SensoGram 30: Scales Of Justice: Measurement in Sensory Evaluation   There is as much confusion as ever in the commercial world about scales of measurement in sensory evaluation.  Should we use a 5-point scale, a 9-point scale, or is it better not to use categories at all?     The choice of scale in sensory evaluation has been largely capricious despite the store-house of knowledge on this topic in psychophysics (where it is technically called ‘scaling’).  Scaling is critically important to the validity of all behavioural science, sensory evaluation being merely one application.  We take precision and validity for granted in physical measurement – you wouldn’t build a house without a proper tape measure - but it is usually overlooked in sensory measurement.   Let’s contemplate what we are trying to accomplish.  When we use a scale in sensory evaluation, we seek to measure a subjective attribute such as sweetness or saltiness or hardness.  By ‘subjective’ we mean that the sensation we seek to measure is registered ‘inside’ (i.e. is experienced in the brain or on the tongue – wherever we localize the sensation).  Thus, by having a respondent make an overt response on some kind of scale, we are attempting to measure a private experience that is not directly accessible.      Now, the key to valid scaling is to be sure that the overt respondent’s score reflects the covert sensation inside.  When we achieve a straightforward mapping of the internal sensation on to the external response on the scale, we have ‘scale validity’, or what psychophysicists call a ‘linear response scale’.  (Note that ‘linear’ in this sense does not mean a straight line; it means, rather, that the overt responses on the scale linearly reflect the gradations of sensation that the respondent is experiencing.)   So which scale can do this?  The scales most prone to bias and nonlinearity – scales which cannot be guaranteed to do this - are those which consist of a small number of categories (e.g. a 5-point scale).        These do not allow respondents sufficient scope to report all the gradations in their perception and often are subject to ‘end effects’.  Scales with more points are better (eg 9-, 11-, or 13-point scales):     However, scales with no points, no categories, and no numbers are better still!            Suppose you are asked to rate the sweetness of a number of products.  Note first that you do not, naturally, carry a category scale of sweetness in your head - you do not go about calling products a ‘3’ or a ‘7’ or a ‘9’.  Thus, numbers on a rating scale are at best redundant and at worst confusing.  Category scales have a long history in psychological measurement, but their popularity is due mainly to convenience and mathematical tractability; humans do not think in discrete categories.  More likely, you will use words such as “this is slightly sweet” or “this is very sweet” or “this is excruciatingly sweet”.    So what are you really doing when you rate the sweetness of a product?  You are considering this level of sweetness as it relates to your past experience.  And your past experience will have set up anchors (extremes) in your sensory memory.  You know what no sweetness tastes like, and you will have experienced an extraordinarily sweet sensation.  Hence, when you rate the sweetness of a product you are in fact rating the position of this sweetness on a continuum of sweetness, held in memory.    Realizing that we are rating a position is a conceptual breakthrough.  It means the scale that best reflects the internal sensation is one with no categories at all, no numbers at all, but which has well-defined extremes at either end.  For example, the scale may have the words ‘Not at all sweet’ at one end and ‘Extremely sweet’ at the other.  The respondent will mark a position on this continuum that best reflects the sweetness in question.   Of course, numbers must be attached to the scale to obtain quantitative data.  But this can be done after the data have been collected.  You may calibrate the scale as 1-10, 0-50, or 0-100, as you prefer.   Interestingly, recent advances in neuroscience have identified structures on the cortex which appear to behave like these scales.  Thus, the rating scale, with its well-defined extremes but with no intervening categories or numbers, may be mimicking the process by which sensory intensity is recorded in the brain.