Emulating Real-Life Situations with a Play Task to Observe Parenting Skills and Child Behaviors

Abstract

Play tasks that use standardized procedures and materials are a practical way to assess parenting skills, child behaviors, and the ways in which parents and children interact. We describe a systematic process for developing the Parent–Child Play Task (PCPT) to assess mother–child interactions for a randomized controlled trial on a video-based parenting program. Participants are 307 mothers and their 3-through 6-year-old children who present oppositional and disruptive behavior challenges. The validity of the PCPT was investigated by testing (a) the extent to which the tasks elicit the specific parent and child behaviors of interest, (b) the consistency of individuals’ behavior across the play tasks, and (c) the concurrent associations of the PCPT observed child behaviors and mother reports of child behavior. The different tasks elicited the mother and child behaviors that they were designed to elicit. Behavior consistency across tasks for individual mothers and children was fair to good, with the exception of two task-specific behaviors. Mothers’ guidance (provision of instructions to foster a skill) during the teaching task and children’s interruptions while mother was busy during the questionnaire task were highly task specific. Modest associations were found between observed children’s noncompliance and inappropriate behaviors, and mother-reported conduct problems and oppositional behaviors. Implications for clinical and research assessments are discussed.

Understanding the quality of interactions between parents and young children can provide important information about the development of behavior problems and social competencies. Researchers and clinicians have long grappled with discovering the most practical and valid way to assess parenting skills, child behavior, and the ways in which a parent and child interact. Valid assessments are essential for intervention planning, evaluating individual progress toward intervention goals, and evaluating the efficacy and effectiveness of programs to enhance parenting skills and improve child behaviors. Social learning and social interaction theories support the importance of parental influence on the development of behavioral patterns in young children, positing that individuals learn from one another through the process of socialization that involves observation and modeling, as well as positive and negative reinforcement (Bandura, 1986; Bronfenbrenner, 1979; Patterson, Reid, & Dishion, 1992). Direct observations of parent–child interactions have been largely utilized to assess and understand the formation of such patterns (e.g., Gardner, 2000). These theories informed our work in developing the Parent–Child Play Task (PCPT), such as the selection of activities that promote interaction and utilization of a real-time coding system to assess the mother–child behavioral sequences. The selected coding system also captures positive and negative reinforcement.

These theories also provide the basis for relevant empirical work regarding parent–child interaction assessments. The influences during interactions between a parent and child are bidirectional: parent actions influence child responses and vice versa (Snyder et al., 2006). The bidirectional influence of parent–child interactions is supported by a number of observation studies. Mother–child mutually responsive orientation (shared cooperative interactions and positive affect) predicted children’s understanding of and compliance to rules (Kochanska, 1997). Greater parent–child mutuality was also associated with fewer child behavior problems (Deater-Deckard & Petrill, 2004). Similarly, greater levels of joint play between a mother and her 3-year-old child predicted fewer conduct problems one year later (Gardner, Ward, Burton, & Wilson, 2003). Reciprocal influence patterns unfold over time, leading to the development of long-lasting behavior patterns. Of particular interest is the extent to which a pattern of reciprocal interchange between a parent and child is predictive of children’s developmental outcomes (e.g., the coercive process during childhood predicting antisocial behavior during adolescence; Patterson et al., 1992). The challenge to researchers and clinicians alike comes in selecting observation procedures and coding systems that are relevant and appropriate for their goals.

A number of decisions need to be made regarding the selection of observation procedures. Foremost is utilizing procedures that provide a valid representation of the parent and child behaviors of interest. A secondary yet important consideration is one of pragmatics: selecting procedures that can be carried out with limited resources. These two considerations are often at odds with one another. For example, McWilliam and Ware (1994) found such high variance from observation to observation of an individual child that an unrealistic number of observations would be required to get a valid representative sample of preschool children’s engagement in the child care setting. Similarly, Ferguson, Briesch, Volpe, and Daniels (2012) found that two 30-minute or three 15-minute in vivo observations of student academic engagement in the classroom provided sufficient representation for monitoring student progress, yet even more observation time (e.g., three 45-minute observations) was needed for drawing important diagnostic decisions about a student. Many times, behaviors of interest that are typically low base rate are not elicited during an unstructured observation period. Although unstructured observations in natural settings may seem to be the most ecologically valid, given these challenges, they may not be practical for obtaining a valid representation of behavior, particularly of young children. As an alternative, semistructured lab task observation assessment procedures have often been utilized to capture parent–child interactions (for a review see Gardner, 2000). Few studies have examined the validity of lab task procedures or have provided sufficient details of the procedures for replication purposes.

Typically a lab task assessment involves the parent or parents and their child coming into the clinic or research offices. The lab task play room is set up with standardized toys for a set of activities that the parent and child engage in. Although a lab task can be designed for all developmental ages and assessment purposes, we focus on procedures for early childhood (ages 3 to 6) for assessing behavioral problems. We refer to the lab task procedures as a play task.

