experimental design i - wordpress.com · experimental design i russ poldrack m285. what do...
TRANSCRIPT
![Page 1: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/1.jpg)
Experimental design I
Russ PoldrackM285
![Page 2: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/2.jpg)
What do scientists do?
• Use experiments to test hypotheses• What is a hypothesis?
– A claim about the effects of some particularmanipulation or factor (independent variable) on asome particular measureable dependent variable
• What is an experiment?– A procedure for manipulating the IV and measuring the
DV
![Page 3: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/3.jpg)
Correlational vs. experimentalresearch
• Experimental research manipulates a factorof interest
• Correlational research examines the effectsof existing variations in a factor
![Page 4: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/4.jpg)
Examples
• Experimental or correlational? What are theIV and DV?– Does gingko biloba result in better memory?– How do the brains of liberals and conservatives
differ?– Does listening to Mozart in utero lead to higher
intelligence?
![Page 5: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/5.jpg)
Confounding variables
• Both experimental and correlational designsare open to confounding effects from othervariables that are correlated with thevariable of interest
• Examples– Does ice cream consumption cause violent
crime?
![Page 6: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/6.jpg)
Different types of hypotheses forneuroimaging
![Page 7: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/7.jpg)
Example hypothesis
• Hypothesis: Stresscauses reducedmemory by affectingactivity in thehippocampus
• How can we testthis?
![Page 8: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/8.jpg)
fMRI experimental design: A basic plan
Define mental processto examine
Define tasks to manipulatethat process
Measure fMRI dataduring tasks
Compare fMRI databetween tasks
Replace “fMRI data” with “RT”and you have cognitivepsychology!
![Page 9: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/9.jpg)
Conceptual and methodogicalaspects of experimental design
• There are two aspects of fMRI design that areimportant to distinguish
• Conceptual design– How do we design tasks to properly measure the
processes of interest?• Aguirre & D’Esposito discuss this
– The issues here are very similar to those in cognitivepsychology
• Methodological design– How can we construct a task paradigm to optimize our
ability to measure the effects of interest, within thespecific constraints of the fMRI scanning environment?
– The book is concerned primarily with this aspect
![Page 10: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/10.jpg)
The subtraction method• Acquire data under two
conditions– These conditions
putatively differ only inthe cognitive process ofinterest
• Compare brain imagesacquired during thoseconditions
• Regions of differencereflect activation due to the“subtracted” process ofinterest
Petersen et al., 1988
![Page 11: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/11.jpg)
Task analysis for Petersen et al., 1988Presented subjects with two conditions:
Letter perception
Word identification
Spoken output
Letter perception
Word identification
Semantic retrieval
Spoken output
Verb generation (“experimental”)Word naming (“control”)
VG is like word naming with semantic retrieval “inserted”
![Page 12: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/12.jpg)
Petersen et al. (1988)
• Found several regionsmore active duringverb generation thanduring word naming
• Can you think of anyreasons why thismight NOT reflectsemantic retrieval?
![Page 13: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/13.jpg)
The pure insertion assumption
• Subtraction requires a strong assumption of “pureinsertion”– Insertion of a single cognitive process does not affect any of the
other processes (no interactions)• Failure of PI means that the results cannot be interpreted
with regard to the specific cognitive process of interest• PI must hold at both neural and cognitive levels
![Page 14: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/14.jpg)
Failure of the PI assumption: Thecognitive level
• Egeth, Marcus, & Bevan(1972)– Compared choice RT to “C-
reaction” in which subjectresponds in one condition andwithholds in the other (i.e.,go/no-go)
– PI assumption: Choice RTreflects insertion of a responsedecision stage over the C-reaction task
– Response condition affected thenature of the memory searchfunction
![Page 15: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/15.jpg)
Failure of the PI assumption: Theneural level
• Jennings et al. (1997)– Compared semantic and
letter judgment tasks withthree different responsemodalities (mouse, vocal,and silent mental)
– Task and response modalityinteracted in left prefrontalcortex
• We don’t know whetherthis reflects cognitive orneural interaction
LIPC
![Page 16: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/16.jpg)
The task analysis assumption
• Subtraction assumes that the task analysis iscorrect– In particular, no other processes are implicitly engaged
by the baseline task, even if they are not required• Example
– Subtraction of word naming from verb generation– What if people automatically engage semantic
processes even if the naming task doesn’t require them?– Can you think of other examples?
![Page 17: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/17.jpg)
The woes of subtraction
• A majority of fMRI studies still use some form ofsubtraction design
• Few of them provide a task analysis to determineexactly which processes are being inserted– Need to use cognitive theories, not folk psychology!
• Few of them deal with the issue of interactionsbetween processes
• How can we overcome these problems?
![Page 18: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/18.jpg)
Conjunction analysis (Price &Friston, 1997)
• Perform several parallel subtractions– Each of which isolates only the process of
interest• Find regions that show common activation
across all of these
![Page 19: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/19.jpg)
from Price & Friston, 1997
![Page 20: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/20.jpg)
Problems with conjunction analysis(Caplan & Moo, 2003)
• Implicit processing– Subjects may engage processes that are not necessary
for the task– Can you think of examples?
