Edited by: Eva G. Krumhuber, University College London, United Kingdom
Reviewed by: Anna K. Doering, University of Westminster, United Kingdom; Albert Lee, Nanyang Technological University, Singapore
This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The purpose of the present study was to assess the stability of locus of control (LOC) scores over time using data gathered from tests constructed to be consistent with Rotter’s definition of LOC. We compared LOC scores of parents (measured prior to the birth of the index child and at 6 and 18 years later) and their offspring (at ages 8 and 16) to explore how stable adult and child LOC was over time and to see how parental LOC was associated with the LOC of the child aged 8 and again at 16. Locus of control was measured using modified versions of adult (ANSIE,
Over a half century ago,
By Rotter’s definition, LOC is a
We emphasize Rotter’s unique definition of LOC because, since he introduced it, researchers have employed various terms interchangeably with Rotter’s, or used “locus of control” to refer to constructs other than the one offered by Rotter. In fact, 30 years after Rotter’s article, not only did
Although two decades have passed since Skinner’s comprehensive review,
To the two conceptual fallacies offered by Kelly, we add a third, “the jumble fallacy” in which researchers use constructs like self-efficacy (
Compounding the confusion created by the existence of so many LOC terms are the difficulties originating in the 100s of different tests used to measure them. Unfortunately, there is not a review of LOC tests comparable to the one Skinner completed for “terms.” Part of the measurement problem is that researchers often do not present construct validity evidence for their LOC test as Rotter did with his (see
Unfortunately, other LOC researchers have not always been as thorough in presenting evidence of their tests’ construct validity. In one case, a single test item was used (
In summary, past researchers have not always made it clear how their control conceptualizations and the tests used to assess them relate to Rotter’s initial construct or with one another. This makes it difficult to assess how results using one test generalize to those found with others.
Pertinent to this point,
Because of the confusion caused by different conceptualizations of LOC and the paucity of construct validity data of tests employed to assess them, in the present study we return to Rotter’s original definition of LOC and use data collected from tests developed consistent with his definition. Much previous research gathered about the stability of individuals’ LOC over time was gathered from cross-sectional studies using small numbers of non-representative participants. There are exceptions. The British Child Health and Education Study (CHES) was a large cohort study that included participants from the United Kingdom born during a 1 week (
As mentioned above,
Schneewind had parents and children take LOC tests as part of a longitudinal study of family relationships. The initial test took place when children were age 10 and parents’ mean age was 36. Sixteen years later Schneewind was able to retest 100 parents and their children. He found that over the 16 years, parent LOC correlated in the 0.50s while children’s scores were in the 0.20s. In addition, parent/child LOC associations, though significant, were small.
In the present study we use tests constructed to be consistent with Rotter’s concept to examine the stability of adult LOC measured before the child was born (mean maternal age 26; paternal age 29) and measured again when the child was age 6 (adult mean parental ages 32 and 35) and 18–20 (parental mean ages 44 and 49). In addition, we assess the stability of children’s locus of score between the ages of 8 and 18.
No previous study has included both mothers’ and fathers’ LOC obtained prenatally and 6 and 18–20 years later as well as measures of their children’s LOC during childhood and adolescence. This structure not only allows us to examine the stability across time of different age groups, but also to assess the associations between children and their parents during childhood and adolescence. Previous information regarding the stability or change of LOC as measured by Rotter defined LOC tests during a life time has primarily come from cross-sectional studies.
We are tracking the generalized LOC of reinforcement across significant developmental periods for children and adults. When developing the children’s Nowicki and Strickland LOC scale the authors stated outcomes needed for the scale to obtain preliminary construct validity: (1) scores will become more internal with increasing age, (2) scores will be related to achievement with internals achieving more than externals, (3) scores will not be related to measures of social desirability, (4) scores will be related to scores from other tests of LOC. Support was found for the predicted outcomes (see section “Materials and Methods” for more specific construct validity information).
There are significant psychological and physiological changes that take place between age 8 and age 16 that may, in turn, produce changes in LOC. In reality, because of their increasing physical, cognitive and psychological maturation, children gain more control over outcomes with time. Cross-sectional data suggest children become more internal with age, but do not tell us if that is what takes place within individuals as they move from childhood into adolescence.
