Reading ease: A two-way repeated measures mixed-model analysis of variance (2 Age groups X 8 Fonts) was conducted on the data for rated Reading Ease. There was a difference among the various fonts tested (F(7,16) = 3.48, p <.05) but no age effect (F(1,22) = 0.07, p >.05), nor an interaction between the age and font (F(7,16) = 0.79, p >.05), on the ease of reading measure. Geisser-Greenhouse correction was used for repeated measures analysis to correct for violations of the sphericity assumption.
Visual inspection of the Mean Subjective Reading Ease for each font (see Figure 7) indicated two font groupings. These were tested using pairwise ANOVA comparisons, adjusted for the family of significant differences at the .05 level. Since nine comparisons were significant, Bonferroni correction for 9 tests was applied. The font that was judged as being the least easy to read (i.e., Century Old Style) differed significantly from Sabon (t(22) = 3.02, p < .006) and Times New Roman (t(22) = 3.56, p < .006). The Sans was significantly different from Bell Gothic (t(22) = 3.05, p < .006) and Times New Roman (t(22) = 3.00, p < .006). No other pairwise comparisons were significant.
Figure 7: Mean overall subjective reading ease of print (1 = Very Easy; 7 = Very Difficult) as a function of font type.
Reading discomfort: A two-way repeated measures mixed-model analysis of variance (2 Age groups X 8 Fonts) was conducted on the reading discomfort data. A significant difference among fonts was observed (F(7,16) = 5.20, p <.05). There was no age effect (F(1,22) = 0.02, p >.05), nor an interaction between age and font (F(7,16) = 0.74, p >.05). Again, Geisser-Greenhouse correction for repeated measures analysis was used to correct for violations of the sphericity assumption.
Visual inspection of the Mean Subjective Reading Discomfort for each font (see Figure 8) suggested two font groups. Pairwise ANOVA comparisons adjusted using the Bonferroni correction for the number of significant differences at the .05 level were conducted. Ten comparisons reached significant at the .05 level and served as the basis for the Bonferroni correction. The font that was judged as producing the greatest discomfort (i.e., Garamond Condensed) was significantly different from the two fonts rated as having produced the least discomfort (Times New Roman, t(22) = 3.56, p < .005; and Sabon, t(22) = 3.24, p < .005). The font judged as having caused the next most discomfort (i.e., Univers Condensed) was also significantly different from Times New Roman (t(22) = 3.54, p < .005), and Sabon (t(22) = 3.92, p < .005). Century Old Style was rated as causing more discomfort than Times New Roman (t(22) = 4.05, p < .005), Sabon (t(22) = 3.70, p < .005) or Bell Gothic (t(22) = 3.08, p < .005). No other pairwise comparisons were significant.
Figure 8: Mean overall subjective reading discomfort of print (1 = Very Little Discomfort; 7 = Very Much Discomfort) as a function of font type.
Letter clarity: The letter clarity data were analyzed using a two-way repeated measures mixed-model analysis of variance (2 Age groups X 8 Fonts). While there was a difference among the various fonts tested (F(7,16) = 3.51, p <.05), there were no age effects (F(1,22) = 1.02, p >.05), nor an interaction between age and font (F(7,16) = 0.60, p >.05). Geisser-Greenhouse correction was used for repeated measures analysis to correct for violations of the sphericity assumption.
Visual inspection of the subjective Mean Letter Clarity for each font (see Figure 9) also indicated two font groupings. These were tested using pairwise ANOVA comparisons, adjusted by Bonferroni correction for the number of significant pairwise differences (alpha = .05). Eleven significant comparisons were observed and used as the basis for the Bonferroni correction. Garamond Condensed was found to be significantly different from Times New Roman (t(22) = 3.26, p < .005). Century Old Style was significantly different from Bell Gothic (t(22) = 3.39, p < .005) and Times New Roman (t(22) = 3.19, p < .005). No other pairwise comparisons were significant.
Figure 9: Mean overall subjective letter clarity of print (1 = Very Clear; 7 = Very Unclear) as a function of font type.
Simplicity/complexity: The simplicity/complexity data were tested using a two-way repeated measures mixed-model ANOVA (2 Age groups X 8 Fonts). There was a significant difference among the various fonts tested (F(7,16) = 11.11, p <.05), and an interaction between age and font (F(7,16) = 5.31, p <.05). The age difference was not significant (F(1,22) = 4.26, p >.05). Geisser-Greenhouse correction was used for repeated measures analysis to correct for violations of the sphericity assumption. Follow-up tests of simple main effects found that complexity differed among fonts for young (F(7,5) = 16.45, p <.025), but not elderly observers (F(7,5) = 2.47, p >.025, See Figure 10). The Geisser-Greenhouse correction for repeated measures was applied to the data of the elderly observers due to violations in the sphericity assumption. T-tests were used to analyze age differences on each of the fonts (Bonferroni corrected for eight tests). An age difference was seen for only two fonts, Sabon (t(22) = 3.77, p <.0063) and Times New Roman (t(22) = 3.34, p <.0063).
Figure 10: Age differences in subjective simplicity/complexity of print (1 = Simple; 7 = Complex) as a function of font presented in order of increasing complexity for young participants.
Although no real font differences emerged among the elderly on the Simplicity/Complexity measure visual inspection suggested two font groupings for younger participants (see Figure 10). Pairwise ANOVA with Bonferroni correction for the number of significant pairwise differences were conducted using a .025 level. Since 16 comparisons were significant at the .025 level, a Bonferroni correction for 16 tests was applied (p =.0016). The font rated as most complex (i.e., Garamond Condensed) was significantly different from all of the sans serif fonts (see Table 10). Century Old Style, Times New Roman and Sabon were significantly different from the three least complex fonts (i.e., Univers Condensed, Bell Gothic and The Sans). No other fonts were significantly different from each another.
Table 10: Significant T Values for Pairwise Comparisons of Simplicity/Complexity Ratings for Young Participants (p<.0016)
| Font |
Garamond Condensed |
Century Old Style |
Times New Roman |
Sabon |
| Times New Roman Sans | 5.48 |
|||
| The Sans | 7.09 |
6.27 |
4.86 |
5.74 |
| Bell Gothic | 5.98 |
4.48 |
4.05 |
4.57 |
| Univers Condensed | 6.46 |
5.40 |
4.38 |
4.55 |
Preference: Given that the preference measure was based on ordinal data they were analyzed using a Friedman two-way ANOVA. It yielded a significant difference among the fonts (chi2(7, N = 24) = 57.90, p <.05). Follow-up ANOVA's demonstrated that there were differences for both young (chi2(7, N = 12) = 26.97, p <.025) and older participants (chi2(7, N = 12) = 38.72, p <.025). Mann-Whitney tests to assess age differences however, showed no significant differences for any single font. Visual inspection of the mean preference for each font indicated two or three possible font groupings (see Figure 11). These were tested using pairwise ANOVA comparisons corrected for the number of significant differences at the .05 level. With nineteen comparisons significant at the .05 level, a Bonferroni correction for 19 tests was used (p = .0026). The least preferred font (Univers Condensed) was significantly different from the four most preferred fonts (Times New Roman, Bell Gothic, Sabon and Times New Roman Sans, see Table 11). The second and third least preferred fonts (The Sans and Garamond Condensed) were significantly different from Times New Roman. The next least preferred font (Century Old Style) was significantly different from Times New Roman and Bell Gothic. No other pairwise comparisons were significant.
Figure 11: Mean overall subjective preference of print (1 = Most Preferred; 8 = Least Preferred) as a function of font type.