Longitudinal Changes in Young Children’s Strategy Use for Emotion Regulation

Abstract

The development of strategies that support autonomous self-regulation of emotion is key for early childhood emotion regulation. Children are thought to transition from predominant reliance on more automatic or interpersonal strategies to reliance on more effortful, autonomous strategies as they develop cognitive skills that can be recruited for self-regulation. However, there are few longitudinal studies documenting age-related changes in different forms and dimensions of strategies. The current study tested predicted age-related changes in strategy use in a task requiring children to wait for something they want. Specifically, we examined the longitudinal trajectories of three strategies commonly observed in delayed reward tasks: self-soothing, seeking attention about the demands of waiting (bids), and distracting oneself. We followed a sample of 120 children (54% male, 93.3% white, from semi-rural and rural economically strained households) from ages 24 months to 5 years who participated in a waiting task each year. Using growth curve modeling, we found declines in self-soothing, rises and then declines in bidding, and increases in distraction from 24 months to 5 years. Next, we investigated whether strategy use trajectories predicted adult ratings of children’s emotion regulation during the task, i.e., whether children appeared calm and acted appropriately while waiting. Growth in duration and dominance of distraction use predicted judgments that children were well-regulated by age 5 years, whereas growth in dominance of bidding use negatively predicted being rated as well-regulated. We discuss implications for the understanding of strategy development and future directions, including understanding strategy effectiveness.

A key aspect of emotion regulation (ER) development is strategy use, i.e., behaviors and thoughts that have potential to change emotions, including to forestall, reduce, or resolve negative emotions (Cole et al., 2019; Gross & Thompson, 2007; Kopp, 2009). Broadly, ER is a dynamic process by which emotions are monitored, evaluated, and modified (e.g., through use of strategies) to meet situational demands and sociocultural expectations (Cole, Ram, & English, 2018; Gross & Thompson, 2007), and its development is regarded as central to academic success, socioemotional competence, and mental health (Graziano et al., 2007; Halberstadt et al., 2001; McLaughlin et al., 2011). It is thought that the development of putative ER strategies in early childhood involves transitions from a predominance of relatively automatic (e.g., gaze aversion, self-soothing behaviors like thumb sucking) or caregiver-reliant behaviors (e.g., bids for adult attention) to more effortful, autonomous behaviors (Kopp, 1982; Rothbart et al., 1992; Sroufe, 1996). Cross-sectional evidence indicates that the frequency of such strategies changes as early as infancy, with gaze aversion decreasing and verbally seeking attention or support increasing from 3 to 13 months of age (Rothbart et al., 1992). These changes in the form and use of strategies clearly depend, in part, on young children’s recruitment of cognitive and language development and socialization experiences (Kopp, 1982). This developmental framework is largely accepted; however, there is a paucity of longitudinal evidence documenting these expected age-related changes in observable strategy use. Thus, the primary purpose of the present study was to investigate developmental changes in strategy use in early childhood.

Because young children are limited in describing what they do to manage their emotions, behavioral observations are an essential tool for assessing ER (Ramsook et al., 2018). Much of the observational research on ER in early childhood has employed tasks designed to provoke frustration by way of removing or blocking access to desired objects, such as toys, snacks, and gifts (Buss & Goldsmith, 1998; Vaughn et al., 1984; Cole et al., 2011). These wait tasks provide a means of observing the capacity to delay gratification and tolerate blocked goals, skills associated with school readiness skills (Blair, 2002; Raver, 2004), and self-discipline skills in adolescence and adulthood (Eigsti et al., 2006; Shoda et al., 1990). For example, young children are often required to wait for food, recess, or for others to finish their work in school and at home. Delayed reward tasks often elicit strategies such as self-soothing, bidding to adults for support or information, and redirecting attention, i.e., distraction (Cole et al., 2011; Gilliom et al. 2002; Grolnick et al., 1996). When children engage in strategies such as distraction that involve switching attention from a desirable object and redirecting it toward another activity, they wait longer and forestall potential frustration or disruptive behavior (Cole et al., 2011; Peak et al., 2002). However, the extent to which growth in purportedly optimal strategies such as distraction relates to children’s overall ability to appear well-regulated and appropriately tolerate a wait has not yet been investigated.

Evidence documenting developmental changes in strategy use is limited in two important ways. First, most studies are cross-sectional and do not test within-person changes in strategy use that represent individual development. Second, few studies assess children’s strategies on more than one occasion and those that do are often limited to two observations or closely spaced (Ekas et al., 2013; Graziano et al., 2011; Supplee et al., 2011), limiting conclusions about how developmental change occurs. One longitudinal study, however, observed children at four age points across early childhood, showing that distraction occurred more quickly and for longer duration between 24 and 36 months (Cole et al., 2011). A limitation of that study was that it did not focus on age-related changes in other strategies that are observed in waiting tasks and did not address changes in frequency or dominance of strategies in the task.

In sum, longitudinal analysis aimed at testing age-related changes in multiple dimensions of strategy use in early childhood is warranted. We examined trajectories of self-soothing, bidding to mother, and distraction use at ages 24, 36, 48 months, and 5 years, in terms of the frequency, average duration, and dominance of strategy use. In addition, we cannot assume that changes in every dimension of strategy use have utility for predicting children’s ER (Cole et al., 2004). Thus, we assessed whether growth in frequency, duration, and dominance of strategies across early childhood related to adult perceptions of children as well-regulated in the task, i.e., remaining predominantly calm and behaviorally appropriate, at age 5 years.

Development of Strategies in Early Childhood

Developmental frameworks identify early childhood as an important period of transition, marked by a shift from infants primarily relying on others to resolve distress to preschool-aged children increasingly deploying autonomous strategies to regulate negative emotions (Kopp, 1982, 1989; Sroufe, 1996). In line with this transition, caregivers expect their toddlers and preschool-aged children to learn to comply with directives and begin to manage their negative emotions with increasing control and independence (Calkins & Hill, 2007). Further, it is thought that the emergence of children’s autonomous use of putative strategies occurs because they are developing internal cognitive resources, such as executive attention, memory, and language, that can be recruited to help them meet caregiver expectations (Bell & Calkins, 2012; Cole et al., 2010; Kopp, 1982; Posner & Rothbart, 2000).

Self-soothing behaviors, such as thumb-sucking or rocking, appear by 3 months and are thought to be relatively rudimentary and automatic, requiring little engagement of cognitive resources (Kopp, 1982; Rothbart et al., 1992). However, self-soothing is associated with contingent decreases in negative emotion in 3- through 10-month-olds (Ekas et al., 2013; Stifter & Braungart, 1995). The frequency of these self-soothing behaviors increases through 12 months, then decreases by 18 months (Mangelsdorf et al., 1995; Rothbart et al., 1992) and again between 24 and 48 months in boys (Supplee et al., 2011). However, there is relatively little evidence in the toddler to preschool years, and the implications of this normative decline in self-soothing for perceptions of children’s ER is not evident.

Actively bidding to adults for attention, support, and information occurs nonverbally as early as age 6 months and continues into toddlerhood in the form of verbal bidding (Buss & Goldsmith, 1998). Bidding to adults is thought to increase as children gain expressive language but still represents reliance on adults (Kopp, 2009). Verbal bids during a frustration task is the only strategy associated with contingent decreases in 3-year-old boys’ anger (Gilliom et al., 2002). However, developmental changes in bidding as a strategy are infrequently documented. Calm bids (but not angry bids) increase in frequency at 36 months relative to 18 or 24 months (Cole et al., 2011). Both language skill level and growth predict a decline between ages 18 and 48 months in angry reactions in a wait task (Roben et al., 2013). Therefore, bidding to adults should increase from toddlerhood to early preschool age; however, how bidding changes in the later preschool years is less clear.