The Adult Nowicki and Strickland Internal-External control scale (ANSIE,
Typical adults continue to change toward internality until they reach older age (Nowicki, 2018a, unpublished
There is surprising little research upon which to base predictions for the LOC associations within a family (see
The Avon Longitudinal Study of Parents and Children (ALSPAC) was designed to determine the environmental and genetic factors that are associated with the health and development of the study offspring (
The mothers and offspring have been followed throughout, but partners were only included initially with the permission of the mothers. Mothers were given a questionnaire which they could pass to their partner if they wished him to participate; partners were given their own reply-paid envelope in which to return their completed questionnaires to avoid potential bias and protect confidentiality.
Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee [ALEC; IRB00003312] (registered on the Office of Human Research Protections database as U Bristol IRB #1), and the Local Research Ethics Committees. The Committees agreed that consent was implied if questionnaires were returned, and informed written consent was obtained from all participants who underwent certain invasive procedures undertaken during the hands-on assessments (which were optional to attend), and for all biological samples prior to analysis (see
Parents completed an abbreviated version of the ANSIE (
Basically, the philosophy of construct validation implies that a new measure of a construct should show a significant relationship with well-established measures of that construct. An example of such a procedure is the correlating of a new measure of intelligence with the Stanford–Binet or with the Wechsler scales. If, however, the authors of a new measure assume the new measure adds something unique or measures the construct more accurately than the established measure, then the resulting relationship with the established measure should be somewhat less than identity. This is important to our present purpose. Since Rotter and others who have used his scale have amassed a large amount of data consistent with theoretical predictions from social learning theory, favorable comparison with this scale is indicated. It is predicted, therefore, that if the ANSIE scale is measuring the same construct as the Rotter scale the two should be significantly related. However, if the ANSIE is accounting for a unique portion of variance, then correlations between the measures should be positive, but should fall in the moderate range.
To assess the association between the ANSIE and the Rotter scales,
Nowicki (2018b, unpublished
The ANSIE appears to meet the minimal requirements necessary for its use as an appropriate measure of LOC in adults. Further work is reported in
The ANSIE scale used in the present study was developed specifically for ALSPAC; it comprised 12 questions taken from the original 40 question scale. The ANSIE was chosen over health-related scales because it was a generalized scale consistent with Rotter’s definition and has the potential to relate to a wider range of outcomes. This shortened ANSIE was validated on a sample of 135 pregnant women prior to use in ALSPAC. It was administered within self-completion questionnaires posted to the mothers during pregnancy, and subsequently 6 and 18 years later. In parallel, during pregnancy and 6 years post-delivery, the mothers were sent questionnaires for their partners to complete with identical LOC questions. When the study offspring were 20 years old, fathers were invited to a clinic, where they responded to a computerized questionnaire which included the identical set of LOC questions. Scores were computed by adding the number of external type answers; they ranged from 0 to 12 with higher scores indicating greater externality.
The children’s LOC measure used in the present study was an adaptation of the CNSIE (
The 40-item scale was administered to children from the third through the 12th grade to obtain reliability estimates, demographic measures and construct validity information. The sample consisted of 1017 elementary and high school students most of whom were Caucasian. All schools were in a county bordering a large metropolitan school system.
Nowicki (2018b, unpublished2) reports results of factor analyses of children that suggest the scale has a general “helplessness” factor of about 0.30.
A prime goal of those who construct LOC scales is to keep social desirability at a minimum.
In terms of convergent validity support for the CNSIE,
In addition, results of studies support the theoretical assumption that internality is associated with higher and externality with lower academic achievement as well as to those behaviors associated with academic achievement, such as persistence. For example,
The CNSIE form used in the present study originated from an administration of the 40-item test to a sample of 120 8 year-old-children and the 12 items with the best item-total correlation were chosen for inclusion in the final form administered to ALSPAC children when they were tested at 8 years of age. The questions were read aloud to the child by the examiner to eliminate variance due to reading ability. The child was asked to respond with a yes/no answer. The tester made clear that there were no right or wrong answers and the items reflected how people thought and felt about different things. A similar scale was sent to the study children in a self-completion questionnaire at age 16. Scores were computed adding the number of external type answers; they ranged from 0 to 12 with higher scores indicating greater externality.
