Abstract
The evidence for this study suggests excellent internal reliability and good cross-cultural validity for the Scale for Parental Playfulness Attitude (PaPA), an important tool for occupational therapists working with Latino-American families to measure parent playfulness in the context of parent–child play.
Parent–child play is a meaningful co-occupation for many parents and children and has numerous benefits for both (e.g., Gil, 1994; Waldman-Levi et al., 2019). Play supports children’s biological, mental, and social development (Rolim Gomes et al., 2018) and encourages adults to use adaptive, stressor-focused coping strategies (Magnuson & Barnett, 2013). Play is generally associated with a particular activity, and assessments are therefore difficult to generalize, whereas playfulness refers to an approach to any activity, suggesting a more widespread application of playfulness assessment to occupational therapy. Skard and Bundy’s (2008) Test of Playfulness (ToP) operationally defines playfulness as intrinsically motivated, internally controlled, free from unnecessary constraints of reality, and well framed. Playful parent–child interactions require both players to achieve these four elements simultaneously. This can be challenging for any dyad, but the presence of barriers such as disability, illness, economic difficulties, or social injustices could make participation in playful parent–child interactions even more challenging. To support participation in playful parent–child interactions through therapeutic interventions, clinicians must be able to assess both parent and child playfulness.
Although valid and reliable assessments of children’s play and playfulness exist, there are few assessments of parental playfulness. The Parental Playfulness System (PPS; Atzaba-Poria et al., 2014) is an unpublished observation tool used to code parents’ expressions of playful characteristics while they are playing with their child. Categories of coding in the PPS are consistent with those of other tools that examine adult playfulness more generally (i.e., creativity, imagination, humor, pretense, and curiosity). The Parent Play Questionnaire (PPQ; Ahmadzadeh et al., 2020) evaluates the frequency of parent–child play, frequency of digital media use, and parents’ attitudes toward play with infants ages 3.6 mo to 2.5 yr via parent report. The Parent/Caregiver’s Support of Young Children’s Playfulness (PSYCP; Waldman-Levi & Bundy, 2016) is a criterion-referenced observation tool for use with children ages 6 mo to 6 yr and looks at the quality and frequency of parent behaviors during parent–child play. The PSYCP is based on the Model of Playfulness, operationally defined in Skard and Bundy’s (2008) ToP. Román-Oyola et al. (2019) applied the same model of playfulness to create a self-report instrument, the Scale of Parental Playfulness Attitude (PaPA), to evaluate parent playfulness in the context of play with children ages 2.5 to 7 yr. Unlike the PPS, PPQ, and PSYCP, the PaPA focuses on parent playfulness and was developed in Spanish, making it beneficial for practitioners working with Latino and Latina clients. 1
The U.S. Latino population is large (60.6 million) and rapidly growing (Gonzalez-Barrera, 2020), suggesting that occupational therapy practitioners frequently service Latino families. This population encompasses different cultural, socioeconomic, and educational backgrounds (Noe-Bustamante, 2019). Although English proficiency and literacy are on the rise, 73% of Latinos report speaking Spanish at home, and 32% do not speak English very well or at all (Krogstad et al., 2015). Therefore, a Spanish-language measure of parent playfulness is needed that yields valid and reliable data for diverse Latino parent–child dyads. The PaPA is a Spanish-language survey that has already been administered in Puerto Rico to a fairly homogeneous group of Latino parents; it has yet to be administered among Latino parents of diverse backgrounds in the mainland United States.