Distraction, or purposefully shifting the focus of attention, emerges between age 24 and 36 months and demonstrates children’s emerging attention control and recruitment of memory and language (Kopp, 1982; Rothbart et al., 1992). Effortful distraction is associated with momentary decreases in anger intensity in toddlers (Buss & Goldsmith, 1998) and reduced probability of contingent increases in anger in 3-year-old boys (Gilliom et al., 2002). Evidence for developmental change in distraction is quite limited, even though distraction is the most prevalent observed strategy during frustration tasks with preschool age children (e.g., Gilliom et al., 2002; Silk et al., 2006). Planful strategies, including distraction, increase in frequency from 24 to 48 months in boys (Supplee et al., 2011) and are used more quickly and for longer periods from 24 to 48 months (Cole et al., 2011). Studies of early executive attention show a developmental change around age 30 months (Rothbart et al., 2011; Rueda et al., 2005); thus, it follows that distraction use should increase from toddlerhood to preschool age.

Dimensions of Strategies

Measuring multiple dimensions of age-related changes in strategy use has the potential to yield distinct insights into ER development (Cole et al., 2004; Thompson, 1994). Most studies that include assessment of strategies in early childhood focus on how frequently they are used. However, additional information can be gained from considering how long children sustain a strategy or how often they rely on one strategy compared to others. For example, two children may use the optimal strategy, distraction, with the same frequency over the course of a wait. One child may engage in longer episodes of distraction while the other distracts only briefly, which may be important dimensions of strategy use to capture. Additionally, children are likely to draw from a repertoire of strategies during an ER task, and the dominance (i.e., use of one strategy relative to total strategy use) may change with child age. For example, it may be expected that children rely more on strategies such as self-soothing and bidding to adults when they are younger and shift to more effortful strategies such as distraction as more complex cognitive abilities emerge (Cole et al., 2011; Gilliom et al., 2002). However, to our knowledge, no studies have investigated dominance of a strategy. We focus on age-related changes in the frequency, duration, and dominance of self-soothing, bidding to mother about the task demands, and distraction, and how they relate to perceptions of children’s ER.

Adult Judgments of Child ER

In early childhood, a central marker of ER is to manage emotions in behaviorally appropriate and increasingly independent ways, enabling children to meet parental and teacher expectations (Blair, 2002; Kopp, 1989; Rimm-Kauffman et al., 2000). It stands to reason that developmental change in strategy use should enable children to demonstrate competence in these domains (Thompson & Goodman, 2010). However, we know little about how developmental change in strategy use relates to adult perceptions of children’s ER. We addressed this gap by investigating associations between strategy trajectories and adult judgments of children tolerating the wait calmly and appropriately. Moreover, we examined how specific dimensions (i.e., frequency, duration, and dominance) of strategy trajectories predict well-regulated behavior.

As noted, from ages 2 through 5 years, self-regulation of emotion should emerge as a function of both cognitive growth and the internalization of social standards (Kopp 1989, 2009). Evidence supporting this framework is illustrated by temperament research, documenting that the executive attention network begins to operate around the third year, which then supports the ability to engage in effortful control (Posner & Rothbart, 2000). This may explain why distraction appears to be one of the most optimal strategies for young children’s delay of gratification and tolerance for waiting (e.g., Cole et al., 2011). Because increases in the duration of distraction co-occur with decreases in the duration of anger in this age range (Cole et al., 2011), it may be that adult observers judge children to be better regulated if they use distraction more frequently, longer, and more dominantly.

In contrast, bidding to an adult or self-soothing might lead adults to judge children as less well-regulated because they need more assistance. Notably, one study found bidding for more information as the only strategy associated with declines in 3-year-old boys’ anger (Gilliom et al., 2002). Thus, early bidding may be appropriate and expected, but increasingly relying on this strategy in the older preschool years may signal an over-dependence on adult support. We have little evidence elucidating how change in self-soothing is associated with children’s ER. This strategy normatively declines toward the end of infancy (Ekas et al., 2013), suggesting that increasingly deploying self-soothing may be negatively associated with children’s ability to regulate appropriately.

Current Study

Although strategies are often considered central to ER (Kopp, 1982), little research explicates the typical course of strategy development and its relation to ER ratings during early childhood. Such research is needed to identify whether observed strategy use reflects purported developmental shifts from caregiver-reliant, rudimentary ER to autonomous, effortful ER across early childhood. The context of tolerating frustration in a wait task in a particularly relevant one to observe children’s strategies in, due to its associations with school-readiness and self-discipline skills (Blair, 2002; Raver, 2004). Thus, we investigated 1) longitudinal change in observed strategy use from 2 to 5 years in a task requiring children to wait for a desirable gift, and 2) whether these patterns of change predicted ER at age 5, measured by adult judgments of children’s ability to meet developmental expectations to remain calm and behave appropriately during the wait.

Notably, Kopp (1982, 1989) offers a framework for stages of strategy development up to age 3 years but does not propose clear hypotheses about the course of developmental change in strategies in early childhood. Thus, we first used growth curve modeling to examine linear and non-linear patterns of age-related change in strategy use and provide age-adjacent comparisons that allowed for specific descriptions of change. We hypothesized that: a) dimensions (i.e., frequency, average duration, dominance) of self-soothing should decrease from 24 months to 5 years; b) dimensions of bidding should initially increase, then decrease later; and c) dimensions of distraction should increase from 24 months to 5 years.

Next, we used regression models to examine links between strategy trajectories and adult judgments of children’s ER at 5 years. We hypothesized that growth in dimensions of distraction from age 24 months to 5 years should be associated with judgments that children more well-regulated. There is less evidence regarding other strategies, but we posited that growth in dimensions of bidding and self-soothing would be associated with judgments that children were less well-regulated.

Method

Participants

Data were derived from an IRB-approved (Development of Toddlers Study, The Pennsylvania State University, Protocol #18993) archived longitudinal study of ER (Cole, Crnic, Nelson & Blair, 2000). Participants were recruited from semirural and rural neighborhoods in mid-Atlantic region of the United States with flyers, community engagement, and birth announcements. Recruitment additionally targeted economically strained families, i.e., household income above federal poverty threshold but below national median for number of household members, using census data.

At ages 24 months (M = 24.39, SD = 1.30 months), 36 months (M = 36.44, SD = .80 months), and 48 months (M = 48.33, SD = .67 months), 120 children (65 boys) participated in planned visits. At age 5 years (M = 68.20, SD = 2.47 months), 96 children (52 boys) participated in an additional visit. Children were identified by their parents as white (93.3%) or biracial (6.7%) and mainly from two-parent households (97.6%). Reported household income at enrollment was on average $40,502.94 (SD = 14,480.73). Families who did not participate in the final visit (n = 24) did not differ from those who had completed the final visit on most demographic characteristics, except that parents who completed the last visit (Mothers M = 31.72, SD = 5.62; Fathers M = 33.26, SD = 6.35) were older than those who did not (Mothers M = 27.90, SD = 4.66; Fathers M = 30.38, SD = 4.73), ts > 2.22, ps < .05. Nine children who attended the visit at 24 months were excluded from analyses due to abnormal task administration (child opening prize immediately, experimenter error); three children missed the visit at 36 months and three children missed the visit at 48 months due to illness or scheduling issues.

