terry9 wrote:
"Unfortunately, that process is amongst the most intellectually difficult enterprises known to man. The trouble is, everyone thinks he can do it.
Speaking crudely, there are four levels of measurement: nominal (in which different numbers represent different names), ordinal (in which bigger numbers represent relatively bigger effects), interval (in which the ordinal property holds and the difference between 1 and 2 is the same as the difference between 101 and 102), and ratio (in which the interval property holds and zero represents none of the property)."
My Reply:
I didn't want to burden the conversation with too much technical detail, but the psychometric techniques I mention above are based upon Semantic Differentials. As you may well know, Semantic Differential rating scales are bipolar and have an inherent "zero point" in the middle. With properly chosen verbal anchors for the rating points (as per Charles Osgood and his colleague) Semantic Differentials provide ratio level measurement. I typically use Factor Analysis to analyze such data to provide a subset of psychological dependent measures. When correlated with acoustic metrics using Multiple Regression all of this provides a predictive model indicating which acoustic metrics are most strongly associated with each psychological metric. You can literally choose a desired level of perceptual experience and determine what physical metrics are associated with it.
I will often begin a research study by asking participants to imagine their "ideal" experience and rate it using the rating scales. This assumes that participants have some sense of what "ideal" is, which is not usually a problem with experienced participants. With such data you have an "ideal" target against which to compare actual listening data. It should be noted that listeners' "ideal" might not sound realistic.