Transcript File - topics in the study of mind

Evolution of Brain and Language
II
Linguistic Structure
The system that had to evolve to
make it possible for us to speak
Course website
• www.mind-study.org/evolution
2
Outline of Topics for the Course
1. Introduction
1.
2.
How evolution works
The human brain
2. Linguistic structure: What had to evolve
3. Human evolution from 5 million BP to 1 million BP
How/why the brain grew so large
4. Early stages of language evolution:
From 3,000,000 BP to 100,000 BP
5. Later stages: From 100,000 BP to 1,000 BP
Language spread and diversification
The Indo-European family and other families
6. The last few hundred years
The exponential progress of evolution
3
2 – Linguistic Structure
• We need to know what it is that had to evolve
• What is linguistic structure?
• How is it represented in our brains?
4
Major anatomical-functional dichotomies
• Left hemisphere vs. Right hemisphere
– Left
• Analytical, linguistic, digital
• Maintains existing beliefs
– Right
• Metaphorical, artistic, analog
• Open to new data and ideas
• Front (anterior) vs. Back (posterior)
– Front
• Action and planning of action
• Process oriented
– Back
• Perception
• Perceptual integration
• Object oriented
5
Left hemisphere vs. right hemisphere
 Left hemisphere
–
–
–
–
–
–
Language
Analytical thinking
Exact
Digital
Heightened contrast
Proof
 Right Hemisphere
–
–
–
–
–
–
Art, music
Holistic thinking
Metaphorical
Analog
Fuzzy boundaries
Hunches, intuition
6
Cerebral dominance for language
• Linguistic abilities are subserved by the
left hemisphere in about 97% of people
– 99% of right-handed people
– A majority of left-handers
• But this is just a first approximation
7
How does all this complex structure work?
• Representation of information in the brain is . .
– quite unlike that in computers
– in fact unlike anything else we have ever known
• Extraordinarily complex
– The cortex has billions of neurons
– Trillions of interconnections
• How can we make sense of it?
• Brain function is revealed by linguistic structure
• “Language is the window of the mind”
8
The linguistic system of the brain
• It does not contain words
– The brain is not a system that stores words
• Rather, it is the system which
– Produces words
– Comprehends words (if incompletely)
• Likewise, it does not contain rules (e.g., for syntax)
– Rules would be linguistic products
9
Relationship of toy and boy to their phonemic expressions
boy
toy
Structure that produces
or recognizes /boy/
t-
b-
a.
-oy
Relationship of toy and boy to their phonemic expressions
t-
b-
a.
-oy
boy
toy
boy
toy
t-
b-
b.
-oy
Relationship of toy and boy to their phonemic expressions
t-
b-
a.
-oy
t-
toy
boy
toy
boy
toy
b-
b.
-oy
t-
boy
b-
c.
-oy
–oy: further details
toy
boy
boy
toy
-oy
t-
b-
-oy
t-
bo
-y
Add phonological components
boy
toy
-oy
t-
b-
-y
o
Vl
Ap
Cl
Lb
Ba
Vo
Sv
Fr
Alternative catalysis
boy
toy
boy
toy
t-
o
b-
-y
-oy
t-
Vl
b-y
o
Vl
Ap
Cl
Ap
Lb
Ba
Vo
Sv
Fr
Cl
Lb
Ba
Vo
Sv
Fr
Irregular past tense
PAST
TAKE
take
took
-d
t-
e
-y
u
-k
Add complex lexemes
UNDERTAKE
OVERTAKE
TAKE
PAST
take
over
under
took
-d
t-
e
-y
u
-k
Some complex noun lexemes
NEW
NEW-AGE-MUSIC
NEW-AGE
AGE
new
age
MUSIC
music
HORSE-BACK-RIDE
HORSE-BACK
BACK
HORSE
horse
back
ride
RIDE
Synonymy and Polysemy
HUMAN-BEING
MALE
synonymy
polysemy
person
human being
man
Syntactic Constructions:
Variable complex lexemes
CONNECT-THE-DOTS
DO-SMTHG-TO-SMTHG
TRANSITIVE VERB
connect
the
dots
THING
More Syntax
DO-SMTHG-TO-SMTHG
BE-SMTHG
INTRANS VERB
TRANS VERB
BE VERB
LOC
QUALITY
THING
Statements and Yes-No Questions
ASK
DECLARE
Examples:
DECLARE
Johnny can swim
PRED
ASK
Can Johnny swim?
SUBJ
FINITE
What are meanings?