Using Skard and Bundy’s (2008) Model of Playfulness as a base, Román-Oyola et al. (2019) created the PaPA items through interviews with 24 parents and children and consultation with external experts on playfulness and test development. This process yielded evidence for the PaPA’s content validity. Subsequently, Román-Oyola et al. (personal communication, January 24, 2020) administered the PaPA to 522 parents dwelling in Puerto Rico; the resulting data yielded evidence for its internal reliability (Cronbach’s α = .90) and construct validity (factor analysis revealed five factors that roughly aligned with the Model of Playfulness and accounted for 55.3% of variance; Román-Oyola et al., personal communication, January 24, 2020). Román-Oyola et al. (2020) named three factors Motivation, Suspension of Reality, and Framing; the other two factors seemed to reflect shared control and personal control and were combined under Control. The apparent multidimensionality of the PaPA, derived from factor analysis, contrasts with the unidimensionality of the ToP, derived from Rasch analysis. The reasons for this difference are not totally clear; however, the inclusion of items reflecting extrinsic motivation in the PaPA may have contributed to it. The Model of Playfulness operationally defined by the ToP includes items reflecting only intrinsic motivation; however, parents commonly choose to play with their children for extrinsic reasons (e.g., to teach them something). This suggests that parents themselves may not always be truly playing or playful when they engage in play activities with their children or, perhaps, that adult playfulness is more multifaceted than child playfulness.
Unidimensionality is an essential feature of construct validity (Bond et al., 2020). When instruments are multidimensional, the meaning of test results for a given client can be unclear (Bond et al., 2020). Because previous analyses of PaPA data revealed multidimensionality, further research is warranted. An explanation of the differences between the ToP’s and PaPA’s dimensionality could be the different types of measurement tools that were used (observation based vs. self-report) or differences inherent in the statistical approaches used to develop the two assessments. As noted, Román-Oyola et al. (2020) tested the construct validity of the PaPA using factor analysis, which assumes multiple factors exist in the data. The ToP was developed using Rasch analysis, a latent-trait psychometric model that assesses construct unidimensionality by examining the extent to which data meet two expectations: (1) Easy survey items will be easy for all parents to endorse and (2) parents who are more playful will be more apt to endorse difficult items. To distinguish between easy and hard items and more and less playful participants, the Rasch model uses the degree to which playful characteristics are endorsed. When data conform to these expectations, one can assume that they represent a unidimensional construct (Bond et al., 2020). In this study, we analyzed PaPA data using Rasch analysis to assess its unidimensionality. We also examined internal reliability and whether Latino parents living in Puerto Rico and the mainland United States responded to PaPA items in the same way. In short, the purpose of this study was to explore the evidence for the PaPA’s construct validity (i.e., unidimensionality), internal reliability, and cross-cultural validity.
Method
We used a quantitative exploratory design applying Rasch, a latent-trait psychometric model. This study received approval from Colorado State University’s Institutional Review Board (Protocol 20–10200H).
Participants
Fifty U.S.-dwelling Latino parents participated; participants were ages ≥18 yr, literate in Spanish, and primary caregivers to a child between ages 2.5 and 7 yr. Most (88%; n = 44) were mothers ages 24 to 47 yr (M = 34.8 yr) with diverse educational backgrounds, 54% (n = 27) had a bachelor’s, master’s, or doctoral degree, and 40% (n = 20) had a high school diploma or less. Most (82%; n = 41) were first-generation Americans; participants were born in Mexico (54%; n = 27), the United States (16%; n = 8), Argentina (8%; n = 4), Venezuela (6%; n = 3), El Salvador (4%; n = 2), Puerto Rico (2%; n = 1), Cuba (2%; n = 1), Uruguay (2%; n = 1), Honduras (2%; n = 1), Japan (2%; n = 1), and Peru (2%; n = 1). About half (52%; n = 26) were not employed. Participants’ children were ages 29 to 85 mo (M = 47.3 mo), and 6 had a disability (autism, n = 2; attention deficit hyperactivity disorder, n = 1; Down syndrome, n = 1; other, n = 2).
Participants completed an online survey describing play with one child. The most common types of play for dyads were physical (72%; n = 36) and structured (64%; n = 32) activities, with the most common barriers to play being time (52%; n = 26) and energy (28%; n = 14). Most participants (64%; n = 32) reported an increase in play with their child during the coronavirus disease 2019 (COVID-19) pandemic. Data from U.S. participants were used to examine evidence for construct validity and internal reliability. In examining cross-cultural validity, we compared U.S. parent data with the data of 50 participants from Román-Oyola et al.’s study (personal communication, January 24, 2020). We selected these Puerto Rico–dwelling parents by matching their data with those of the U.S.-dwelling participants using child age, disability status, and child–parent gender. We then examined systematic differences in item difficulty between groups.