Procedure

During annual videotaped lab visits, children participated in a series of tasks, which alternated between those designed to tax ER and provide relief (e.g., free play). The present analyses focused on the 8-minute waiting task (Vaughn et al., 1984). Prior to administration, mothers were briefed and asked to tell their children that they had to wait to open a gift while their mother answered some questions. Mothers were instructed to do what they would normally do at home when their children needed to wait for their mother’s attention. Mothers and children were instructed to sit at different tables. Research assistants (RAs) gave mothers questionnaires to complete, saying, “There is the work I told you about.” With children, RAs placed a shiny wrapped bag at the child’s table, saying, “This is a surprise for you.” Then RAs handed the child a boring toy (cloth cymbals, car with missing wheels, horse with one missing leg) and said “And here is something for you to play with. I’ll be back in a few minutes.” As RAs left the room, they signaled for mothers to give the child the instruction to wait to open the gift until the mother finished working. After the procedure, mothers let their child open the gift and the RA returned.

Measures

Strategies

Strategies were coded from videotapes in one-second intervals (Grolnick et al., 1996), indicating whether each strategy was present (1) or absent (0) per second. Most codes were not mutually exclusive (except for attention re-direction: distraction versus attention focused on gift). Descriptions of codes remained the same at each wave. Each coder trained to 80% accuracy with a master coder and reliability was calculated for a randomly selected subset of cases (15%); omnibus κ = .78 at 24 months, .84 at 36 months, .86 at 48 months, and .83 at 5 years. Across all waves, the three strategies selected represented 92% of possible strategic behavior coded.

Self-soothing was coded when children engaged in physical movements such as putting their head on table or sucking on their thumb. Bids were coded when children verbally or nonverbally sought adult attention or support about the delay, e.g., “I want the prize,” or sought more information regarding the wait, e.g., “Why do I have to wait?” Distraction was coded when children became absorbed in an alternate, appropriate activity, e.g., playing with boring toy or attending to objects in the observation room. All strategies were coded as self- or mother-initiated, but only self-initiated strategies were used.

Three variables – frequency of bouts, duration of bouts, and proportion of bouts relative to other strategy bouts (our operationalization of dominance) – were created for each strategy. To calculate these three variables, we first defined strategy bouts as periods of contiguous second(s), (at least 1 second duration) of strategy use. Bout frequency was calculated as the number of separate bouts for each strategy. Average bout duration was calculated as the average number of seconds per bout across total number of bouts (Cole et al., 2011). Dominance was represented by a calculation of the proportion of bouts during which the strategy was used over the total numbers of bouts during which other strategies were used.

ER Judgments

An independent coding scheme was used to rate ER in the wait task at ages 24 months and 5 years (adapted from Zahn-Waxler et al., 1994). Trained coders separately rated each child’s primary emotion tone as calm or negative, based on facial and vocal expressions, and primary behavior as on-task (e.g., engages with given toy appropriately or resists in a socially acceptable way; does not persistently engage with surprise and accepts the wait; occupies self appropriately without dependence on mother, or accepts temporary loss of mother’s attention) or off-task/disruptive (e.g., dependence on mom to keep oneself occupied; does not accept loss of mother’s attention or wait; destructive, aggressive, or defiant). Inter-rater reliability was calculated for 15–20% of cases; κs for emotion ratings ranged from .58 to .79 and for behavior from .71 to .76. Codes were assigned for each of two 4-minute segments based on coder feedback that it was difficult to assign a single rating to an 8-minute task. Emotion and behavior codes were combined to indicate whether a child was calm on-task versus any other emotion/behavior combination. The segment codes were summed, resulting in a rating that ranged from 0 to 2.

Analytic Strategy

To examine age-related changes in dimensions of strategy use, multilevel models were fit in R version 3.6.1 (R Core Team, 2017), nlme package (Pinheiro et al., 2017). Data were considered missing at random (based on attrition analysis, lack of identified factors that systematically accounted for missingness in study variables, and visual inspection of missing data patterns) and handled with the maximum likelihood method. Models were estimated with polynomial contrasts for age. The final model was selected by comparing fit statistics for nested models. Linear and quadratic terms reflected rate of change from 24 months to 5 years. To help interpret and describe age-related changes, we used the emm package (Lenth et al., 2018) to extract estimated marginal means and consecutive age contrasts, reported with Tukey method adjustments for multiple comparisons. To test predictions about change in strategy use predicting adult ER ratings, we extracted each child’s model-predicted intercepts (score at 24 months) and linear and quadratic trajectories from multilevel models (fixed + random coefficients). These parameters were then regressed on adult ratings at 5 years, controlling for ratings at age 24 months.

Results

Means and standard deviations for study variables are presented in Table 1. Spearman’s rho correlations are presented in Table 2. Child gender and temperament (surgency and negative affectivity as rated by mothers on the TBAQ-R; Goldsmith, 1996) were largely unrelated to strategy use in this sample and therefore were not included in models. Although related to overall attrition, mother and father age were largely unrelated to strategy use and therefore were not included in models. Including parent age as covariates in preliminary multilevel models also did not change overall results.

Table 1.

Means and Standard Deviations of Study Variables

	Bout Frequency			Bout Average Duration			Bout Dominance
	M	SD	Range	M	SD	Range	M	SD	Range	N
Strategy Use
Self-Soothing, 24m	2.80	3.43	0–19	10.35	37.50	0–389	.21	.26	0–.98	111
Self-Soothing, 36m	1.68	2.51	0–12	7.24	32.65	0–350	.10	.17	0–.84	117
Self-Soothing, 48m	1.22	2.03	0–13	5.93	12.22	0–80	.07	.14	0–.66	117
Self-Soothing, 5y	.67	1.36	0–6	3.20	7.02	0–48.40	.03	.08	0–.67	96
Bidding, 24m	7.50	5.50	0–25	2.82	1.75	0–11	.17	.15	0–.65	111
Bidding, 36m	11.01	6.30	0–33	2.92	1.27	0–8.67	.28	.23	0–1	117
Bidding, 48m	12.31	6.87	0–34	3.18	1.27	0–12.80	.24	.19	0–1	116
Bidding, 5y	12.15	7.15	0–35	2.75	.97	0–6.11	.17	.12	0–.5	96
Distracting, 24m	5.66	3.18	0–15	17.11	16.01	0–161	.62	.27	0–1	111
Distracting, 36m	6.68	4.30	0–20	18.10	11.85	0–69	.62	.25	0–.99	117
Distracting, 48m	6.89	3.32	0–15	25.90	20.22	0–198	.69	.22	0–.98	117
Distracting, 5y	10.68	4.32	0–29	22.55	9.00	0–51.57	.79	.16	0–1	96
ER Ratings
ER rating, 24m	.20	.48	0–2	--	--	--	--	--	--	120
ER rating, 5y	.53	.73	0–2	--	--	--	--	--	--	96

Note. ER = emotion regulation; 24m = 24 months, 36m = 36 months, 48m = 48 months, 5y = 5 years. Time intervals calculated in seconds.

Table 2.

