Family relations exert a unique influence on infant social and cognitive development (Feinberg, 2003; McHale, 2007; Favez et al., 2009). The concept of “family” has been used extensively in the developmental literature to encompass diverse social systems whose common denominator is to be “larger than dyads.” Several models of family functioning have been proposed, such as the Circumplex model (Olson et al., 1989), the family competence model (Beavers and Hampson, 1990), the McMaster model (Epstein et al., 2003), and the Security in the Interparental Subsystem model (Cummings and Davies, 2010). Most of these models are inspired by and in continuation of the structural approach of Minuchin (1974). These models highlight family-level dimensions such as cohesion (i.e., the emotional closeness between family members) or adaptability (i.e., the aptitude of the family system to respond adequately to internal and external demands); moreover, they allow one to conceptualize the interrelation between family dynamics and individual functioning. However, they were not specifically designed for the early family; an understanding of the relational dynamics of a system that includes an infant necessitates taking into account the particularities of the relations and interactions with a very young child (in particular, an emphasis on non-verbal behaviors, as the baby does not have language). The family alliance (FA) model has been conceptualized to this aim (Fivaz-Depeursinge and Corboz-Warnery, 1999). To understand the specificity of the model, one must be aware of its background: on the one hand in order to situate it in the continuation of studies that led to progressively enlarging the view of the significant relational context for the infant from the mother-infant dyad to multiperson systems (including the father and other close relatives such as siblings), and on the other hand in order to consider the interdependence between two levels of “relational reality,” that is, the individuals’ representations of the relational world and the behaviors implemented in the interactions with others (Stern, 1985; Reiss, 1989; Favez, 2010).

Historically, the emphasis of developmental studies was first put on the mother–infant relationship, the mother being in most cases the main caretaker in occidental societies. These studies have shown the interrelation between the mental state of the mother, her parenting behaviors, and the social and affective development of the child (Grolnick and Gurland, 2002). First, maternal representations of the relational world (in particular of herself as a mother, of the baby, and of herself in relation to her baby) influence the mother’s aptitude to be emotionally available, to understand the behaviors of her infant, and to respond appropriately to her infant’s needs (Benoit et al., 1997; Mayseless, 2006; Biringen and Easterbrooks, 2012). Maternal representations may be to a certain extent “distorted” as a consequence of her relational history, or depending on her personality or possible psychopathology; globally, distorted refers to a difficulty in the mother to form a representation of the baby as autonomous and different from her (Slade et al., 2005). Second, studies have shown that daily mother–infant interactions are highly patterned: visual and affective exchanges begin, for example, with a “salutation” phase, followed by a “dialog” phase with several bouts of engagement and pause, and end with a “termination” phase. Within this structure, mutual behaviors of the mother and the infant are contingent on one another (i.e., each behavior of a partner is an answer to the behavior of the other, and behaviors are produced within a proper time frame); moreover, the mother continuously adjusts her stimulations of the child to the child’s affective state. Finally, the interactions are organized around a theme such as a peek-a-boo game (Brazelton, 1974; Stern, 1977). Distorted representations may induce alterations in the structure of the interaction: a mother with depression tends, for example, to have a representation of the baby as not being interested in interacting with her; she may then enact overstimulating behaviors in order to force the interaction (i.e., by not respecting the necessary pauses in the interaction that the baby needs to take in order to avoid excessive arousal), or, in contrast, she may show understimulating behaviors, as she lacks the motivation to interact. As a consequence, the baby may withdraw quickly from the interaction by not looking at the mother, which will be understood by her as confirmation of the lack of interest of the baby. This dysfunctional pattern thus tends to repeat itself, which has an impact on the development of the infant’s emotional regulation skills: the infant may progressively generalize withdrawal to all social situations (Anders, 1989; Tissot et al., 2011). Disturbances in daily relationships were identified as the main mediating variable by which emotional difficulties or a psychopathology affecting the mother have an impact on the child (Goodman and Gotlib, 1999).