A play task is more efficient to implement than home observations and thus more feasible for clinics working with large numbers of families and for research studies with many participants. Efficiency is gained not only from the time it takes to travel to family’s homes but also by (a) developing a play task that elicits the target behaviors of interest, which reduces the length of observation time needed to capture those behaviors, and (b) using semistructured activities with a standard set of toys or materials, which will likely reduce the variability between observations with the same parent–child dyad from that found in naturalistic, unstructured settings, thus reducing the number of observations needed to obtain a representative sample of behaviors. An additional advantage of standardization is that it allows for a comparison of an individual’s behaviors and skills over time for monitoring progress and examining changes prior to and following an intervention. It also allows for a comparison between individuals, for example, a comparison of intervention versus control groups in randomized controlled trials (RCT), or comparisons between mothers and fathers or between subgroups of children (boys vs. girls, clinical vs. nonclinical).

The goal of a play task is to emulate real-life situations relevant to young children, to observe the phenomena of interest and the pattern of antecedent and consequent events that occasion the child’s behavior. Tasks for the PCPT were derived from prior empirical work that focused on child behavior and on the quality of parent–child interactions. For example, clean-up tasks have measured child noncompliance to parent directives, parent–child conflict, and parent use of power in disciplining their child (e.g., Dishion et al., 2008; Gardner, Sonuga-Barke, & Sayal, 1999; Kochanska, 1997). Teaching tasks have measured the ways in which parents try to teach their children new skills and children’s performance and behavioral responses with novel tasks (e.g., Fagot & Gauvain, 1997). Tasks have also been developed that emulate situations in which the parent is busy to assess how children occupy themselves when they do not have attention from their parent, the strategies parents use with their child when they are busy, and how parents deal with interruptions from their child (Gardner, Burton, & Klimes, 2006; Sanders, Dadds, & Bor, 1989). Although the tasks of the PCPT are derived from this prior work, the combination of selected tasks is unique. Moreover, little work has been done to test the validity of a play task in eliciting the child or parent behaviors of interest.

The prior work in validating play tasks has involved comparing assessments in unstructured situations or natural settings to structured or lab settings. For example, play tasks were used to compare children’s behavior in unstructured, nondirected play situations versus structured situations (Barkley, 1989; Danforth, Barkley, & Stokes, 1991). We are aware of only two small-scale studies that have compared the behaviors of parents and children in semistructured lab settings to naturalistic observations in the home. Although parent interactions with their child (use of directives, praise, and criticism) and children’s noncompliance and problem behaviors were correlated in both settings, parent use of directives and praise were greater and levels of child problem behaviors were lower in the lab setting compared to the home (Webster-Stratton, 1985). Dadds and Sanders (1992) found that for mothers with depressed children, the mothers’ sad affect during problem solving in the lab setting was related to her aversive behavior (e.g., criticism, disagreement) toward the child in home observations, and for mothers with children who have conduct problems, mothers’ angry affect in the lab setting was related to her aversive behavior at home. Given that studies with observations of parent–child interactions in both home and lab settings are rare and not very feasible with larger samples, alternative ways to test the validity of play tasks are needed.

This paper reviews the considerations and process for developing a play task assessment and introduces the PCPT that was developed for an RCT of the efficacy of a video-based parenting intervention. The purpose of this present study was to describe the procedures of the PCPT as an illustrative example of play task development and to test the validity of the PCPT in three ways. First, the validity for each play task activity to elicit the intended parenting skills and child behaviors was examined. Second, we tested the degree of consistency of individuals’ behaviors across the play task situations. Third, we tested convergent validity by examining task-specific associations of observed child behavior with mother-reported child behavior.

Development of the Parent–Child Play Task

There are a number of decisions to be made when developing a play task that stem from the outcomes of interest (see Figure 1). The outcomes of interest define who is invited to participate, what situations are emulated, and how to document the behaviors assessed. Specifically, the coding scheme, method, technology, and summarization methods used depend on the outcomes of interest. The number of participants in a given play task, the situations selected to emulate real-life situations, and whether the data summaries are for clinical or research purposes will also guide decisions about documentation. Our development process for the PCPT followed this process.

Figure 1.

Decisions for Developing a Play Task

Defining the outcomes of interest

The outcomes of interest for the PCPT are related to parenting skills for children who present challenging behaviors. The parenting outcomes of interest were to increase parents’ conversations, teaching, and positive interactions with their child and to decrease angry and harsh interactions and the use of inappropriate discipline techniques. The outcomes of interest for children were to increase positive social interaction skills and following directions and to decrease conduct problems (e.g., rule breaking, aggression) and oppositional behaviors.