Spearman’s rho Correlations for Study Variables

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.
Bout Frequency
1. Self-Soothing, 24m	--
2. Self-Soothing, 36m	.039	--
3. Self-Soothing, 48m	−.180+	.080	--
4. Self-Soothing, 5y	.105	.176+	.209*	--
5. Bidding, 24m	−.075	.039	.017	−.002	--
6. Bidding, 36m	.060	−.092	−.014	.036	.075	--
7. Bidding, 48m	.032	−.083	−.126	.087	.066	.216*	--
8. Bidding, 5y	.021	.106	.024	.029	−.051	.101	−.025	--
9. Distraction, 24m	−.419*	.043	.092	−.034	.264*	−.056	−.029	−.064	--
10. Distraction, 36m	.108	.003	−.063	−.016	−.141	.083	−.125	.005	.092	--
11. Distraction, 48m	.024	−.136	−.065	.037	−.095	−.013	−.139	.037	−.092	.075	--
12. Distraction, 5y	.003	−.003	−.044	−.006	−.096	.018	.086	.194+	−.028	.109	.129	--
13. ER rating, 24m	−.117	.137	−.082	.051	−.061	−.145	−.112	.127	.332*	.052	−.041	−.132	--
14. ER rating, 5y	−.004	−.010	.024	.244*	.104	.164+	−.109	−.389*	.055	.222*	.145	.047	−.037
Bout Average Duration
1. Self-Soothing, 24m	--
2. Self-Soothing, 36m	−.007	--
3. Self-Soothing, 48m	−.082	.061	--
4. Self-Soothing, 5y	.139	.150	.156	--
5. Bidding, 24m	−.093	.163+	−.030	−.147	--
6. Bidding, 36m	−.048	−.100	.096	.139	−.089	--
7. Bidding, 48m	.071	−.041	−.133	−.020	.065	−.069	--
8. Bidding, 5y	.016	−.214*	−.085	−.138	−.010	.065	.120	--
9. Distraction, 24m	−.098	.049	.183	−.133	.109	.030	.008	−.017	--
10. Distraction, 36m	−.024	−.020	−.088	−.028	.105	.050	.190*	.072	.148	--
11. Distraction, 48m	.018	−.037	−.014	−.027	.008	.114	.116	−.024	.030	.105	--
12. Distraction, 5y	−.018	−.052	.041	.063	.198+	.132	.194+	.085	−.052	.091	.215*	--
13. ER rating, 24m	.104	.126	.008	.042	−.035	.004	.178+	−.024	.063	.030	.120	.019	--
Bout Dominance
14. ER rating, 5y	.058	−.037	.036	.246+	.186	.137	.071	−.079	−.060	.020	.097	.343*	−.037
1. Self-Soothing, 24m	--
2. Self-Soothing, 36m	−.054	--
3. Self-Soothing, 48m	−.182+	.112	--
4. Self-Soothing, 5y	.180+	.264*	.212*	--
5. Bidding, 24m	−.235*	.053	.021	−.087	--
6. Bidding, 36m	−.006	−.229*	.079	.046	.135	--
7. Bidding, 48m	.010	−.036	−.169	−.055	.097	.196*	--
8. Bidding, 5y	−.049	.013	.026	−.124	.022	.115	.130	--
9. Distraction, 24m	−.766*	.048	.228*	−.126	−.302*	−.148	−.094	.036	--
10. Distraction, 36m	−.022	−.385*	−.166+	−.202*	−.165+	−.696*	−.162+	−.149	.130	--
11. Distraction, 48m	.096	−.087	−.355*	−.149	−.125	−.226*	−.764*	−.192+	−.030	.302*	--
12. Distraction, 5y	.036	−.103	−.069	−.231*	−.063	−.131	−.148	−.823*	.036	.208*	.287*	--
13. ER rating, 24m	−.067	.125	−.030	.034	−.177+	−.115	−.061	.193	.208*	.032	.017	−.201*	--
14. ER rating, 5y	.009	−.041	.015	.171+	.173+	.001	−.148	−.467*	−.078	.066	.179+	.424*	−.037

Note. Correlations presented within measurement dimension of ER; ER = emotion regulation;

⁺p < .10,

^*p <.05;

24m = 24 months, 36m = 36 months, 48m = 48 months, 5y= 5 years.

Age-Related Change in Strategy Use

Self-Soothing

Model selection procedures are summarized in Table 3 (left panel) and results from the best multilevel models examining age-related change in bout frequency, average duration, and proportion (reflecting dominance) of self-soothing are summarized in Table 4. The model examining age-related change in bout frequency of self-soothing was best fit by accounting for both linear and quadratic change (Figure 1, left panel). Across children, there were significant linear decreases with age, b = −1.077 (.414), p < .01, and non-significant convex curvature, b = −.135 (.116), p = .249. Estimated marginal means indicated significant decreases occurred between 24 months (M = 2.782) to 36 months (M = 1.664), estimate = −1.118 (.340), p < .01, and marginal decreases occurred from 48 months (M = 1.231) to 5 years (M = .648), estimate = −.584 (.261), p = .07. However, there was individual variation in children’s intercepts (SD = 2.974), linear change (SD = 3.539), and quadratic change (SD = .900).

Table 3.

Multilevel Model Selection Procedures

	Self-Soothing			Bidding			Distraction
	Intercept	+Linear	+Quadratic	Intercept	+Linear	+Quadratic	Intercept	+Linear	+Quadratic
Frequency
AIC	2092.572	2035.273	1997.083	2934.475	2900.460	2895.963	2522.835	2456.977	2446.718
BIC	2104.839	2042.063	2042.063	2946.742	2929.083	2940.942	2535.102	2485.601	2491.697
Log Likelihood	−1043.286	−1010.637	−987.542	−1467.237	−1443.230	−1436.981	−1258.417	−1221.489	−1212.359
χ2 from previous	--	65.298*	46.190 *	--	42.015*	12.497 *	--	73.858*	18.259 *
Average Duration
AIC	4129.917	4068.672	4004.405	1527.252	1520.478	1510.295	3671.771	3662.966	3652.886
BIC	4142.177	4097.279	4049.360	1539.519	1549.101	1555.274	3684.038	3687.500	3697.866
Log Likelihood	−2061.958	−2027.336	−1991.203	−760.626	−753.234	−744.147	−1832.886	−1825.483	−1815.443
χ2 from previous	--	69.245*	72.266 *	--	14.774*	18.183 *	--	14.805*	20.079 *
Dominance
AIC	−213.268	−287.500	−366.544	−228.524	−222.666	−264.548	−11.546	−48.638	−60.310
BIC	−201.008	−258.893	−321.590	−216.257	−198.132	−219.568	.721	−20.015	−15.331
Log Likelihood	109.634	150.750	194.272	117.262	117.333	143.274	8.773	31.319	41.155
χ2 from previous	--	82.232*	87.044 *	--	.142	51.882 *	--	45.092*	19.672 *

Note:

^*p < .05;

Bolded models were selected based on improved model fit from previous model.

Table 4.

