Task 6 Flashcards
Cognitive subtraction
= a type of experimental design in functional imaging in which activity in a control task is subtracted from activity in an experimental task
=> by comparing the activity of the brain in a task that utilizes a particular cognitive component (e.g. the visual lexicon) to the activity of the brain in a baseline task that does not, it is possible to infer which regions are specialized for this particular cognitive component.
The method of cognitive conjunction requires that one is able to identify a set of tasks that has a particular component in common. One can then look for regions of activation that are shared. This reduces the problem of interaction because the interaction
terms will be different for each pair of subtractions.
Parametric designs
In a paramertic design, as opposed to a categorical design, the variable of interest is treated as a continuous dimension rather than a categorical distinction.
In intuitive terms, one is measuring associations between brain activity and changes in the variable of interest, rather than measuring differences in brain activity between two or more condition
We meaure associations between brain activity rather than differences
Baseline task is not necessary
Correlate brain activity with cognitive/bhevaioural measure
Example: Price at al. (1992) – listeining to different rates of spoken words (0 words/min – 90/min)
- Change in activity in various regions can be correlated with the rate of speech
- Activity increased with speech rate but then decreased again as it got faster (inverted U), suggesting that every regions has an optimal level at which it functions
Resting state paradigms
A technique for measuring functional connectivity in which correlations between several regions (networks) are assessed while the participant is not performing any tasks
brain regions that are functionally connected the noise levels, that are found in the resting state, tend to correlate together
helps to identify regions for which fluctuations in activity tend to be shared (i.e. default network)
Block design
Stimuli from a given condition are presented consecutively together
stimuli that belong together in one condition are grouped together
has more power in fMRI (can detect significant but small effects) then event related
useful of the response takes longer to occur
sometimes, we cannot know how events should be grouped, so we must perform an event-related design (i.e. tip-of-the-tongue phenomenon, spontaneous occurring hallucinations, oddball paradigm, go-no go task, …)
sometimes there is no way of knowing in advance how events shoulder be grouped therefore block design is impossible ( control words, normal words)
Event-related design
Stimuli from two or more conditions are presented randomly or interleaved
different stimuli or conditions are interspersed with each other and subsequently separated for the purpose of analysis
enable a wider range of experimental designs and are more closely related to the typical design structure of most cognitive psychology experiments (certain experiments can only be conducted this way i.e. tip-of-the-tongue phenomenon).
Why do functional imaging data sometimes disagree with lesion data?
There are 2 possible disagreements that can take place:
1) Imaging + & lesion – : Imaging data implies that A region is used and essential but lesion data does not. Possible reasons for the disagreement are:
- Activated region reflects a strategy used by the participant and is not essential
- Region reflects a general cognitive resource (attention, …) that is not specific to the task
- Region is being inhibited rather than excited
- Lesion studies were not powerful enough to detect the importance of the region
2) Imaging – & lesion + : imaging implies a region is not used but lesion suggests it is. Resons:
- Experimental and baseline depend on the region and comparison produces a null result
- Difficult to detect activity in the region (small, location is different in different subjects, …)
- Impaired performance after lesion reflects damage to tracts passing through the region rather than the synaptic activity in the gray matter of the region itself (activity is more spread and does not only depend on the region where the lesion is in)
Multi-voxel pattern analysis =
An fMRI analysis method in which distributed patterns of activity are linked to cognitive processes
With multi-voxel pattern analysis (multivariate analysis) we can examine the pattern of activation over a distributed set of voxels to enable a more fine grained approach. There is also another analysis (univariate) where we just look at the activation value of one voxel for itself and compare it to other voxels.
Substraction assumptions
Subtracting the experimental condition from the control/baseline condition. The BOLD signal difference then represents all regions involved in the performance of the task.
Assumptions:
- conditions can de added (pure insertion)
- there are no interactions among the components of the task
even though the assumptions are often violated, the results can still be useful, especially of used with a blocked design.
Factorial assumptions
This technique relies on neurophysiological evidence as well as complementary behavioural data for the precise definition of the tasks components and allows tests for interaction between each component.
The subject performs a task where cognitive components / dimensions are intermingles in one moment and separated in another.
Assumptions:
- Linearity (between BOLD responses resulting from the conditions – if not the findings might be skewed due to non-predicted interactions)
Parametric (amaro & Barker)
Includes increasing the cognitive demand associated with a particular task. An accompanying increase in BOLD response then implies an association between the region and the parameter. However, increasing a parameter over a certain limit can cause the involvement of other cognitive processes that are not crucial for the task itself but only for the increased attention etc. devoted to it.