Selecting participants

Given the outcomes of interest, the focus of the play task is on parent–child interactions. PCPT task assessments can include the mother, the father, both parents, or other caregivers in the household (e.g., grandmother). For this study, the decision was made to invite the mother and child to the play task, as the mother is typically the parent who spends the most time with the child in a caretaking role and is most likely to participate in a parenting intervention. Although father–child play task data would clearly provide valuable information about the nature of father–child interactions, the decision to include only one parent in the play task assessments was made for feasibility reasons.

Defining situations to emulate

Parents report that problem behaviors occur with their young children in many diverse situations or contexts, such as when they are busy (e.g., talking on the phone or to another adult, completing paper work), during shopping trips, and when the child does not want to do something (e.g., cleaning up their toys, going to bed, or stop doing something they are enjoying; Sanders et al., 1989). Positive interactions between parent and child are expected to occur often when they are engaged in play together, and parents’ teaching is expected to occur when their children are given tasks that are novel and a little difficult. In developing the tasks for the PCPT, it was challenging to create a lab-based scenario to emulate shopping or bedtime routines, so tasks emulating the mother being busy and a clean-up task were created. A free-play task and a teaching task were also created to elicit positive interactions and parent teaching. Table 1 shows the PCPT tasks, timing, and materials, and the behaviors each task intended to elicit.

Table 1.

Description and Goals of the Parent–Child Play Task Activities

Activity	Time	Materials	Goal: Behaviors to elicit
1) Questionnaire	20 min	One Where’s Waldo book.	Child’s activity when mother is occupied. How mother deals with child interruptions when she is busy.
2) Free play with phone interruption	10 min total. Phone interruption is 4 min.	Colorful small blocks. 4 small people figures. Plastic farm animals and fencing. 3 colored bins for toys. Telephone.	Warm up activity for observing spontaneous interactions. Interactions when there are competing demands for mother’s attention during phone call.
3) Clean up with wait time	10 min. maximum cleanup plus 5 min. wait time.	Same as above. Instruction card for mother.	Mother directions. Child compliance. Mother–child interactions when waiting with nothing to do.
4) Adult-led teaching	5–10 min.	12–48 piece puzzle (depending on child age).	Mother teaching. Child response to teaching and to novel task.

Documentation of parent–child interactions

As there was an interest in examining parent–child interactions and behavioral sequences that could identify contingencies of interaction, a sequential coding system was selected. Since the tasks selected were expected to elicit specific interactions and behaviors at a somewhat frequent rate, a computerized coding system was selected so that frequent behaviors could be coded in real time. The play task sessions were video-recorded for later coding. The behaviors of the mother and child dyads were coded sequentially as they occurred in one pass through the video. The coding system identified the initiator, action, and affect. Specific codes were derived from a number of sequential coding systems used in prior research and are described further in the Measures section.

Pilot testing

The play task methods were piloted and adjustments were made as needed. Six different mother–child dyads participated in the pilot. Participating children were between 3 and 6 years old. The purpose of the pilot was to determine the extent to which (a) the instructions for the tasks were understood, (b) the times allotted for a task were realistic, (c) the task seemed to elicit the behaviors of interest, (d) the room was set up well, and (e) the materials matched the children’s developmental levels. The pilot was done in an iterative manner: after three mother–child pairs participated, adjustments were made and then the remaining mother–child pairs participated to allow review of how well the adjustments worked. Examples of adjustments made were to childproof the room better (e.g., shorten the cords of the phone, move the microphone higher out of reach) and to add a waiting time after the toys were cleaned up to assess parent–child interactions when there is nothing to do. This change was added because some children cleaned up rather quickly (e.g., in three minutes), leaving a rather short period for that task. It also emulated real-life situations in which parent and child need to wait together with little to do (e.g., waiting in line at the grocery store, waiting for an appointment).

Study Aims

This study aimed to test the validity of the PCPT. The first aim was to test the extent to which the PCPT elicited the parenting and child behaviors of interest. We expected (a) the Questionnaire Task, in which mother is busy, to elicit interruptions and demands from the child; (b) the Free-Play Task to elicit more conversation between the mother and child than the other tasks; (c) the Clean-Up Task to elicit more directives from the mother, which would be followed by child compliance or noncompliance; and (d) the Teaching Task to elicit mothers’ guidance.

The second aim was to test the consistency of individuals’ behavior across the play task situations. Although each task was designed to bring about particular behaviors, we expected relative consistency of an individual’s behavior across the different tasks. We expected fair to good consistency for the mother and child behaviors.

The third aim was to test the convergent validity between the observed child behavior problems (noncompliance and inappropriate behavior) during the play task with mother reports of conduct and oppositional behaviors. We expected significant yet modest relationships between the observations and mother reports, as is typical (Achenbach, McConaughy, & Howell, 1987; Kerr, Lunkenheimer, & Olson, 2007). We tested the associations of mother-reported problem and oppositional behaviors with the task-specific behaviors, such as noncompliance during clean up, as well as with noncompliance and inappropriate behavior across the entire play task assessment. We expected that the task-specific behaviors might result in stronger associations with the mother reports of child behavior due to their specificity during tasks designed to emulate real-life situations.

