Existing research has focused less on the component actions of play and more on children's approach to the activity, the dispositional characteristics of play. Krasnor and Pepler (1980) characterized play according to both behavioral criteria Flexibility, Positive Affect, and Nonliterality, and a motivational criterion Intrinsic Motivation and noted that the more of the four criteria that were present, the more likely it was that an activity would be defined as play. Rubin, Fein, and Vandenberg (1983) also proposed a multifaceted approach to defining play in terms of (a) the inferred disposition of the player, (b) the actions that make up the activity, and (c) the context in which the behavior takes place. The criteria of Rubin et al. refer to both behavioral features, Nonliterality and Positive affect, and motivational features, Intrinsic motivation, Voluntariness, and Process-prevails-over-product. Consistent with this latter criterion, Pellegrini and Smith (1998) also remarked upon the apparent purposelessness of play.

In one of the few studies that has attempted to assess the definitional utility of existing play criteria, Smith and Vollstedt (1985) had both experienced participants (e.g., trainee teachers) and participants inexperienced in observing children's play, rate video excerpts of preschool children's behavior as either play or not play. Ratings were made against five criteria, four from Krasnor and Pepler's (1980) definition, Flexibility, Positive Affect, Nonliterality and Intrinsic motivation plus the criterion of Rubin et al. (1983), Process-prevails-over-product, labeled Means/Ends by Smith and Vollstedt. Results demonstrated the utility of Krasnor and Pepler's Nonliterality, Positive Affect, and Flexibility criteria and Rubin et al.'s Means/Ends criterion to distinguish play from nonplay but not Krasnor and Pepler's Intrinsic Motivation criterion. When multiple criteria were attributed to a play episode, it was designated "play" more consistently thereby supporting Krasnor and Pepler's additive hypothesis, and pointing to the multidimensional nature of play. These findings, however, offer little understanding of what criteria participants themselves might identify in the absence of criteria against which they could rate video excerpts. Additionally, all play excerpts in Smith and Vollstedt's (1985) study were of young children (3-4 yrs) in preschool settings. There is a need still to present participants with examples of typical play of children of different ages and across different play contexts.

Other approaches to categorizing children's play involve the identification of different subtypes of play. Pellegrini and Smith (1998) identified two subtypes: firstly, Activity play, more common in boys, and exemplified by games during middle childhood, and rough-and-tumble play (R&T) in middle childhood and early adolescence, and secondly, Pretend play, more common in girls and occurring most frequently between 3-5 years. Preschoolers engage in nonliteral behavior, generating pretend roles and scenarios, either alone (solitary pretence) or in mutual pretence, occurring in groups of two or more children. Classifications of subtypes of play imply (a) common play criteria, and (b) additional behavioral criteria unique to each play type and demonstrate both age and sex differences in different play types.

The present study was aimed at identifying criteria by which knowledgeable adults categorize children's play. Apart from the study by Costabile et al. (1991), this research differs from previous research, because adult participants' own responses to video examples of different forms of 5-12-year-old children's play were used as the basis to develop criteria of acceptable interrater agreement. Criteria thus developed will be used in future research as a basis for comparison with children's and other adults' responses to video excerpts of play.

The following questions were addressed in the present research.
1. When asked to characterize children's play according to their own criteria, do adults knowledgeable about children's behavior use the same criteria as those identified in existing categorizations of play?
2. What aspects of the observed behavior do they attend to when they report the criteria upon which they themselves categorize the children's play?

Overall criteria κ = .69; Behavioral criteria, κ = .64; Motivational criteria, κ = .61 and Contextual criterion, κ = .82. According to Fleiss (1980) there is good interrater agreement for Behavioral and Motivational categories, where κ >. 60, and excellent interrater agreement for Contextual criterion, where κ > .75.

Play criteria derived from the participants' own responses to the video episodes were mainly consistent with findings from previous research. Behavioral criteria of Positive Affect and Nonliterality are similar to criteria proposed by Krasnor and Pepler (1980) and Rubin et al. (1983). The Behavioral criterion of Unrestrained/Unrestricted is comparable with the Process-prevails-over-product criterion in the categorization of Rubin et al. and Smith and Vollstedt's (1985) Means/Ends criterion. These findings reconfirm the reliability of all three criteria to categorize play. Methodological differences notwithstanding, results of the present study thus offer limited validity to 3/4 criteria, Positive Affect, Nonliterality and Means/Ends identified by Krasnor and Pepler and Smith and Vollstedt. Findings of the present study also support the contention of Rubin et al. that play is best categorized by attending to the disposition of the player. Participants in the present study referred frequently to the children's approach to the activity. For example, "got a happy look on their face" (Positive Affect/Episode 5) and "[the children were] only hitting softly" (Nonliterality/Episode 5). Consistent also with the findings of Rubin et al. and Pellegrini and Smith (1998), participants remarked upon the apparent purposelessness of activities. For example, "it's a bit kill-die, no obvious game" (Unrestrained/Unrestricted/ Episode 7). Communication was the most commonly attributed motive of the players, and it is noteworthy that participants attributed effective communication between players to children's apparent familiarity with each other. For example, "seem like friends, both know the song off by heart" (Episode 6). Howes et al. (1992) and Wachs (1985) have noted that familiarity and friendship are most likely to promote play. Children themselves also report that activities that occur between friends are more likely to be play (e.g. R&T) than not play (e.g. aggression) (Costabile et al., 1991). Practice was another motive attributed often to players in the video episodes, and this is consistent with Piaget's (1951) assertion that play is an assimilative behavior that allows children to consolidate existing mental representations (schemata). Cooperation, a motive attributed less frequently, has also been discussed as a potential developmental function of play by Pellegrini and Smith (1992), who consider whether or not shared pretence might aid in the development of social skillfulness and cohesion among children.

The presence of toys or props, exemplified by the response, "each child's got a different costume, all got different props, and these determine what each do with the character" (Contextual/Episode 7), was identified often in this study yet is not included in existing categorizations of play. Rubin et al. (1983) and Wachs (1985), however, have stressed the importance of the physical context, including toys and props, when categorizing an activity as play.

Participants identified different clusters of criteria when characterizing different play episodes and such categorization offers support for subtypes of play. For example, Episode 1, an example of mutual pretence was characterized by positive affect, nonliterality, lack of restraint, practicing, communicating, cooperating and using toys and props whereas Episode 5, an example of R&T play, was characterized by positive affect, nonliterality, lack of restraint, and cooperation.

Limitations of the present study include its small sample size and the fact that all video excerpts were of indoor group play. For future research, there is still a need to include even more diverse examples of children's play such as solitary play, outdoor play, and play without toys and props.