Research and theory on play have recognized that it is a significant activity that promotes children’s socio-emotional and cognitive development, enhancing also their psychological adjustment (Morrison, 2001). According to their age, children engage first in functional and exploratory play behaviors, then they shift to symbolic or pretend play (Lillard, 2001; Cartwright, 2004). Pretend play considered children’s exploration and interpretation of the world with symbols, fantasy, make-believes, expression of emotions and capacity to explore different circumstances in an imaginary context. In pretend play, children have also the chance to act out personal inner and social situations facilitating emotional and cognitive understanding of their experiences (Fein, 1987; Moore and Russ, 2006; McAloney and Stagnitti, 2009). As Marcelo and Yates (2014) pointed out, pretend play is a powerful mechanism that includes the interplay of various dimensions, all crucial for a healthy development in childhood. The use of imagination and affects expression is strictly connected with children’s gains in cognitive and representational abilities (Bornstein et al., 2002), self-regulation (Berk et al., 2006), coping (Goldstein and Russ, 2000), psychological adjustment, problem solving (Singer and Singer, 1990) and perspective taking (Fisher, 1992; Seja and Russ, 1999). Pretend play is the result of cognitive and affective integration which is responsible for children’s development and acquisition of all the abilities reported above (Russ, 2004, 2006).

In order to observe and evaluate these numerous dimensions related to children pretend play, research in this field underscored the need for standardized measures to assess children’s cognitive and affective processes that occur together in play sessions (e.g., Russ, 1993, 2004; Kaugars and Russ, 2009). However, studies often devised tools devoted to separately assess cognitive and affective domains in play; moreover, most of these instruments did not define clear standard task or administration procedures (e.g., Gitlin-Weiner et al., 2000; Chessa et al., 2012). To date, Russ (2004) was one of the first authors who tried to satisfy these conditions in devising the Affect in Play Scale (APS), a standardized instrument of pretend play in school-age children (6 to 10-years-old). It consists of a videotaped participants’ 5-min play task with two puppets and some colored blocks, scored on cognitive (Organization, Elaboration, Imagination, Comfort) as well as on affective components (the Frequency and Variety of affective themes and Positive versus Negative affective themes) of pretend play. This measure showed good psychometric properties as internal consistency, reliability and validity (Seja and Russ, 1999; Russ, 2004 for a review). A confirmatory factor analysis (CFA) in an Italian sample showed the existence of two correlated factors: a cognitive and an affective one (Delvecchio et al., 2016a). More recently, Kaugars and Russ (2009) validated an adapted form of the APS for children from 4 to 5 years, the Affect in Play Scale-Preschool version (APS-P). Differences between APS and APS-P concerns play material- stuffed toys versus puppets- and instructions, in order to motivate and engage younger children in the play session. Moreover, there are slightly differences also in the coding procedure to better adapt the coding with preschool children way of play (Delvecchio et al., 2016b). Also for APS-P, interrater reliability, internal consistency, as well as internal and external validity resulted robust (Kaugars and Russ, 2009; Fehr and Russ, 2014; Marcelo and Yates, 2014). Fehr and Russ (2014) reported also an exploratory factor analysis on APS-P that included two correlated factors. Delvecchio et al. (2016b) evaluated APS-P external and construct validity in Italian preschoolers. A two-factor model CFA approach led to two correlated factors: a cognitive and an affective one.

Although no theoretical issues might lead to the hypothesis that toys fit better for preschoolers than for school-age children, and vice-versa, the two version of the scale maintained initially two separated pathways.

Later, APS and APS-P were preliminary compared in the Italian context with the simultaneous administration of the two versions of the scale group of children aged 4-to-10 years (Mazzeschi et al., 2008). No significant differences between the two version of the play task across ages were found, excepting for Comfort: both preschool and school-age children seemed more comfortable with toys than with puppets. Authors suggested that maybe in the Italian culture children are not so familiar with puppets and, moreover, that plastic and stuffed toys might elicit a wider range of emotional expressions (Mazzeschi et al., 2008). Following these preliminary ideas, the extension of APS-P, defined as APS-P Extended Version was firstly validated in a large sample of 6-to-10 years old children and later a study using APS-P and APS-P Extended Version was run to assess pretend to play in preschool and school Italian children aged 4–10 (Delvecchio et al., 2016a,c). Findings of the two studies confirmed the good psychometric properties of this extension and the adequacy of the original APS-P structural model across the ages. The opportunity to have the same set for play session across the ages seems to give a good chance to monitor developmental changes in children assessment and psychotherapy (Cordiano et al., 2008). Moreover, in real life, school-age children use to play with toys similar to the ones proposed in the APS-P task.