Method

Participants

Participating families were from a midsized county (population approximately 350,000) and an urban county (population approximately 750,000) in Oregon, U.S.A. The study was advertised using multiple methods, such as flyers and brochures posted during community events and in preschools and medical and dental offices, and announcements on the radio, television, child care and school newsletters, and on the internet. The advertisement specified that participation was for parents who were experiencing behavioral challenges with their child, such as disobedience, aggressiveness, or temper tantrums. This population was selected for this study to inform parenting interventions aimed at reducing and preventing disruptive behavior problems. Eligibility criteria were that the child was ages 3 to 6, the child lived with the mother, mother could speak English, and the family was currently not receiving any parenting support or child mental health services. Interested parents called the offices and were then screened using a subset of 15 items from the Intensity scale of the Eyberg Child Behavior Inventory (ECBI; Eyberg & Ross, 1978). The subset of 15 ECBI items used in this study were selected on the basis of being highly correlated with the original Intensity scale of the ECBI (r = .94; Cronbach’s alpha = .91). Parents of children with scores of one or more standard deviations above the mean on the ECBI were eligible to participate. Once determined eligible, the parents and their child were invited to the offices to learn more about the study, provide informed consent, to complete the baseline questionnaire, and to participate in the play task. All procedures and research activities were reviewed and approved by the Institutional Review Board (IRB).

A total of 318 families participated in the study, and baseline observation data were available for 307 mothers and their children (97%). Thirty-one percent of the participating mothers were single parents, and 72% were Caucasian, 10% African American, 5% Hispanic, 2% Asian or Pacific Island Native, 1% Native American, and 10% were of mixed race. The median household income of participants was in the $25,000 to $29,999 range.

Research Design and Procedures

The PCPT was developed for an RCT of a video-based version of the Triple P–Positive Parenting Program (Sanders, 2012). The baseline assessments occurred prior to randomization of families into one of three conditions: standard intervention (to receive the Triple P video series), enhanced intervention (to receive the Triple P video series enhanced with content designed to increase parents’ implementation of the practices), and a waitlist control group. The present study utilized baseline play task observation data and parent questionnaires, which were collected prior to randomization and intervention.

Measures

Mothers and their participating child were invited to come into the research offices for an assessment visit. Child care was provided for nonparticipating children if needed. The play task occurred in a dedicated assessment room with a small child-sized table with two chairs in the center of the room. A larger table was against the wall for the telephone and where the face-to-face interview occurred during the Transition and Interruption Task. A discreet motorized camera, which was monitored and operated from a separate video room, and microphone were installed out of children’s reach.

Per the PCPT protocol (available upon request from the first author), a research assistant announced instructions for a task and then left the play room while that task occurred (materials and descriptions for each task are in Table 1). During the assessment visit the mother completed the first section of the questionnaire as part of the play task (the Questionnaire Task). During this task the child was provided with a Where’s Waldo? book to look at while his or her mother filled out the questionnaire. This first section of the questionnaire included the full 36-item ECBI; thus, mothers completed the ECBI before the mother–child interaction portion of the PCPT. No questions about parenting skills appeared on the first half of the questionnaire, so as not to potentially bias the mothers’ behavior during the play tasks.

For the Free-Play Task the researcher brought out three toy bins and opened them up, saying, “…we would like you and your child to play together. Here are some toys for you to play with.” The mother was also alerted that the researcher was going to phone her while they were playing to ask some questions and was prompted to answer the phone when it rang. During the Clean-Up Task the mother was asked to follow an instruction card that said, “Get your child to help pick up the toys” and specified in which colored bin to put each type of toy. For the Teaching Task mothers are told, “Here is a puzzle that I would like you to put together. It may be a little difficult for your child to do by her/himself. Go ahead and help your child by offering suggestions, but do not do the puzzle for her/him.” The PCPT took approximately 55 minutes.

The second section of the questionnaire was then completed after the play task was over. After this baseline assessment visit, mothers then reported on their child’s behavior during three separate telephone interviews.

The Parent–Child Play Task Observation System (PCPTOS)

The video recordings of play task sessions were coded on computers using a systematic sequential coding system. Both mother and child behaviors were coded simultaneously in real time with a 4-digit coding system. The first digit indicated who was doing the behavior, the next two digits represented the content of the behavior (e.g., approval, directive), and the fourth digit indicated the emotional affect. As a number of reliable parent–child interaction coding systems exist, the codes and definitions were selected and derived from the following coding systems for coding parent or caregiver and child interactions, behaviors, and emotional affect: Family Observation Schedule Revised (Sanders, Waugh, Tully, & Hynes, 1996), Family Child Care Observations (Rusby, Smolkowski, Marquez, & Taylor, 2008), Interpersonal Process Code (Rusby, Estes, & Dishion, 2000), Family Process Code (Dishion et al., 1987), and Living in Family Environments (Hops et al., 1990).