The father–child relationship has been studied in a second step, as the father was long considered to exert only a distal influence on the development of the infant (Frascarolo and Favez, 1999). However, the important social changes that Western societies went through at the end of the 1960s and the beginning of the 1970s (with, for example, the removal of “paternal power” from civil codes in countries such as France in 1970 or Switzerland in 1977) gave rise to a generation of “new fathers,” who began to claim responsibilities in the care of their children, even in the child’s early years. Following these changes, numerous studies have been dedicated to the development of the father–child relationship, replicating the protocols that have been used in studies on mother–child relationships. The results have shown similarities between mothers and fathers in the aptitude to understand the emotional signals of their babies (for example, both parents are equally sensitive to their child’s needs, and they are both able to adjust their stimulations to the affective state of the child); moreover, father–infant interactions are patterned similarly to mother–infant interactions, but with differences in the preferred themes of interactions: fathers are more prone to use physical play and to be unpredictable in their stimulations of the child (see Parke, 1996, and Lamb, 2010, for comprehensive reviews). Finally, father psychopathology also has a negative impact on the development of the child (Connell and Goodman, 2002; Ramchandani et al., 2008).

The results of these studies have thus shown that both parents are significant partners for the child, which brought to the fore the question of the coordination between mother and father concerning their respective beliefs and expectations about parenting, as well as their daily interactions with the infant. In this regard, empirical evidence has shown a true emergent effect when the family interacts as a triad. Parke and O’Leary (1976) have, for example, observed that the interactions between one parent and the infant are different when they happen in the presence or absence of the other parent; these researchers coined the term “second-order effect” to describe the influence of one relation on another: for example, both mothers and fathers show less negative emotion and are less involved in the interaction when they interact with the child in the presence of the other parent than when they are alone with the child (Johnson, 2001; Udry-Jørgensen et al., 2016). It thus became obvious that a comprehensive assessment of the social context of the development of the child could not be done by only taking into account each dyad separately. Stated differently, the dyadic systems (mother-infant and father-infant) cannot be added to one another; the triadic level – the way that interactions are organized when mother, father, and infant are together – is a level in its own right, offering a particular context in which the child lives a unique experience.

The acknowledgment of the importance of the triadic level also occurred as a result of studies that have shown the impact of the relationship between the parents on the child. Studies on interparental conflict have shown the psychological and behavioral negative consequences for school-aged children who are raised in a conflicted home (Cummings and Davies, 2010). Similarly, studies on the consequences of divorce have shown that an enduring post-divorce interparental conflict is the main variable that explains children’s difficulties in coping with the separation (Amato and Keith, 1991; Amato, 2001); in contrast, cooperation between the ex-spouses tends to temper the negative effects of divorce (Amato and Rezac, 1994; Amato et al., 1995; Hetherington and Stanley-Hagan, 2002; Adamsons and Pasley, 2006). In infancy, interparental conflict also affects children. First, there is a direct effect: repeatedly witnessing intense and unresolved conflicts between the parents has consequences on the emotional regulation skills of the infant. Atypical vagal tone has thus been observed that tends to be generalized to any stressful situation; the child may then have a tendency to social withdrawal when facing novelty (Crockenberg et al., 2007; Lindblom et al., 2014). Second, there is an indirect effect: parent–infant relationships are in themselves affected by the disturbances in the parent–parent relationship. This phenomenon was described as the “spillover effect” (Fainsilber Katz and Gottman, 1996; Crockenberg and Leerkes, 2003a): parents engaged in a conflict show more anger and irritation in their interactions with the infant, have less empathy and emotional availability, and display controlling and rejecting behaviors – all in all, behaviors that are similar to those described in parents with a psychopathology (Webster-Stratton and Hammond, 1999; Davies et al., 2002). The representations the parents have of the infant are also modified – or distorted; for example, parents describe the child as being more difficult than external informants do (Fainsilber Katz and Gottman, 1991, 1996).