• Interactions between processing stages– Conjunction only gets rid of interactions if they do not
activate the same regions to the same degree acrosstasks
• Can we ever verify this?– E.g., dual-route model
![Page 21: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/21.jpg)
Other kinds of fMRI designs
• Parametric designs• Factorial designs• Priming/adaptation designs
![Page 22: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/22.jpg)
Parametric designs
• Employs continuous variation in a stimulus/taskparameter– E.g., working memory load, stimulus contrast
• Inference:– Modulation of activity reflects sensitivity to the
modulated parameter
![Page 23: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/23.jpg)
Boynton et al., 1996
![Page 24: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/24.jpg)
Cohen et al., 1996
![Page 25: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/25.jpg)
Assumptions of parametric designs
• Assumption of pure modulation– Each level of the task differs quantitatively in
the level of engagement of the process ofinterest, rather than qualitatively
– How might this fail?
![Page 26: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/26.jpg)
Factorial design
• A factorial design involves multipleconcurrent subtractions
• Allows for testing of interactions betweencomponents
• Still requires pure insertion assumption andtask decomposition– But additivity can be tested for the specific
factors that are manipulated
![Page 27: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/27.jpg)
Wagner et al., 2001
• Is the left inferior prefrontal cortex moresensitive to the amount of semantic retrievalor the degree of control needed forretrieval?
![Page 28: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/28.jpg)
Wagner et al., 2001: factorial design
Which target is most related to the cue?
beaver bag mousewindow table
ASSOCIATIVE STRENGTH
STRONG WEAK
beaver window xxxx xxxx mouse
beaver table bag dam window
beaver dam window xxxx xxxx NUMBER
OFTARGETS
2
4
![Page 29: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/29.jpg)
z = +24 z = –9
A. Posterior LIPC (~BA 44/6)
B. Anterior LIPC (~BA 47/45)
STRONG–2
STRONG–4
WEAK–2
WEAK–4
T6
4
2
0
A
B
![Page 30: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/30.jpg)
Priming/adaptation designs
• Presentation of an item multiple times leads to changes inactivity– Usually decreased activity upon repetition
• Inference:– Regions showing decreased activity are sensitive to (i.e. represent)
whatever stimulus features were repeated• Requires version of pure modulation assumption
– Assumes that processing of specific features is reduced but that thetask is otherwise qualitatively the same
![Page 31: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/31.jpg)
Wagner et al. (2000)
![Page 32: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/32.jpg)
Can adaptation fMRI characterize neural representations?
• A voxel containing neurons that respond to allpoliticians, irrespective of party
• A voxel containing some specificallyDemocratic neurons, and other specificallyRepublican neurons.
Two stimuli: can neurons tell thedifference?
From R. Raizada
![Page 33: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/33.jpg)
Responses to individual stimulido not show whether neurons can tell the
difference
• Different sets ofneurons are activewithin the voxel,but overall fMRIresponses areindistinguishable
From R. Raizada
![Page 34: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/34.jpg)
Neural adaptation to repeated stimuli does show the difference:What counts as repetition for neurons in a voxel?
It’s a politician Same neurons, adapting:It’s a politician again
It’s a Republican Different, fresh neurons:It’s a DemocratFrom R. Raizada
![Page 35: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/35.jpg)
Example: Raizada, 2002
• Examined “categorical perception” ofphonemes
1-4/4-1 2-5/5-2 3-6/6-3 4-7/7-4 5-8/8-5 6-9/9-6 7-10/10-70
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pooled stimulus type
Pro
porti
on o
f diff
eren
t res
pons
es
1 2 3 4 5 6 7 8 9 100
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Stimulus type: ba=1, da=10
Prop
ortio
n of
da
resp
onse
s
![Page 36: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/36.jpg)
Categorical perception
60 msec
![Page 37: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/37.jpg)
Phoneme category adaptation• Looked for differential
adaptation for repeating itemsthat cross the category boundaryversus those that do not cross theboundary– Same acoustic difference in each
case– Differential adaptation reflects
the degree to which brain areashear those phonemes as different
• Temporo-parietal cortex showsgreater loss of adaptation whenthe stimulus crossed the categoryboundary than when it did not
![Page 38: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/38.jpg)
The functional characterizationapproach
• Many studies take a “functional mapping”approach– The final goal is to say that area X is active in relation
to process A• Another approach is to first identify an area of
interest– Often based on a subtraction design, along with other
knowledge from neuroscience• Then, use multiple task manipulations to
characterize the function of the area
![Page 39: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/39.jpg)
Parahippocampal place area(Epstein and Kanwisher 1998)
∑ Blocks: Faces, objects, houses, scenes∑ Intact and scrambled∑ Analysis:
• ROIs = pixels in target area from an independent scan• Compute signal diff for scram. vs. intact• Diff sig > for scenes than for others
∑ Alternatives and control exp.s• Can you think of other alternatives or controls?