For the present study, the mother behaviors of interest included praising, guidance, conversing, and clear start, clear stop, and vague directives. The child behaviors included conversing, demanding, interrupting, inappropriate rule-breaking behaviors, and compliance and noncompliance to mother’s directive. Mother and child positive, neutral, and negative affect were also coded. Coding categories were mutually exclusive, and coding hierarchies were defined to prevent coding a single behavior into more than one category. Code descriptions are provided in Table 2. The score for each coded behavior is the rate of the behavior per minute.

Table 2.

Content Codes for Mother and Child

Code Name	Code Definition
Mother Behavior Codes
Approval	Verbal praise or positive recognition of the child’s behavior.
Guide	Directions, instructions, useful information, or suggestions which teach a behavior or specific skill in order to foster independence with a task.
Converse	General conversation, chit-chat about routine matters, relating stories, conversation about past or present, questions and answers, or verbal acknowledgements of child’s statement.
Clear Start Directive	Clear and firm directives for behavior change for the child to do something in the immediate future.
Clear Stop Directive	Clear and firm directives for behavior change for the child to stop doing something in the immediate future.
Vague Directive	An unclear but firm directive for behavior change to either start or stop a behavior. The desired behavior change is not clearly articulated.
Child Behavior Codes
Converse	General conversation, chit-chat about routine matters, relating stories, conversation about past or present, questions and answers, or verbal acknowledgements of mother’s statement.
Demand	Aversive or unpleasant verbal instructions directed to the mother.
Interrupt	Any attempt to interrupt the mother’s activity (i.e., when mother is filling out a questionnaire, talking on the phone, or being interviewed).
Inappropriate	Other inappropriate, rule-breaking behaviors that are not included in the other codes.
Compliance	Clearly obeying mother’s directive. The child starts complying within 5 seconds of a directive to either start or stop doing something.
Noncompliance	Child does not follow mother’s directive for a behavior change to either start or stop doing something within 5 seconds.
Affect Codes for Mother and Child
Positive Affect	The mother or child is expressing happiness or affection.
Neutral Affect	The mother or child uses even-tempered, matter-of-fact, or conversational tone.
Negative Affect	The mother or child is expressing distress, discomfort, irritability, anger, or overt sadness.

Six research assistants were involved in coding the mother–child interactions from the videos of the play task; four coders had three to more than eight years of experience and two were new to observational coding. The coders underwent training which involved learning the definitions, examples, and nonexamples of each code by practicing with flashcards and watching videotapes of parent–child interactions. Training took six to eight weeks (two weeks learning the codes and four to six weeks practicing) for coders to achieve 80% interrater agreement on the coding of practice videotapes. Research assistants were randomly assigned to code the PCPT videos, and randomly assigned interrater reliability checks were conducted on 15% of the lab tasks. All coders were kept masked regarding participant experimental condition and whether or not it was a reliability check. Time unit interrater agreement was computed using GSEQ v5.1 with a tolerance window of six seconds (Bakeman, Quera, & Gnisci, 2009). The interrater agreement for content codes at baseline was 83% with a Kappa of .78 and for affect was 90% with a Kappa of .85.

Mother Questionnaire of Child Behavior

During the assessment visit, the mother completed the full version of the ECBI (Eyberg & Ross, 1978). The frequency of 36 different child misbehaviors was rated by participating mothers on a response scale of 1 to 7 (never to always). Questions on the ECBI include defiance, anger, fighting, attention difficulties, and mealtime and bedtime problems. This scale has good internal consistency, test–retest reliability, discriminant validity, and construct validity (Eyberg & Ross, 1978; Robinson, Eyberg, & Ross, 1980; Schuhmann, Durning, Eyberg, & Boggs, 1996). In the present study, the internal consistency alpha for this scale was .89.

Parent Daily Report

In addition, the Parent Daily Report (PDR; Chamberlain & Reid, 1987) was collected. Mothers reported on their child’s behavior in the past 24 hours during a set of three phone interviews during each assessment time point for the project. Good participation rates were achieved; 90% of mothers completed all three phone interviews, 4% completed two of the interviews, 4% completed only one phone interview, and 3% did not participate in any phone interviews. Mothers reported on the frequency of each behavior on a scale of 0 to 6 (from never to 6 or more times in the past 24 hours). The subscales for oppositional behavior and conduct problems were computed by averaging scores across subscale items for each separate PDR, and then averaging those subscale scores across the three baseline PDRs to calculate a single baseline score for each subscale. Oppositional behavior comprised six items: arguing with an adult, being disobedient, resisting bedtime, misbehaving while shopping, having temper tantrums, and yelling at someone (α = .68). Conduct problems comprised six items: being aggressive, being destructive, fighting with siblings, lying, teasing others, and stealing (α = .78).

