Geography Reference
In-Depth Information
Fig. 1 An example of counting landmarks on a sketch map: (
1
) passage, (
2
) structure, (
3
)
landform, (
4
) water landmark, (
5
) rock landmark, (
6
) tree or a part of tree, (
7
) sign and (
8
) land
cover
video recording of the drawing participants. In case of the few differences that
occurred in the classifications, the classifications were synthesised through the
group work. Every marked feature was regarded as a landmark and separate
sections of continuous landmarks were treated as individual landmarks in the
“Passages”, “Waters”, “Land cover” and “Landforms” landmark groups (Fig.
1
).
We did not count landmarks that were mentioned during the thinking aloud while
drawing if they were not also actually drawn on the sketch map.
Statistical Calculations
In order to identify landmark concepts that the participants used in significantly
different frequencies between conditions, we ran two-tailed Wilcoxon rank-sum
tests for each landmark using the SciPy Python package (Jones et al.
2001
). In all
the statistical calculations, we used significance level
0.05.
In order to detect differences in the use frequencies of landmark groups between
different conditions, we ran permutational multivariate ANOVA (PERMANOVA;
Anderson
2001
) using the R software (R Core Team
2013
) to test the main effects of
time of day and task as well as their interaction effect (50,000 permutations). We
chose the non-parametric PERMANOVA because our samples are small and the
normality of landmark frequency distributions is doubtful: Shapiro-Wilk multivar-
iate normality test showed non-normality for the thinking aloud data from night
α
¼