The development of the child is thus affected by the cooperation and the conflict between the parents. This cooperation, or lack of it, has been conceptualized as the coparenting relationship – that is, the support parents bring to each other, or not, in their relation with the child at instrumental and emotional levels (McHale and Lindahl, 2011; Beaton et al., 2013, for a review; Minuchin, 1974). Cooperative coparenting comprises support and warmth between the parents; disagreements are resolved by negotiation (McHale, 1997). On the other hand, when there is an underlying and enduring conflict, parents may show hostility to each other: each parent may try to undermine what the other has done, or be critical and disparaging of the parenting work of the other. Another possible issue of conflict is skewed coparenting: one of the parents withdraws from family life; thus, there is no overt conflict, but no real coparental relationship either. Finally, the conflict may have as a consequence a lack of positive affects and emotional support, even though the two parents are still cooperating at an instrumental level (McHale et al., 2002). All of these variations in coparenting are non-optimal contexts for the development of the child; if they become chronic, they are linked with adaptive problems and even psychopathology in the child (McHale and Rasmussen, 1998; McHale and Fivaz-Depeursinge, 1999; Schoppe et al., 2001; Schoppe-Sullivan et al., 2009; LeRoy et al., 2012; see the meta-analysis by Teubert and Pinquart, 2010).

It is of note that the emphasis has been put on each parent and on their collaboration; the child should not, however, be considered merely a passive partner in the interactions, even though her/his contribution has often been overlooked (McHale and Crouter, 2003). First, studies have shown that the infant has very early (at as early as 3- to 6-months-old) the attentional and behavioral aptitudes to interact at a triangular level: on the one hand in person-person-object situation (PPO), by sharing attention toward an object with the adult or by sharing intention for action (Trevarthen and Hubley, 1978; Butterworth, 1998; Rochat and Striano, 1999; Striano et al., 2007; Moreno-Núñez et al., 2015; Reddy, 2015), and on the other hand in person-person-person situations (PPP), for example by visually following the exchange between the parents (Fivaz-Depeursinge et al., 2005, 2007), or by being engaged in the interaction with two peers at a time (Selby and Bradley, 2003). Interestingly, studies which have compared the two types of situations have shown that the PPP situations elicit a greater behavioral arousal in the infant (for example, more smiles or hand movements) than the PPO situations (Tremblay and Rovira, 2007). These triangular aptitudes are used by parents when they compete for the attention of the child in order to distract themselves from an ongoing conflict (Fivaz-Depeursinge and Favez, 2006), a process that may be the origin of the triangulation process described by Minuchin (1974) in families with older children, where children are used as a scapegoat or go-between to regulate the tensions between the parents. Second, the temperament of the child (in terms of disposition to regulate emotions) has an effect on the parenting behaviors of the parents and also on the interparental relationship. Parents of a so-called difficult child (slow to warm up, irritable, difficult to calm or fussy) develop a style of parenting that is all the more responsive during the first months, but in the end, their unsuccessful efforts to calm the child can lead them to show a parenting style that is colder and more distant (Crockenberg and Smith, 1982; van den Boom and Hoeksma, 1994; Sanson and Rothbart, 1995). Similarly, the probability of interparental conflict is higher (Stifter, 2003), especially when there is preexisting dissatisfaction in the marital relationship (Crockenberg and Leerkes, 2003b; McHale et al., 2004; Schoppe-Sullivan et al., 2007). In our own study on the transition to parenthood, we have seen that the child’s temperament at 3 months and the coparental interactions during the pregnancy are the two best predictors of the quality of the triadic interactions at 18 months (Favez et al., 2013).

Several findings of the studies we have reviewed may be highlighted, as they constitute building blocks for the FA model. First, studies have shown that every partner brings a significant contribution to the family interaction, including the infant. Moreover, even when one parent is not active in the interaction, the parent’s mere presence is sufficient to introduce modifications into the interaction between the other parent and the child. This points to the importance of considering the inclusion of each and every partner of the family in the interaction. Second, a successful interaction implies organization of the behaviors of the partners: there should be contingency between the behaviors, respect of turn taking, and no competitive behaviors – especially in the stimulations of the parents toward the child. Third, interactions have a theme: even at a non-verbal stage, a play has a narrative plot around which the behaviors are articulated. Finally, the affective or emotional tone of the exchanges are of paramount importance; studies have shown, for example, that coparental coordination that is only instrumental and devoid of affects does not constitute a functional coparenting relationship.

