Midterm Flashcards
what is cognitive neuroscience
study of psychological processes and how they’re implanted in the brain
3 keys to conducting an fMRI study
- Design a behavioral task that isolates the psychological processes of interest in the brain
- Modify the behavioral task so that it’s suitable for an fMRI study
- Analyze the behavioral and fMRI data
ethical considerations of studies
informed consent
debriefing (tell purpose of study at the end)
privacy/confidentiality
fraud (making up data)
Four goals of scientific research
Describe: observe a behavior, make a prediction
Predict
Determine causes
Explain mechanisms
3 types of studies
descriptive: observational, notice patterns in the world and report them
correlative: notice a pattern a between two different variables and try to determine how they’re related
true experiments: random assignment, manipulate a variable
within vs between subject design
between: groups receive different treatments
within: each subject receives all treatments (makes it harder for 3rd variable to affect results
within subjects design limitation
carry over effects: effects from one experimental condition carry over to the next
timing/order of treatments matter
example of confounds
clever hans: cues from experimenters affected horse’s reaction
bigfoot
congruency effect
reaction times are faster when the target and distractor are congruent than when they are incongruent
Congruency sequence effect
(aka conflict adaptation)
the congruency effect is smaller when the previous trial is incongruent than when congruent
Which process drives the CSE?
- Selection for action (botvinick)
- shift attention toward the relevant target
- Response inhibition (ridderinkhof)
- inhibit the response engendered by the distractor
- Learning and memory confounds (mayr)
botvinick et al.
selection for action (previous theory): ACC focuses attention on stimuli we want to act upon
conflict monitoring (botvinick): ACC activated when conflict is detected, dorsolateral prefrontal cortex (DLPFC) activated, then attention is increased
Mayr et al.
- Mayr et al predicted:
- CSE happens when target and flanker are repeated
- No CSE when either target or flanker change
- Found that conflict adaptation occurred only in repetition trials, contradicting the conflict monitoring model
- CSE comes back when stimulus is repeated in n-2 to n
- Reduced ACC activity on iI (vs cI) trials reflects repetition priming not conflict monitoring
selection for action (attention adaptation) triggered by:
perceptual expectation (expectations about what comes next)
conflict monitoring
negative affect (almost pressing wrong button makes you frustrated)
cI vs iI trials in botvinick et al.
cI trials: higher conflict
iI: high selection for action
cI trials had greated ACC response
learning and memory processes that may influence the CSE include:
feature repetitions (mayr et al.)
contingency learning biases (schmidt)
Mayr et al used what kind of task
2-AFC flanker task
to avoid feature repetitions, some researchers have…
employed tasks w/ larger stimulus sets (ex: ullsperger et al.)
contingency learning confounds/biases
Ullsperger: 9-AFC (alternative forced choice)
with larger stimulus sets (to avoid feature repetitions), congruent and incongruent trials are presented equally, even though there are fewer unique congruent stimuli (shown 8x often)
flankers (ex 11 11) appeared w/ congruent target (ex 1) 50% of the time but w/ incongruent targets (22 22, etc.) only 6.25% of the time
lead to CSE
describe both experiments in mayr et al.
experiment 1: standard flanker task, but analyzed trials by whether they were a change or repetition
experiment 2: removed repetition by alternating b/w up/down and left/right arrows
what does AFC stand for
what do classic experiments use
mayr et al
ullsperger
alternative forced choice
classic: 2-AFC (left/right)
mayr: 4-AFC (right/left/up/down)
ullsperger: 9-AFC (9 possible responses)
how does a contingency learning bias lead to a CSE
particpants want to respond fast/accurately, so learn strategy of making response congruent w/ flankers
after congruent trial (strategy worked!), contingency bias increases (faster RT cC trials, slower to cI)
mental rotation confounds
Kunde and Wuhr
observed a CSE in a 4-AFC, even in trials w/o exact feature repetition or contigency learning biases)
however, all arrow stimuli were mental rotations of the same arrow stimulus
Prime probe word task (Schmidt and Weissman)
avoided feature repetitions, contigency learning, and mental rotation confounds
- 4-AFC task (left-right vs up-down)
- Alternated b/w these two tasks every trial (no response repetitions)
- Presented congruent/incongruent stimuli equally (no contigency learning)
conclusion of schmidt and weissman
CSE can be observed independent of confounds
a control process influences the CSE, but does not reveal which one:
selection for action (aka attention adapation) or response inhibition
how does prime probe word task of in class experiment differ from schmidt and weissman?
