We investigated how accurate observations of canonical babbling (CB) were and explored predictive babbling measures in children with and without medical diagnoses of conditions that can lead to speech and language problems.

From 2012 to 2014 this Stockholm‐based study recruited 38 children aged nine months to 21 months with medical diagnoses and 30 children aged 10 months without diagnoses and included 21 previously studied 12‐month‐old children without medical diagnoses. CB and consonant sound production were directly observed by video recording natural play with a caregiver. The percentage of CB was calculated from each recording, and a validated observation form was used. How accurately the children with and without CB were classified was investigated with sensitivity and specificity. The groups were compared using predictive babbling variables.

The observation method identified children with and without CB well, with a specificity of 0.89 and sensitivity of 0.93, respectively. Children with predictive babbling measures were identified in the clinical group (specificity 0.93–0.97), and a lack of these measures indicated a risk of being in the clinical group (odds ratios > 10). The sensitivity was low (0.32–0.42).

Observation effectively identified a lack of CB and supported the importance of assessing babbling measures.