measurement issues are the nine temperament categories (Activity, Rhythmicity, Approach/Withdrawal, Adaptability, Threshold of Responsiveness, Intensity of Reaction, Quality of Mood, Distractibility, Attention Span and Persistence) proposed by Thomas and Chess (1977). These nine categories were established by content analysis of the parent interview protocols in the infancy period of the first 22 subjects of Thomas and Chess’s New York Longitudinal Study (NYLS). The pediatrician Carey and his associates, among others, developed subsequently on the basis of this work a series of parental rating temperament questionnaires for children of different age groups, from infancy to adolescence: e.g. for 8-12 years the “Middle Childhood Temperament Questionnaire, MCTQ” (Hegvik, McDevitt & Carey, 1982). The aim of this study is to explore the factor structure and psychometric properties of a German translation of the “Middle Childhood Temperament Questionnaire MCTQ” (Hegvik et al., 1982). METHODS

Subjects The parents of 307 German elementary-school fourth-grade children (age in yr: X = 10.3, SD = 0.4; 133 girls, 174 boys) completed the temperament questionnaire?. Instrument The Middle Childhood Temperament Questionnaire (Hegvik et al., 1982) is a parental report questionnaire consisting of 99 six-point scaled items (answers ranging from “almost never” to *This article is based partly on a presentation held at the 5th Conference of the EAPP, Ariccia-Genzano/Rome, Italy, 1990. tThese children (154 gifted, 156 nongifted) constitute the original sample of Phase II of the research project “Life Environmental Analysis of Gifted Elementary School Children” (Project director: D. H. Rost) carried out at the Psychology Department of Philipps University, Marburg, Fed. Rep. Germany and funded by the German Ministry for Education and Science (FKZ: B 3607.00 87). 205

206

TATIANA CZESCHLIK

“almost always”) designed to assess the nine temperament categories in 8 to lZyear-old children. The category Biological Rhythmicity is replaced here by Predictability. Hegvik et al. (1982) report internal consistencies of the categories ranging from 0.71 to 0.86 and test-retest reliabilities (74 days) ranging from 0.80 to 0.93 (for scale development see Hegvik et al., 1982; for critique see Windle, 1988). Procedure

The parents of the 307 children were asked to complete during home visits several questionnaires concerning various aspects of themselves and their children. One of these instruments was the German translation (by the author of this article) of the MCTQ. The parents were asked to base their ratings on the child’s recent and current behavior (in the last 4-6 weeks). RESULTS The inspection of the frequency distributions of the 99 items showed no noteworthy from the normal distribution.

deviations

Internal consistency

The internal consistencies for the nine a priori scales are satisfactory. Table 1 compares the internal consistencies for the nine categories found by Hegvik et al. (1982) and by this study. There are no major deviations between the reported internal consistencies. This aspect of the work of Hegvik et al. can be replicated. Scale intercorrelations

The intercorrelation matrix of the scales showed overall moderate to low intercorrelations, with few exceptions: the 36 intercorrelations of the nine a priori scales range between r = 0.05 (min), r = 0.30 (median) and r = 0.70 (max). The correlation matrix is significantly different from the identity matrix. Principal component analysis

A varimax-rotated PCA was computed for the MCTQ-items. Marker variables were defined as follows (cf. Rost & Schermer, 1986, p. 132): 1. Substantial loadings: a 2 0.30. proportions of the communality: a2/h2 2 = 0.50 (Fiirntratt, 1969). 3. Substantial differences in loadings: (aI2 - a,‘)/h’ 2 0.25 (Rost & Haferkamp, 1979).

2. Substantial

Table 2. Distribution of marker variable? in the original scales (Varimax-rotated nine-dimension PCA for the Middle Childhood Temperament Questionnaire items) Marker variables Table I, Middle Childhood Temperament Questionnaire: comparison of internal consistencies of the nine temperament categories for the orieinal scale comuositions Cronbach’s OL Category Activity Adaptability Approach/Withdrawal Distractibility Intensity Mood Persistence Predictability Threshold -

Hegvik ef al. (1982)

Czeschlik (1990)

0.86 0.79 0.87 0.74 0.83 0.81 0.85 0.74 0.71

0.82 0.65 0.83 0.75 0.76 0.78 0.83 0.74 0.53

Component

Total

Scale

Cl

16

c2

8

c3

10

1 6 3 6 I I 9

c4

9

C5

2

C6 c7 CA

5 4 6

cs

4

I

6 3 I I 5 4 6 4

Original MCTQ scale (and original number of items) Mood (12) Adaptability (I 1) Intensity Activity (12) Persistence Intensity Approach (I 2) Predictability Predictability (I I) Persistence (12) Predictability Adaptability Distractibility (8) Activity Threshold (10) Intensity (1‘1)’

‘See definition in the text. Bold face numbers indicate that these marker variables are items correctly loading in the corresponding category.