Instrument
Participants responded to Román-Oyola et al.’s (2020) original version of the PaPA (PaPA–O). The PaPA–O contains 42 items, each rated on a 5-point scale ranging from 1 (totally disagree) to 5 (totally agree), and it takes approximately 15 min to complete. The items measure intrinsic and extrinsic motivation (Mot), internal control (Cont), suspension of reality (Sus), and framing (Fram).
Procedure
We uploaded the PaPA–O through an end-to-end encrypted application service (Secure Sockets Layer) and collected and stored data via an encrypted cloud platform hosted by Microsoft Azure. Organizations serving Latino families sent a link to participants, who completed the survey on an electronic device. Parents were instructed to think about play with one child. Completion of the survey indicated consent and entitled participants to receive a $10 gift card.
Data Analysis
We analyzed data according to Rasch with Winsteps Version No. 4.4.4 (rating scale model; Linacre, 2021a). Rasch analysis converts ordinal-level observations (Person A is more playful than Person B) to interval-level scores (Person A is 3 logits more playful than Person B). Rasch does this by placing items and people along the same hierarchy according to item difficulty (i.e., how hard was the item for people to endorse?) and parent playfulness (i.e., how easy was it for a given parent to endorse difficult items?). We looked for evidence of construct validity, internal reliability, and cross-cultural validity using indicators and criteria listed here and further defined and explained in Table 1. As evidence of construct validity, we examined point-measure correlations, inlier-sensitive fit (infit) mean-square statistics, principal-components analysis (PCA) of linearized residuals, progression of the rating scale, spread of items, and the logic of the item hierarchy. For evidence of internal reliability, we examined person-reliability index and strata. For evidence of cross-cultural validity, we inspected Rasch–Welch probability and differential item functioning (DIF; Linacre, 2021b).
Rasch Indicators and Criteria for Questions of Construct Validity, Internal Reliability, and Cross-Cultural Validity
Note. Indicator criteria obtained from Fisher (2007) and Linacre (2021b). PaPA = Scale of Parental Playfulness Attitude.
Results
We conducted an iterative analysis to identify items and ratings that did not align with the expectations of the Rasch model and to make changes to the PaPA accordingly. Our analyses led us to combine ratings 1 (totally disagree) and 2 (disagree) for all items and to remove one control item with a negative point-measure correlation and all 13 extrinsic motivation items, yielding 28 items (Figure 1) rated on a 4-point scale ranging from 1 (disagree) to 4 (totally agree). Overall, data gathered using this version of the instrument (the PaPA–2) provided evidence for adequate construct validity, excellent internal reliability, and good cross-cultural validity.
All PaPA–2 items included in the final analysis, arranged by their respective element of playfulness, as found in the Model of Playfulness (Skard & Bundy, 2008).
Construct Validity
We found adequate evidence of construct validity for data collected with the PaPA–2. Indicators and criteria are defined in Table 1. All point-measure correlations for the items were positive, and infit statistics for 26 (93%) items were within the desired range (Table 2).
PaPA–2 Items: Relative Difficulty Measure, Infit MnSq, and PMC Values
Note. Items range from hard to easy. Values in boldface are outside a priori determined acceptable values for inlier-sensitive fit mean square (Infit MnSq) and PMC. Infit MnSq is the extent to which items conform to expectations of the Rasch model; PMC indicates whether items correlate positively with total scores. Cont = internal control; Fram = framing; Mot = intrinsic motivation; Sus = suspension of reality; PaPA–2 = Scale of Parental Playfulness Attitude–2; PMC = point-measure correlation.
