ABSTRACT The validity of the standard ideal L2 self scale has increasingly been called into question. This paper reports both quantitative and qualitative investigations into whether the ideal L2 self items tap into the intended construct of an actual–ideal discrepancy. Study 1 involved an experimental approach manipulating the items to explicitly refer to ability beliefs. Data from 1,362 participants across three countries (Austria, China, and Saudi Arabia) showed a lack of discriminant validity between original and manipulated items. Study 2 used cognitive interviewing to examine the thought processes of 24 Japanese university students as they responded to ideal L2 self items. Thematic analysis revealed that responses were dominated by reflections on current ability and expectations about using the language in specific situations, rather than envisioning an idealised future self. The findings of Studies 1 and 2 converge to indicate that the standard ideal L2 self scale does not successfully operationalise the intended theoretical construct of actual–ideal self discrepancies. Instead, responses are mostly driven by beliefs about ability to achieve the states described in each item. The results therefore challenge the validity of this widely-used scale, calling for a reinterpretation of its findings in the L2 Motivational Self System literature.