increase time separating the prime (distractor) from the probe (target) from 33 ms to 800 ms
logic of class experiment (increasing time b/w distractor and target)
overall congruency effect should vanish
however, subjects may still use their memory of what the previous trial was to predict what the next trial will be
therefore, should observe a CSE even though there is no overall congruency effect
class experiment: selection for action (attention adaptation) theory
after an incongruent trial and w/ long time b/w distractor and target, subjects can shift all attention to the target when it appears
if they shift all their attention to the target, there won’t be any attention left to identify the distractor
thus, there should be no congruency effect after incongruent trials
class experiment: response inhibition theory
w/ a long time b/w target and distractor, subject inhibit response signaled by the distractor after incongruent trials
if they do, there should be a negative (reverse) congruency effect after incongruent trials (faster on iI than iC)
CSE graph label
p value
probability that null hypothesis is false
null hypothesis = no significant difference b/w two groups
one tailed t test
test if sample mean is larger or smaller than population mean
significant when the difference b/w means are large enough that it’s unlikely to occur by chance
5% in one tail
two tailed t test
test if sample mean is larger or smaller than population mean
2.5% in each tail instead
paired (one-sample) t-test
two sample means come from the same subjects
compare subjects to themselves
unpaired (two-sample) t-test
the two sample means come from different subjects
main effect
dependent variable changes based on the independent variable
interaction
the effect of one independent variable on the dependent variable varies based on another independent variable
unparallel lines
simple effects
limit analysis to one of the independent variables to see the what is driving an interaction
only calculated when an interaction is significant
clinical applications of fMRI example
pateints with traumatic brain injury (TBI) still had some brain activity when imaged with fMRI
basic science and fMRI: movies and brain activity
people shown movies while in fMRI
reconstruct movie based on brain activity in the primary visual cortex
dogs and fMRI example
do dogs have similar brain activity as humans?
using fMRI, found that area of reward in dogs correlates to same area in humans
first brain imaging experiment
Angelo Mosso
humans lay on seesaw, thought that fi people think hard, blood would rush to head and tip the seesaw
Blood Oxygen Level Dependent (BOLD) signal
an indirect measure of neural activity
increased neural activity → increased blood flow → flow of oxygenated blood (oxy/deoxyhemoglobin ratio increases) → increased BOLD signal
typical BOLD signal looks like what
initial undershoot: due to neural area using up oxygen before getting more oxy blood
oxy/deoxy ratio increases
oxy/deoxy ratio goes below baseline (use up oxy blood)
what does conducting an fMRI study involve
put subject in scanner
have subjects perform cognitive tasks
record the BOLD signal
how is BOLD signal recorded
multiple horizontal slices taken, then divided into many small cubes (voxels)
high spatial resolution
TR
time it takes to get a whole set of slides
cognitive subtraction
2 conditions only differ in one respect, parts of brain that are more active is due to that difference
assumption of pure insertion
assumption for cognitive subtraction
adding new component to a task doesn’t change the basic processes that were already in the task (can add w/o changing pre-existing processes)
interpreting the bold signal
increased bold signal does not mean more neurons are firing
means more metabolic activity is occuring, but could be excitatory or inhibitory activity
fMRI vs normal MRI
MRI (structural MRI): used to study brain anatomy
fMRI (functional): used to study brain function
how does structural MRI work
normally: magnetic fields of proteins initially random
in scanner: some protons align
radio wave pulse: orients protons and produces signal
eventually protons relax back to original orientation
diffusion tensor imaging (DTI)
MRI based method for imaging the axon tracts that form the white matter of the brain
DTI measures density and motion of water that travels along myelin-covered axons
water normally diffuses in all directions (isotropic), but diffusion along myelinated axon is limited (anisotropic), so water moves only along length of axon bc myelin creastes lipid boundary that influences water movement
sports and head injury research
concussions associated w/ chronic traumatic encephalopathy (CTE)
tau protein builds up in brain
problems w/ memory, concentration, depression
3 main parts of the brain
neocortex, limbic, brainstem/cerebellum
Layer 1 of the brain
brain stem: breathing, swallowing, bladder, HR
cerebellum: balance, coordination
Layer 2:
limbic system
amygdala: emotion
hippocampus: memory
cingulate gyrus/cortex: conflict, emotion, pain
layer 3: neocortex
parietal lobe: touch
frontal lobe: higher order thinking, planning, decisions
temporal lobe: hearing, memory
occipital lobe: vision
anatomical directions
rostal: toward nose
caudal: back of head
ventral: bottom
dorsal: top
parts of cingulate cortex
Parvizi et al.