![Page 40: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/40.jpg)
Epstein et al: Localizer
![Page 41: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/41.jpg)
Epstein et al: Functionalcharacterization
![Page 42: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/42.jpg)
Designing an fMRI scan
• Once the task is chosen, we have to devise a wayto present the task in order to measure the effectsof the manipulation
• In an RT experiment, we would just average theresponse time for each trial from each conditionduring the experiment
• In an ERP experiment, we would average theresponse occurring in the period just after the trial(~ 800 milliseconds)
• Why won’t these approaches work for fMRI?
![Page 43: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/43.jpg)
Designs for fMRI• Many people (the text’s authors included) treat
blocked designs and event-related (ER) designs asvery different beasts
• This is largely an historical accident– In the early days, blocked and ER designs required
different analysis tools• The linear systems/general linear model approach
encompasses both within the same framework• It is better to think of designs as falling on a
continuum of general features– How often does each task occur?– How rapidly do tasks change?– How predictable are these changes?
![Page 44: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/44.jpg)
Factors to consider
• Different types of designs vary in a numberof ways:– Statistical power to detect differences between
conditions (detection)– Statstical power to characterize the shape of the
hemodynamic response (estimation)– Effects on behavior– Simplicity to implement
![Page 45: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/45.jpg)
PET designs
• PET pre-dated fMRI as the main tool forfunctional brain imaging
• In PET, the subject has to perform each taskfor 60-90 seconds while the label is takenup in active regions
• Early fMRI studies took a similar approach,presenting different tasks in separate longblocks
![Page 46: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/46.jpg)
Blocked Language Paradigm:Word completion vs. visual fixation
STIPRE
COU
+ + + +
0 30 60 90 120 150 180 210
GRE
STIPRE
PRE STR
STIPRE
ABS
TIME (SEC)
From R. Buckner, HBM2001
What doesthis identify?
![Page 47: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/47.jpg)
438
439
440
441
442
443
444
445
446
447M
R SI
GNA
L
0 30 60 90 120 150 180 210
TIME(sec)
Typical Response
STIPRE
COU GRE
STIPRE
PRE STR
STIPRE
ABS
From R. Buckner, HBM2001
![Page 48: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/48.jpg)
“Castle” paradigm: multiplecondition blocked design
Task A
Task B
Task C
![Page 49: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/49.jpg)
“loose” vs. “tight” comparisons
• Comparing a task of interest to a “low-level”baseline task results in a “loose” comparison– The regions identified by the comparison correspond to
a large set of cognitive processes• Other researchers use closely matched tasks
designed to isolate a single cognitive process(“tight” comparison)– The regions identified by the comparison relate to a
specific cognitive process, given the assumptions of thedesign
• Why/when would you use one or the other?
![Page 50: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/50.jpg)
Baseline issues
• Interpreting increases or decreases in fMRIsignal requires that we understand what thebaseline is against which those changesoccur
• Two questions:– What is the brain’s natural baseline?– How do we chose control tasks?
![Page 51: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/51.jpg)
How do we interpretdeactivations?
• Many regions are deactivated duringperformance of a cognitive task comparedto a resting state– IE, they are less active during the task than
during rest• What does this reflect?
– Active increase during the resting state– Active decrease during the task
![Page 52: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/52.jpg)
Task-independent decreases
From Gusnard & Raichle, 2003
![Page 53: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/53.jpg)
A natural baseline?• Raichle and colleagues
have argued that oxygenextraction fraction (OEF)reflects the brain’s naturalbaseline– % of oxygen sent to brain
that is utilized metabolically– Activation results in
decreased OEF, whereasdeactivation results inincreased OEF
• OEF is spatially uniform inthe cortex– Reflects long-term
equilibrium between activityand blood flow
![Page 54: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/54.jpg)
Raichle et al., 2002
![Page 55: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/55.jpg)
What is the right baseline task?
• Early imaging studies of memory did notfind activation in the hippocampuscompared to baseline– But we know that the hippocampus is necessary
for memory• Why did this happen?
![Page 56: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/56.jpg)
Stark & Squire, 2002
![Page 57: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/57.jpg)
How long should blocks be?
Below about 8 seconds, the HRF cannot return to baseline
![Page 58: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/58.jpg)
Fourier analysis of blocked designs
ƒ =
Power
Frequency (Hz)
![Page 59: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/59.jpg)
The HRF is a low-pass filter
Convolution with the HRF selectively reduces high-frequencysignals
![Page 60: Experimental design I - WordPress.com · Experimental design I Russ Poldrack M285. What do scientists do? •Use experiments to test hypotheses •What is a hypothesis? –A claim](https://reader034.vdocuments.net/reader034/viewer/2022052012/6027d54ae53c1f38c669f56c/html5/thumbnails/60.jpg)
Power
Frequency(Hz)
There is a fundamental tradeoff between the noise characteristicsof fMRI data and the bandpass characteristics of the HRF