Assessment of cognitive and affective abilities in pretend play in preschool and school years should be seen as beneficial for researchers and clinicians (Bergen, 2002; Pellegrini, 2010). The use of APS and APS-P, in its different versions, allowed to include tools devised on an evidence-based approach in the assessment field, since that they are standardized and validated measures of pretend play (Kazdin, 2005; Mash and Hunsley, 2005). Recent attention has been directed toward the development of Brief versions of the APS [Affect in Play Scale–Brief Rating (APS–BR); Cordiano et al., 2008] and the APS-P [Affect in Play Scale–Preschool–Brief Rating (APS-P-BR); Pearson et al., 2008; Fehr and Russ, 2014]. These measures would support and encourage to use play assessment in a wider variety of research and clinical areas because they are less invasive for patients, not requiring videotaping but in vivo coding and are easier to use for experts since they do not involve extensive scoring training (Cordiano et al., 2008).

The APS-BR and APS-P-BR and its Extended Version continue to involve the observation of a 5-min standardized play task, with puppets or toys to be used respectively with school and preschool children. Instructions and task prompts are the same of the original measures, except that the instructions for the brief versions do not mention videotaping. Some studies explored psychometric properties, in terms of reliability and validity, of APS-BR (Cordiano et al., 2008) and APS-P-BR and its Extended Version (Fehr and Russ, 2014), in two small samples, comparing the relationship between standard and brief version using both videotaped and in vivo coding. The authors of both studies concluded that the APS-BR and the APS-P-BR and its Extended Version strong correlations with the original version and with external criterion demonstrated that they are worthy measures and might be used in clinical settings in combination or in substitution to the original APS and APS-P. However, the use of two different versions for preschool and school children does not allow the continuity of measure of play across age.

The first aim of this paper was to fill this gap by validating the brief version of the APS-P and its Extended Version in an Italian sample aged 4–10 years. The second aim of this paper was to compare APS-P and its Extended Version, and APS-P-BR and its Extended Version to verify if the brief form would easily substitute the original one while maintaining the integrity of the original scale. Finally, another aim was to contribute to the external validity of the APS-P-BR and its Extended Version by relating the scale with a measure of divergent thinking.

As a preliminary result, interrater reliability was expected to be excellent, since it always resulted so in previous studies on different versions of APS-P in Italy (Delvecchio et al., 2016a,c) and APS-P-BR (Fehr and Russ, 2014). In Fehr and Russ (2014) study on the APS-P-BR and its Extended Version, just an exploratory factor analysis (EFA) was carried out yielding a two correlated factors structure: the first one included organization, imagination, and positive affect and the second one was characterized by undefined affect, comfort, and negative affect. For this reason, the first purpose was to assess the structural validity of APS-P-BR and APS-P-BR Extended Version through CFA, comparing two different models, the classical APS theoretical model (Model 1; Cordiano et al., 2008) and the one proposed by Fehr and Russ (2014) (Model 2). In Model 1, the four cognitive scores are included in the cognitive factor; frequency and positive/negative tone of affective themes load on the affective one; in Model 2 the first factor include organization, imagination, and positive affect and the second one is characterized by undefined affect, comfort, and negative affect. Furthermore, the structural invariance across age groups was investigated by mean of the multiple confirmatory factor analysis (MCFA) approach. Gender as well as age effects were explored. According to previous studies on the different versions of the APS-P, no significant gender differences were expected, however, significant differences were expected between preschool and school age children (Kaugars and Russ, 2009; Delvecchio et al., 2016a,c). It was expected that school-age children score higher both in cognitive and affective variables, demonstrating a higher ability to elaborate and organize their stories and to express a wider range and frequency of affects (e.g., Denham et al., 1994; Stagnitti et al., 2007). The last purpose was to evaluate APS-P-BR and APS-P-BR Extended Version external validity with APS-P and its Extended Version and a measure of divergent thinking, as already used in the original studies. Consistent correlations between corresponding factor scores on the APS-P and its Extended version and APS–P–BR and its Extended Version were expected (Cordiano et al., 2008; Fehr and Russ, 2014). Correlations with divergent thinking were expected significant for both the cognitive and affective factors, although with a medium to low effect size (Russ and Schafer, 2006; Cordiano et al., 2008; Delvecchio et al., 2016a,c).

