Lecture 1 Definitions: Perception Intro and Neuroimaging

sensation

the immediate, basic experience generated by external stimuli

perception

the interpretation of sensations, giving them meaning and organization

bottom-up processes

previous experience shouldn’t matter or affect how you process stimuli

top-down processes

processing of info is influenced by expectations which can influence and cause different perceptual experience

thresholds

finding the limits of what can be perceived

signal detection theory

measuring difficult decisions

sensory neuroscience

the biology of sensation and perception

action potential

signal that travels down the length of the axon

refractory period

harder to have an action potential during this time

neuroimaging

set of methods that generates images of the structure and/or function of the brain

computerized tomography (CT)

produces slices of images of the brain based on x-rays passing through the body

magnetic resonance imaging (MRI)

imaging technology that uses the responses of atoms to strong magnetic fields to form images of structures like the brain. The method can be adapted to measure activity in the brain as well

indirect measure

measuring the consequences of activity

direct measure

picking up electrical potentials or activity

blood oxygen level-dependent (BOLD) signal

the ratio of oxygenated to deoxygenated hemoglobin that permits the localization of brain neurons that are most involved in a task; used in fMRI

positron emission tomography (PET)

imaging technology that enables us to define locations in the brain where neurons are especially active by measuring the metabolism of brain cells using safe radioactive isotopes

electroencephalography (EEG)

measures brain electrical activity at the scalp from the rapid post-synaptic potential changes of pyramidal cells in the cortex

pyramidal cells

highly concentrated at upper portion of cortical surface

event-related potentials (ERP)

a measure of electrical activity from a subpopulation of neurons in response to a stimuli. created by averaging together many EEG trials time-locked to a specific event

magnetoencephalography (MEG)

a technique similar to EEG that measures changes in magnetic activity across populations of many neurons in the brain

functional magnetic resonance imaging (fMRI)

a variant of magnetic resonance imaging that makes it possible to measure localized patterns of activity in the brain. Activated neurons provoke increased blood flow, which can be quantified by measuring changes in the response of oxygenated and deoxygenated blood to strong magnetic fields

rate coding

measuring the response of a neuron by summing the number of times it fires in some interval

spike timing

neurons produce spikes that vary rhythmically with the input signal

population coding

combining the responses of many interacting neurons

computational models

the use of mathematical language and equations to describe steps in psychological and/or neural processes (often implemented with a computer)

statistical optimization models

a computational account describing how a perceptual system uses the statistics of past experience to improve its current performance

efficient coding model

theoretical and/or computational models that explain neural processing by assuming that sensory systems become tuned to predictability in natural environments in ways that economically encode predictable sensory inputs while highlighting inputs that are less predictable

maximum likelihood estimation

makes the best use of multiple sources of information about the same physical property of the world in order to estimate its quantity

bayesian models

theoretical and/or computational models that employ bayesian statistical methods to generate an internal model of the source of sensory inputs based on prior experience

artificial neural networks

computational methods that consist of networks of nodes with weighted connections between them. connection weights increase and decrease following experience in ways that resemble the organization of biological neural networks

deep neural networks (DNNs)

a type of machine learning in AI in which a computer is programmed to learn something. Has large number of layers and nodes with millions of connections. Network is trained with known answers, and can subsequently provide answers from input it has never seen before