Multilevel Models for Self-Soothing

	Bout Frequency			Bout Average Duration			Bout Dominance
Fixed Effects	Estimate (SE)		95% CI	Estimate (SE)		95% CI	Estimate (SE)		95% CI
Intercept	2.724 (.316)**		2.104, 3.344	9.992 (3.334)**		3.456, 16.529	.203 (.024)**		.157, .249
Linear Slope	−1.077 (.414)**		−1.888, −.265	−2.500 (4.215)		−102.764, 5.764	−.106 (.031)**		−.168, −.045
Quadratic Slope	.135 (.116)		−.094, .363	.112 (1.251)		−2.340, 2.565	.017 (.009)+		−.001, .034
Random Effects	SD	Corr.	Corr.	SD	Corr.	Corr.	SD	Corr.	Corr.
Intercept (SD)	2.974			33.337			.231
Linear Slope (SD)	3.539	−.837		39.522	−.744		.290	−.866
Quadratic Slope (SD)	.900	.743	−.988	11.433	.530	−.961	.075	.796	−.992
Residual	1.616			15.559			.103
Model Fit
AIC		1997.083			4004.405			−366.544
BIC		2042.063			4049.360			−321.590
Log Likelihood		−987.542			−1991.203			194.272

Note. N = 120 children, 441 observations;

⁺p < .10,

^**p < .01;

SE = Standard Error; SD = Standard Deviation; CI = Confidence Interval.

Figure 1. Predicted Trajectories of Self-Soothing.

Note. Red line represents fixed estimates; black lines represent individual variation based on random estimates.

The model examining age-related change in average duration of self-soothing was best fit by accounting for both linear and quadratic change (Figure 1, center panel). Across children, there were no significant patterns of change, including in consecutive age contrasts. There was large individual variation in children’s intercepts (SD = 33.337), linear slopes (SD = 39.522), and quadratic slopes (SD = 11.433).

The model examining age-related change in proportion of self-soothing was best fit by accounting for linear and quadratic change (Figure 1, right panel). Across children, there were significant linear decreases with age, b = −.106 (.031), p < .001, and marginal convex curvature, b = .017 (.008), p = .05. Estimated marginal means indicated significant decreases between 24 months (M = .211) and 36 months (M = .100), estimate = −.112 (.025), p < .001, and between 48 months (M = .071) and 5 years (M = .028), estimate = −.044 (.018), p < .05. There was individual variation in children’s intercepts (SD = .231), linear slopes (SD =.290), and random slopes (SD =.074).

Summary.

Supporting hypotheses, most dimensions of self-soothing showed declines with age. Bout frequency showed negative linear change from 24 months to 5 years. Dominance of self-soothing (i.e., proportion of self-soothing relative to other strategies) showed negative linear change and marginally positive quadratic change from 24 months to 5 years. Unexpectedly, there was no change in average duration of self-soothing from 24 months to 5 years. There was also individual variation in these trajectories, particularly for average duration.

Bidding

Model selection procedures are summarized in Table 3 (center panel) and results from the best multilevel models examining age-related change in bout frequency, average duration, and proportion (reflecting dominance) of bidding are summarized in Table 5. The model examining age-related change in bout frequency of bidding was best fit by accounting for both linear and quadratic change (Figure 2, left panel). Across children, there were significant linear increases with age, b = 4.163 (.877), p < .001, and significant concave curvature, b = −.862 (.291), p < .01. Estimated marginal means indicated significant increases between 24 months (M = 7.48) and 36 months (M = 10.98), estimate = 3.504 (.765), p < .001. There was individual variation in children’s intercepts (SD = 1.346), linear slopes (SD = 3.498), and quadratic slopes (SD = 1.269).

Table 5.

Multilevel Models for Bidding

	Bout Frequency			Bout Average Duration			Bout Dominance
Fixed Effects	Estimate (SE)		95% CI	Estimate (SE)		95% CI	Estimate (SE)		95% CI
Intercept	7.536 (.516)**		6.524, 8.547	2.780 (.158)**		2.470, 3.091	.180 (.015)**		.151, .210
Linear Slope	4.163 (.877)**		2.443, 5.883	.387 (.211)+		−.027, .802	.128 (.026)*		.077, .179
Quadratic Slope	−.862 (.291)**		−1.433, −.291	−.125 (.061)*		−.245, −.005	−.044 (.009)*		−.062, −.027
Random Effects	SD	Corr.	Corr.	SD	Corr.	Corr.	SD	Corr.	Corr.
Intercept (SD)	1.346			1.370			.051
Linear Slope (SD)	3.498	.848		1.503	−.842		.134	.896
Quadratic Slope (SD)	1.269	−.864	−.916	.354	.710	−.973	.048	−.923	−.997
Residual	5.416			1.034			.153
Model Fit
AIC		2895.963			1510.295			−264.548
BIC		2940.942			1555.274			−219.568
Log Likelihood		−1436.981			−744.147			143.274

Note. N = 120 children, 441 observations;

⁺p < .10,

^*p < .05,

^**p < .01;

SE = Standard Error; SD = Standard Deviation; CI = Confidence Interval.

Figure 2. Predicted Trajectories of Bidding.

Note. Red line represents fixed estimates; black lines represent individual variation based on random estimates.

The model examining age-related change in average duration of bidding was best fit by accounting for both linear and quadratic change (Figure 2, center panel). Across children, there were marginal linear increases with age, b = .387 (.211), p = .07, and significant concave curvature, b = −.125 (.061), p < .05. Estimated marginal means indicated significant decreases between 48 months (M = 3.18) and 5 years (M = 2.75), estimate = −.423 (.166), p < .05. There was individual variation in children’s intercepts (SD = 1.370) and linear slopes (SD = 1.503), and some variation in quadratic slopes (SD = .354).

The model examining age-related change in proportion of bidding was best fit by accounting for both linear and quadratic change (Figure 2, right panel). Across children, there were significant linear increases with age, b = .128 (.026), p < .01, and significant concave curvature, b = −.044 (.009), p < .001. Estimated marginal means indicated significant increases from 24 months (M = .173) to 36 months (M = .283), estimate = .110 (.022), p < .001, marginal decreases from 36 months (M = .283) to 48 months (M = .240), estimate = −.044 (.021), p = .09, and significant decreases from 48 months (M = .240) to 5 years (M =.174), estimate = .−.066 (.024), p < .05. There was little individual variation in children’s intercepts (SD = .051), linear slopes (SD = .134), and quadratic slopes (SD = .048).

Summary.

Consistent with hypotheses, bidding showed initial increases, then decreases with age. Frequency, average duration, and dominance of bidding exhibited an inverted U-shape pattern of change with age (representing positive linear and negative quadratic trajectories).

Distraction

Model selection procedures are summarized in Table 3 (right panel) and results from the best multilevel models examining age-related change in bout frequency, average duration, and proportion of distraction are in Table 6. The model examining age-related change in bout frequency of distraction was best fit by accounting for both linear and quadratic change (Figure 3, left panel). Across children, there was significant convex curvature, b = .675 (.179), p < .001. Estimated marginal means indicated marginal increases from 24 months (M = 5.65) to 36 months (M = 6.67), estimate = 1.019 (.466), p = .08, and significant increases from 48 months (M = 6.88) to 5 years (M = 10.65), estimate = 3.770 (.526), p < .001. There was individual variation in children’s intercepts (SD = .929), linear slopes (SD = 1.598), and quadratic slopes (SD = .734).

Table 6.

