A method for comparing perceptual distances and areas with multidimensional scaling.

Area of Science:

• Psychology
• Linguistics
• Statistics

Background:

• Multidimensional scaling (MDS) is commonly used to visualize perceptual data, but typically lacks inferential statistical treatment.
• Standard MDS visualizations present average results, overlooking crucial inter-participant variation in perceptual judgments.
• The absence of statistical rigor in MDS limits the reliability of interpreting distances and group structures.

Purpose of the Study:

• To develop a statistical method for enhancing multidimensional scaling (MDS) analyses.
• To enable reliable statistical comparisons of perceptual distances between speech sound categories.
• To address the limitation of ignoring inter-participant variation in traditional MDS.

Main Methods:

• A novel technique is presented to add statistical power to traditional MDS solutions.
• The method allows for the statistical comparison of distances between individual objects (e.g., speech sounds).
• It also provides a means to compare the dispersion or area occupied by groups of objects in multidimensional space.

Main Results:

• The proposed method increases the statistical power of MDS, allowing for more reliable comparisons than visual inspection alone.
• It facilitates the comparison of distances between specific segments within the perceptual space.
• The technique enables robust comparison of the dispersion of object groups, accounting for inter-participant variability.

Conclusions:

• This approach significantly enhances the inferential capabilities of multidimensional scaling.
• Researchers can now statistically compare perceptual distances and group dispersions in speech sound categories.
• The method offers a flexible framework for incorporating statistical analysis into MDS, improving the interpretation of perceptual data.