The functional architecture of mother-infant communication, and the development of infant social expressiveness in the first two months

By two-three months, infants show active social expressions during face-to-face interactions. These interactions are important, as they provide the foundation for later emotional regulation and cognition, but little is known about how infant social expressiveness develops. We considered two different accounts. One emphasizes the contingency of parental responsiveness, regardless of its form; the other, the functional architecture account, emphasizes the preparedness of both infants and parents to respond in specific ways to particular forms of behaviour in their partner. We videotaped mother-infant interactions from one to nine weeks, and analysed them with a micro-analytic coding scheme. Infant social expressiveness increased through the nine-week period, particularly after 3 weeks. This development was unrelated to the extent of maternal contingent responsiveness, even to infant social expressions. By contrast, specific forms of response that mothers used preferentially for infant social expressions-mirroring, marking with a smile- predicted the increase in these infant behaviours over time. These results support a functional architecture account of the perceptual and behavioural predispositions of infants and parents that allow young infants to capitalize on relatively limited exposure to specific parental behaviours, in order to develop important social capacities.

2 on each occasion. To ensure these interactions were as natural as possible, and were conducted when the infant was in an alert state, the mother was asked to feed the infant if appropriate, and then to change their nappy, followed by her being asked to engage with the infant socially as she normally would, without using objects. Filming started when the mother first attempted to engage with the infant. If the infant became distressed and could not be consoled, filming stopped, and was attempted again once the infant had calmed.

Coding
Data for coding were available for 91 of the 100 planned assessments: in 4 instances mothers were not available during the time frame required; in 3 cases there were technical problems with the videos; and in 2 instances infants were distressed throughout. Mother-infant interaction videos were event-coded on a one second time base, using purpose built software for identifying associations between maternal and infant behaviour (details available from LDP). The coding scheme was based on that of Murray and colleagues 1,2 (coding manual available from LM/LB), and included key infant and maternal events described in the literature on early infant social development and mother-infant interactions [3][4][5] . Coders were graduate students who were blind to study aims and hypotheses, and to background information concerning the infants and their families.
Infant behaviour. Infant behaviours were mutually exclusive, and were clearly discernible, discrete, events with definite onset, and therefore readily identifiable by the mother in live time. Since mothers direct their attention almost exclusively to the infant's face during interactions in the first few months 6 , we coded behaviours involving infant facial movements and emissions of sound. Behaviours in group A required the infant to be gazing at their mother's face, as looking at the interactive partner is one of the characteristics of infant communication at this age 6,7 , whereas the other groups of infant behaviour were scored independently of the direction of infant gaze.

A. Social behaviours
(i) 'pre-speech' 3 : active movements of lips and tongue (e.g., tongue pushed into the bottom lip, moving it forward, or protruded beyond the lips), and of open mouth shaping (e.g., into an 'O', or pursed, as during cooing, even though unvoiced) that appear to be directed at the mother (Supplementary Figure S1); (ii) smiles; (iii) neutral-positive vocalisations, such as cooing.

B. Non-social mouth movements
Mouth movements that appear undirected (e.g., chewing or sucking movements, rolling lips together), or else mouth movements clearly directed to a non-social goal (e.g., rooting to the infant's fist).

Contingency:
Given that infants of this age cannot detect events as contingent if they occur 3 seconds or more after their own behaviour 9 and that mothers' responses almost always occur within 2 seconds of infant expressions 10 , maternal contingent responses were coded as events occurring within two seconds of the infant events above (and they were, therefore, also 'contiguous'), as in 1,2,10 .

A. Mirroring
Responses of the same valence and intensity 4,11 as the infant's behaviour that are either exact matches, or that match the principal features with some minor modification, i.e., 'enriched' mirroring in which some element is added (such as a vocalisation to a clear mouth opening), 'partial' mirroring in which some element is omitted (such as the facial expression of a cry being imitated but not the sound), or 'modified' mirroring in which the form is slightly changed, often in conventionalised ways (e.g., responding to an infant 'ooo' vocalisation with 'goo') (Supplementary Figure S2).

B. Marking
Responses of the same valence and intensity as the infant's behaviour that single out and 'mark' an infant behaviour with 'attention-attracting' cues 5 , without mirroring it.
(i) Positive, with smiles: e.g., infant makes a tongue protrusion and the mother raises head and eyebrows, then nods and clearly smiles, saying 'Is that your tongue?' (Supplementary Figure S3); (ii) Neutral, without smiles: e.g., infant smiles, and mother responds by raising and lowering her head, and then says 'ooh' with raised brows and an expression of interest (Supplementary Figure S4).

C. Negative responses
i) Rejecting responses: responses that criticize, reject or mock the infant (e.g. the infant's lip is pouted in distress and the mother frowns, shakes her head, and says "oh no, we can't have that", or else responds with a sneer); ii) Mis-attuned responses -valence: Responses where maternal affect is markedly discordant with the infant's behaviour (e.g. the infant shows sign of distress and the mother smiles broadly and laughs); 5 Mis-attuned responses -intensity: a) hyper-responding: responses that are a clear and extreme exaggeration in relation to the infant behaviour (e.g., the infant shows minimal mouth opening, and the mother swiftly and strongly moves her head up and down and opens her own mouth very wide); b) hypo-responding: responses that clearly downplay the infant behaviour (e.g., the infant makes a strong positive vocalisation, or gives a strong, 'joyful' smile, and the mother responds with a flat, dull, vocalization, or minimal, weak smile).

Prominence:
Maternal non-responsive behaviours were also coded as events; they include spontaneous vocalisations, smiles, tongue protrusions, mouth openings, head nodsbehaviours that often occur in attempts to engage with the infant, but that are not contingent on infant events. The number of these behaviours was used to determine the prominence of maternal responses, calculated as maternal behaviours that were responses to the infant as a percentage of all maternal behaviours coded (i.e., responses plus non-responsive behaviours).

Coding Inter-rater reliability
Videos were coded by two researchers, who both independently coded the same 20% of the total sample, including one interaction for each mother-infant dyad. Reliability for infant events was as follows: social behaviours κ = .92; non-social mouth movements κ = .92; negative affect κ = .85; biological events κ = .86. Reliability for maternal responses was as follows: mirroring κ = .90; positive marking κ = .80; neutral marking κ = .80; negative responses κ = .83. For maternal non-contingent behaviours, the intra-class correlation was 0.87.

Infant events and maternal responses were each investigated through Principal Components
Analysis (PCA), using Parallel Analysis, to determine the number of components to extract, 6 and Simplimax rotation. We used a generalised linear mixed modelling (GLMM) framework to address the study questions. been highlighted principally for its referential function 18 . At two-three months, this is often 7 phatic-i.e., it references the participants' state of engagement, as when the parent greets the infant when eye contact is made 19 , but it has also been noted to attribute significance, or meaning, to infant behaviour and expressions 18,[20][21][22][23][24]1