Analytic Techniques

The coded behaviors were computed into rates per minute and affect codes were computed as percentages of time spent in that affect to account for the different lengths of the play tasks. Paired t-tests were used to assess the difference in specific mother and child behaviors between the task of interest and the other tasks. Using the procedures and guidelines recommended by McGraw and Wong (1996), one-way random effects intraclass correlations (ICCs) were computed to assess the individual consistency of the mother and child behaviors across tasks. An ICC of .600 and greater is considered good consistency, .400 through .599 is fair, and less than .400 is poor (Shrout & Feiss, 1979). Associations between observed and mother-reported child behaviors were evaluated using bivariate correlations. Specifically, Spearman’s rho was used to appropriately account for the distribution of the data.

Results

The means, standard deviations, and range of values for the behavioral codes and mother report of child behavior are shown in Table 3. Descriptive statistics for the measures are reported in Table 3. Low base rates (an average frequency of less than every 15 minutes) were found for two observed behaviors: mothers’ use of stop directives and children’s demands. Converse was the most frequent behavior of mothers and children, averaging about two per minute. The ratio of compliance to noncompliance to mothers’ directives was slightly more than 2 to 1. On average, children expressed negative affect about once every four minutes. Both observed and mother-reported behavior had good variance. For example, children’s inappropriate behavior during the PCPT ranged from zero to two inappropriate behaviors every three minutes.

Table 3.

Descriptive Statistics of Coded Behavior and Mother Reports of Child Behavior

Variable	Mean	SD	Minimum	Maximum
Mother Behavior Codes
Start Directive	0.36	0.23	0.01	1.61
Stop Directive	0.06	0.06	0.00	0.36
Vague Directive	0.18	0.18	0.00	1.11
Guidance	0.77	0.32	0.03	1.92
Approval	0.29	0.17	0.00	0.91
Converse	2.02	0.67	0.16	4.19
Child Behavior Codes
Child Demand	0.08	0.10	0.00	0.63
Child Interrupt	0.36	0.19	0.00	1.20
Child Compliance	0.39	0.23	0.02	1.75
Child Noncompliance	0.16	0.16	0.00	1.25
Child Converse	2.15	0.70	0.14	4.22
Child Inappropriate	0.17	0.21	0.00	1.55
Child Negative Affect	0.27	0.41	0.00	0.52
Mother Reports
ECBI Behavior Problem Frequency	167.91	23.60	89.00	238.00
PDR Oppositional	1.87	1.02	0.11	5.25
PDR Conduct Problems	1.11	0.88	0.00	5.58

Note. Behavior codes are reported as rate per minute. The possible range for the ECBI is 0–252 and for the PDR is 0–6.

The Extent to Which Tasks Elicited the Behaviors of Interest

The comparisons of means for the behaviors of interest for each task to the means for the behaviors across all the other tasks are provided in Table 4. As expected, each task elicited greater levels of the intended mother and child behaviors compared to the other tasks. Many of the differences were substantial; mother guidance occurred over 10 times more often during the Teaching Task compared to the other tasks, and child interruptions occurred about seven times more often during the Questionnaire Task compared to the other tasks.

Table 4.

Code Rate Per Minute Elicited By Specific Tasks

Task and Codes	Average for Intended Task	Average Across Other Tasks	t-test	p-value
Questionnaire Task
Child Interruptions	.82	.13	28.52	<.001
Child Demands	.13	.06	7.06	<.001
Free-Play Task
Mother Converse	3.18	2.18	16.44	<.001
Child Converse	2.68	2.54	2.52	.012
Clean-Up Task
Mother Clear Start Directives	.89	.32	17.16	<.001
Mother Clear Stop Directives	.17	.05	10.40	<.001
Mother Vague Directives	.32	.15	8.39	<.001
Child Compliance	.86	.35	19.25	<.001
Child Noncompliance	.36	.12	9.63	<.001
Teaching Task
Mother Guidance	3.77	.35	36.66	<.001

A post hoc analysis investigated whether findings regarding the validity of the tasks held true for both girls and boys. The tasks elicited the expected behaviors in models split by gender. Across the 10 behavior-by-task comparisons for each gender (20 tests), only two of the behaviors were not elicited more frequently in the specific relevant task, compared to other tasks for boys; the Free-Play Task did not elicit significantly greater converse and the Clean-up Task did not elicit greater compliance for boys. Only one task-specific behavior was not significant for girls; the Clean-up Task did not elicit more frequent vague directives from mothers to their daughters compared to other tasks. As these analyses were conducted post hoc, the few gender differences found require further empirical support.