Based on these findings, the FA model has been designed to assess the degree of family engagement and coordination in any joint activities (Fivaz-Depeursinge and Corboz-Warnery, 1999). It allows one to describe relational dimensions specific to “beyond-the-dyad” systems and simultaneously takes into account all the members of the family who are taking part in an interaction. Besides the studies mentioned earlier, the model is inspired by the symbolic interactionism approach (Blumer, 1969) and by the ecosystemic model (Keeney, 1979), which put non-verbal communication and contextual information at the core of human interactions. According to the model, FA depends on four interactive functions: participation, organization, focalization, and affect sharing. These functions are hierarchically embedded; each is a necessary condition for the achievement of the next.

The functions are implemented in the interaction according to the spatial and temporal features of the partners’ behaviors. The distance between the partners and the orientation of their bodies form a transactional space that allows emotional exchanges (Hall, 1966; Argyle, 1972; Kendon, 1977). Moreover, the positions of the different body parts inform about the partner’s readiness to interact and determine the partner’s level of engagement in the interaction. Communication is clear and engagement is high when all the body parts are congruent (face, torso, and pelvis are all oriented toward the partner); incongruence between body parts results in ambiguity about the engagement (e.g., the face is oriented toward the partner, while the torso and pelvis are oriented away from the partner; Scheflen, 1973). Participation thus implies that all partners signal their readiness to interact by their body positions and orientation, especially the most basic level of the pelvises: a geometric figure could be drawn between the center points of each partner’s pelvis; the torsos are oriented toward this transactional space. Organization is most often indicated by the inclination of the torsos (e.g., the speaker leans forward while the listener is seated straight or leaning back on the chair; each change in roles is marked by a change in body position), by the gestures in congruence with the role (e.g., the speaker uses specific hand gestures), and by the contingency in the behaviors of the partners (e.g., the speakers hold the floor each in turn and not at the same time). Focalization is determined by joint visual attention (e.g., when all partners look at the same object, or when all listeners look at the speaker) and by the development of a common theme (e.g., a consistent topic through the verbal or non-verbal exchanges). Finally, affect sharing is determined by facial expressions, verbal exchanges, and vocal tone (e.g., the emotional expressions of all partners are attuned).

In order to allow the assessment of FA through systematic observation of interactions, we have developed several observational situations for the laboratory. The two main situations are the Lausanne Trilogue Play (LTP), specifically designed for triadic interactions and aimed at assessing how the family interacts in several predetermined configurations, and the PicNic Game (PNG), specifically designed for families with several children and aimed at assessing how the family organizes itself spontaneously. These situations are windows into family functioning; their basic premise is the fact that being able to resolve a small task (to follow the instructions and to cope with the stress of being observed) in a laboratory setting is representative of the aptitude of the family to solve more important tasks (Beavers and Hampson, 1990; Kerig, 2001).

The systematic assessment of the alliance is carried out with the help of validated assessment tools, such as the Family Alliance Assessment Scales (Favez et al., 2011) in the LTP and the Revised-PicNic Assessment Scales (Favez et al., 2016) in the PNG – both of these instruments include dimensions specific to their respective observational situations. This has allowed us to highlight, in the context of longitudinal studies with non-referred families, that the FA is stable in time, from the fifth month of pregnancy (alliance measured in the prenatal LTP) to the end of the child’s second year and until the firstborn child has reached the age of 5 years (alliance measured in the LFP and in the PNG). For example, coparenting dysfunctional processes such as competitive behaviors may be observed at every age in a given family: when the parents-to-be play with a doll during pregnancy, instances of competition may be shown when parents argue angrily about whom the doll is supposed to resemble the most; at 3 and 9 months, the parents might offer divergent stimulations to the baby in order to capture her/his attention; at 18 months and later, the competition may be verbal with mutual contradictions, contradictory instructions given to the child, and angry comments by the parents addressed to one another (Favez et al., 2006a, 2012; Frascarolo et al., 2016).

