The way in which perceived quality affects consumer food choice is complex due to variations in intrinsic and extrinsic product attributes that interface with personal factors and the socio-cultural context. Taking the example of salmon fillets sold in the USA and Japan, this study uses a think-aloud protocol (TAP) as a data collection method and a corpus (text data)-driven approach as an analytical tool for uncovering how consumers infer quality cues on salmon fillets. The study aims to help in designing representative studies and may help in explaining survey results. The findings indicate that consumers infer quality under the influence of multiple factors, including individual knowledge of food production, processing and cooking, personal preferences based on culinary heritage and the socio-cultural environment. This study also shows how a TAP can be used in combination with text mining to provide richer insights into consumer choice.