A Reliable Conversation Probe Measure for Assessing Children’s Conversational Skill

Once children move beyond early childhood, social conversation becomes the primary means of social connection and collaboration. Individual differences in conversational skill have well-established consequences for peer acceptance. Weaker social communication skills are associated with poorer mental health and increased behavioural difficulties.

Despite its importance, conversational ability is difficult to assess systematically. In contrast to domains such as vocabulary, phonology, or morphosyntax—and even inferencing—there are few reliable tools that allow children’s conversational skills to be elicited and compared efficiently. Existing approaches typically rely on parent or teacher report, or on lengthy transcription and coding of naturalistic conversations.

Across a programme of research, we have developed a ‘conversation probe’ method designed to elicit and rapidly score children’s ability to maintain conversational topics. During an apparently naturalistic interaction, the examiner introduces a series of pre-planned, topic-appropriate statements (probes; e.g. “I think big dogs are quite scary”). The conversation is gradually guided to contexts in which each probe can be naturally introduced. Children’s responses are then identified via automatic transcription and coded, following a manual, as contingent (relevant to and extending the topic), off-topic, minimal, or null. This approach equates conversation topic across children, avoids the need to listen to entire conversations, and allows for efficient scoring.

The present study examined both inter-rater and test–retest reliability of the conversation probe measure. Seventy-two Year 4 children participated (43% female; 15% multilingual; 17% with a developmental diagnosis), all attending mainstream state primary schools. Each child completed a 5–7 minute one-to-one conversation with a researcher on two occasions approximately two weeks apart. Inter-rater reliability was excellent (κ = .92), and test–retest reliability was high (r = .55, p < .001). Score distributions indicated that the measure (proportion of contingent responses) was well-suited to Year 4 children.

Conversation probes provide a reliable and efficient tool for assessing conversational skill with clear potential utility for both research and clinical practice. Our previous work has shown that contingent responding is significantly associated with the Children’s Communication Checklist–2 pragmatic composite. Here we outline a ‘how to’ guide for using the conversation probe procedure.