The soft laddering (SL) data were analysed in the same manner as Reynolds and Gutman (1988). Interview tapes were transcribed and transformed into a series of hierarchical ladders containing the key elements, from attribute to highest level of abstraction. The two interviewers separately assigned a construct category code to the elements of each ladder produced by every participant. To create a meaningful comparison between SL and the two hard laddering methods elements were coded to the same construct categories as those presented in the hard laddering procedure. Otherwise, new categories were created. Differences in coding were resolved by discussion and consensus between the two interviewers. Interviewers then separatelycoded all elements and comparisons were made and differences resolved through discussion. Negative elements were reversed to become positive for example, phrases such as ‘‘not get sick’’, ‘‘not sick all the time’’ and ‘‘not be weak’’ were categorised into ‘‘goodhealth’’. We recognise that such treatment may not precisely reflect recent observations that doing well at moving toward an incentive is not necessarily the same experience as doing well at moving away from a threat (Carver, 2001; Carver & Scheier, 1998). Within one methodology designed to elicit why product attributes are important to consumers, this issue would beimportant. However, our methodological study specifically addressed comparability of output across three methods necessitating consistent coding.