Attention-sensitive signalling is the pragmatic skill of signallers who adjust the modality of their communicative signals to their recipient’s attention state. This study provides the first comprehensive evidence for its onset and development in 7-to 20-month-olds human infants, and underlines its significance for language acquisition and evolutionary history. Mother-infant dyads (N = 30) were studied in naturalistic settings, sampled according to three developmental periods (in months); [7–10], [11–14], and [15–20]. Infant’s signals were classified by dominant perceptible sensory modality and proportions compared according to their mother’s visual attention, infant-directed speech and tactile contact. Maternal visual attention and infant-directed speech were influential on the onset and steepness of infants’ communicative adjustments. The ability to inhibit silent-visual signals towards visually inattentive mothers (unimodal adjustment) predated the ability to deploy audible-or-contact signals in this case (cross-modal adjustment). Maternal scaffolding of infant’s early pragmatic skills through her infant-directed speech operates on the facilitation of infant’s unimodal adjustment, the preference for oral over gestural signals, and the audio-visual combinations of signals. Additionally, breakdowns in maternal visual attention are associated with increased use of the audible-oral modality/channel. The evolutionary role of the sharing of attentional resources between parents and infants into the emergence of modern language is discussed.