Inter-rater reliability was assessed on 40 randomly selected videos of APS-P and its Extended Version scored with the APS-P-BR and its Extended Version scoring system. Ten independent raters (Ph.D. level and graduate students in clinical psychology), one of which trained by the author of the scale, rated the 40 protocols independently. The average ICCs showed adequate to excellent range: organization: 0.82–0.88, elaboration: 0.86–0.90, imagination: 0.81–0.86, comfort: 0.88–0.94, frequency of affects: 0.84–0.90, and positive/negative tone of affect expression: 0.86–0.89. The remaining protocols were coded independently by the raters (about 50 protocols each).

Means and standard deviations for the total sample, preschool children (4 and 5 years old), school-aged children (from 6 to 10 years old) as well as boys and girls are reported in Table 1.

Fit indices comparing the classical APS-P theoretical model (Model 1; Cordiano et al., 2008) and the model proposed by Fehr and Russ (2014) (Model 2) are displayed in Table 2. Model 1 included the four cognitive scores loading in the cognitive factor and frequency and positive/negative tone of affective themes loading in the affective one; whereas Model 2 included organization, imagination, and positive affect in the first factor and undefined affect, comfort, and negative affect in the second.

The parsimonious index was lower for Model 1 suggesting that the theory-driven model fits the data more adequately than the Model 2. Figure 1 displayed the factor loadings of Model 1; they were all significant and comprised between 0.67 and 0.80. The inter-factor correlation between the cognitive and affective factors was 0.82.

Table 3 reports the fit indexes for both preschool and school-aged children as well as the results of the MCFA. Fit indexes referring to the subsample of school-age children showed an excellent fit, as well as for preschool children with the exception of RMSEA that showed slightly higher value than expected. Configural model (M1) indices displayed adequate fit. Looking at the metric invariance (M2), ΔCFI was equal to 0.01 and BIC was lower in comparison to M1 (Table 3). Testing for scalar invariance (M3), ΔCFI showed greater value than the suggested cut off and BIC resulted higher than for M2 (Table 3). Moreover, M3 showed a poor fit. These findings advocated for a weak measurement invariance of the hypothesized model across preschool and school-aged children. Thus, following Van de Schoot et al. (2012) recommendation, it is suggested to proceed testing for partial invariance by freeing one item at the time. As proposed by Dimitrov’s (2010), it is acceptable freeing less than 20% of parameters, which correspond to less than 2 variables for APS-P-BR and its Extended Version. Examination of the modification indices highlighted item 1 (“organization”) as the most problematic, suggesting to free its intercepts across groups in order to reach a better fit for the model. Therefore, the partial scalar invariance model (M3a) was run. It showed generally an adequate fit. Results displayed evidence of partial invariance between the groups showing values of ΔCFI below the cut off and a lower BIC value (Table 3) between M3a and M2.

The analysis of variance was run to assess age and gender differences. Mean levels of the APS-P-BR and APS-P-BR Extended Version factor scores, for the total sample, preschool and school-aged children, as well as boys and girls, are displayed in Table 4.