MCTQ: Table

3. Middle

factor

structure

207

Childhood Temperament Questionnaire: number and distribution variables for the components of diverse solutions

of marker

PCAs for three to nine components Components

3

4

5

6

7

8

9

Cl c2 c3 c4 c5 C6 Cl C8 c9

33 17 16

24 16 12 13

23 20 12 8 9

17 13 12 10 9 8

18 10 9 8 6 I 8

14 10 10 11 7 5 4 7

16 8 10 9 2 5 4 6 4

z

66

65

72

69

66

68

64

% Explained variance

24.4

28.1

31.1

33.9

36.3

38.5

40.5

Table 4. Middle Childhood Temperament Questionnaire: consistencies and explained variance (after Varimax-rotation) eight-component solution Number of items

Comoonent 1. 2. 3. 4. 5. 6. 7. 8.

Mood Approach Persistence Activity Distractibility Intensity Organisation Threshold

Cronbach’s

14 10 10 11 7 5 4 7 68

01

Total % variance

0.83 0.83 0.76 0.83 0.71 0.73 0.78 0.63 f

internal for the

= 0.76

7.1 5.4 5.4 5.4 4.2 3.8 3.8 3.6 Z = 38.5

Nine components (because of the postulated nine a priori categories) could not replicate the structure proposed by Hegvik et al. for the 99 items of the MCTQ. Only 65% of these are marker variables, and only 44% of these load on seven of the nine components as allocated by Hegvik et al. (bold face in Table 2). Twenty-one items were marker variables scattered over diverse components. The remaining 35 items could not be allocated to a specific dimension (see Table 2). Thus, the a priori nine categories structure of the MCTQ in its present form is not supported by factor analysis. Twenty-nine latent roots are e > 1 (accounting for 67.7% of unrotated variance). Though the kink in the Scree test is not clearly discernible, it definitely lies before 11 components. Varimax-rotated principal component analyses were conducted for three to eight components. The number and distribution of the marker variables (as defined above) for these different solutions is shown in Table 3. The five-component solution presents the highest number of marker variables, followed by the six- and eight-component solutions. Inspection of the item contents for the scales of the five-, sixand eight-component solutions suggests that the eight-component solution is the one which most appropriately describes the structure of the MCTQ: the dimensions of the other solutions are to heterogeneous and cannot be interpreted. The eight components of the MCTQ Internal consistencies and explained variances (after Varimax rotation)* for the eight temperament scales are presented in Table 4. Table 5 shows the three highest loading marker variables for each of the eight components. DISCUSSION

This study examines Hegvik et al.‘s assertion that their “Middle Childhood Temperament Questionnaire MCTQ” measures the temperament categories proposed by Thomas and Chess. *A copy of the table showing

the loadings

of the 99 items in the eight-component

solution

is available

from the author.

208

TATIANA CZESCHLIK Table 5. Middle Childhood Temperament Questionnaire MCTQ: the three highest loading items for each scale of the eight-comoonent solution 1. Dimension: Mood 27. Moody when corrected for misbehavior 88. Responds intensly to disapproval (shouts, stamps foot, etc.) 93. Reacts strongly (cries or complains loudly) to a disappointment

or failure

2. Dimension: Approach /Wilhdrawal 17. Approaches new visitors in the home 55. Avoids (stays away from, doesn’t talk to) new guests or visitors in the home 2. Avoids (stays away from, doesn’t talk to) a new sitter on first meeting 3. Dimension: Persistence 89. Has difficulty doing things on time (homework, keeping appointments, 78. Remembers to do homework without being reminded 21. Stays with homework until finished 4. Dimension: Actiuily 22. Stands or sits calmly while waiting (for snacks, turn in game, parent’s 96. Bursts loudly into the room when entering 67. Fidgets when he/she has to stay still (car ride, restaurant, etc.) 5. Dimension: 19. Looks up 12. Looks up 18. Reacts to

etc.)

attention,

etc.)

