Brain structural correlates of individual differences at low-to high-levels of the language processing hierarchy: A review of new approaches to imaging research

In the domain of language and audition, studies have shown large individual differences, within
the normal range (i.e. in healthy, non-expert individuals), in performance on tasks involving
speech sound processing, vocabulary knowledge, and reading, these in both monolingual and
bilingual participants and in native and non-native language contexts. These individual differences
have often been related to individual differences in brain structure. Evidence for structural
differences is especially striking since brain structure can be assumed to be more stable, or
less malleable, than brain function. Brain function, on the other hand, can be expected to
change, or be plastic, after only very short periods of training/learning. The present paper
provides a review of studies that have investigated the brain structural correlates of normative
individual differences in aspects of language-related performance, these spanning a hierarchy in
terms of the underlying complexity of processing and brain networks involved. Specifically, the
review is structured so as to describe work examining the following domains, which involve
progressively increasing levels of complexity in terms of the posited perceptual/cognitive
sub-functions involved: 1) lower-level acoustic processing; 2) phonetic processing, including
non-native speech sound learning, learning to use pitch information linguistically, non-native
speech sound articulation, and phonetic expertise; 3) working memory for verbal and for pitch
information; 4) semantics, in the context of lexical knowledge and of semantic memory; 5)
reading; 6) syntax, both natural and artificial; 7) bilingualism; and finally 8) executive control
of language in the contexts of fluency and of speech-in-noise processing. Results are discussed
and synthesized in the context of lower to higher-level brain regions thought to be functionally
involved in these respective domains, which are very often, if not always, the very ones that
structurally partly predict domain-specific performance.