Multilevel Models for Distraction

	Bout Frequency			Bout Average Duration			Bout Dominance
Fixed Effects	Estimate (SE)		95% CI	Estimate (SE)		95% CI	Estimate (SE)		95% CI
Intercept	5.916 (.315)**		5.299, 6.533	15.841 (1.561)**		12.781, 18.901	.611 (.025)**		.562, .660
Linear Slope	−.552 (.525)		−1.582, .478	5.584 (2.655)*		.378, 10.790	−.013 (.035)		−.081, .055
Quadratic Slope	.675 (.179)**		.325, 1.025	−.957 (.779)		−2.484, .570	.024 (.153)**		.004, .045
Random Effects	SD	Corr.	Corr.	SD	Corr.	Corr.	SD	Corr.	Corr.
Intercept (SD)	.929	--	--	12.347	--	--	.206	--	--
Linear Slope (SD)	1.598	.947	--	21.221	−.853	--	.246	−.644	--
Quadratic Slope (SD)	.734	−.924	−.904	5.587	.838	−.999	.065	.479	−.978
Residual	3.314			11.764			.171
Model Fit
AIC		2446.718			3652.886			−60.310
BIC		2491.697			3697.866			−15.331
Log Likelihood		−1212.359			−1815.443			41.155

Note. N = 120 children, 441 observations;

^*p < .05,

^**p < .01;

SE = Standard Error; SD = Standard Deviation; CI = Confidence Interval.

Figure 3. Predicted Trajectories for Distraction.

Note. Red line represents fixed estimates; black lines represent individual variation based on random estimates.

The model examining age-related change in average duration of distraction was best fit by accounting for both linear and quadratic change (Figure 3, center panel). Across children, there were significant linear increases with age, b = 5.584 (2.655), p < .05, and non-significant concave curvature, b = −.957 (.779), p = .220. Estimated marginal means indicated significant increases from 36 months (M = 18.00) to 48 months (M = 25.70), estimate = 7.665 (1.74), p < .001). There was moderate individual variation in children’s intercepts (SD = 12.345), linear slopes (SD = 21.221), and quadratic slopes (SD = 5.587).

The model examining age-related change in proportion of distraction was best fit by accounting for both linear and quadratic change (Figure 3, right panel). Across children, there was significant convex curvature, b = 9.588 (3.243), p < .01. Estimated marginal means indicated significant increases from 36 months (M = .617) to 48 months (M = .689), estimate = .072 (.024), p < .05, and 48 months (M = .689) to 5 years (M = .789), estimate = .100 (.027), p < .001. There was individual variation in children’s intercepts (SD = .206), linear slopes (SD = .246), and quadratic slopes (SD = .065).

Summary.

Supporting hypotheses, dimensions of distraction increased with age. Bout frequency showed positive quadratic change with age, average duration showed positive linear change with age, and dominance (i.e., proportion) showed positive quadratic change with age. Notably, results revealed increasingly positive change in bout frequency and dominance with age. There was individual variation in trajectories, particularly for average duration.

Associations with ER

To determine whether strategy development was related to adult judgments of children’s ER, children’s predicted intercepts (estimated score at 24 months, controlling for group-level change) as well as linear and quadratic slopes (reflecting predicted rate of change from 24 months to 5 years) for each mixed effect model were extracted and used as predictors in regression models predicting adult ER ratings at 5 years, controlling for ratings at 24 months. Results are summarized in Table 7.

Table 7.

Regression Models Testing Associations between Strategy Trajectories and Emotion Regulation Ratings

	Bout Frequency		Bout Average Duration		Bout Dominance
	Estimate (SE)	95% CI Lower, Upper	Estimate (SE)	95% CI Lower, Upper	Estimate (SE)	95% CI Lower, Upper
Self-Soothing
Adjusted R 2	−021		−.021		−.024
F-statistic	.391		.407		.323
Intercept	.325 (.364)	−.396, 1.046	.186 (.563)	−.928, 1.301	.529 (.164)	.203, .855
ER rating, 24m	−.125 (.146)	−.413, .163	−.107 (.148)	−.399, .185	−.113 (.146)	−.401, .176
Soothing, 24m	.817 (1.062)	−1.287, 2.921	.086 (.130)	−.170, .343	18.689 (31.809)	−46.306, 83.684
Soothing, Linear	3.121 (3.984)	−4.771, 11.014	.227 (.335)	−.437, .890	72.394 (125.210)	−175.647, 320.435
Soothing, Quadratic	10.125 (12.919)	−15.467, 35.718	.623 (.913)	−1.186, 2.432	232.523 (400.934)	−561.705, 1026.752
Bidding
Adjusted R 2	.174		.058		.177
F-statistic	7.203 *		2.818 *		7.331 *
Intercept	.301 (6.897)	−13.362, 13.964	.412 (1.550)	−2.658, 3.482	17.730 (10.480)+	−3.035, 38.502
ER rating, 24m	−.051 (.132)	−.313, .210	−.092 (.140)	−.370, .186	.381 (.135)	−.230, .305
Bidding, 24m	.102 (1.098)	−2.073, 2.277	.026 (.582)	−1.127, 1.179	−.114.000 (73.950)	−260.537, 32.446
Bidding, Linear	−.361 (.232)	−.822, .100	−.718 (1.909)	−4.500, 3.065	−.357.500 (111.600)*	−578.553, −136.467
Bidding, Quadratic	−1.123 (.572)	-2.257, .010	−2.765 (6.576)	−15.792, 10.262	−1107.000 (381.000)*	−1861.538, −352.177
Distraction
Adjusted R 2	.058		.091		.105
F-statistic	2.826 *		3.946 *		4.476 *
Intercept	21.246 (37.463)	−52.968, 95.459	−7.944 (2.832)*	−13.555, −2.333	−22.793 (7.652)*	−37.952, −7.634
ER rating, 24m	−.149 (.148)	−.441, .144	−.164 (.138)	−.437, .109	−.035 (.139)	−.311, .241
Distraction, 24m	−3.228 (5.850)	−14.817, 8.361	.204 (.067)*	.071, .337	27.194 (8.854)*	9.654, 44.733
Distraction, Linear	1.632 (2.424)	−3.171, 6.435	2.516 (.858)*	.815, 4.216	100.811 (33.162)*	35.117,166.505
Distraction, Quadratic	−1.020 (2.286)	−5.548, 3.508	9.167 (3.147)*	2.933, 15.400	328.668 (109.497)*	111.755, 545.581

Note. ER = emotion regulation;

^*p < .05;

SE = Standard Error; CI = Confidence Interval; 24m = 24 months.

Self-soothing

Intercepts, linear slopes, and quadratic slopes for self-soothing were not significantly associated with adult ratings in terms of bout frequency, ps > .373, average duration, ps > .471, or proportion, ps > .385.

Bidding

Intercepts and linear slopes for bout frequency of bidding were not associated with adult ratings, ps > .124; however, quadratic slopes for bout frequency of bidding was marginally inversely associated with adult ratings, b = −1.123 (.572), p = .05. Average duration of bidding was not associated with adult ratings, ps > .513. For proportion of bidding, both linear slopes, b = −.036 (.011), p < .01, and quadratic slopes, b = −.011 (.038), p < .01, were inversely associated with adult ratings.

Distraction

Intercept, linear slopes, and quadratic slope for bout frequency of distraction were not significantly associated with adult ratings, ps > .317. For average duration of distraction, intercepts, b = .204 (.138), p < .01, linear slopes, b = 2.516 (.858), p < .01, and quadratic slopes, b = 9.167 (3.147), p < .01, were associated with adult ratings. Similarly, for proportion of distraction, intercepts, b = 27.194 (8.854), p < .01, linear slopes, b = 100.811 (33.162), p < .01, and quadratic slopes, b = 328.668 (109.497), p < .01, were positively associated with adult ratings.