Distracribility when someone walks past the doorway right away from play when telephone or doorbell rings sudden noises (doorbell, siren) while playing a game with friends

6. Dimension: Intensity 9. Shows strong reactions (yells, shouts, etc.) when pleasantly I. Runs to get where he/she wants to go 3. Easily excited by praise (laughs, claps, yells, etc.)

surprised

7. Dimension: Organisation 44. Keeps her/his room neat and orderly 35. Quits routine household chores before finished 66. Misplaces possessions (clothes, toys, school papers, etc.) 8. Dimension: Threshold 94. Notices and comments on textures (rough, soft, smooth) of materials (clothing, upholstery, etc.) 33. Notices differences in taste or consistency of food (different brand, recipe, etc.) 23. Reacts to temperature of water (in bath, swimming, etc.)

surfaces,

Principal component analyses reveal that the structure of the nine a priori scales as designed by Hegvik et al. are not reproducible in a German sample. This is neither due to the specific sample nor to the use of a translation, since the structure of another temperament questionnaire (also translated and administered to the same sample as the MCTQ), the “Teacher Temperament Questionnaire” (short form by Keogh, Cadwell & Pullis, 1982) is the same as the one proposed by the authors. A careful examination of the distribution of the marker variables and of the content of the components showed that an eight-dimensional solution best describes the structure of the German version of the MCTQ. Of the original nine categories proposed by Hegvik et al. (1982), six (Mood, Approach, Activity, Distractibility, Intensity, Threshold) could be supported only partly (with shorter scales composed only partly by the corresponding original items), three categories (Predictability, Persistence, Adaptability) were not replicated and two new factors emerged (Persistence, with items belonging to five different original MCTQ categories, and Organisation). These results are not surprising. Though the Carey-questionnaires (and their translations) designed to measure the nine temperament categories are being very widely used (Burk, 1980; Goldsmith & Rieser-Danner, 1990; Kyrios, Prior, Oberklaid & Demetriou, 1989; Matheny, 1989; Rennen-Allhoff & Reinhard, 1988) they have been also heavily criticized, mainly because of their poor psychometric properties (Hubert, Wachs, Peters-Martin & Gandour, 1982; Martin, 1988; Rowe & Plomin, 1977; Windle, 1988). The construction of these series of questionnaires was not based on factor analyses. Attempts to replicate the nine categories by means of factor analyses (mainly of the Carey questionnaires for infants and toddlers) failed (Rowe & Plomin, 1977; Windle, 1988; Matheny, Wilson & Nuss, 1984; Prior, Sanson & Oberklaid, 1989). The number of dimensions varies between three and seven. However, the studies reporting factor-analytical replication attempts all suffer from methodological problems (such as samples too small in relation to the number of items; not reporting criteria for determining the number of factors to be extracted, insufficient definitions of marker variables). Furthermore, examination of the descriptions of the temperament categories

MCTQ: factor

structure

209

(Chess & Thomas, 1984, pp. 42-43) lends additional support to collapsing the categories Mood and Adaptability. Adaptability is defined as “Responses to new or altered situations. One is not concerned with the nature of the initial responses, but with the ease with which they are modified in desired directions.” Quality ofMood is described as: “The amount of pleasant, joyful and friendly behavior, as contrasted with unpleasant, crying and unfriendly behavior.” Adaptability is recognizable by the positive or negative quality of mood responses to alterations in the environment of a child. Therefore, Adaptability is already included in Mood, and does not represent an independent characteristic. Indeed, in other studies these two categories are highly intercorrelated. McClowry (1990) reports for the MCTQ intercorrelations between Adaptability and Mood for two measurement times of 0.67 and 0.79 indicating a conceptual overlap between the categories Adaptability and Mood. However, the results of the preceding factor analysis suggest that intercorrelations for the MCTQ have to be interpreted with caution, e.g. their stability is questionable, since the nine scales are not factor-analytically reproducible. Internal

