Electrocardiograms (ECGs) are electrical signals measured from the heart through sensors. ECGs provide valuable clinical information and insights about a patients cardiac status and are routinely used for the screening, diagnosis, and monitoring of cardiovascular conditions. As artificial intelligence (AI) methods and large language models (LLMs) continue to advance in the medical domain, there is increasing need to develop multimodal AI systems capable of combining ECG signals with natural language understanding and reasoning for better in-domain AI assessments. This project aims to investigate the multimodal nature of developing AI models using extracted information from ECG signals as a basis for clinical question answering. In particular, the project aims to investigate effective ways to encode ECG waveform information and how such information can be translated to useful grounds for LLM reasoning. The research will explore approaches for extracting clinically meaningful characteristics from ECGs, including cardiac rhythm, conduction characteristics, waveform morphology, and other diagnostically relevant features, and subsequently integrating these representations with language models for question answering. The overall objective is to establish methods for connecting physiological signal analysis with language-based clinical reasoning. The resulting methods aims to provide a foundation for future multimodal clinical AI systems capable of interacting with ECG data through natural language questions while maintaining a connection to the underlying physiological measurements.