Conjunction analysis
Subtle deviation from the factorial design in that 2 or more conditions share the cognitive component of interest. Thus, this approach wants to find commonalities instead of differences between regions.
Blocked
Based on maintaining cognitive engagement by presenting stimuli from one condition in between those of another condition (T – C – T - …)
AB block = the alternation of 2 conditions in which a “cycle” corresponds to 2 epochs of each condition
Epoch = moment
This has received criticism because of the assumptions etc. but still produces valuable results
Event-related
This technique implied event-related fMRI (erfMRI) in that it records activity that is locked to one specific event. We can vary the time between the events in order to keep the subjects attention and to shorten the duration of the experiment. This strategy is then called rapid erfMRI / jittered. Even though this strategy enhances the statistical power (because of the increased number of stimuli presented) the ability to estimate the HRF are reduced. The use a slow or rapid design depends on our study and what we want to find.
A minimum of 4 sec should be in between consecutive stimuli
The main advantage is the ability to detect variations in hemodynamic responses, allowing us to detect the HRF. It also allows for analysis related to individual responses (i.e. see the effects of judgement, emotional content, …).
Event-related fMRI is less sensitive to head movements.
mixed designs
This combines block- and event-related designs to compare maintained and transient neural activity. It allows for extracting brain regions exhibiting an item-related pattern of information processing (transient), or a task-related information processing (sustained).
In the blocks, we detect the responses given through an event-related design
There are however numerous assumptions and the researcher will face issues connected to poorer HRF shape estimation and post-hoc analysis.
Behaviourally driven fMRI
Here, the subject just lies in the scanner and does nothing. We observe variations in the BOLD response related to spontaneous activity.
Limitations:
- linearity properties of overlapping HRF,
- depends on each subject performance, and the inter-subject variability, number of events per condition and consequently the statistical power of the study are unknown beforehand, even if a pilot study is performed
While blocked designs allow compari- son between
encoding conditions that yield, on average, higher or lower levels of subsequent recollection, they do not allow a direct trial-by-trial comparison between specific encoding trials that lead to subsequent remembering and those that lead to subsequent for- getting.
Results from event-related potential (ERP) studies, which allow for trial-by-trial analysis, suggest that
(WAGNER)
the neural signature during verbal encoding differs for subse- quently remembered and subsequently forgotten experiences, with remembered experi- ences being associated with a greater posi- tive-going response over frontal and parietal regions
However, ERP studies are char- acterized by
limited spatial resolution
Our studies, together with previous results (2), suggest that what makes a verbal expe- rience memorable partially depends on the extent to which
(Wagner)
left prefrontal and medial
temporal regions are engaged during the ex- perience.
our experiments demonstrate that left parahippocampal gyrus is more ac- tive during the encoding of verbal experienc- es that are later remembered relative to those later forgotten, even though these two classes of experiences were equally novel within the context of the experiment
cognitive subtraction problems
+the assumption of pure insertion
+choosing wich other factor to take into account t as it might cause interaction
+hard to festlegen on a baseline
cognitive conjunction
The method of cognitive conjunction requires that one is able to identify a set of tasks that has a particular component in common. One can then look for regions of activation that are shared. This reduces the problem of interaction because the interaction terms will be different for each pair of subtractions
functional specialisation
Functional specialization implies that a region responds to a limited range of stimuli/conditions and that this distinguishes it from the responsiveness of other neighboring regions. It
functional integration
Functional integration, on the other hand, refers to the way in which different regions communicate with each other
The basic approach of functional integration is to model how activity in different regions is interdependent. This is used to infer the effective connectivity or functional connectivity between regions when performing a task
Example: Friston & Frith (1995) – imaging study where shizopheric people and non schizophrenic people had to generate words beginning with A or words beginning with repeating letters
- Similar activity of brain regions between groups but lack of correlation between the activity in these regions in schizophrenics (functional disconection
decision tree
1= parametric, cognitive substraction
2,3=factorial design=> wich method?
=> PET=blocked design
=>fmri=>does the hypothesis require to present events randomly or not ?
=>no= block
=>yes => event related design
There are two broad scenarios in which functional imaging data and lesion-deficit data can disagree
isagreement 1: Imaging data imply that a brain region is used in a given task, but lesion data suggest that this region is not essential to the task (imaging +, lesion –)
Possible reasons for disagreement:
• The activated region reflects a particular strategy adopted by the participants that is not essential to performing the task.