This stability suggests that triadic (and polyadic) interactions are patterned, just as dyadic interactions are. This confirms that the observation of interactive behaviors is a sound methodology to assess early family relationships; in addition, it allows us to include infants and young children for whom other assessment strategies are not relevant. Because they are patterned, interactions observed in the laboratory may be considered meaningful and to reflect real-life interactions of the family. From a developmental point of view, the stability of interactions implies that FA constitutes a stable relational context in which the child learns social skills; this relates to the concept of “practicing family” proposed by Reiss (1989), namely, the transmission of values and ways of behaving through the usual interactive patterns of the family. These contexts may be more or less favorable to the cognitive and emotional development of the child: data from our longitudinal studies show, for example, that children from families with a dysfunctional alliance are less efficient in theory of mind tasks and in the understanding of complex emotions at age 5 years. Interestingly, these families are also the families in which parents report less social support from their environment (Favez et al., 2006a,b, 2012).

Dysfunctional alliances are predominant in families referred for parental psychopathology (e.g., postpartum depression), and alliance assessment has a good discriminant value between referred and non-referred families (McHale and Fivaz-Depeursinge, 1999; Fivaz-Depeursinge et al., 2000). The hierarchical structure of the model has been empirically validated by observation of interactions in non-referred families (Frascarolo et al., 2004): when a function of higher order is fulfilled (e.g., focalization), the functions of lower orders are always fulfilled (participation and organization), but the reverse is not true (e.g., participation does not imply organization, focalization, or affect sharing). Clinical practice has shown that the more severe the psychopathology in parents (e.g., mothers with a diagnosis of schizophrenic disorder, or with a major depression), the more difficult it is for the family to fulfill even the lower order functions. Accordingly, interventions may target the dimensions, beginning with participation in families in which including everyone is difficult, and then organization if participation is fulfilled but attributing roles generates conflicts, and so on (Favez et al., 2009). Interventions may be performed by using the situations of observation in several ways: first by direct intervention during the situations themselves, based on the family’s interactive behaviors and by suggesting alternative and non-problematic patterns, or by using the information stemming from the situation to prescribe rituals to be carried out at home between two sessions. Second, the films allow the use of video feedback, which is a powerful way to enhance family experience of relationships by supporting awareness of positive and negative interactive patterns (McDonough, 1993; Fivaz-Depeursinge et al., 2004).

Developmental research findings have shown the importance of the family level for a comprehensive understanding of the child’s development. FA is a model designed to conceptualize family interactions in the early family, including one (or more) infants or young children; based on observable interactions, the qualification of family relationships in terms of alliances according to four interactive functions is an effective and valid way of making a diagnosis of family functioning. Two main situations of observation have been designed according to this model: the first one, the LTP (with its variations), has been used for research (e.g., McHale et al., 2004; Favez et al., 2006a; Hedenbro et al., 2006; Feldman, 2007; Von Klitzing et al., 1999a,b), as well as for clinical purposes (Fivaz-Depeursinge et al., 2000; Harrison, 2005) from the beginning of the 1990s. The second one, the PNG, has been used in research settings (Favez et al., 2012) and has begun to be used in clinical settings (Frascarolo and Favez, 2009; Bullens and D’Amore, 2013). These situations were mainly used in laboratory settings, but they may be also used at home, with some adjustments; this is especially true for the PNG as it has a less structured format than the LTP’s. The data from studies using these situations of observation show that the FA model is efficient to describe and assess family dynamics according to specific family-level dimensions – that is, dimensions that would be impossible to account for in assessing only individuals or dyads. Additional studies are now needed to assess the extent to which the FA model and its tools may be generalized across cultures – and specifically to non-Western ones. First, the situations of observation and their respective instructions may not be relevant to elicit patterned family interactions in any culture. Second, the behavioral indicators of a functional alliance may vary from one culture to another, similar to the cultural variations that have been observed in mother–infant interactions (e.g., Posada et al., 2002; Kärtner et al., 2010). Finally, the use of those situations of observation and their coding systems require a training, whether for clinical or research purposes. Workshops are organized on a regular basis by the Center for Family Studies in Lausanne (CEF); information can be obtained by contacting the authors of this article.

The studies mentioned in this paper were approved by the Ethical Committee of the State of Vaud, Switzerland.

NF: Wrote the main parts of the manuscript. FF: Reviewed the manuscript and wrote specific parts regarding the PNG. HT: Reviewed the manuscript and wrote specific parts regarding the LTP.

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.