PCA of residuals revealed one contrast of 4.46 eigenvalues, involving approximately 4.5 items, or 16% (see Table 1, Item 1c, for an explanation of PCA). All items had loadings >0.5. One cluster with five Mot items pulled away from a second cluster with two Fram items. These data suggest that the PaPA items may be measuring more than one dimension. Because PCA suggested the possibility of a second dimension, we examined the relationship among clusters using disattenuated correlations (i.e., corrected for error). Large disattenuated correlations suggest that clusters reflect two threads of the same dimension rather than multiple dimensions. The disattenuated correlation was .38 (moderate relationship). A simulated data set using data generated via probability based on anchored item, person, and rating-scale parameters (Linacre, 2021b) yielded a much stronger disattenuated correlation of .77 (strong relationship), implying that the two clusters reflect two threads of the same unidimensional construct of parental playfulness rather than two separate dimensions.
The 4-point rating scale progressed in the expected order for most items. There were no gaps between items greater than 0.5 logits. However, the spread of item difficulty was not sufficient to capture the playfulness of the sample; parents’ mean playfulness score (1.95) was much greater than the mean item difficulty (0), and the range of parent scores (from −0.38 to 6.08; Figure 2, left side) was greater than the range of items (from −1.44 to 1.14; Figure 2, right side). Items formed a logical hierarchy from easy to difficult, with items hardest to endorse reflecting active negotiation of control (Cont1–3) and the easiest items reflecting motivation to play due to simple enjoyment (Mot 2–4; Figure 2 and Table 2).
Internal Reliability
PaPA–2 data demonstrated excellent evidence for internal reliability. The person-reliability index was 0.85, and the strata value of 3.55 suggests that the PaPA–2 reliably discriminated more than three different levels of playfulness.
Cultural Validity
Three items differed between parents living in the mainland United States and those living in Puerto Rico. Fram6 was easier for participants in the mainland United States (DIF measure = −0.36) than for participants in Puerto Rico (DIF measure = 0.44, p = .016, DIF contrast = 0.80). Cont1 was easier for participants in the mainland United States (DIF item measure = 1.06) than for those in Puerto Rico (DIF measure = 1.67, p = 0.026, DIF contrast = 0.61). Cont6 was easier for participants in Puerto Rico (DIF measure = −0.83) than for those in the mainland United States (DIF measure = −0.07, p = .043, DIF contrast = −0.76).
Discussion
Because of the importance of parent–child play and the growing diverse Latino-American population, practitioners need a valid and reliable measure of parent playfulness. Thus, we investigated evidence for the internal reliability, construct validity, and cross-cultural validity of the PaPA. We found evidence of excellent internal reliability. The PaPA–2 yielded a high person reliability coefficient, suggesting replicability of participant placement on similar items, and it distinguished 3.55 levels of playfulness, suggesting that the PaPA–2 may be sensitive to change from intervention. These findings echo the internal reliability data (Cronbach’s α = .90) from Román-Oyola et al.’s study (personal communication, January 24, 2020) with Spanish-speaking parents living in Puerto Rico.
Playfulness in children has long been established as a measurable and unidimensional construct (Skard & Bundy, 2008). We found preliminary evidence that the PaPA–2 also measures a unidimensional construct. Goodness-of-fit statistics for 26 of 28 items (93%) conformed to the Rasch model. This is slightly less than the desired 95% but reflects the small number of items; to meet the desired criterion with only 28 items, only 1 item could fail to fit. The fact that 93% of PaPA–2 items conformed to the expectations of the Rasch model provides evidence for the unidimensionality of parental playfulness as a construct. A unidimensional construct allows researchers and practitioners to quantitatively distinguish between varying amounts of that construct. Parental playfulness could also be described qualitatively, but determining that parental playfulness is unidimensional allows practitioners to differentiate between more and less playful parents and to quantify the change that occurs through therapeutic intervention.