consistency

and unidimensionality

The results of this article emphasize the difference between internal consistency (e.g. the degree of interrelatedness among the items belonging to a scale, measured by coefficient alpha) and homogeneity of a factor (e.g. the set of items of a scale all measure a single dimension): two different concepts which nevertheless are very often confounded. For example, McDevitt and Carey (1978, p. 250) discussing the properties of the Behavioral Style Questionnaire BSQ, write: “Item homogeneity, when present, provides statistical evidence that the items within a category are measuring the same trait.” This confusion leads to the erroneus assumption that a scale with a high internal consistency is also factorially homogeneous. Windle (1988) discusses in more detail this fallacy. From this study one can conclude that (1) the original MCTQ is not an adequate instrument for assessing the proposed temperament categories (2) studies employing the original MCTQ and reporting differences between different (clinical, cultural, etc.) groups of children do not tell us very much, since it is not clear what really is being measured. This study indicates that eight components (Mood, Approach/Withdrawal, Persistence, Activity, Distractibility, Intensity, Organisation and Threshold) adequately describe the structure of the MCTQ in a German sample. Replications of the present study are necessary to establish the factorial invariance of the MCTQ for diverse samples. The procedure described here for the MCTQ applies also of course, to the other Carey-questionnaires, and probably also to the questionnaires developed on the basis of the Carey-items by other research teams. If temperament research is to exist as a serious area of psychological research, sound measuring methods are badly needed. REFERENCES Bates, J. E. (1989). Concepts and measures of temperament. In Kohnstamm, G. A., Bates, J. E. & Rothbart, M. K. (Eds), Temperament in childhood (pp. 3-26). Chichester: Wiley. Burk, E. (1980). Relationship of temperamental traits to achievement and adjustment in gifted children. Unpublished doctoral dissertation, Fordham Univ. Chess, S. & Thomas, A. (1984). Origins and evolution of behavior disorders. From infancy to early adult life. New York: Brunner/Mazel. Czeschlik, T. (1990). Middle Childhood Temperament Questionnaire (MCTQ): Factor structure of the German version. Presented at 5th Conference of the European Associalion for Personality Psychology, Ariccia-Genzano-Rome, Italy. Fiirntratt, W. (1969). Zur Bestimmung der Anzahl interpretierbarer gemeinsamer Faktoren in Faktorenanalysen psychologischer Daten. Diagnosfica, 15, 62-75. Goldsmith, H. H. & Rieser-Danner, L. A. (1990). Assessing early temperament. In Reynolds, C. R. & Kamphaus, R. W. (Eds), Handbook of psychological and educational assessment of children. Personality, behavior, and context (pp. 245-278). New York: Guilford Press. Hegvik, R. L., McDevitt, S. C. & Carey, W. B. (1980). Middle Childhood Temperament Questionnaire. Unpublished manuscript. Hegvik, R. L., McDevitt, S. C. & Carey, W. B. (1982). The Middle Childhood Temperament Questionnaire. Developmental and Behavioral Pediatrics, 3, 197-200. Hubert, N. C. & Wachs, T. D., Peters-Martin, P. & Gandour, M. J. (1982). The study of early temperament: Measurement and conceptual issues. Child Development, 53, 571-600. Keogh, B. K., Pullis, M. & Cadwell, J. (1982). A short form of the Teacher Temperament Questionnaire. Journal of Educational Measurement, 29, 323-329.

TATIANACZESCHLIK

210

Kyrios, M., Prior, M., Oberklaid, F. & Demetriou, A. (1989). Cross-cultural studies of temperament: Temperament in Greek infants. International Journal of Psychology, 24, S-603. Martin, R. (1988). The Temperament Assessment Battery for Children: Manual. Brandon, VT: Clinical Psychology. Matheny Jr, A. P. (1989). Temperament and cognition: Relations between temperament and mental test scores. In Kohnstamm, G. A., Bates, J. E. & Rothbart, M. K. (Eds), Temperamenf in childhood (pp. 263-282). Chichester: Wiley. Matheny, A. P., Wilson, R. S. & Nuss, S. N. (1984). Toddler temperament: Stability over ages and across settings. Child Development,

55, 1200-1211.

McClowry, S. G. (1990). The relationship of temperament to pm- and posthospitalization behavioral responses of school-age children. Nursing Research, 39, 30-35. McDevitt, S. C. & Carey, W. B. (1978). The measurement of temperament in 3- to 7-year-old children. Journal of Child Psychology

and Psychiatry,

19, 245-253.

Prior, M. R., Sanson, A. V. & Oberklaid, F. (1989). The Australian Temperament Project. In Kohnstamm, G. A., Bates, J. E. & Rothbart, M. K. (Eds) Temperament in childhood (pp. 537-554). Chichester: Wiley. Rennen-Allhoff, B. & Reinhard, H. G. (1988). Erprobung deutscher Versionen der Temperamentfragebogen von Carey und Mitarbeitem. Zeitschrtft fur Kinder und Jugendpsychiatrie, 16, 61-66. Rost, D. H. & Haferkamp, W. (1979). Zur Brauchbarkeit des AFS (Angstfragebogen fur Schiiler). Eine empirische Analyse und eine vergleichende Darstellung vorliegender Untersuchungen. Zeitschrift fur Empirische Paaagogik, 3, 183-210. Rost, D. H. & Schermer, F. J. (1986). Strategien der Priifungsangst-Verarbeitung. Zeitschrrft fur Dtjerentietle und Diagnostische

Psychologie,

7, 1277143.

Rowe, D. & Plomin, R. (1977). Temperament in early childhood. Journal of Personality Assessment, 41, 150-156. Thomas, A. & Chess, S. (1977). Temperament and development. New York: Brunner/Mazel. Windle, M. (1988). Psychometric strategies of measures of temperament. A methodological critique. International Journal of Behavioral Development,

11, 171-201.