Experts and novices use the same factors--but differently--to evaluate pearl quality

Abstract

Well-trained experts in pearl grading have been thought to evaluate pearls according to their glossiness, interference color, and shape. However, the characteristics of their evaluations are not fully understood. Using pearl grading experiments, we investigate the consistency of novice (i.e., without knowledge of pearl grading) and expert participants' pearl grading skill and then compare the novices' grading with that of experts; furthermore, we discuss the relationship between grading, interference color, and glossiness. We found that novices' grading was significantly less concordant with experts average grading than was experts' grading; more than half of novices graded pearls the opposite of how experts graded those same pearls. However, while experts graded pearls more consistently than novices did, novices' consistency was relatively high. We also found differences between the groups in regression analyses that used interference color and glossiness as explanatory variables and were conducted for each trial. Although the regression coefficient was significant in 60% of novices' trials, there were fewer significant trials for the experts (20%). This indicates that novices can also make use of these two factors, but that their usage is simpler than that of the experts. These results suggest that experts and novices share some values about pearls but that the evaluation method is elaborated for experts.

Figure 1. Schematic diagram of internal structure of cultured pearl.

The translucent nacre surrounds the nucleus. The nacre comprises hundreds to thousands of nacreous layers. The thickness of each layer is approximately 300–500 nm.

Figure 2. Measured values of interference color and glossiness.

Each point shows the average values of 15 measures of a pearl. We used five pearls for each rank (“A” rank shown by star and “B” rank shown by circle) in both sets. Higher measured values seem to be associated with higher ranks (“A” rank) in both interference color and glossiness. (A) Relationship between interference color and pearl ranks in each stimulus set. (B) Relationship between glossiness and pearl ranks in each stimulus set. (C) Relationship between interference color and glossiness. Colors represent stimulus set. Pearson’s r of the set in orange is 0.208, and that shown in green is 0.128.

Figure 3. Photographs of the experimental conditions.

(A) A snapshot of “Mobile-Labo” truck. The experimental space is beyond the inner door. For this photograph, we parked on a street in front of the office in the courtyard of the laboratory and conducted the experiment. (B) Experimental setting for the experts. The person sitting in front is the participant. The white board above the black desktop is the diffuser. The participant is asked to sort the pearls according to a subjective criterion in the small box. The experimenter received the box and recorded the order.

Figure 4. Comparisons between experts and novices.

(A, B) Individual averages of concordance rates for each pearl set (blue and red). In A, the participants are sorted into high to low concordance. The asterisks signify that the difference between concordances for each pearl set was significant (*: p<0.05; **: p<0.01). (C) The average concordances of both groups are shown. As shown, there is a significant difference between them. (D) For consistency, the values of experts and novices were significantly higher than chance and the 95% upper limit. The red solid and dashed lines indicate the expected chance value of W and the values corresponding to the 95% upper limit of chance value of W, respectively, which was calculated from a randomized resampling (n = 1,000,000). The difference between them is statistically significant (*: p<0.05). (E) The averages of all ρ coefficients in both groups. The difference between them is statistically significant (*: p<0.05). The error bars in graphs refer to the standard deviations.

Figure 5. Regression analyses by measured variables.

In all panels, the upper insets show the best (left) and worst (right) regression results. The vertical axes show the rank orderings, and the horizontal axes show either the explanatory variable (in A, B, D) or the prediction from the regression equation (in C). The lower graphs show the results of all the regressions. In all graphs, each point represents one trial, and each row represents one participant. The numbers written in parentheses are the numbers of trials for which the regression is significant (shown by red points) and the total number of trials. The vertical axes show the coefficients of determination (R²). (A and B) Results of simple linear regression analyses on interference color and glossiness, respectively. (C) Results of multiple regression analyses on both variables separately. (D) Results of simple linear regression analyses on the product of both variables.