The criterion for logic of the item hierarchy was met (see Figure 1 for corresponding PaPA items and Figure 2 for the item hierarchy). It is logical that the Mot items (measuring intrinsic motivation) that reflect a desire to play because of simple enjoyment (Mot2–4) were easier to endorse than items reflecting a desire to play because of full engagement in play (Mot18, parent loses track of time; Mot20, parent forgets about stressful things). Similarly, the hierarchy distinguished between internal control items that probe parents’ abilities to relinquish personal control (e.g., Cont6, acting silly; Cont8, allowing child to lead) and those that probe their ability to attain cooperative control (e.g., Cont1–3, agreeing and establishing rules together). Negotiating control requires parent and child equality and is more difficult than relinquishing personal control. The hierarchy also differentiated between independent and joint suspension of reality. It is easier for parents to act out characters or be imaginative on their own (Sus1–3) than to join the child’s make-believe or imaginary play (Sus4 and Sus5). Finally, the hierarchy reflects how parents more easily received (Fram4) than gave (Fram3) cues when framing play. Even though parents felt it was difficult for children to read their play cues, it was easy for them to engage their children in play (Fram5). Parents may possibly think of cues as subtle indicators that differ from more obvious ways of initiating play.
Person–item hierarchy: Wright map depicting participant ability (left) from more to less playful and PaPA–2 item difficulty (right) from harder to easier to endorse.
Two concerns regarding construct validity require monitoring. First, the mismatch between parent playfulness and item difficulty revealed that much of the sample demonstrated greater playfulness ability than could be tested with the PaPA items. There are at least three possible explanations for this. The first, and most straightforward, is that the items are too easy; many were too easy for any parent, and most parents (>70%) passed all items (see Figure 2). If items are too easy for most families, then future researchers should create additional, harder items. Second, we cannot discount the possibility these self-selected participants completed the PaPA–O because they highly value play with their children and are very playful. Because we did not administer an external measure of playfulness, we cannot evaluate that explanation. Finally, our sample may have answered most questions in the affirmative because they deemed positive responses to be socially acceptable.
The second concern regarding construct validity arises from the PCA of residuals, which yielded a contrast with an eigenvalue of 4.46 (i.e., the strength of approximately 4.5 items). In this contrast, one group of participants awarded unexpectedly high ratings to Mot items and low ratings to Fram items, and a second group did the reverse. The reason for this contrast is unclear and requires monitoring. Disattenuated correlation findings were similarly unclear, but a correlation calculated with simulated data (.77) suggested two threads of the same dimension rather than multiple dimensions (Linacre, 2021b). This preliminary evidence indicates that the PaPA–2 does in fact measure a unidimensional construct, which improves on the findings from Román-Oyola et al.’s study (personal communication, January 24, 2020) using the PaPA–O. One major difference between the PaPA–O and the PaPA–2 is that we excluded extrinsic motivation items. Guided by Rasch fit statistics and theoretical underpinnings, we eliminated all extrinsic motivation items from our final analysis because they appeared to form a separate dimension and did not fit within the unidimensional construct of playfulness found repeatedly in studies analyzing the ToP (Skard & Bundy, 2008). The PaPA–2 measures only intrinsic motivation, control, suspension of reality, and framing and shows preliminary evidence of unidimensionality.
Given the diversity of the Latino population, evidence of cross-cultural validity of data gathered with the PaPA is a priority. Because our sample was small and heterogeneous, we could not examine invariance of the item hierarchy across subgroups. Thus, we drew PaPA data from a sample of Román-Oyola et al.’s (personal communication, January 24, 2020) Puerto Rico–dwelling parents. Three items differed significantly, and literature suggests culturally based explanations.
Cont6 (“I don’t mind acting silly when playing with my child”) was significantly easier for parents in Puerto Rico than for those in the mainland United States. Surveillance practices imposed “unequally on poorer, browner, and blacker communities” (Gellman & Alder-Bell, 2017, pp. 2–3) may explain why acting silly was harder for mainland U.S. Latino parents. Parents who feel watched may be less inclined to act silly for fear of others’ perceptions. However, Puerto Rico–dwelling parents may play with less inhibition, surrounded by neighbors who likely share cultural values and ethnic traits.
