Studies on the language of very preterm children under the age of two are mostly based on spontaneous interaction data, analyzed through developmental scales or criterion-based forms (Benassi et al., 2017). For older children, assessment is preferentially done using standardized tests (Zimmerman, 2017). As pointed out by Imgrund et al. (2019) for 4-year-old preterm children, we assume that a combination of these two methods would provide a more accurate picture of the characteristics of the communicative and language development of very preterm toddlers. Our study therefore aims at examining the relevance of combining elicited data from a standardized test with data from recordings of spontaneous interactions. This work is based on the data from 6 children aged 12 and 18 months. These children were tested through an early communicative development test (ECSP; Guidetti and Tourette, 2009), based on the Early Social Communication Scales (Seibert and Hogan, 1982). They were also filmed while interacting with a parent during a free play. The test focused on children’s skills in social interaction, joint attention and behavior regulation. The analysis of the spontaneous data also focused on joint attention, and in addition, on the interlocutors’ dialogic positions and on the typology of communicative modalities. The comparison of the outcomes for the ESCP test and for the interaction data highlighted the value of using a combination of these two methods to provide a detailed description of children’s skills. The interaction data with a parent gives a dynamic illustration of the difficulties and skills first identified through the elicited data. They appear to be of significant benefit to work on the assessment of children’s skills. They give access to what children can do in interaction with a familiar adult, in familiar interactional modalities, and also provide information on parental input and parental language support practices.