Results showed that school-aged children scored higher than preschoolers in the cognitive factor (F_(1,534) = 71.29, p < 0.001, ηp2 = 0.118) as well as in the affective factor (F_(1,534) = 8.06, p < 0.001, ηp2 = 0.015). No gender differences [cognitive factor: (F_(1,534) = 1.83, p = 0.176, ηp2 = 0.003); affective factor: (F_(1,534) = 0.08, p = 0.774, ηp2 = 0.000)] as well as no interaction effects [cognitive factor: (F_(1,534) = 0.06, p = 0.800, ηp2 = 0.000); affective factor: (F_(1,534) = 0.93, p = 0.336, ηp2 = 0.002)] between gender and age were found.

Correlations between APS-P-BR and APS-P-BR Extended Version factor scores and the factor scores of the divergent thinking measure were all significant (Table 5). However, different patterns emerged. APS-P-BR and its Extended Version cognitive factor showed higher positive correlations with both the cognitive and affective factors of APS-P and its Extended Version, whereas the affective one displayed a medium correlation with the cognitive APS-P and its Extended Version factor and a lower value for the affective one. In line with what expected, fluency and flexibility (divergent thinking) showed significant positive correlations with both cognitive and affective domains of play.

This paper evaluated the construct and external validity of this APS-P-BR and its Extended Version through age, in a wide sample of preschool and school age Italian children. The coding system of the brief version of the APS-P was applied to the videotapes of children play sessions recruited in the Delvecchio et al. (2016a) study, in which APS-P and its Extended Version standard procedure was administered. CFAs results supported the best data fit for the theoretical two-correlated-factor model with one factor related to cognitive dimension and one factor to affective dimension (Cordiano et al., 2008; Kaugars and Russ, 2009). Previous papers focusing on the original form of the APS-P and its Extended Version in Italian children confirmed empirically the same structure, both for preschooler and school-age children (e.g., Delvecchio et al., 2016a,c). On the same direction, MCFA showed some evidence of measurement consistency across preschool and school-aged children; metric invariance was partial and obtained after freeing organization variable. Although most of the fit indices confirmed this adequacy, some others did not, advocating for further studies on structural validity and metric invariance of APS-P-BR and its Extended Version. Moreover, though Fehr and Russ (2014), run an EFA on APS-P-BR in USA, as far as we know, this was the first attempt to conduct a CFA on the APS-P-BR and its Extended Version. As for age, significant differences were found for both cognitive and affective factors. As expected, also for the APS-P-BR and its Extended Version school-age children reported higher scores then preschooler (Kaugars and Russ, 2009; Delvecchio et al., 2016a,c). This finding is in line with previous literature on play, which underlined how older children showed a more sophisticated way of play, in terms of their abilities to structure and organize their stories and to comprehend and express a wider range of affective themes (e.g., Stagnitti et al., 2007). According to gender, no significant differences were found for the cognitive and the affective factors. These findings were in line with expectations and with Kaugars and Russ’ (2009) and Delvecchio et al. (2016a) studies in which no gender differences emerged.

As expected, significant correlations between cognitive and affective factors of the brief version of the APS-P and its Extended Version and the correspondent factors of the original version were found (Cordiano et al., 2008; Fehr and Russ, 2014). However, while observing correlations patterns in detail, Cohen’s d was high for the one between the two cognitive factors, instead it was low for the correlation between the affective factor of the brief version and of the original one. This data suggests that the coding systems of APS-P-BR and its Extended Version and APS-P and its Extended Version for cognitive variables are substantially able to detect the same aspect of play, maybe because the scoring of the four cognitive variables consists of a global score. In this case, the possibility to rewind the tape in the APS-P and its Extended Version does not add information about the cognitive scores that can be easily detected also in the in vivo coding. On the other hand, the affective variables could be affected by the possibility to watch the tape more than one time. The affective variables score is based on counting each word, expression or object in the story conveying affective themes, and during the in vivo coding some of them can be lost. Moreover, it is interesting to note that the APS-P-BR and its Extended Version cognitive factor correlates with high Cohen’s d with both cognitive and affective factors of the APS-P original version and its Extended Version, showing also a correlation between cognitive factor of the brief version and affective factors of both brief and original version of APS-P and their Extended Version. On the other hand, APS-P-BR and its Extended Version affective factor showed medium correlations with both cognitive and affective factors of the APS-P original version and its Extended Version, confirming that the affective factor has a quite different trend in the brief version, for the reasons hypothesized before. Significant positive correlations between APS-P-BR and its Extended Version cognitive and affective factors and divergent thinking were found. Children with higher scores in cognitive and affective components of play reported higher score in both Fluency and Flexibility. The same pattern was found in studies involving the APS-P original version and its Extended Version, suggesting that both cognitive and affective dimension of pretend play are associated with improved divergent thinking skills (e.g., Russ and Schafer, 2006; Marcelo and Yates, 2014; Delvecchio et al., 2016a).

