Triadic Interaction among Newly Acquainted 2-year-olds

Two-move Sequences. Parke et al. (1979) focused on the concept of influence patterns within families; in particular, they demonstrated how a parent influences the other parent and their child. In a dyadic interaction, there are only two possible patterns of influence: A can influence B, or B can influence A. However, the situation becomes much more complex if a third person enters into the interaction (Table 1). In the system developed by Parke et al. (1979), the primary influence source is the person who initiates the first move in the interaction (e.g., ‘A’ might be a father, as depicted in the example provided by Parke et al. (1979), which is presented in Table 1). The primary recipient is the member of the triad who is the recipient of that behaviour (‘B’ who might be the mother or ‘C’ who might be the infant). The secondary influence source initiates the second move (in Parke et al.'s (1979) example, this might be the mother, B, or the infant, C). Finally, the secondary recipient is the recipient of the second move.

Table 1. . Patterns of Social Influence in Triads (from Parke et al., 1979)

Within this model, different types of influence patterns were identified. Direct interaction occurs when A initiates an action towards B, who then returns some behaviour to A; in other words, interaction takes place within a dyad, and the third member of the group is not involved. For clarity's sake, we refer to this pattern as dyadic throughout this article. Other influence patterns involve all members of the triad. Triadic patterns may be transitive, for example, where A initiates action towards B, who then directs action towards C. Alternatively, they may be parallel, for example, where A initiates action towards B, and then C also directs some behaviour towards the same recipient, B; or circular, for example, where A initiates an action towards B, and C then directs a behaviour back to the primary initiator A.

In the present study, we have recorded interaction amongst three peers, using an observational coding system that comprises specific actions and reactions shown by any member of the group, in response to any other; the stream of actions and reactions can be divided into alternating moves within episodes of interaction (Hay et al., 1991). Thus, a two-move sequence can be any permutation of two choices from the set of six possible combinations of initiator and recipient {A→B, B→C, C→A, B→A, A→C, C→B}. It is instructive to place the set of four interaction patterns introduced by Parke et al. (1979) on to a 6 × 6 transition matrix (Table 2).

Each of the four Parke et al. (1979) patterns corresponds to six cells of the matrix, accounting for 24 of the 36 cells altogether. It is then immediately obvious that the remaining 12 of the 36 cells in the matrix in Table 2 are unaccounted for. It is possible to complete the matrix by adding two additional sequential patterns, each of which accounts for six of the remaining 12 cells. These are a repetitive pattern where, for example, A directed a move to B and then did so again, and a new recipient pattern where, for example, A directed a move to B and then followed that with a move directed to C. In other words, in the context of a triad, a particular individual can either repeat overtures to one recipient or switch attention to another recipient. The repetitive pattern might be seen as a triadic overture but, because neither recipient responds, it does not immediately result in social interaction.

The extended set of six patterns of dyadic, transitive, circular, parallel, repetitive, and new recipient covers all possible two-move sequences. Thus, in addition to being mutually exclusive, the extended set of six patterns is exhaustive as well. The two additional two-move patterns do not by themselves constitute interaction, but, in the context of a longer episode of interaction, they may have social meaning.

Three-move Sequences. Parke et al.'s (1979) coding scheme categorizes sequences of two consecutive moves. However, it is important to examine whether 2-year-olds can sustain longer sequences of triadic interaction. Thus, we have extended this analysis to examine sequences of three consecutive moves. Particularly salient to a study of triadic interaction is the fact that some of the possible three-move sequences involve each of the three children initiating a move. This means that their involvement in the triadic interaction extends beyond merely being the passive recipient of another child's actions. A three-move sequence can be any permutation of three choices from the set of six possible combinations of initiator and recipient {A→B, B→C, C→A, B→A, A→C, C→B}, giving 6³ or 216 possible permutations altogether.

Each cell of the resulting transition matrix represents a unique pattern for three sequences of three consecutive moves, so there are 36 three-move patterns altogether. It is clearly not sensible to give individual names to each of 36 three-move patterns. However, it is possible to assign meaningful names to the patterns grouped into categories, based on the total number of children involved in the three-move interaction sequence, the number of children who are initiators of moves, and the number of children who are recipients of moves. The 36 three-move patterns can be grouped into seven categories: monadic, dyadic, triadic^2–1, triadic^2–2, triadic^2–3, triadic^3–2, and triadic^3–3. The five triadic categories have names of the form triadic^i–r, where ‘i’ is the number of initiators (2 or 3) and ‘r’ is the number of recipients (1, 2, or 3). Table 3 summarizes the seven categories of monadic, dyadic, triadic^2–1, triadic^2–2, triadic^2–3, triadic^3–2, and triadic^3–3 sequences.

