Children’s differential susceptibility to parenting: An experimental test of “for better and for worse”

Highlights

• Within-subject experiments are a powerful way to test differential susceptibility.

• Negative feedback increased negative affect and decreased positive affect.

• Positive feedback did not have robust effects on affect or behavior.

• Effects of feedback did not depend on negative emotionality.

Abstract

Differential susceptibility theory proposes that a subset of individuals exist who display enhanced susceptibility to both negative (risk-promoting) and positive (development-enhancing) environments. This experiment represents the first attempt to directly test this assumption by exposing children in the experimental group to both negative and positive feedback using puppet role-plays. It thereby serves as an empirical test as well as a methodological primer for testing differential susceptibility. Dutch children (N = 190, 45.3% girls) between the ages of 4 and 6 years participated. We examined whether negative and positive feedback would differentially affect changes in positive and negative affect, in prosocial and antisocial intentions and behavior, depending on children’s negative emotionality. Results show that on hearing negative feedback, children in the experimental group increased in negative affect and decreased in positive affect more strongly than children in the control group. On hearing positive feedback, children in the experimental group tended to increase in positive affect and decrease in prosocial behavior. However, changes in response to negative or positive feedback did not depend on children’s negative emotionality. Moreover, using reliable change scores, we found support for a subset of “vulnerable” children but not for a subset of “susceptible” children. The findings offer suggestions to guide future differential susceptibility experiments.

Introduction

Differential susceptibility theory suggests that children vary in their general susceptibility to environmental influences, with some being more strongly affected than others by both negative (risk-promoting) and positive (development-enhancing) experiences (Belsky, 1997b, Belsky, 2005, Belsky et al., 2007, Boyce and Ellis, 2005, Boyce et al., 1995, Ellis et al., 2011). Thus, the very characteristics that make children disproportionately vulnerable to negative experiences might also make them disproportionately likely to benefit from positive experiences and vice versa (“for better and for worse”). Past research has not been able to directly test this key assumption, however, because participants in differential susceptibility studies have not been exposed, experimentally, to both negative and positive environmental conditions. The resulting lacuna presents a challenge to differential susceptibility theory because if the assumption of “for better and for worse” was falsified, a central assumption of the theoretical framework would be called into question. Using an experimental research design, the current study was designed to test this assumption by exposing the same individuals to both negative and positive social contexts. The study introduced a new and powerful way to test for differential susceptibility, thereby serving as an empirical test of differential susceptibility as well as a methodological primer.

The differential susceptibility model differs from the traditional diathesis–stress model (Zuckerman, 1999). Whereas the latter emphasizes the disproportionate vulnerability to negative environments of some individuals, the former highlights the disproportionate susceptibility to both the negative effects of harsh environments and the beneficial effects of supportive environments in the same individuals. Differences in susceptibility are hypothesized to have lasting impact on children, with susceptible children experiencing sustained developmental change (for better or for worse) based on the environment they encounter (Ellis et al., 2011). It is important to recognize that our experiment was designed, instead, to test short-term reactions to minor changes in the environment, that is, differential reactivity. Whether reactivity also implies developmental susceptibility, and vice versa, is an open question (Stamps, 2016). However, given the need to manipulate children’s environment for better and for worse, a focus on reactivity instead of susceptibility was ethically preferable. The strength of experiments like these lies in providing a “test of principle.”

To investigate which children are more or less susceptible, previous studies have tested genotypic variations, physiological reactivity, and temperament traits as indicators of differences in susceptibility. Following early research on differential susceptibility (Belsky, 1997a, Belsky, 2005, Belsky et al., 1998), in this study we focused on temperament traits as markers of susceptibility. Correlational studies suggest that children higher on negative emotionality (defined as the tendency to be easily distressed; Rothbart, Ahadi, Hershey, & Fisher, 2001) are more susceptible to parenting and other environmental influences (for reviews, see Belsky and Pluess, 2009, Ellis et al., 2011, Pluess and Belsky, 2010, Slagt et al., 2016). Compared with their counterparts lower on negative emotionality, these children showed more behavior problems and lower social and academic adjustment when parenting quality was low and showed fewer behavior problems and better adjustment when parenting quality was high (e.g., Roisman et al., 2012).

Although correlational studies have provided support for differential susceptibility, experimental evidence for differential susceptibility remains limited (Ellis et al., 2011, van IJzendoorn and Bakermans-Kranenburg, 2012). Experimental tests of differential susceptibility studies have several advantages over correlational studies (Bakermans-Kranenburg and van IJzendoorn,, 2015, van IJzendoorn and Bakermans-Kranenburg, 2012). First, in correlational studies, children’s environment (E) and their score on the susceptibility marker (P) can be correlated; child characteristics can evoke parenting (evocative rPE), children with certain characteristics can seek out certain environments (active rPE), and characteristics shared by parents and their children can underlie associations between parenting and child development (passive rPE) (Rutter, 2006). Furthermore, environmental conditions can shape susceptibility factors (Boyce and Ellis, 2005, Del Giudice et al., 2011). In experimental studies, the environment is manipulated in standard ways and randomization of participants to conditions breaks down the possibility of rPE. Second, experimental studies prevent the oftentimes highly skewed distributions of environmental measures by manipulating the environment and having similar numbers of participants in each environmental condition. Third, manipulation of the environment creates standardized, clear, and targeted measures of environmental stimuli. Such measures decrease “noise” in the assessment of environmental stimuli and increase the power to detect interactions if present. In sum, experimental examination of differential susceptibility affords the most solid basis for causal inference.