Data were used as continuous when calculating correlation coefficients, and as binary when comparing external with internal orientation. For these analyses, an external LOC (ELOC) was defined as a score greater than the median and an internal LOC (ILOC) as equal to, or less than, the median. We chose to dichotomize both the parent and child data into ELOC and ILOC to facilitate easier interpretation of results. Because of the likelihood of collinearity, stepwise logistic regression was used with p to enter of 0.10, rather than multiple linear regression which assumes a linear association. Analyses were repeated for boys and girls separately. Pseudo-R2 was used as a measure of Goodness-of-fit (GOF).
The basic data for LOC distributions at each time point are shown in Table
The means, standard deviations, and medians for the study parents at three-time points, and for the offspring at two time points measured from birth.
Individual | Time-point | Mean | Median | ||
---|---|---|---|---|---|
Mother | Pregnancy | 12604 | 4.37 | 2.11 | 4 |
+ 6 years | 8633 | 3.83 | 1.99 | 4 | |
+18 years | 3758 | 3.48 | 2.01 | 3 | |
Father | Pregnancy | 8738 | 3.60 | 2.30 | 3 |
+6 years | 4507 | 3.28 | 2.06 | 3 | |
+20 years | 1855 | 2.83 | 1.86 | 3 | |
Offspring | Age 8 years | 6374 | 5.99 | 2.08 | 6 |
Age 16 years | 4770 | 3.20 | 2.12 | 3 |
For 3487 mothers the LOC score was available at each of the three-time points. Correlation coefficients are presented here as they have the advantage of showing the relationships between the LOC measures between the different family members as well as over time. They were 0.55 and 0.54 for comparisons of pregnancy LOC with those 6 and 18 years later respectively. Similarly, the correlation between the measures of the mother at 6 and 18 years was strong (0.56). Although there were fewer fathers with measures at the three-time points (
The correlations between parental LOC scores at the three ages with the child’s LOC scores are shown in Table
Correlation coefficients between parents and children.
Individual | Time point | Child at 8 years | Child at 16 years | |
---|---|---|---|---|
1859 | ||||
Pregnancy | 0.166 | 0.169 | ||
6 years | 0.142 | 0.160 | ||
18 years | 0.171 | 0.203 | ||
726 | ||||
Pregnancy | 0.193 | 0.157 | ||
6 years | 0.160 | 0.189 | ||
20 years | 0.142 | 0.143 |
Examination of results for boys and girls separately indicated that correlations between maternal LOC and 16-year-old children tended to be slightly higher (range: 0.18 to 0.20) than found for the 8-year-olds, whereas the correlations between paternal LOC pregnancy score tended to be higher with child LOC at 8 (Table
Ways in which the binary estimates of the externality of the parents contribute toward the externality of the child are shown in Table
Stepwise logistic regression analyses to determine whether specific parental externalities were independently associated with externality of the 8 year old child.
Individual time point | Unadjusted |
Adjusted |
||||
---|---|---|---|---|---|---|
OR[95%CI] | OR[95%CI] | |||||
M – pregnancy | 5902 | 1.46[1.31,1.62] | ∗∗∗∗ | 2620 | 1.31[1.09,1.57 | ∗∗ |
M – 6 years | 5227 | 1.48[1.32,1.67] | ∗∗∗∗ | 2620 | 1.23[1.01,1.49] | ∗ |
F – pregnancy | 4403 | 1.64[1.45,1.85] | ∗∗∗∗ | 2620 | 1.41[1.18,1.70] | ∗∗∗ |
F – 6 years | 2958 | 1.55[1.33,1.80] | ∗∗∗∗ | 2620 | 1.30[1.08,1.55] | ∗∗ |
M – pregnancy | 2960 | 1.45[1.24,1.68] | ∗∗∗∗ | 1958 | 1.26[1.03,1.55 | ∗ |
M – 6 years | 2660 | 1.46[1.23,1.72] | ∗∗∗∗ | 1958 | 1.30[1.05,1.61] | ∗ |
F – pregnancy | 2202 | 1.59[1.34,1.90] | ∗∗∗∗ | 1958 | 1.45[1.20,1.76] | ∗∗∗ |
F–6 years | 1503 | 1.40[1.13,1.73] | ∗∗ | DNE | ||
M – pregnancy | 2942 | 1.47[1.26,1.70] | ∗∗∗∗ | DNE | ||