Summary

Results indicated that growth in bout frequency, average duration, and dominance (i.e., proportion) of self-soothing, growth in bout frequency and average duration of bidding, and growth in bout frequency of distraction were unrelated to adult ER ratings at age 5. Consistent with hypotheses, children who used more bidding over time relative to other strategies were rated by adults as less well-regulated at age 5. Children who used more distraction over time relative to other strategies and who used longer distractions over time were also rated by adults as more well-regulated at age 5.

Discussion

Although developmental frameworks assert shifts in early childhood strategy use from automatic, caregiver-reliant behaviors to effortful behaviors that draw on internal cognitive resources (Kopp, 1982, 2009; Thompson & Goodman, 2010), there is limited longitudinal evidence of developmental change in observed strategy use. Moreover, few studies examine multiple dimensions that characterize strategy deployment in a task, i.e., their frequency, duration, and dominance. Our findings address these gaps, showing distinct developmental trajectories in dimensions of children’s self-soothing, bidding to mother about task demands, and distraction from age 24 months to 5 years during a wait task. Additionally, there is no prior evidence regarding whether such developmental change has predictive utility for adult expectations of children’s ER. Thus, this study is the first, to our knowledge, that documents associations between dimensions of strategy use trajectories from age 24 months to 5 years and adult ratings of children as well-regulated in the task.

Age-related Changes in Strategies

Prior studies documented declines in the frequency of self-soothing toward the end of the first year of life (Ekas et al., 2013) and from 24 to 48 months (Supplee et al., 2011). Our findings extend this evidence, suggesting that self-soothing use declines from 24 months to 5 years along several dimensions. Children engaged in self-soothing less often, and relatively less than bidding and distraction, as they grew older; declines in self-soothing occurred from 24 to 36 months and again from 48 months to 5 years. However, the average amount of time children spent self-soothing did not change with age, and there were large individual differences in these trajectories. Future research may identify whether such individual differences are related to the specific type of self-soothing behavior or to meaningful differences in child characteristics, such as temperament. In our sample, negative affectivity and surgency at 18 months were largely unrelated to strategy use. However, temperament dimensions such as negative affectivity should decline as children age (Posner & Rothbart, 2000), and these declines can explain variance in children’s use of distraction and bids to their mothers (Tan et al., 2012). Thus, future studies are warranted that account for how changes in negative affectivity influence variation in how long children soothe themselves (and perhaps how long it takes to effectively self-soothe).

Bidding to mothers about the demands of waiting – not being able to open the gift, having only a boring toy, and mother not paying attention to the child – yielded a complex pattern of developmental change. From 24 to 36 months, children’s bidding increased in frequency, average duration, and dominance relative to other strategies, consistent with the view that language development contributes to the use of bidding (Cole et al., 2011; Kopp, 2009; Roben et al, 2013). However, children’s bidding declined in each of these dimensions in subsequent years. Age-adjacent comparisons suggested that bidding peaked around 36 months of age. In the context of language development, by age 3 years, most children have acquired more structurally complex and efficient language skills, such as multi-phrase questions and statements (Brown, 1974; Conti-Ramsden & Durkin, 2012). These advances may enable children to increasingly communicate with their mothers in the early preschool years, and according to prior research, bidding appears helpful for reducing frustration around age 3 years (Gilliom et al., 2002). Moreover, as children enter the preschool age period, parents often expect their children to increasingly communicate with words rather than nonverbal emotional expressions (Brownell & Kopp, 2007), which may contribute to early increases in verbalization. Future research should consider how the quality and content of bids may shift and function differently across development as well. For example, there is a shift in the emotional quality of bids documented in the preschool years, from angry to calm bids (Cole et al., 2011). However, as children progress beyond age 36 months, a trend toward more autonomous strategies is expected; in tasks with blocked or delayed rewards, the optimal strategy for such young children is distraction.

We observed increases in the frequency, duration, and dominance of distraction as children grew older. The evidence for increases in how often and for how long each distraction occurred is consistent with previous reports (Cole et al., 2011; Supplee et al., 2011). Our findings add further information in documenting that children spent more time distracting themselves compared to other strategies as they grew older. Notably, how often children used distraction and to what extent they relied on distraction relative to other strategies grew rapidly in the later preschool years. Although the increase in frequency was marginal from 24 months to 36 months, it increased significantly from 48 months to 5 years. Distractions were very brief initially and increased significantly from 36 to 48 months. Finally, the dominance of distraction relative to self-soothing and bidding increased moderately from 36 to 48 months and most markedly from 48 months to 5 years. Taken together, these changes in strategy use may be related to the growing maturity of executive neurocognitive processes, such as sustained attention and inhibitory control, that appear around age 3 years (Rothbart et al., 2011; Rueda et al., 2005). From 24 to 36 months, children may initiate distractions more frequently, but they were relatively brief. With maturation in sustained attention capacity around age 3 years, children appear to begin to sustain longer distractions. It may also be that around this age, parents increasingly encourage use of distraction or teach their children to engage this strategy during a wait or other frustrating events (Morris et al. 2007).

Our findings document trajectories of strategy use during early childhood, with children increasingly deploying strategies that draw on internal, cognitive resources as they age (Kopp, 1989, 2009; Thompson & Goodman, 2010). General declines in self-soothing and general increases in distraction further support expectations that more automatic strategies such as self-soothing should wane as more effortful and independent strategies such as distraction should rise (Calkins & Hill, 2007; Kopp, 1982). Our findings with bidding suggest that strategy development may not just entail, on average, increases or decreases with age; instead, bidding exhibited a more complex inverted U-shaped pattern. It may be that early language skills support bidding as a more prevalent strategy before attentional control emerges later in early childhood (Rothbart et al., 2001; Rueda et al., 2005). Examining specific age contrasts in the dominance of strategies further supports this pattern. From 24 to 36 months, distraction was stable, but children relied less on self-soothing and more on bidding. From 36 to 48 months, reliance on self-soothing remained stable, and children grew to rely more on distraction and marginally less on bidding. From 48 months to 5 years, however, children relied even less on self-soothing and bidding, and markedly more on distraction. Future studies should investigate whether these patterns of seemingly replacing more automatic or dependent strategies with more cognitive and autonomous strategies are related, as posited, to children’s maturing neurocognitive processes and socialization experiences.

Associations between Strategies and Adult Ratings of ER

Next, we investigated whether changes in strategy use from age 24 months to 5 years were associated with adult judgments of children’s overall ER during the wait task. Specifically, adults judged whether children were more well-regulated (i.e., predominantly calm and behaved appropriately) compared to less well-regulated (i.e., predominantly distressed, off-task, or disruptive). Consistent with expectations that distraction reflects the optimal strategy for managing a wait in the preschool years, we found that age-related growth in distraction was associated with ratings that children were more well-regulated at age 5. This finding adds to evidence that latency to distraction decreases and latency to anger increases during this age period (Cole et al., 2011). Notably, it was not the frequency of distraction that mattered; ideally, the longer children are able to occupy themselves with an appropriate activity while waiting, the more successful this strategy should be. Thus, age-related growth in the average duration of distractions and their dominance among strategies were associated with adults’ ratings of child ER. Future research should investigate the extent to which this time spent in distraction is effective in helping children modulate or forestall bouts of frustration during the task.