Although Puerto Rico–dwelling parents found it easier to act silly, they also found it more difficult than mainland-dwelling Latino parents to play in any location (i.e., Fram6, “My child and I can play anywhere [house, park, patio, supermarket, etc.]”). The cultural value of respeto (respectful behavior; Colón et al., 2019) may mean that Puerto Rican families designate certain spaces for play and other places for respectful behavior. Respeto is also associated with deference to parental authority (Colón et al., 2019), which may explain why mainland-dwelling Latino parents rated Cont1 (“My child and I agree easily on what we want to play”) as easier than Puerto Rico–dwelling parents. Desire for parental authority and respect may make it difficult for parents in Puerto Rico to share control with their child. Future researchers should consider whether to retain these items because, although Fram6 provides unique information, Cont6 and Cont1 may be redundant with Cont7 and Cont2, respectively. Even if these three items are retained, the overall cross-cultural validity of the PaPA–2 remains fairly strong, because 25 of 28 items did not differ significantly in relative difficulty for participants in the mainland United States and Puerto Rico.
Limitations and Future Research
Our small sample means we may not have captured the full range of parent playfulness, and it did not allow for conclusions about item functioning for subgroups of Latino Americans. Moreover, bias is inherent in all self-report instruments; when used in practice, an observational assessment might yield different results because the two approaches capture different yet valid perspectives. Finally, convenience sampling and the unique temporal context of COVID-19 may limit the PaPA–2’s generalizability.
Future researchers should administer the PaPA–2 to parents with a wide range of playfulness, targeting populations hypothesized to have both poor and excellent joint play. This would also help researchers establish what PaPA score constitutes “playful enough.” To explore the predictive validity of the PaPA–2, future researchers should administer it alongside external measures of playfulness.
Implications for Occupational Therapy Practice
The results of this study may be used to guide assessment and intervention related to optimizing quality participation in parent–child play and have the following implications for occupational therapy practice: ▪ Given the preliminary evidence presented in this study, practitioners can use the PaPA–2 as a valid and reliable measurement of caregiver playfulness for Spanish-speaking clients with children ages 2.5 to 7 yr in the mainland United States and Puerto Rico. ▪ Practitioners can use the PaPA–2 as a pre- and postintervention outcome measure, given that it can distinguish among at least three levels of playfulness (e.g., more playful, moderately playful, less playful). ▪ To guide client intervention, practitioners may administer the PaPA–2 and compare item responses with the item hierarchy shown in Figure 2. Therapists can provide the just-right challenge by focusing on skills described in the items that are near the participant’s level of playfulness and progress to more difficult skills.
Conclusion
The evidence suggests that caregiver playfulness, as defined by the PaPA–2 items, is a unidimensional construct that is not significantly influenced by contextual differences. Contextual factors such as culture may greatly influence the structure and perceived function of play, but playfulness as an approach to occupation crosses two Latino cultures (mainland United States and Puerto Rico) without significant variation. In addition, playfulness as a unidimensional construct does not seem to include extrinsic motivation. Parents may be motivated to play with their children for extrinsic reasons (e.g., to teach a child new skills, to gain respect, to bond); however, this type of motivation does not contribute to the parent’s own playfulness.
There is good preliminary evidence that the PaPA–2 can establish a baseline for caregiver playfulness for Spanish-speaking clients in the mainland United States and Puerto Rico to help occupational therapy practitioners identify targets for intervention and measure the resulting change. Such an instrument is useful for occupational therapists seeking to promote a parent–child dyad’s participation in play.
Footnotes
1In a national survey of Latinos in 2019, 47% of respondents preferred to use their country of origin to describe identity (e.g., Mexican, Cuban), whereas 39% preferred Hispanic or Latino (Gonzalez-Barrera, 2020). To respect these preferences and remain succinct, we use Latino in the remainder of this article to refer to those who would identify their family’s country of origin as within Latin America.
Acknowledgments
Special thanks to Dr. Rosa Román-Oyola, Axel Gierbolini Bermúdez, Dr. Magdalena Serpa, and Dr. Tricia Grady-Dominguez for their contributions to this study. Funding for participant gift cards was provided by Wanda Mayberry Research Funds. Kayley Goertzen is a White woman and does not belong to the Latino population of participants. This could influence equitable presentation of data. To minimize cultural and ethnic bias, Goertzen consulted Latino researchers throughout all stages of this study.