M – 6 years | 2567 | 1.50[1.27,1.78] | ∗∗∗∗ | 1287 | 1.29[1.00,1.66] | ∗ |
F – pregnancy | 2201 | 1.69[1.42,2.00] | ∗∗∗∗ | 1287 | 1.39[1.08,1.80] | ∗ |
F – 6 years | 1455 | 1.73[1.39,2.14] | ∗∗∗∗ | 1287 | 1.58[1.22,2.04] | ∗∗∗ |
Unadjusted odds ratios provide the difference between the straightforward results for external compared with internal individuals; adjusted odds ratios compute the difference after taking account of the other factors that may explain the relationship. For the risk of the 16-year-old being externally oriented, we first present the unadjusted odds ratios [95% confidence intervals], then the results of stepwise logistic regression offering the four parent measures (Adjustment A in Table
Stepwise logistic regression analyses to determine whether specific parental externalities were independently associated with externality of the 16 year old offspring, first analyzing just for parental ELOC (model A), and then additionally for the child’s ELOC at age 8 (model B).
Individual time point | Unadjusted |
Adjusted (A) |
Adjusted (B) | ||||
---|---|---|---|---|---|---|---|
OR[95%CI] | OR[95%CI] | OR[95%CI] | |||||
M – pregnancy | 4441 | 1.65[1.45,1.87] | ∗∗∗∗ | DNE | DNE | ||
M – 6 years | 4033 | 1.69[1.47,1.94] | ∗∗∗∗ | 1.59[1.30,1.94] | ∗∗∗∗ | 1.43[1.14,1.79] | ∗∗ |
F – pregnancy | 3473 | 1.54[1.34,1.77] | ∗∗∗∗ | 1.26[1.03,1.54] | ∗ | DNE | |
F – 6 years | 2485 | 1.75[1.48,2.08] | ∗∗∗∗ | 1.54[1.26,1.88] | ∗∗∗∗ | 1.56[1.28,1.92] | ∗∗∗∗ |
Child at 8 | 3283 | 1.79[1.54,2.06] | ∗∗∗∗ | – | 1.58[1.29,1.93] | ∗∗∗∗ | |
( |
( |
||||||
M |
1826 | 1.45[1.18,1.79] | ∗∗∗ | DNE | DNE | ||
M – 6 years | 1707 | 1.59[1.28,1.98] | ∗∗∗∗ | 1.61[1.19,2.18] | ∗∗ | 1.83[1.28,2.61] | ∗∗∗ |
F – pregnancy | 1452 | 1.46[1.17,1.83] | ∗∗∗ | DNE | DNE | ||
F – 6 years | 1083 | 1.88[1.44,2.44] | ∗∗∗∗ | 1.75[1.34,2.30] | ∗∗∗∗ | 1.75[1.28,2.40] | ∗∗∗ |
Child at 8 | 1377 | 1.77[1.41,2.23] | ∗∗∗∗ | – | 1.56[1.14,2.14] | ∗∗ | |
( |
( |
||||||
M – pregnancy | 2615 | 1.74[1.48,2.04] | ∗∗∗∗ | 1.38[1.07,1.78] | ∗ | 1.43[1.09,1.89] | ∗ |
M – 6 years | 2326 | 1.74[1.46,2.09] | ∗∗∗∗ | 1.29[0.99,1.70] | (∗) | DNE | |
F – pregnancy | 2021 | 1.57[1.31,1.88] | ∗∗∗∗ | DNE | DNE | ||
F – 6 years | 1402 | 1.66[1.33,2.07] | ∗∗∗∗ | 1.56[1.24,1.95] | ∗∗∗ | 1.47[1.13,1.92] | ∗∗ |
Child at 8 | 1906 | 1.77[1.47,2.14] | ∗∗∗∗ | – | 1.63[1.25,2.12] | ∗∗∗ | |
( |
( |
The following summary of the results provides a foundation for the discussion to follow. Using a comparison of correlation coefficients, the present study found: (1) there was stronger evidence for continuity of parents’ LOC over time than there was for their offspring; (2) the correlations between the parents’ LOC and that of their offspring were significant but low at both 8 and 16 years of age. Further analysis categorizing external from internal individuals showed that: (3) each external parent was associated with an independent increased risk of their 8-year-old offspring being external, but the increased risk was <50%; (4) each parent who was external when the child was aged 6 was independently associated with the child’s LOC at age 16; (5) the external child at age 8 was at increased risk of being external at age 16 independent of the contribution of their parents’ LOC; (6) there were differences between the sexes in the patterns of association –
Parent LOC was associated with their children’s LOC, but only in a very limited manner at either point in time, regardless of which pairings were used, be they of same gender, mother prominent, or father prominent. Correlations ranged from near zero to the low 0.20s indicating that other factors besides parent LOC are involved in determining children’s orientation.
