This paper challenges the exclusive reliance on speaker personality traits in experimental language attitude research. The most explicitly articulated motivation for this restriction is Kristiansen’s (2009) contention that the use of scales pertaining to the investigated varieties themselves could make listener-judges aware of the research purpose, as a result of which their attitudes would be consciously offered, shallow perceptions instead of deeper conceptualizations. On the basis of a free response experiment (in which respondents were invited to explicitly articulate their attitudes towards labelled varieties of Netherlandic Dutch), we demonstrate first that language attitudes are made up of speaker personality traits, but also, and unmistakably, of speech-related perceptions of the investigated varieties. A follow-up experiment (with scaled responses to unlabelled speech stimuli) confirms that the inclusion of speech traits in a speaker evaluation experiment does not affect the nature and structure of the attitudes observed. Building on these data, we argue against the idea that specific attitude measurement techniques should correlate with specific attitude consciousness levels, and we make a plea for a multi-modal approach to language attitudes.