Summary with the article: Perceptual attributes of voice: development and use of rating scales van Gelfer (1988)

Voice judgements of a listener are not easy. In a study that compared different studies on voice description terms, only 2 of the 27 terms on total, ‘horse’ and ‘nasal’, were used in all of the studies. Defining a normal voice is especially hard, and researchers busy with this definition almost never pay attention to listeners judgements.

There are a few scales available for measuring “normal” voices. These are:

• System by Darley, Aronson and Brown
  Sophisticated but hardly used for nonpathological speakers.

• F.B. Wilson Voice Profile
  Used by clinicians and also quite reliable, but same problem as mentioned earlier: hardly used for nonpathological speakers

• Osgood Semantic Differntial Technique
  Listeners use a bipolar scale to judge a persons voice. Works pretty well, but some features are not very applicable to voice, like “hot-cold” or “sunny-couldy”

The author wants to create a bipolar scale like the Osgood Semantic Differential Tehchnique, but with adjectives that are more useful in terms of voice description. This was done by first creating a list of adjectives that researchers frequently use for describing voices. After that, a bipolar scale was made and this scale was tested.

Selection of terms

57 adjectives were chosen out of literature on voice. A group of 18 volunteers (6 from speech pathology, 6 from vocal music and 6 without special training on voice) rated these adjectives in terms of how confident tell felt using such a word for describing a voice. Then the best adjectives were selecting by looking at their total scores on confidence. In the end, 22 words were chosen. Examples are ‘high’, ‘low’, ‘pleasant’, ‘rough’ and ‘soft’. The whole list can be found in the article.

Selection of opposites

After the first phase, the selection of terms, the researcher wanted to find out if these words were opposites. Graduate students in speech pathology, vocal music students and non-experts were selected. They were granted a list of the chosen words and had to write down the opposite. If they thought two words meant the same, they had to underline the word they’d prefer to use. Then the words that were used most as an opposite were selected. This resulted in a list of 17 bipolar terms. Examples of these opposites are loud-soft and pleasant-unpleasant, but again the whole list can be found in table 3 in the article.

Use of scale

To test the scale, expert (in speech pathology) and non-expert participants were told to listen to 20 female voice samples. They had to rate the samples with the 17 earlier selected terms. After this, the agreement between judges was calculated (with Kendalls Coefficient of Concordance, W). This W can differ from 0 to 1, with 1 indicating full agreement between judges. In the article, a table is provided which shows all the W scores. Both untrained and trained listeners rated high pitch versus low pitch best, but there were also differences. Even though there was only a W of 0.17 among untrained listeners on the steady-shaky scale, this one was still adopted in the scale the author eventually developed.