This paper contributes to the validation of APS-P-BR and its Extended Version, a measure that would encourage the use play assessment in a wider variety of research and clinical areas, not requiring videotaping. This aspect makes the measure less invasive for patients, also increasing their confidentiality. Results suggest that APS-P-BR and its Extended Version through age, is a reliable and valid measure in terms of construct and external validity, retracing the main results obtained also in the APS-P and its Extended Version (e.g., Kaugars and Russ, 2009; Delvecchio et al., 2016a,c).

The APS-P-BR and its Extended Version do not completely overlap the original form, especially for the affective dimension. As discussed before, the affective factor of the brief version did not exactly retrace the one of the original version. So, on the one hand, the APS-P-BR and its Extended Version could be easily used in clinical settings, not requiring videotaping, on the other hand, APS-P and its Extended Version could be more suitable to have a complete affective profile of the play session. For sure the APS-P-BR and its Extended Version is a promising tool for assessing children pretend play that would easily substitute or, in some cases, used together with the APS-P and its Extended Version through age.

The present study showed also some limitations. First, APS-P-BR is not administered directly to children, but its scoring system was applied to tapes of children play sessions following APS-P and its Extended Version standard procedure. Future studies need to be carried out using APS-P-BR and its Extended Version procedure and without videotaping. Then, raters who scored the tapes with the brief version procedure were already familiar with the APS-P and its Extended Version coding system. This aspect could have affected their way of scoring in terms of make the two coding processes more similar. Then, although the study involved a large sample of children from 4 to 10 years old, generalizability is limited to non-clinical sample and to Italian culture, stating that culture could affect children way of play (Chessa et al., 2012). Future research should involve different kinds of clinical and cross-cultural samples. Finally, APS-P- BR and its Extended Version affective factor would need specific attention. Training should deepen the scoring system of affective variables of the brief version because they seem to require a moderate amount of time to be practiced. Expertise in play session scoring resulted strictly connected with an increasing availability of the raters and accuracy in detecting cognitive, but primarily, affective variables. Future development of the APS-P-BR and its Extended Version scoring procedure could include an affective theme checklist the raters could use during the in vivo administration to be guided in detecting the highest numbers of affective themes.

Comitato Etico Area 17, Università degli Studi di Padova. As explained in the manuscript, in these studies videotapes of already recruited children play sessions were coded. Following, I reported the consent procedure used for the original recruitment. This study was conducted in compliance with the ethical standards for research outlined in the Ethical Principles of Psychologists and Code of Conduct (American Psychological Association, 2010). Participation in the study was solicited via leaflets. School approval and parents written signed informed consent to participate in the study were obtained before data collection. Children were asked to provide their own oral consent. No incentives were awarded and voluntary participation was emphasized. Administration was proposed during scheduled classes, according to the standard administration procedures. Confidentiality was assured by replacing children’s personal information with a numeric code. According to their teachers, all children were developing typically. Before starting the task, preschoolers’ cognitive and verbal abilities were assured by the WPPSIIII (Wechsler, 2008) vocabulary and bloc design subtests. A score equal or greater than was required in order to participate in the study. At the same time, for school -aged children, cognitive and verbal abilities were confirmed by teachers’ reports on the basis of Italian language and maths tests.

DDR supervised the rater who coded the videotapes. She wrote the theoretical introduction and the discussion of the results. SS supervised the objective/hypothesis and procedure sessions. AL supervised the sample recruitment and the discussion of the results. ED supervised the sample recruitment and carried out statistical analysis.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.