TASK 6 - DESIGN & ANALYSIS Flashcards
experimental design
= consists of manipulation the subject’s experience and behaviour that is likely to produce a BOLD response
- data acquisition, results, interpretation all depend on thorough experimental design
- what to manipulate: stimulus kind + properties, stimulus timing, subject instructions
- goal is to test specific hypotheses
cognitive subtraction
= compare brain activity in task that uses a particular cognitive component (A+B) to brain activity in baseline task that doesn’t (A) –> infer which region is specialised for this particular component (B)
- assumes pure insertion: addition of B does not change the character of A
- (A+B)-A=B –> stimulus condition-control condition = change in conditions
- statistical analysis of each voxel
cognitive subtraction
- application
- useful especially in combination with blocked designs, allowing for simple modelling of BOLD response, resulting in robust + reproducible results
cognitive subtraction
- limitations
- pure insertion: if violated, we would have an interaction effect that makes imaging data ambiguous at best
- choice of baseline task: strong impact on data that is obtained
- -> requires good cognitive theory about the elements that comprise the task
- ideally: baseline should be as similar to the experimental task as possible
cognitive conjunction
= do multiple subtractions; use more than one control condition
- identify a set of tasks that has a particular component in common (look for regions of activation that are shared across different subtractions (rather than relying on a single)
- two different tasks that have one thing in common (what area is activated by both tasks, shows us our goal (face processing of apes vs. humans))
1) (A+B+AB)-A=B+AB
2) (C+B+CB)-A=B+CB - -> different than subtraction (one thing different, rest the same)
- remedy for pure insertion (cognitive subtraction)
factorial design
- can study main + interaction effects
- main effects: e.g. effect of attentional load, irrelevant of motion –> (A+B)+(C+D)
- interaction effects: e.g. is the difference between motion (vs. no motion) different under low attentional load (vs. high load)? –> (A-B)-(C-D)
cognitive conjunction + factorial design
- assumptions
- doesn’t assume pure insertion: terms will be different for each pair of subtraction
- -> technique is useful for studying cognitive interactions
- linear relationship between BOLD response resulting from the conditions, otherwise some finding may be contaminated by non-predicted interactions
parametric design
= one feature of the stimulus (parameter) stimulus is varied in a continuous way
- measure association between brain activity + changes in the variable of interest
- allows to isolate rhythms of brain, level of activity
- low level of stimulus can be control condition of higher levels –> similar conditions
parametric design
- problems
- increasing a parameter over a certain limit can involve recruiting other cognitive processes that are not necessarily present at lower levels of performance of the neural system
functional specialisation
= implies that a region responds to a limited range of stimuli, distinguishing it from neighbouring regions
- most studies focus on functional specialisation
- -> not necessary to assume that the region is solely responsible for performance on a given task
- -> not necessary to assume that other regions may not also respond to same stimuli
functional integration
= the way in which different brain regions communicate with each other
- model how activity in different regions is interdependent
- used to infer effective/functional connectivity between regions when performing a task
- correlate different regions of brain activity with each other
resting state paradigm
= behaviourally driven fMRI = no task, just chill
- fluctuations in brain activity are little more than noise
- BUT, in brain regions that are functionally connected, the noise levels tend to correlate together
- -> allows to identify sets of networks
stimulus timing
= determine presentation/timing of stimuli to get optimal design
stimulus timing
- block design
= repetition of relatively long blocks with stimuli that are homogenous within one block
- stimuli that belong together in one condition are grouped together
- blocks have duration that usually exceeds the length of the hemodynamic response (15-30s)
- blocks invoke cognitive state; consist of multiple trials sampled from same condition in rapid succession
- BOLD response to condition A = collective average of trial responses
- -> cannot distinguish between differences within the block, because we aim to see the effect of whole block/condition
block design
- Trojano study
- study whether mental imagery and perception make use of same brain areas
a) mental clock test: requires mental imagery of two clocks and mental comparison (only auditory presentation of clocks)
b) perception task: does not require mental imagery and comparison because clocks are visually presented
c) syllable counting task: control task with same attentional load as mental clock test but not requiring mental imagery
Trojano study
- cognitive subtraction
- subtractions: imagery vs. perception
- -> differential activation pattern during mental imagery and perception
Trojano study
- experiment 1
=
Trojano study
- experiment 2
=
block design
- advantages
- more power (= better able to detect a significant but small effect) because of high BOLD signal (because trials are averaged)
- required for state-based processes (processes that require brain to take several trials to adapt)
- suitable for subjects that perceive switch tasks as disruptive
- relatively easy to analyse
block design
- disadvantages
- don’t get HR for single trials, no estimate of time course of HR
- some events cannot be blocked because the task requires that they are unexpected and occur infrequently
- may promote unintended subject strategies, anticipation and habituation
- minimally compatible with behavioural methods (RT, SDT)
- require pre-specification of comparisons and subtractions
stimulus timing
- event-related design
= erfMRI = consists of trials that are usually sampled randomly from the set of experimental conditions or with random inter-trial spacing
- stimuli that belong to different conditions are intermingled –> subsequently separated out for the purpose of analysis
- trials may be spaced out in time to allow for resolution of hemodynamic response
event-related design
- slow-ER
= present 1 stimulus, wait for HR to drop to baseline and then present next stimulus
- BOLD responses are smaller than the blocked design
- record single-trial responses: 1 BOLD for 1 stimulus –> you can directly compare stimuli
- -> very slow, sometimes not representative of neural processing
event-related design
- rapid-ER
= present stimulus in a sequence that doesn’t wait for the HR to drop
- jittering: timing between stimuli is always different
- -> important for analysis of rapid-ER
- assumes linearity: response we receive, is the sum of the two individual stimuli
- -> if the responses sum up linearly, we can deconvolute and separate the original responses
- deconvolution: mathematical abbreviation to ‘unmix’ the several stimuli
rapid-ER
- Wagner study
- rapid presentation of words
- postscan whether words were memorised (or forgotten)
- -> post-hoc scans are important difference to blocked design - compare responses to trials of remembered words and forgotten words
- -> look at areas that were more active during either one of the conditions
rapid-ER
- advantages (over short-ER)
- enhances statistical power
- reduces ability to estimate HRF properties of a single stimulus
- reduces ability to estimate problem of linearity vs. non-linearity of BOLD interaction in overlapping HRFs
event-related design
- oddball ER
= processes that can’t be blocked
event-related design
- advantages
- allow for random intermixing of trial types
- can provide estimate of HR time-course
- separations of HR from artefact events
- multiple post-hoc experimental contrasts possible
- study unusual events
- design compatible with other methodologies (similar to ERP)
event-related design
- disadvantages
- have reduced statistical power
- cannot be used for state-based processes
- may be unsuitable for inflexible subjects
- more difficult to analyse
mixed design
= combination of block design and event-related design
+ can provide info about maintained vs. transient neural activity during paradigm performance
+ extract brain regions exhibiting an item-related pattern of information processing (transient) or a task-related information processing (sustained)
– involves more assumptions than the others