As expected, and in contrast to distraction, changes in bidding and self-soothing were not clearly associated with adult judgments of children as well-regulated at age 5 years. First, for self-soothing, there was no relation to ratings of child ER. The importance of the relatively automatic strategy of self-soothing for reducing negative emotion has been documented in the first year of life (Ekas et al., 2013; Stifter & Braungart, 1995). Given the normative decline in use of self-soothing in this sample, our findings may suggest that use of self-soothing holds diminishing importance for ER by age 5. However, it is worth noting that self-soothing may continue to be a valuable tool for children and adults in different forms. For example, intentional and controlled self-soothing, such as breathing exercises in mindfulness interventions, are thought to improve ER (Arch & Craske, 2006). This form of self-soothing is likely different from the automatic deployment of self-soothing observed in this study. However, future research is needed to assess the nature of when self-soothing is effective and when it is limited across development.

The degree to which children increasingly relied on bidding was related to ratings of children as less well-regulated at age 5 years. In the wait task context, increasingly relying on bids to the mother likely reflects a lack of sufficient autonomy. A limitation of our work is that we did not distinguish among types of bids. Some types of bids, such as complaining, may be judged more poorly than other bids, such as information-seeking, e.g., “and when you are finished your work, then I can open it, right?” (Gilliom et al., 2002). Kopp (2009) discussed that during this age period, children’s development of self-efficacy and the ability to draw on memory, language, and reasoning should assist their transition from reliance on adults to more self-reliance in coping with negative emotions. Future research that directly examines the roles of these cognitive processes and the content of bids may better illuminate the specific ways in which bidding can be unhelpful for regulation.

Dimensions of Strategies

As suggested in prior work (Cole et al., 2004), consideration of how to measure strategy use has potential to reveal the complexity of ER development in early childhood. For self-soothing, how often children used it and how much they relied on it relative to other strategies changed with age; how long they used it did not. For bidding and distraction, we documented that bout frequency, average duration, and dominance all exhibited patterns of age-related change. Further, we found that trajectories of bout frequency for each strategy did not predict children’s ability to be well-regulated, even though frequency is a common measurement of strategies. Although average duration and dominance are less common, how long children engaged in distraction and how much children relied on bidding and distraction relative to other strategies explained children’s ER. Other dimensions (e.g., repertoire; latency, flexibility; Thompson, 1994) may also yield meaningful insights into strategy development.

Limitations and Future Directions

This study used observational tasks to assess strategy use. Although this method is widely used in early childhood ER research (Adrian et al., 2011), observations may not easily capture internalized strategic attempts that young children can make. For example, distraction may not always reach the level of observable behavior. This remains a difficult issue for understanding ER in young children given developmental limitations in the ability to self-report. Future research that combines observations with other methods, e.g., physiological and self-report, may be important for improving our understanding of strategy development (Ramsook et al., 2018). Relatedly, the wait task was designed to assess frustration (Vaughn et al., 1984); however appropriate strategies are likely task-specific (Dennis et al., 2009). Thus, other task contexts may show different developmental patterns in terms of age-related change and functionality, or even qualitatively in strategies displayed. For example, there have been mixed results for the role of toddlers’ attentional skills in fear regulation during fear tasks (Kiel et al., 2019). The role of strategies for modulating positive emotion is yet another area for future research, including the degree to which distractions may elicit positive feelings that, in turn, help forestall frustration.

Although we found significant associations between trajectories of average duration and dominance of distraction and dominance of bidding and ER ratings, strategies explained relatively little variance in ER ratings. This suggests examining the contribution of other factors involved in ER (socialization, neurocognitive skills) along with developmental trajectories in children’s strategy use may be important. Further, investigating age-related changes in strategy use in relation to children’s emotion may yield important insights into strategy effectiveness (Cole et al., 2004). For example, considering the contingent emotional responses to strategy deployment or examining developmental changes in sequences of strategies, on-task behaviors, and emotions would be fruitful ways to examine this question. Relatedly, children often deploy strategies in social contexts. A limitation of the current study is that we did not measure parenting behaviors or parents’ responses to child behavior. Measuring variation in caregiver responses to children’s strategies and socialization of strategies, particularly communicative ones such as bidding, is another important area for future research (Morris et al., 2007).

Finally, expectations for ER and strategy use are embedded within larger sociocultural norms and values. This sample adds to the extant literature by including boys and girls from an understudied demographic of economically-strained rural and semirural communities, but it will be important to include other samples, including more racially, ethnically, and economically diverse ones. As reviewed by Raver (2004), there is a need to determine whether emotion regulation can be measured and modeled similarly across sociocultural contexts, such as the impact of economic and cultural factors. Although the goal of the study was not to compare age-related changes in these children’s strategy use with any other group, we note that their strategy use was largely consistent with the predictions that are based mainly on studies of predominantly White children from more economically diverse families (e.g., Ekas et al., 2013; Rothbart et al., 1992) or more racially diverse children from families with lower reported incomes (e.g., Supplee et al., 2011). In this way, the current study fills a gap by including children who are less frequently studied. Given that this sample differs in sociocultural context from many previous studies of emotion regulation, these issues are key to consider in interpretation and generalizability of results. Research that directly assesses families’ expectations for their children will be a stronger test of whether developmental changes in strategy use align with indicators of emotional competence. For example, including caregiver or teacher ratings of ER would offer measurements beyond the laboratory context, perhaps capturing more of the competencies expected of children in their daily lives. Our measures of global ER served to highlight associations between the micro-coding of strategies and children’s overall behaviors within the same task. However, there is also a potential limitation of on- or off-task behaviors having some overlap with strategic behaviors that can be displayed in the task (e.g., dependence on the mother in global ratings may overlap to some degree with bidding to the mother). Parent or teacher ratings could also be useful for future research to determine whether associations relate to other forms of children’s ER behaviors across more varied areas of functioning. Additionally, studies with children at higher risk for ER difficulties may yield important insights about how and when deviations from normative trajectories may have longer-term implications for child functioning.

Conclusions

In support of claims that strategy use changes in early childhood (Kopp, 1982; Thompson & Goodman, 2010), we documented age-related changes in children’s self-soothing, bidding about task demands, and distraction during a wait task. Some trajectories in strategy use predicted adult judgments of children’s ER in the task. Using a longitudinal design, we found that self-soothing generally declined, bidding exhibited an inverted U-shape, and distraction generally increased from 24 months to 5 years. Further, we found that growth from 24 months to 5 years in the dominance of bidding relative to other strategies predicted ratings of children as less well-regulated. In contrast, growth from 24 months to 5 years in the duration and dominance of distraction predicted better ratings of children as more well-regulated. Since the ability to handle ordinary frustrations and disappointments, such as tolerating a wait, is a key aspect of school readiness skills and self-discipline (Blair, 2002; Raver, 2004), findings highlight the course of strategies young children develop to manage this developmental task and the importance of distraction in particular. Findings additionally provide evidence for claims that across the toddler to preschool-age years, children increasingly draw on emerging internal, cognitive resources and decreasingly rely on more rudimentary strategies. Furthermore, our findings suggest that trajectories that are in accordance with this developmental shift may allow children to be well-regulated in the overall task. Although additional evidence is needed, continuing to identify strategy use development and its relevance for children’s ER has meaningful implications for understanding children’s normative emotional development and, in the long-term, can help to identify impactful deviations.