Despite
numerous studies on functional mapping of language, characterization of
language related areas is far from complete. Whether there is a map is a
contentious issue in the first place. I don’t want to go into the philosophical
debate, since I think it’s becoming clearer now that both sides (1) people who
wants you to believe that brain is made of modules that can be labeled using
terms defined in preexisting theories, and (2) people who think that that the
cerebrum is just a homogeneous sheet of processing units are both wrong.
I do
believe in the reality of linguistic units (phoneme, morpheme, phrases, etc.)
and I am fond of learning various syntactic theories (government and binding,
minimalism, HPSG, etc.), but I think it’s naive to assume that there are brain
modules dedicated to levels of language representations and syntactic
operations. I am very skeptical of Marr’s levels of design (computational /
algorithmic / implementation) when it comes to the brain. Recent development in
deep neural networks provides a good counter-example: we can design a
processing system in which we cannot label the function of each component
clearly, at least in terms of preexisting theories.
Having
said that, I do not think that the cerebrum is just a homogeneous sheet. The
most important structure obviously emerges from inputs and outputs as well as
long and short range connecting fibers. Genes can modulate numerous parameters
related to the formation of neo-cortex, although it does not look as
heterogeneous as “old” organs (like the heart or subcortical areas) since there
has not been nearly as much time historically to tweak these parameters.
So
in short, I’ll start with inputs and outputs (for the nature of inputs and
output see the appendix), and try to carefully characterize the working of
local areas, paying full attention to the fact that traditional way of
characterizing language areas may be totally wrong. Main source of information
are the review papers by Friederici (2011), which summarizes numerous studies
in the linguistic brain are summarized, and Hickok and Poeppel (2007), written
by two of the powerhouse of knowledge and critical thinking in language.
Language
related event related brain potentials (ERP) provide valuable insights, since ERP
research has a rich history of critical works and also it provides timing
information.
·
N100, associated with acoustic and phonological
process, is localized around the auditory corterx.
·
ELAN (150mS), associated with syntactic category
assignment, is localized around anterior
temporal / inferior frontal gyrus (IFG).
·
LAN (400mS), associated with morphosyntactic
features for argument structures, is localized around anterior temporal gyrus/IFG.
·
N400, associated with lexical access load, is
localized around mid-posterior superior temporal gyrus /IFG.
·
P600, associated with syntactic reanalysis (but can
be semantically motivated), is localized around middle temporal gyrus/basal
ganglia.
Results
of MRI studies are harder to summarize. For one thing, stimulus conditions vary
from one study to another. Also labels like “semantic”, “multimodal”, etc. mean different things in different studies.
For instance semantics as in thematic role assignment and word disambiguation
are totally different.
So I
focus focus on inputs, outputs, long-range connections, and the axes in the
frontal and temporal areas along which the quality of information changes most
dramatically. The auditory input to the
temporal lobe and the motor output to language related muscle control (see
appendix) in the frontal lobe are clear. The output from the speech processing
is less clear, seemingly go into higher area including the area 45. The input
to the speech motor control is also unclear, which may consist of knowledge to
be expressed and the motivation to speak. The main axis of qualitative
variation, to me, appears to be the time-range. As the position in the language
related cerebral cortex approaches the posterior end of the prefrontal cortex,
or closer to the primary auditory cortex in the temporal lobe, the information seems
to be more short term (e.g. a simple motor command or a short duration of
certain speech spectrum). As the position approaches frontal end of the
prefrontal cortex or furthest end from the primary auditory cortex, the
information becomes long term (or even static), (e.g. an intended message
content or a sentence or even a paragraph).
I
will probably keep revising the diagram, but the figure below is the one I got
for now.
Appendix:
Subcortical language pathways
What
kind of auditory signal does the primary auditory cortex (area 41) receive? The
auditory pathway from the cochlea to the primary auditory cortex consists of
Ventral and Dorsal Cochlear Nucleus, Superior Olivary Complex, Lateral
Lemniscus, Inferior Colliculus, and Medial Geniculate Body. So by the time the
signal reaches is area 41, it is already processed for directional information as
well as for some speech related features such as complex spectrum and onset
times.
Speech
motor control requires exquisite coordination of many muscles, so the output
from the speech related primary motor cortex (inferior part of area 4, which in
turn receives from area 6) goes out to many nerve tracts, including cranial
(V:Trigmantal, VII: Facial, IX/X: Glosspharyngeal/Vagus, XII: Hypogrossal) and
Laryngeal nerves. Besides these nerves from the primary motor cortex called
pyramidal tracts, there are extrapyramidal speech tracts that goes from the
cerebellum, premotor cortex via basal ganglia, and output to thalamus, which
nevertheless do not directly innervate the lower motor neurons that control
speech related muscles. Unlike the auditory signal pathway, coordination of
these multiple output does not involve many lower level ganglions.
Caplan
D. The neurobiological basis of language. Brain. 2007 May 1;130(5):1442-6.
Damasio
AR, Geschwind N. The neural basis of language. Annual review of neuroscience.
1984 Mar;7(1):127-47.
Friederici
AD. The brain basis of language processing: from structure to function.
Physiological reviews. 2011 Oct 1;91(4):1357-92.
Hickok
G, Poeppel D. The cortical organization of speech processing. Nature Reviews
Neuroscience. 2007 May 1;8(5):393-402.
Stowe
LA, Haverkort M, Zwarts F. Rethinking the neurological basis of language.
Lingua. 2005 Jul 31;115(7):997-1042.
Tallal
P, Miller S, Fitch RH. Neurobiological basis of speech: a case for the
preeminence of temporal processing. Annals of the New York academy of sciences.
1993 Jun 1;682(1):27-47.
