2. Neuroscience and Connectionism Flashcards
phrenology
- first to offer thoery of localized function
- regions of the brain performed particular activities
- destruction of a given area decreases/eliminates function
- individual’s defined on the basis of the activities of numerous regions
endocast
was assumed that if we can’t open the skull, skull shows shape of the brain
failure of phrenology
- postulated that the skull refelcted the extent to which each function was developed (e.g. larger bumps =more developed faculty)
- regions were not as well-defined as phrenology believed
- more generally, intensely local account of brain function also appears to be incorrect; processes are distributed and plastic (flexible - if brain is damaged, we can still recover function)
levels of neural function
neuronal
- a single, or small set of neurons
neural systems
- large set of neurons that perform specialized function
- typically referred to be a collective name (visual cortex, amygdala, dorsolateral prefrontal cortex)
neuronal communication refresher
- presynaptic neuron transmits signal
- postsynaptic neuron receives signal
- excitatory synapse increases likelihood of firing
- inhibitory synapse decreases likelihood of firing
- action potential
visual system
- rods and cones = activated as photons hit the back of the eye
- bipolar cell receives the signal and transmits it to the ganglion cell
- ganglion cell receive info from many receptor cells
eyes and brain - signals from rods and cones sent through optic chiasm to primary visual cortex
- sensation from one side of the eye = transmitted to the contralateral side of the brain
language
Wernicke’s area
- comprehend meaning of words
Brocas’s area
- production of words
Wernicke and Broca’s area create a circuit together; stimulus (speech) received and processed, usually generating a response (speech)
celia
ear cell
representing information in the brain
- redundancies are built in to neuronal representations
- stable pattern of activation create LTM
- there is no grandmother cell
- patterns give our world meaning
role of neural redundancies
- allow set of neurons to perform multiple tasks at once
- different patterns of neural activity generate the same outcome
levels of function: neuron
- small information processors
- use input-output
- can be highly plastic
levels of function: systems
- perform specific function efficiently
- some innate, some can be learned
levels of function: cognition
numerous systems function togeher to give rise to attention, memory, and decision making
firing patterns of cognitive function
- if patterns of activation store info, we can examine them to determine what cognitive processes are associated with regions of the brain
- look at _____ to find a consistent story
1. accuracy
2. RT
3. patterns of activation (can be temporal or spatial)
a bit of physics (radiation, substances, magnetism)
- radiation entails that energetic particles move through space
- different substances have different densities with certain materials absorbing (or reflecting) energy differently
- magnetism: dependent on the electric charges/currents running through an atom; magnetic fields can be altered with the introduction of current
neuron waste
- neurons expel waste
- allows to examine patterns of activation and their proxies
- these regularities can be used to examine what regions of the brain are active during a task
- more activity = more blood to that region
neuropsychological evidence for cognition
electrical signals
- Galvani’s experiments with muscles
- ERP (event related potential)
brain damage
- Phineas gage: lack of ability to control emotions, impairment of executive function, change in personality
cerebral blood flow (CBF) and oxygen use (BOLD)
ERPs
- electrical signal from the brain caused by neurons firing
- regions fire at different rates depending on stimuli that is presented
- can be measured with EEG (electroencephalography) or MEG (magnetoencephalography)
- EEG mainly for temporal (not spatial) because our skull is so thick
CT scans
- X-Ray Computed Tomography uses electromagnetic radiation
- uses property of certain matter in your body to block x-rays, creating “gaps” in the resulting 2D images
- these gaps represent structures and the 2D images can be assembled into a 3D image of your brain
- assess where the underlying brain damage is
- use is limited because x-rays are bad for you
PET scans
Positron Emission Tomography
- uses biologically active molecule (similar to glucose)
- tracer emits positrons which are detected by a CT scanner
- from this, 3D model of mental activity at a given point in time can be created
MRI
Magnetic Resonance Imaging assumes
- atoms have tehir own magnetic field, in combination with radio waves this field can be reversed
- by examining changes in this field over time, we can obtain a picture at a given moment of the behaviour of atoms within neurons
- given that no radioactive material is required, procedure is a noninvasive way of obtaining a picture of the brain
fMRI
functional Magnetic Resonance Imaging
- has spacial and temporal resolution
- researchers found that hemoglobin was affected by a magnetic field in a different way depending on whether it was oxygen-rich or depleted of O
- given use of O by a cell = dependent on that cell activity , we can use blood flow as a measure of the activation of a given region during a certain task and see the changes over time
- has a lot of false positives if not done properly
connectionism
- one of the most successful models of human cognition
1. collection of units (not necessarily neurons)
2. units have 2 discrete states
3. feedback used from outcome
4. learning rate adjustment determines speed of acquisition of info - biologically more realistic than other systems, like functionalism
basics of connectionism
- for any given task, some layer of nodes (neurons) is activated when presented with a stimulus
- pattern of activation becomes associated with the stimulus through feedback
- the weights between units (strength and association) are adjusted each time the model completes a run
- learning rate can be altered to simulate difficulty of task, length training, etc.
- refer to slide 35
neural properties of connectionism
benefits of connectionism stem from its similiarity to brian B
- complex network of associations of small units
- units can be inhibitory or excitatory
- units = either on or off (they fire or they don’t)
parallel distributed processing
- units = either on or off (they fire or they don’t)
- e.g. bus coming - shape, colour, depth, etc. if it works all at once, RT increases
- many neural connections may be active at the same time
concerns with connectionism
- units = too simplistic; they cannot represent info like the brain does
- has generic units whereas brain has specialized cells and areas for processing
- mechanism that produces feedback in connections is not supported by neurological studies
neural synchronicity and memory
- important for learning, memory and consicousness: global property
- was hypothesized that a given global pattern of firing could be associated with an item in a memory list
- refer to slides 38-39
connectionism: interim conclusions… for now
- connectionist models provide useful tools for understanding human learning and decision making
- we must understand their assumptions and compatibility with the task we wish to model
- these are limited in their ability to explain B
Hebb rule
- when an axon of cell A is near enough to excite a cell B and repeatedly takes part in firing it, some growth process/metabolic change takes place in one/both cells
- such that A’s efficiency increases, as cell B is firing
- idea that connection between two neurons takes place only if both neurons are firing at about the same time
- has greatly influenced subsequent theorizing
connectionism: identity theory
problem = multiple realizability is difficult
connectionism: functionalism
- problem = syntax does not lead to semantics
- problem = not tolerant to damage
serial processing
only one neural activity may take place at any one time