For example, DOG
In the Mind
The world
outside
Conceptual
properties
of dogs
Perceptual
properties
of dogs
All those dogs
out there and
their properties
The concept DOG
• We know what a dog looks like
– A visual subnetwork, in occipital lobe
• We know what its bark sounds like
– An auditory subnetwork, in temporal lobe
• We know what its fur feels like
– A somatosensory subnetwork, in parietal lobe
• All of the above..
– constitute perceptual information
– are subnetworks with many nodes each
– Are interconnected into a larger network
24
The concept DOG as a network
A – Auditory
C – Conceptual
M – Memories
P – Phonological
T – Tactile
V - Visual
T
A
P
C
V
M
Each node in this diagram
connects to a subnetwork
of properties
25
Some nodes of the cortical net for fork
T
M
PP
C
P
PA
V
26
Some nodes of the cortical net for fork
T
M
PP
C
P
PA
V
27
A word network with two subnets partly shown
T
C
PP
PR
PA
M
C – Cardinal concept node
M – Memories
PA – Primary auditory
PP – Phonological production
PR – Phonological recognition
T – Tactile
V 28– Visual
V
Visual features
Ignition of a word network from visual input
T
C
PR
Art
PA
M
29
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
30
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
31
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
32
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
33
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
34
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
35
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
36
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
37
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
38
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
39
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
40
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
41
V
Ignition of a word network from visual input
T
C
PR
Art
PA
M
42
V
Speaking as a response to ignition of a net
T
C
PR
Art
PA
M
43
V
Speaking as a response to ignition of a net
T
C
PR
Art
PA
M
44
V
Speaking as a response to ignition of a net
T
C
PR
Art
PA
M
From here (via subcortical
structures) to the muscles
that control the organs of
speech articulation
45
V
An MEG study from Max Planck Institute
Levelt, Praamstra, Meyer, Helenius & Salmelin, J.Cog.Neuroscience 1998
46
Timing of neural pathway travel
• Neuron-to-neuron time in a chain (rough estimate)
– Neuron 1 fires (say, @ 100 Hz)
• Time for activation to reach ends of axon
– 10 mm @ 10 mm/ms = 1 ms
• Time to activate post-synaptic receptor – 1 ms
– Neuron 2
• Activation reaches firing threshold – 4 ms (??)
– Hence, overall neuron-to-neuron time – ca. 6 ms
• Time required for spoken identification of picture
– Subject is alert and attentive
– Instructions: say what animal you see as soon as you see
the picture
– Picture of horse is shown to subject
– Subject says “horse”
– This process takes about 600 ms
47
Some types of Meaning
• Conceptual
– Concrete—CAT, CUP
– Abstract—CONFLICT, PEACE, ABILITY
– Qualities/Properties—HELPFUL, SHY
• Perceptual
–
–
–
–
Visual—BLUE, BRIGHT
Auditory—LOUD, MUSICAL
Tactile—ROUGH, SHARP
Emotional—SCARY, WARM
Represented as
nodes in different
areas all over the
cortical network
• Processes
– Material
• Low-Level—STEP, HOLD, BLINK, SEE
• Mid-Level—EAT, TALK, DANCE
• High-Level—NEGOTIATE, EXPLORE, ENTERTAIN
– Mental
• THINK, REMEMBER, DECIDE
• Relations
– Locational—IN, ABOVE
– Abstract—ABOUT, WITH-RESPECT-TO
48
The Role of RH in semantics
• Conceptual information, even for a
single item, is complex
– Therefore, widely distributed
– A network
– Occupies both hemispheres
• RH information is more connotative
– LH information more exact
49
Some connections of the concept node CUP
HANDLE
SHORT
CERAMIC
TAPERED
MADE-OF-GLASS
WITH-SAUCER
CUP
Different lines (connections) have different strengths
Abstract notation and narrow notation
Abstract Notation
Bidirectional
TAKE
PAST
take
took
-d
Narrow Notation
TAKE
take
Downward
took
PAST
-d
TAKE
take
Upward
took
PAST
-d
Abstract notation and narrow notation
Abstract Notation
Bidirectional
ab
a
b
Narrow Notation
ab
a
Downward
b
Upward
a
ab
b
Two speech areas
Primary Oral
Motor Area
Primary Auditory
Area
Primary Oral SomatoSensory Area
Variation in strength of connections
1) Connections (shown in graphs by lines) differ from one
another in strength. A stronger connection transmits
more activation than a weaker one, if both are
receiving the same amount of activation.
2) Nodes have threshold functions, so that outgoing
activation varies with amount of incoming activation;
and different nodes have different threshold functions.