Although modest in size there was a pattern of differences depending on the sex of the child and the gender of the parent that, to a certain extent, supported the dominant parent prediction, but
Identifying antecedents of children’s LOC appears to be more complex than children modeling the LOC of their parents. The relatively low parent, child LOC associations suggests that researchers need to search for other possible intervening variables involved in the development of LOC.
To evaluate the modeling hypothesis, it would be important to know how often and how well parents display clear and explicit examples of internality or externality. Operationalizing behavioral attributes of LOC such as delay of gratification, responsibility, persistence, resistance to coercion or information gathering would be helpful in this effort. Up to now, most researchers have simply assessed parent LOC and assumed they raised their children consistent with their orientation and children somehow “pick up” parents’ tendency to behave in internal or external ways and in turn are motivated to model them. The present study found a significant but small association between parent and child LOC, suggesting there is much more to learn about what else parents do to facilitate the learning of internal/external control expectancies in their offspring.
After reviewing the literature concerning antecedents of LOC,
Locus of control in adults between early to middle adulthood, appears to be relatively stable from before the child was born to when the child was aged 6 and 18–20. Correlations in the 0.50s for both mothers and fathers across all measurement points suggest consistent stability that didn’t seem differentially affected by the birth of the child. Correlations in the 0.50s, though significant, still leave considerable variance unaccounted for, and indicate that many women and men changed their LOC orientation over time. It would be helpful to know what events were associated with stability and changes toward internality or externality. Such information could provide valuable insights into how control expectancies are learned, which in turn could be used to develop interventions to change LOC.
Researchers have identified some factors associated with change in adult LOC as measured by the present tests.
Children’s LOC scores are less stable than those of their parents. Children’s emotional, physical, and cognitive abilities are changing more quickly and extensively than their parents. Based largely on past cross-sectional findings, children’s LOC appeared to become more internal during childhood and into adolescence. However, there are scant data from longitudinal studies to confirm this trajectory within children over time. Although the mean LOC scores of children was more internal at age 16 than at age 8, the correlations between the two ages is only 0.23, revealing significant intra-and inter-child change between the two testing times. In contrast to cross-sectional data which suggest stable movement toward internality during childhood, our longitudinal data suggest considerable volatility and change in children’s LOC between the ages of 8 and 16. Because of its relative instability, finding factors associated with stability and change in children’s LOC are important and should be identified.
Although this study is limited by the gaps between measures of LOC, it is unique in having longitudinal data with a large population sample, utilizing trios of mothers, fathers, and their children. The major limitation is a greater loss of external than internal participants over time.
The present study is among the first to establish consistency of adult LOC scores beginning before children were born and extending for two decades, and child LOC extending from childhood into adolescence using standardized and construct valid tests tied to Rotter’s definition. Adults’ LOC scores are significantly more consistent than children’s. We also found a significant, but small, correlation between parent and child LOC and suggested that other factors, such as child temperament and time parents and children interact, may be associated with the development of control expectancies. We urge future researchers to focus on gathering observational information concerning parent, child interactions, and operationalizing parent behavior that reflects their control expectancies.
SN had the idea. JG planned the analyses. GE and SG undertook the statistical analyses. JG and SN wrote the first draft. All authors contributed equally to writing later drafts, checking, and editing.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses.
Nowicki, S. (2018a).
Nowicki, S. (2018b).