Individual Consistency Across Tasks

ICCs to test individual consistency across tasks are shown in Table 5. Good consistency was found for mother approval, converse, clear start directives, and vague directives; fair consistency was found for mothers’ clear stop directives; and poor consistency was found for mothers’ guidance across tasks. The poor individual consistency for guidance across tasks might be expected given the prior finding of more guidance elicited during the Teaching Task than other tasks. Good consistency was found for child compliance, converse, inappropriate behavior, and negative affect; fair consistency was found for child demands and noncompliance; and poor consistency was found for child interrupt. Child interrupt was shown to be more specific to the Questionnaire Task, in which the mother is busy and the child has little to do.

Table 5.

Consistency of an Individual’s Behavior Across Tasks

Code Name	ICC
Mother codes
Start Directive	.641
Stop Directive	.446
Vague Directive	.736
Guidance	.316
Approval	.601
Converse	.655
Child codes
Child Demand	.541
Child Interrupt	.385
Child Compliance	.700
Child Noncompliance	.573
Child Converse	.688
Child Inappropriate	.629
Child Negative Affect	.630

Associations Between Observations and Mother Reports of Child Behaviors

Associations between observed and mother-reported child behaviors are reported in Table 6. Associations are shown for the observed behavior during both the task that was expected to elicit that behavior (e.g., noncompliance during Clean-Up Task) as well as across all tasks, with mother reports of behavior problems on the ECBI and of oppositional behaviors and conduct problems on the PDR. With regard to noncompliance, small yet statistically significant correlations were found between observed noncompliance (in both the Clean-Up Task and across all tasks) and mother-reported total behavior problems and oppositional behavior (r = .12 to .19). With regard to inappropriate behavior, a modest association was found between observed inappropriate behavior and the ECBI problem behavior scores across all tasks (r = .16). Mother-reported conduct problems on the PDR were not associated with either observed child noncompliance or inappropriate behavior during the play tasks.

Table 6.

Correlatons between Mother-Report and Observed Child Behaviors

Task and Codes	ECBI Frequency	PDR Oppositional	PDR Conduct Problems
Child Noncompliance during Clean-Up Task	.16**	.13*	.07
Child Noncompliance across all Tasks	.19**	.12*	.04
Child Inappropriate across all Tasks	.16**	.05	.04

Note. Spearman’s Rho correlation statistic was utilized to account for nonnormal distribution of the observation data.

^*p < .05.

^**p < .01.

Discussion

When careful systematic processes are used in their development, semistructured play tasks can provide valid information on parent–child interactions, parenting skills, and child behavior. Important steps in the development process are to (1) clearly define the targeted parent and child behaviors that are outcomes of interest, (2) identify the most appropriate participants in the play task, given the outcomes of interest and feasibility constraints, (3) systematically develop play tasks that elicit the parent and child behaviors of interest, (4) pilot test the play task on a few participants that represent the targeted population to ensure that the protocol works well, and (5) create a method for analyzing and summarizing the play task behaviors that matches intervention goals and is feasible to carry out.

The PCPT has been used to illustrate this development process and to examine the extent to which the tasks validly measure parenting behaviors and disruptive behavior of young children targeted for intervention. First, the tasks successfully elicited the behaviors of interest. Many of these findings were quite substantial. For example, the Questionnaire Task, which emulated situations in which the mother is busy and the child has little to do, elicited over six times more child interruptions compared to the other tasks. Likewise, the Teaching Task elicited about 10 times more mother guidance compared to the other tasks. These findings provide support that the tasks emulated real-life situations to measure how a parent and child tend to interact during those situations. Thus, if a parent or child makes improvements in their behavior as a function of an intervention, the PCPT would be expected to capture those changes.

The test of consistency across tasks for individual mothers and children showed fair to good consistency (Shrout & Feiss, 1979). These results indicate that, for example, mothers who tend to give their child vague directives tend to do so in a variety of situations and contexts. The consistency was low for the behaviors that were shown to be highly task specific: child interrupt during the Questionnaire Task and mother guidance during the Teaching Task. These results imply that when examining intervention effects, it may be best to examine improvements in mother guidance during a teaching task, for example. On the other hand, given the consistency across tasks, it appears that it would be best to examine decreases in use of vague directives across all tasks. These levels of consistency for individual mothers and children indicate that the play task is validly measuring an individual’s tendency for a particular behavior. Such consistency is hard to find with observations in natural settings (e.g., McWilliam & Ware, 1994).

The concurrent validity of the PCPT was tested by examining observed child noncompliance and inappropriate behaviors with mother-reported behavior problems. Significant yet small associations were found between observed noncompliance and mother-reported frequency of problem behaviors (ECBI) and oppositional behaviors (PDR), and between observed inappropriate behavior and mother-reported problem behavior frequency. The correlation with mother report of child problem behaviors was slightly stronger for observed noncompliance across all tasks than for the specific Clean-Up Task. Although we expected these associations to be modest (Achenbach et al., 1987; Kerr et al., 2007), they were smaller than expected. Perhaps adding more difficulty to the Clean-Up Task, such as needing to clean up a desired toy that the child has not had much time to play with, would increase the opportunity to capture noncompliance in children who exhibit oppositional behavior.

