demonstrated, with expectancy graphs similar to those that we have generated, that
despite their typically modest relations with important outcomes, such tests can indeed
have practical utility. Expecting tests to contribute substantially more to the prediction of
inherently complex real-life performance is probably unrealistic.
Our conclusion might have been anticipated on a more theoretical basis. It has been
traditional to view language proficiency in terms of the four domains of listening, reading,
speaking, and writing. However, these domains are linked by more basic underlying
processes (e.g., Baker & Baker, 2009). There are specialized terms for these elements
(phonology, morphology, semantics, syntax, and pragmatics), but in general they concern
the following: (1) knowledge of sounds; (2) formation of words; (3) knowledge of
word combinations; (4) grammar; and (5) how language is used in context. An obvious
example of this underlying linkage is the role of semantics in each domain. Knowing the
meaning of words enables listeners and readers to interpret messages appropriately, and
it allows speakers and writers to construct them effectively. Because the four language
domains are related in such intricate ways, a measure of ability in one domain can, when
used in conjunction with the ability in related target domains, add depth, nuance, and
accuracy to the measurement of the target construct.