Anchors

This section explains the anchor sounds used in the comparative evaluation.

The primary role of anchor sounds is to define the extrema of the rating scale. But they also have a more practical role, that is to link together all the ratings, whether they are done on the same or on distinct screens.

The absence of these two sounds would result in biases between different screens, as, according to each randomly defined sound distribution, some screens may include better sounds than others.

These two anchor sounds, often combined with the option to require at least one rating at the top and one at the bottom of the scale, ensure that all sounds are rated on the same scale. That is under the hypothesis that the two anchor sounds are rated consistently by all participants at these extreme positions on the scale, or close to them. In the analysis stage, participants failing to rate these sounds at or close to the extrema of the scale should be removed from the panel (for example, participants with several deviations superior to 10% of the scale range).

This is the reason why the choice of the anchor sounds should not be taken lightly, and must ensure their consensual identification as the rating scale extrema. That is also why Ansys strongly advises to select anchor sounds outside of the sound dataset of interest, as ratings are expected to have little variability. Choosing anchor sounds among the sounds of interest will introduce an experimental bias on their mean ratings.

However, it is impossible to choose generic anchor sounds. First, because the expected position of any sound on the rating scale depend on the question asked. Second, because anchor sounds must remain comparable to the test sounds to be efficient: strongly atypical sounds will result in the other sounds being rated with very little range on the scale, making it difficult to extract valuable and significant rating differences.