3) A connection of a given strength can carry varying
degrees of activation from one moment to the next,
since each node is sending out varying degrees of
activation in accordance with property 2.
54
Learning
• Network structures have to be built for language
– And other kinds of information
• But they are not actually built
• The neural connections are already there
• Those needed for a particular language are selected
– At each of multiple layers of structure
– “Neural Darwinism” (Edelman)
•
•
•
•
Prerequisite: Abundance of connections in the initial state
Then learning is a process of selection
Latent nections become dedicated
Latent connecting lines become established
55
Can we justify the abundance hypothesis?
• Number of neurons in cortex (avg.): ca. 27.4 billion
• Neurons beneath 1 mm2 of surface: ca. 113,000
• Neurons beneath 1 cm2 of surface: ca. 11,300,000
56
Extent of neuronal fibers in the cortex
• Estimated average 10 cm of fibers per neuron
(conservative estimate)
• Avg. cortex has about 27 billion neurons
• 27 billion X 10 cm = 2.7 billion meters
Or 2.7 million kilometers
– About 1.68 million miles
– Enough to encircle the world 68 times
– 7 times the distance to the moon
57
Number of synapses in cortex
• 40,000 synapses per neuron (4x104)
• Times 27 billion neurons (27x109)
• 4x104 x 27x109 = 108x1013
or about 1.1x1015 (over 1 quadrillion)
58
Strengthening of Connections: Learning
C
C
A
A
B
B
If connections AC and BC are active at the same time, and
if their joint activation is strong enough to activate C, they
both get strengthened and the threshold of C is adjusted
Language networks and neural networks
• For this question we have to consider the narrow notation
– It is less abstract
– Closer to the neural substrate
• Comparing Relational Networks and Neural Networks
–
–
–
–
–
–
RN lines and nodes (narrow notation) are unidirectional
Nerve fibers and cell bodies are unidirectional
RN Connections are either excitatory or inhibitory
NN Connections are either excitatory or inhibitory
RN: Inhibitory connections are to either a node or another line
NN: Inhibitory connections are to either a cell body or another axon
60
Connections in RN and NN
Neural Networks
Relational Networks
•
•
•
•
•
Unidirectional
Excitatory or Inhibitory
Two types of inhibitory
Different strengths
Varying activation
•
•
•
•
•
Unidirectional
Excitatory or Inhibitory
Two types of inhibitory
Different strengths
Varying activation
61
Thresholds
•
•
•
•
In RN, every node (of narrow notation) has a threshold
Outgoing activation is a function of incoming activation
In NN, every cell body has a threshold
Outgoing activation is a function of incoming activation
62
But nodes of narrow RN do not correspond to neurons
• Rather, they correspond to cortical columns of neurons
• The cortical column in the functional unit of the cortex
• A column consists of around 100 neurons
– Vertically stacked on top of one another
• In a cortical column . .
–
–
–
–
All pyramidal neurons have the same response properties
Redundancy
But different pyramidal neurons project to different other areas
There are also inhibitory neurons
• They turn off activation as needed
• In same column or in neighboring columns
63
The node of narrow RN notation
vis-à-vis neural structures
• The node corresponds not to a single
neuron but to a bundle of neurons
• The cortical column
• A column consists of 70-100 neurons
stacked on top of one another
• All neurons within a column act together
– When a column is activated, all of its neurons
are activated
Microscopic
views of cortex
Different stains
show different
features
65
Some Properties of the (mini)column
• ROUGHLY CYLINDRICAL IN SHAPE
• CONTAINS CELL BODIES OF 70 TO 110 NEURONS (TYPICALLY 75-80)

ABOUT 70% ARE PYRAMIDAL,

THE REST INCLUDE
OTHER EXCITATORY NEURONS (spiny stellate)
SEVERAL KINDS OF INHIBITORY NEURONS
• DIAMETER IS ABOUT 30–50 M IN, SLIGHTLY LARGER THAN THE
DIAMETER OF A SINGLE PYRAMIDAL CELL BODY
• TWO TO FIVE MM IN LENGTH, EXTENDS THRU THE SIX CORTICAL LAYERS
• IF EXPANDED BY A FACTOR OF 100, THE DIMENSIONS CORRESPOND TO A
TUBE WITH DIAMETER OF 1/8 INCH AND LENGTH OF ONE FOOT
• THE ENTIRE THICKNESS OF THE CORTEX (THE GREY MATTER) IS
ACCOUNTED FOR BY THE COLUMNS
• (BASED ON MOUNTCASTLE 1998)
66
That’ s it
for today
67