To date, a handful of experimental studies—all macrotrials studying developmental susceptibility instead of reactivity—have shown that children high on negative emotionality might be more susceptible to broad parenting interventions. Two of these studies found that infants high on negative emotionality profited more from experimentally induced increases in supportive parenting than less negative infants, as evidenced by their increased attachment security (Cassidy et al., 2011, Klein Velderman et al., 2006). Others showed that children high on negative emotionality profited disproportionately from experimentally induced increases in supportive parenting, as evidenced by their decreased internalizing and externalizing problems and increased cognitive functioning (Blair, 2002, Scott and O’Connor, 2012). Finally, a quasi-experiment showed that girls higher on sensory-processing sensitivity, compared with those low on this trait, benefitted more from an intervention aimed at reducing depression (Pluess & Boniwell, 2015).

The studies discussed above testify to the progress being made in experimentally testing differential susceptibility. Nevertheless, experimental studies to date examined only positive changes in parenting and, for clear ethical reasons, did not look at the experimentally induced effects of negative changes in parenting. Two solutions have been suggested to test this (Ellis et al., 2011). First, animal models could be used to conduct experiments involving both positive and negative changes in the environment of the same subjects (see, e.g., Suomi, 1997). Second, experimentally induced changes in the microenvironment of the same individuals (i.e., minor stimuli in an individual’s immediate surroundings), both for better and for worse, could be used. These “nanotrials” examine the immediate neural or behavioral responses to a small range of positive and negative stimuli (van IJzendoorn & Bakermans-Kranenburg, 2015, p. 153) and, strictly speaking, test for differential reactivity instead of susceptibility. For example, one study used an attention bias modification procedure to train one group of children to pay attention to negative pictures while training another group to pay attention to positive pictures (Fox, Zougkou, Ridgewell, & Garner, 2011). Children with the low-expression form of the 5-HTTLPR gene developed stronger attention bias toward both negative and positive affective pictures than children with the high-expression form of the gene. In another study, children were questioned by either a supportive or non-supportive interviewer (Quas, Bauer, & Boyce, 2004). Autonomic reactivity was associated with increased memory accuracy among children questioned in a supportive manner, but it was associated with decreased accuracy among children questioned in a non-supportive manner.

Following these studies, the current experiment used a “nanotrial” approach. To provide a stringent test of “for better and for worse,” we manipulated children’s environment in two directions, simulating both positive and negative feedback from parents to their children using puppet role-play scenarios (Kamins & Dweck, 1999). Such experimentally induced changes in children’s microenvironment have proven to be effective in changing children’s emotions and behaviors at posttest (Kamins and Dweck, 1999, Zentall and Morris, 2010).

Apart from mostly focusing on exposure to positive environments, previous intervention studies used between-groups designs. That is, children in the experimental group were exposed to one type of environment only. It could be, however, that some children are more susceptible to positive parenting and others are more susceptible to negative parenting, which would oppose the idea that certain individuals are susceptible “for better and for worse.” Within-person designs exposing the same children to both negative and positive changes in their microenvironments are needed to demonstrate that those who profit most from a positive change in the environment also react most to a negative change. To achieve this, we used a novel approach to testing differential susceptibility, combining a between-groups design (i.e., experimental group and control group) with a within-person design. Children in the experimental group received two manipulations, namely positive feedback and negative feedback, whereas children in the control group received no feedback.

In addition, previous intervention studies did not always assess child outcomes ranging from negative to positive. Predicting child outcomes with a restricted range, such as ranging from the presence to absence of behavior problems, enables a test of only half of the “for better and for worse” interaction (Belsky et al., 2007, Roisman et al., 2012). It does not reveal whether highly susceptible children, for example, also show the highest levels of social competence or positive affect under supportive circumstances. Therefore, we assessed child outcomes ranging from negative (negative affect and antisocial intentions and behaviors) to positive (positive affect and prosocial intentions and behaviors).

Furthermore, in none of the aforementioned experiments was random assignment to intervention and control groups stratified according to the susceptibility marker. In our study, children’s temperament was measured during a screening phase preceding the experiment. Children with low or high scores on either negative emotionality or surgency¹ were selected for the experiment because oversampling extreme scores on the susceptibility marker (at both the high and low ends) increases power to detect interaction effects (Preacher, Rucker, MacCallum, & Nicewander, 2005). In this way, we ended up with four temperament groups: children low on both dimensions, children high on negative emotionality and low on surgency, children low on negative emotionality and high on surgency, and children high on both dimensions. The selected children were then randomly assigned to an experimental or control group, stratified according to their temperament group.

In this article, we introduce a new approach to testing differential susceptibility by combining a between-groups experiment and a within-participant design. Focusing on both negative and positive microenvironments and outcomes, the current study provides one of the first experimental tests of a true crossover interaction. If predictions of differential susceptibility theory are correct and apply to situational reactivity, children scoring higher on negative emotionality should show more pronounced changes in affect, intentions, and behavior when receiving positive feedback as well as negative feedback. This offers a pioneering test of the hypothesis that the same children who respond most strongly to negative changes in their environment also respond most strongly to positive changes.

We hypothesized that children in the experimental group would increase in positive affect and prosocial intentions and behavior after the positive manipulation. Likewise, we hypothesized that children in the experimental group would increase in negative affect and antisocial intentions and behavior after the negative manipulation. We expected these changes to be stronger in the experimental group compared with the control group. Importantly, we hypothesized changes in outcome measures on receiving feedback to be moderated by temperament group. If the predictions of differential susceptibility are correct, among children high on negative emotionality, receiving positive as well as negative feedback should produce more pronounced changes in outcomes compared with their counterparts low on negative emotionality.