The first category, monadic, groups together the three-move patterns in which only one child attempts to initiate interaction. There are four three-move patterns in the monadic category, being the four possible two consecutive two-move pattern permutations of the two new patterns, repetitive, and new recipient. This is not coincidence; as with the repetitive and new recipient patterns of two-move sequences, one would only expect the monadic pattern to occur in the context of a longer interaction episode. The second category, dyadic, comprises those three-move patterns in which two children interact, and each child initiates at least one of the three moves. The remaining five categories, triadic, cover patterns in which all three children interact, and at least two children initiate a move. (The case of three children being present, but only one child initiating behaviour, is categorized as monadic, as it does not qualify as interaction.)

Distinguishing five triadic categories is a useful theoretical exercise, but for estimating the amount of triadic interaction engaged in by 2-year-olds, a further simplification is helpful. If one discards the number of recipients, then one can cut the number of triadic categories from five to two. Triadic^2–1, triadic^2–2, and triadic^2–3 merge to form triadic², comprising those three-move patterns in which two children initiate interactions, while the third child is involved only as a recipient. Triadic^3–2 and triadic^3–3 merge to form triadic³, comprising those three-move patterns in which all three children interact, and each child initiates one of the three moves. In summary, using the simplified set of four categories, possible three-move sequences can be characterized as monadic, dyadic, triadic with two initiators, and triadic with all three children initiating behaviour.

In sum, we seek to determine whether 2-year-old children engage in triadic interaction with newly acquainted peers. We ask whether the patterns of dyadic and triadic influence that are shown in family groups also occur amongst young peers.

Two-move Sequences. The model of triadic influence used by Parke et al. (1979), including the additional categories of repetitive overtures and overtures to a new recipient, was applied to two-move sequences (see Table 5). Again, this was done first with respect to minimal interaction, where a third actor might participate by watching one of the others, and then for active interaction, where each actor used movements, gestures, or vocalizations.

Most of the minimal sequences (84 per cent) were dyadic, where a particular recipient responds to the original initiator (Figure 1a). Repetitive overtures by the initiator occurred in 3.1 per cent of the sequences, whereas switching to a new recipient occurred in only .6 per cent. Triadic interaction was observed in 12.4 per cent of the two-move sequences, with transitive influence the most common pattern and parallel influence the least common, χ²₍₂₎ =23.81, p < .01.

A slightly greater proportion of two-move sequences (13.7 per cent) were triadic when only active moves were considered (Figure 1b). Once again, transitive influence was most common and parallel influence least common, χ²₍₂₎=14.49, p < .01.

Three-move Sequences. Combinations of two-move sequences were then examined to examine possible triadic influence in sequences of three active moves (see Table 5). Again, most of the sequences of active moves were dyadic (77.4 per cent); monadic sequences occurred rarely (Figure 1c). The two different types of triadic sequences (those where all three 2-year-old children initiated action and those where two actors were initiators but the other only a recipient) were about equally likely to occur, 50.2 per cent involving two initiators and one recipient, and 49.8 per cent involving all three children as initiators. Taken together, over one-fifth of three-move sequences (22.6 per cent) were actively triadic.

The Coding System. The primary coder transcribed the video records using the PICS, slightly modified from the version used in Experiment 1 (Hay et al., 1999). Moves and episodes were identified as in Experiment 1. The video records were transcribed on to coding sheets. Each move was transcribed on to a single row of the coding sheet. The start time for the move was entered into the first column labelled ‘Onset Time’. The move number within the episode was entered into the second column labelled ‘Move’.

In the version of the coding system that was used in Experiment 1, moves could have been coded as occurring at the same time as other moves. This occurred rarely; only 3 per cent of moves were recorded as occurring simultaneously with other moves. An additional 3.7 per cent of moves were recorded as having two or more recipients. These two forms of co-occurrence complicate the use of Parke et al.'s (1979) method for analysing two- and three-move sequences. A critic might argue that the co-occurrence of moves and the possibility of moves being directed to more than one recipient might artificially overestimate the amount of triadic interaction that actually occurred. As a result, in Experiment 2, we decided to disallow co-occurrence of initiators and recipients. Thus, if two children were acting apparently simultaneously, the observer had to judge who began to act first, and then code a separate move for each child. Separate columns were used to record acts made by each child. The recipients of moves made by each of the three children were recorded in separate columns.