found that the anterior midcingulate cortex (aMCC) plays an important role eliciting determination and motivation within an individual to face challenges
Coleshill et al.
studied unilateral hippocampal electrical stimulation in epilepsy patients; found that that stimulation of the left hippocampus produced recognition memory deficits for words, while right hemisphere was associated with face recognition deficits
longer neural signal leads to _____ BOLD response
taller and longer BOLD response
superposition
fundamental basis for block design
individual responses summate to yield a large response → large BOLD response
how long is ideal for a block
10-20 seconds
allows enough time for the BOLD signal to go back to baseline so we can actually see signal
pros of block vs event related design
block: easier to see overall signal
event related: easier to see shape per trial
block design graph: why is there a dip
undershoots slightly from initial trials
signal from initial trials go back to baseline and drag the sum signal down
owen et al.
infer cognitive function in vegetative state
had patients/controls imagine tennis, then imagine navigation (of living room)
tennis and navigation produce different patterns of activity, and patients and controls had similar activation for both tasks
faces vs houses experiment
fusiform face area (FFA) activated more for faces than houses
parahippocampal place area (PPA) activated more for houses than faces
problems of longer blocks
- sagging caused by undershoots summing together (dip in graph)
- lots of noise at low frequences of stimulation, therefore we want a higher frequency (shorter blocks so that noise is a diff frequency than signal)
houses and faces experiment: what does alternating showing houses and faces do
prevent practice/habituation effects
resulting signal reveals main differences in activity b/w block types
maximizes ability to detect differences in activity at the expense of detecting absolute activity
what if we wanted to detect absolute activity with block design?
include low-level baseline condition (fixation) b/w our two block types of interest
signal reveals activity for each block type and differences in activty b/w the two block types
pros and cons of block designs
pros: big signals (summed activation), large signal relative to noise is easier to detect
cons: if subject makes mistake, hard to get rid of just one trial, cant separate activity (male faces vs female faces)
slow event related designs
present 1 trial every 16-20 seconds to isolate BOLD signal to each stimulus/trial
cued stroop task
double dissociation:
ACC activated when there is conflict/incongruent
DLPFC activated more by attention (ink color)
why would you want to go faster with event-related designs?
subjects get bored
linearity assumption
if things add up, you can go faster
if not, BOLD responses differ depending on whether going fast or slow
degree of sub-linearity referred to as a “refractory effect”
2 design types of event-related design: rapid unjittered
randomized trial order, spacing b/w trials is consistent
can see difference b/w conditions, but not how much it changes relative to baseline since baseline isn’t measured
2 design types of event-related design: rapid jittered
vary spacing between trials to get more baseline activity measurements
wagner et al.: recognition memory
- participants saw 80 word trials in rapid jittered design
- fixation trials added to create jitter
- asked if word is abstract or concrete, then did recognition test late
- results: they could predict which words subjects would remember (words w/ more activity)
functional connectivity
if activity in two regions is correlated across time, then these regions may be communicating
functional connectivity: flanker/ stroop tasks
more connectivity b/w ACC and DLPFC in incongruent trials than congruent trials
preprocessing of fMRI data
gets rid of variability in bold signal that is not due to the task
5 parts of preprocessing
slice-timing
realignment
coregistration
normalization
smoothing
slice timing: different way of collected slices in each TR?
- interleaved: all odd or even slices taken first
- ascending (1,2,3,4,….): start at bottom of brain
- descending (24,23,22,…): start at top of brain
slice timing: SPM assumes…
each scan is instantaneous
green curve: comes from slice collected latest because it is farther along in the BOLD signal sooner than the blue curve
slice timing correction
since later slices are shifted left, need to shift slices right to align everything with first slice
slice timing only needed if
temporal dynapics of BOLD responses are important (event-related designs)
if TR is small enough (< 3 seconds)
realignment
- correct for motion of participants’ head during the scan
different types of motion
pitch, roll yaw
normalization
- stretch and warp each participant’s brain to match a template brain
smoothing
replace the intensity value of a voxel w/ a weighted average of its original value plus the neighboring voxels
why do we smooth
to increase signal to noise ratio
to increase chance of seeing common activation b/w subjects
