Detection of statistical differences between visual textures as a function of mean dot density

Observer sensitivity to differences in the statistical distributions of locally defined element density in visual textural stimuli was investigated. Ten female Ss participated in the experimental study for a total of two half-hour sessions each. The stimuli consisted of six replications of 40 visual displays. Four mean density (M.d.) conditions were used; three were based on an arithmetic progression and one on a logarithmic progression, i.e. mean densities 9, 12, 15 and 12 (log). It was felt that the use of a logarithmic progression in the construction of the patterns would facilitate the detection task. The displays consisted of two patterns which were either similar or dissimilar statistically. Patterns were constructed with the aid of an IBM 1443 computer printer, then photographically reduced on a 35 mm. negative high-contrast film. Both patterns of the similar pairs had the same structuredness or variance values and patterns of the dissimilar pairs had different variance values. Visual displays were back-projected on a screen for a viewing-time of 1.5 seconds and the visual angle subtended by each pattern was 5.7. Ss were instructed to decide whether the patterns presented were similar or dissimilar. The dependent variables were accuracy of response, latency of response and confidence of judgment. Analyses of variance, Pearson product-moment correlations and Receiver Operating Characteristic (ROC) curves were used to obtain the following results: Detection performances for mean density 12 (log) were generally higher over all dependent measures. Accuracy results also indicated higher levels of detection for M.d. 9 at dissimilar stimulus pairs. Though differences did not reach significance, there was evidence of a monotonic decrease in performance across all dependent variables from M.d. 9 to m12 to 15. It is possible that the phenomena of clustering and dual-perceptual processing may account for these results. At mean densities 9 and 12 (log), for accuracy and confidence measures, dissimilar stimulus pairs produced better detection results. These results contradict those reached by Stoloff (1967) and certain explanations are offered. Results for the ten stimulus pairs do not approximate Stoloff's scale of structuredness and it is argued that the scale is classified arbitrarily. Finally, differences yielded from replications are considered in the light of the possible perceptual processes involved. Implications for future research are discussed.

Name thesis:degree_name

Maîtrise ès arts (psychologie)

Level thesis:degree_level

2e cycle

Discipline thesis:degree_discipline

École de psychologie et d'éducation (1964-1965 à 1972-1973)

Grantor dc:publisher

Université de Moncton

Year dc:date.issued

1973

Author dc:creator

• Melanson, Diane C.

Advisor dc:contributor.advisor

• Muise, Gérard

• Perception visuelle

Statement dc:rights

• Author

Language dc:language.iso

iso639-2b, eng

Identifier

oclc: 726429771

Dc Identifier Other

umir:770

OAI identifier oai:identifier

oai:umoncton.scholaris.ca:20.500.14658/8305