The code-labelling exercise is an attempt to apply natural language education techniques for solving the challenge of teaching introductory programming to non-STEM novices in higher education. This paper presents findings from a study exploring the use of natural language teaching techniques in programming education collected in an Action Research cycle. The results support the use of a structural approach to teaching programming to this target audience; particularly, the translation-grammar method seems to integrate well with programming education. The paper also explores the potential underlying reasons. It seems the exercise invokes an assimilation of student's existing cognitive schemata and supports a deep-learning experience. The exercise is an invitation to other teachers to create further iterations to improve their own teaching. It also seeks to enrich the portfolio of teaching activities for solving the challenge of teaching introductory programming to non-STEM novices.