The first author coded the video records of all observational sessions. Two independent observers coded 50 per cent of the sessions. On average, the observers recorded episodes of peer interaction with 99 per cent agreement. They identified the initiator and recipient of particular moves with 81 per cent agreement. Agreement on the specific content of agreed-upon moves was κ=.65, p < .01.

Categorization of Episodes. Active and minimal moves were defined as in Experiment 1. Once again, an episode was considered to be minimally triadic if each member of the triad served as an actor in at least one move within the episode and actively triadic if each member of the triad served as an actor in at least one active move. Independent observers achieved good agreement in recording the minimal move of watching the peer, κ=.81, p < .01.

Episodes were judged to contain conflict when one child objected to something another child was doing, that is, if one child's behaviour had been protested, resisted, or retaliated against by at least one peer (Hay & Ross, 1982). Specific behavioural categories qualified as forms of protest, resistance, or retaliation: vocal and verbal protest; resistance, by moving away, or withdrawing oneself or an object away from the peer's grasp; attempts to regain a disputed object by reaching, contacting or tugging on it; and forceful physical contact against the peer (see Table 4). Moves containing these behavioural elements were designated as conflict moves; and, so, episodes could be classified as being comprised wholly or partially of conflict moves. Two independent coders agreed on the presence of conflict, κ=.91, p < .01.

Two-move Sequences. The different possible sequences of two moves are presented in 2, 2 for minimal and active interaction, respectively. As in Experiment 1, most sequences of moves were dyadic, with one child directing a move to one of the peers who returned a move to the original child.

Nearly one-third of the minimal interaction sequences (31.4 per cent) showed triadic patterns of influence (a qualitative example of a triadic interaction is presented in Table 6). In contrast to Experiment 1, parallel influence was the most common pattern, transitive influence the least, χ²₍₂₎ =15.84, p < .01. Active interaction episodes, on the other hand, contained fewer triadic sequences (15.4 per cent). No pattern of triadic influence occurred significantly more often than any other.

Three-move Sequences. Because the analysis of three-move sequences focuses on the roles of the third actor, only active interaction is considered (Figure 2c). (The pie chart consolidates the monadic and dyadic categories insofar as only 0.6 per cent of sequences were monadic.) The score of 15.5 per cent for triadic³ provides further evidence that 2-year-olds are capable of triadic interaction in which all three toddlers play an active part. Indeed, most of the triadic sequences (60 per cent) involved all three children as initiators.

Two-move Sequences. The proportions of conflict and non-conflict two-move sequences are presented in Table 7. In contrast to McGrew's (1972) findings, triadic interaction was more likely to occur when children were not in conflict. The proportion of conflict sequences that were dyadic was greater than in non-conflict sequences, z=4.83, p < .01. In particular, the parallel pattern was significantly more likely to occur outside of conflict, z=2.63, p < .01.

Three-move Sequences. At the level of three-move sequences, triadic interaction was similarly more likely to occur out of conflict (Table 8). Of the sequences in conflict episodes, 79.7 per cent were monadic or dyadic; whereas, outside of conflict, only 69.2 per cent were monadic or dyadic, z=2.22, p < .05.

Two-move Sequences. When minimally triadic moves were considered, groups of girls and groups of boys did not differ with respect to the extent to which each influence pattern was shown. However, with respect to actively triadic sequences, the circular pattern was observed in only 2.7 per cent of sequences in girl triads, but in 7.3 per cent in boy triads, a significant difference, z=−2.00, p < .05.

Three-move Sequences. At the level of three-move sequences, the girl and boy triads show similar proportions of monadic, dyadic, and triadic³ sequences. However, for the boy triads, 12.4 per cent of sequences involved the triadic² pattern, as opposed to 4.7 per cent of three-move sequences in girls’ triads, z=−2.01, p < .05.

Conflict and Non-conflict Moves. Within the context of conflict, the girls’ groups were more likely to show the repetitive pattern, in which one toddler makes repeated overtures to a particular peer. The girls showed this pattern in 7.8 per cent of the two-move sequences, whereas the boy triads showed this pattern in only 1.5 per cent, z=2.43, p < .05. No other significant gender differences were observed when conflict and non-conflict episodes were examined separately.