In addition, no associations were found between these observed behaviors and mother-reported conduct problems on the PDR. Although the PCPT succeeded in eliciting behaviors of interest, the play task situation may not be suitable for measuring low-rate, high-intensity behaviors, such as aggression, destroying property, lying, and stealing. Other behaviors measured on the PDR conduct problem scale include peer-related problems. If these behaviors were of particular interest to the clinician or researcher, then these behaviors would better be captured by creating a play task with a peer or peers.

The findings in the present study point to the need to carefully match the measurement method with the goals of the clinical program or research and the resources available. Careful attention to this match will enable one to select the most efficient method while retaining the ability to collect important information on intervention progress or outcomes (Metzler, Sanders, & Rusby, 2014). For feasibility purposes, we recommend shortening the PCPT to 45 minutes by shortening the questionnaire task from 20 minutes to 10 minutes. In post hoc analysis no significant differences in the rates of child demands or interuptions were found between the first 10 minutes of the Questionnaire Task and the last 10 minutes, providing support that this shortening would not impact the results.

The four tasks of the PCPT combined provided important information about mother–child interactions in a variety of situations; how the child acts when his or her mother is otherwise occupied, how the mother responds to child interruptions, typical mother–child interactions during play, how the mother provides directions and the child responds to those directions, the strategies the mother uses to teach her child, and how the child responds to learning a novel task. For clinical use, the variety of tasks can help pinpoint when difficulties tend to occur and what parenting strategies would be useful for preventing those diffictulies. On the other hand, if the clinical interest is solely focused on teaching strategies, perhaps conducting the Teaching Task would suffice.

For research, it may be useful to include the entire PCPT or specifically selected tasks in a multiple method measurement plan. The combination of microsocial observation data (such as that collected with the PCPT) and macrolevel data (e.g., parent questionnaires of their practices and children’s behavior) in research evaluating interventions is advocated for obtaining objective assessment and accounting for the situation specificity of behavior (Achenbach et al., 1987; Snyder et al., 2006). Moveover, adding physiological measures to the PCPT could contribute to basic science research investigating the processes that underlie behavior, emotions, and learning. The utility of the PCPT for these purposes has yet to be established.

Study Limitations

This study clearly does not address many other issues that exist regarding objective observations of parenting and child behaviors. For example, we utilized behavior frequencies to provide validity information for a semistructured play task and did not approach more complex data summary, such as sequential data analysis. Such methods will be needed for analyzing changes in parent–child interaction patterns. In addition, a parenting program may target complex changes in parenting practices that are quite challenging to capture in observations. For example, a parenting program may advocate the use of incentives to increase a desired behavior (e.g., sticker charts, privileges), but then recommend fading the incentives over time to maintain the behavior change. In this case, expecting a steady increase in positive incentives may not represent the ideal parenting outcome. Another challenge we experienced was how to detect planned ignoring, in which the parent deliberately ignores annoying but harmless behaviors (e.g., whining) so as not to reinforce them with attention, but which can appear similar to “cluelessness” to observers who are masked to participant condition and details about intervention content (which was the case in the present study).

Given the modest associations between mother reports of child behavior and the observed child behavior from the PCPT, further work is needed to validate the concurrent validity for measuring child behaviors. A study that involves multiple reporters and observations in a variety of settings (e.g., home, school, clinic) would provide further evidence of validity. To more thoroughly establish construct validity, however, it is important to test the discriminant validity of the play task in addition to convergent validity. To test discriminant validity, constructs that are not expected to be related to oppositional and disruptive behaviors would be examined. The challenge for such an investigation is that oppositional and disruptive behaviors often co-occur with other behaviors in young children, such as internalizing and attention problems (Boylan, Vaillancourt, Boyle, & Szatmari, 2007; Thomas & Guskin, 2001).

There are a number of additional aspects of validity that have not been tested in the present study reported here. These include concurrent validity of parenting practices via parent report and observations, test–retest of behavior frequencies, predictive validity of observed parenting practices on observed child behavior, and sensitivity of the PCPT to detect intervention change. Future work could also test the validity for different ages and for different populations. Generalization of the findings to fathers or other caregivers is limited, as this study involved mother–child dyads. Nevertheless, the aspects of validity tested here for a semistructured play task of parent–child interactions are important and there is limited research in this area.

Conclusion

Objective observational assessments play an important role in clinical and research settings. The challenge is finding feasible ways to collect valid observation data, particularly in community settings. This research illustrates that data from carefully developed semistructured play tasks can provide valid information about parenting and child behavior. In this case, for research purposes, a complex coding system has been used to summarize the mother–child interactions that occurred during the play task. Utilization and summary of such a coding system is likely not feasible in community settings, however. A critical next step would be to create a way to efficiently provide summary reports of the information gathered in these validated play tasks for data-based decision making in clinical practice and other community settings.