How does the brain work?

Włodzisław Duch

Computational Intelligence Laboratory,
Department of Informatics,
Nicolaus Copernicus University,

Grudziądzka 5, 87-100 Toruń, Poland.

e-mail: id: wduch, na serwerze fizyka.umk.pl.

WWW: http://www.fizyka.umk.pl/~duch

Understanding mind/brain: the final frontier.
Ghosts are things of the past.
Mind is what the brain does.
Dualism is dead (literarily).

Brain research, behavioral neurophysiology, cognitive neuroscience, psycho-neuro-immuno-logy ... many fields coming together.
Can we understand the brain? Can we understand H₂ ?

Types of questions:

• Why? Why zebras have stripes? Why taboos exist?
• How? How are brain functions working?

Why: examples.

Sociobiology, Evolutionary Psychology, Evolutionary Medicine ...

Why do we have brains?
To control movement! Some see creatures eat their brains when they stop moving!
Why left-right crossing in brain control?
Brain developed from ganglia in multi-segment worms which have a coil response.
Igbo West African cultures raise yam plants and harvest it at the beginning of the rainy season. Religious taboo: yam feast at the end of the rainy season, despite shortage of food. Why? Rainy season: mosquitoes spread malaria; many people have sickle cell anemia, a genetic disorder; become more resistant to malaria; yams contain a compound that combats the sickling of red blood cells.

How questions: at what level?

Each level allows to answer specific "how" questions. Strong coupling between the levels, no systematic approximations leading from one level to another. Basic neuroanatomy and localization of functions. Brain imaging techniques: MRI, fMRI, EEG, MEG, SPECT, combinations. Lesions, diseases and injuries, controlled experiments on animals.

Brain as modules - neuropsychology.

Information flow and analysis explains many amazing phenomena.

Some neuroanatomy.

Triune brain: archipalium (reptilian brain), paleopalium (mammal brain), neopallium (rational brain)

Divided brain: hemispheres and specialization. "Dominant" hemisphere (usually left): language, speech understanding, reading, logical thinking, internal narration and dialog, precise and analytic, conscious. The mute other: spatial orientation, melody and rhythm, complex tasks, general patterns, emotion analysis, simple language without grammar, non-verbal (prosody) cues. In normal brain strong interhemispheric cooperation. Some differences between men-woman, ideographic-alphabetic writing users.

Neocortex - about 10¹⁰ neurons in many specialized areas, about 10¹⁴ connections.

3 basic functions of the cortex:

• Sensory analysis, precise discrimination of sensory data.
• Motor control, planning, coordination.
• Higher cognitive functions: memory, thinking, planning, emotional (value) analysis.

Sensory analysis: primary, secondary (association), tertiary (multimodal) areas.

Motor control: primary, secondary areas + cerebellum - ca. 3 x 10¹⁰ neurons but only about 10 ¹³ connections.
Association cortex - prefrontal, parietal, various gyri.

Brain lobes, 4 views: lateral, dorsal, medial and basal.

BrainlobesX, lateral lobes movie.

Brainlobesmid medial lobes movie.

Planum temporale movie.

Each gyrus (or part) has specific function (sometimes several functions).

How does cortex work?

Two basic mechanisms: topographical maps and population coding.

Sensory analysis: reduce input dimensionality, preserve topographical relations.
Somatosensory, motor, tonotopic, visual maps; maps also in cerebellum, thalamus and other subcortical structures.

Somatosensory topographic maps in postcentral gyrus.
Somatosensory radiation movie.

Recognition: fast thalamo-limbical route and slower cortical, ventral (object recogniton) and dorsal (movement, action) streams.

VPL=Ventral Posterior Lateral Nucleus
VPM = Ventral Posterior Medial Nucleus

Map of motor cortex

Each of the 6 senses has its cortex, primary and association.

Vision

Optic radiation movie. Ungerleider and Mishikn (1982): what and where streams, ventral and dorsal. Millner and Goodale (1995): perception and action.

Retina => Lateral Geniculate Nucleus => already separated, magnocellular (large cells), light intensity/contrast and motion; and parvocellular (small cells), color perception and detail.

Vision in macaque
Vision in humans

Do we need eyes to see? No. People that became blind still see! (cf. O. Sacks, Anthropologist from Mars)
Perception: visual if spatial relations preserved.

Two streams and subcortical: LGN => parietal, temporal.
Affective and cognitive recognition: affective seems to be more fundamental, but may be quite false!

Example: delusional misidentification syndromes (DMS)
Capgras delusion: wife, child or well-known person replaced by a double (one-self also).
Dissociation between affective and cognitive recognition - limbic lobe damages and right hemisphere angular gyrus (parietal-temporal-occipital).

Cotard delusion - I'm dead!

Many neuropsychological syndromes are understandable at the modular level:

• agnosias - inability to recognize something, cf. anosagnosia, denial of paralysis.
• asymbolia - lack of understanding of meaning, cf. pain
• alexia - inability to read
• agraphia - inability to write; you may write but will not be able to read that!
• aphasia - problems with speech understanding or production
• apraxia - inability to make skill movements

Subcortical structures: thalamus, limbic system and brain stem

Allobjects movie.

Thalamus - key structure, gating information that we are conscious of.

Brain as networks

Neural models explain associative memory as a cooperative phenomenon.
Memory: distributed, in strength of synaptic connections.
Minicolumns and microcolumns.

'Synfire chains' - minicolumns talking to each other.
Example of simulations of object recognition by monkeys (for example, by Miyashita).

Link system - hipocampus.
Modulatory system - other limbic structures.

Global features: binding, working memory (WM).

Consciousness as the ability to discriminate and make non-verbal comments on the WM states.
Computer psychiatry: mostly associative memory models.
Simulations possible for ca. 1000 neurons, 1 mln synapses.
Neurophysiological quality simulations for single neurons and few neurons.

Brain as molecules

Why neurons and networks work the way they do?

Psychopharmacology, neurotransmitters, neuromodulators (important NO, CO), many neuroactive substances.
Molecular and genetic mechanisms of learning: LTP, LTD, correlation of activity
Psychiatrical diseases, like Alzheimer, schizophrenia: etiology hard to find
Dopamine and serotonin: produced mostly in the brain stem nuclei.

Radical proposals (Ira Black): brain as biochemical network; instead of electrical potentials concentration of neurochemicals as dynamical variables.

Open problems

Brain Maps will be more and more accurate.

But where is the mind?

How to make simple models connecting neural and psychological levels?
Using psychological spaces! Mind space as arena for mental events.
Some applications to categorization.

Molecular level - still poorly understood, important from medical point of view.

High-dimensional dynamical models and their stable states.
Feeling in artificial system - "dressed global dynamical states" necessary ?

Some neuropsychological syndromes are hard to understand; cf. complex models of neglect.

Binding of brain activity into the stream of consciousness - requires understanding of EEG.
The race to find neural correlates of conscious states is on.
Creation of semantic memory - how episodes change into structured knowledge?
Going from associative memory to sequential reasoning, systematic problem solving.

How can creativity arise in complex systems? Complex representations + fast search.
Some AI programs beat humans in games, theorem proving.
Ex: Robbins 1936 conjecture was proved in 1996 by a program.

Włodzisław Duch