Females are known to excel over males in most reading tasks, but not consis
tently so in tasks that require processing information from maps, tables, c
harts and diagrams, so called `Documents'. The IEA Reading Literacy data pr
ovides possibilities to investigate gender differences across countries in
such tasks in two age groups, 9-year-olds and 14-year-olds. The general que
stion about cultural influences vs. an invariant pattern of gender differen
ces is of great interest for gender research, and central in this study. Th
e aim of the paper is to describe and analyze gender differences on Documen
t tasks, and investigate if and how the pattern of differences varies over
countries. Another aim is to demonstrate the power of using a multivariate
analysis technique by contrasting it against traditional univariate approac
hes. The univariate analysis indicates female advantage as the most common
in the younger group and a mixed pattern in the older. The multivariate ana
lysis indicate that Document tasks are not unidimensional, because both gen
eral and specific dimensions can be extracted from the raw scores. The trad
itional univariate analysis often disguised true patterns of differences in
the data, both in terms of country differences and in terms of the directi
on of the gender differences. Raw score differences between the genders pro
ved to be due to differences in both general and passage specific dimension
s. Ten of the countries showed gender differences in both directions in the
general dimension among 9-year-olds, while an almost consistent pattern of
female advantage was found among 14-years-olds. Many of the specific passa
ge dimensions turned out to favor either males or females. This complex pat
tern varied over both age groups and across countries, although commonaliti
es in the pattern among subgroups of countries were common.