• The activated region reflects the recruitment of some general cognitive resource (e.g. due to increased task difficulty, attention or arousal) that is not specific to the task.
• The activated region is being inhibited (i.e. switched off) rather than excited (i.e. switched on).
• The lesion studies have not been powerful enough to detect the importance of the region (e.g. too few patients, lesion not in correct location, tasks used with patients not the same as those used in imaging).
Disagreement 2: Imaging data imply that a brain region is not used in a given task, but lesion data suggest that this region is critical to the task (imaging –, lesion +)
Possible reasons for disagreement:
• If the experimental task and baseline task both depend critically on this region, then a comparison between them might produce an artifactual null result.
• It might be intrinsically hard to detect activity in this region of the brain (e.g. it is a very small region, it is in different places in different individuals or genuine activity produces a small signal change).
• The impaired performance after lesion reflects damage to tracts passing through the region rather than the synaptic activity in the gray matter of the region itself.
multi-voxel pattern analysis
An fMRI analysis method in which distributed patterns of activity are linked to cognitive processes.
Rather than locating the peak area of activity (as in regular fMRI analysis), one can examine the pattern of activation over a distributed set of voxels to enable a more finegrained approach
firm new lie detector?
For example, Haxby et al. (2001) gave participants pictures from eight different types of category, including cats, houses, faces, and shoes. The neural activity from an individual trial was then compared to the previous known patterns of activity to determine the most probable category that was being viewed. This procedure could predict, given pairwise comparisons, what the person was seeing with 96 percent accuracy.
Other research has shown that activity in theseregions can be used to accurately predict semantic categories when reading words (Mitchell et al., 2008) or when recalling previously seen images from memory (Polyn et al., 2005).
resting state paradigm
Resting state paradigms = A technique for measuring functional connectivity in which correlations between several regions (networks) are assessed while the participant is not performing any tasks
brain regions that are functionally connected the noise levels, that are found in the resting state, tend to correlate together
helps to identify regions for which fluctuations in activity tend to be shared (i.e. default network)
factorial design (amaro & baker)
This technique relies on neurophysiological evidence as well as complementary behavioural data for the precise definition of the tasks components and allows tests for interaction between each component.
The subject performs a task where cognitive components / dimensions are intermingles in one moment and separated in another.
parametric (amaro& BAKER)
Includes increasing the cognitive demand associated with a particular task. An accompanying increase in BOLD response then implies an association between the region and the parameter. However, increasing a parameter over a certain limit can cause the involvement of other cognitive processes that are not crucial for the task itself but only for the increased attention etc. devoted to it.
functional integration
Functional integration = The way in which different regions communicate with each other
Essential for a full understanding of how cognition is linked in the brain
Basic approach: model how activity in different regions is interdependent (used to infer effective connectivity / functional connectivity)
Correlate different brain regions of activity with each other
Example: Friston & Frith (1995) – imaging study where shizopheric people and non schizophrenic people had to generate words beginning with A or words beginning with repeating letters
- Similar activity of brain regions between groups but lack of correlation between the activity in these regions in schizophrenics (functional disconection)
Problems with cognitive subtraction
Assumption of pure insertion (or deletion) = The assumption that adding a different component to a task does not change the operation of other components
Interaction = The effect of one variable upon another
To avoid the problem of interaction, the baseline task should be as similar as possible to the experimental task
- compared reading aloud with viewing of written words could involve multiple stages of processing, which makes the findings ambiguous
big brother
Example: Haxby et al. (2001) – pictures of 8 different categories were shown and the neural activity received was compared to previously known patterns of activity. With pairwise comparisons, we can predict what the person was seeing with 96% accuracy (shoe vs. other object).
Limited, because answers are generated from a closed set of options
Other studies have used the layout of the visual cortex to reconstruct visual images (cells prefer one picture over another are more active when seeing one picture as opposed to another).
Patterns of activity in the prefrontal cortex can be used to predict (even before the person made their response) which of two tasks will be performed.
Studies performed on people in a vegetative state gave them the possibility to “answer” to questions with brain activity (yes = imagine tennis; no = imagine your house)
Evaluation
Brain imaging can be used to infer the type of stimulus that is being processed and simple cognitive decisions, but the specific content of thoughts can not be inferred from fMRI (yet).
ROEFS COMMENTARY ON TROJANO ET AL. – MATCHING TWO IMAGINED CLOCKS: THE FUNCTIONAL ANATOMY OF SPATIAL ANALYSIS IN THE ABSENCE OF VISUAL STIMULATION
o Subjects imagine pairs of times that are presented acoustically, then judge at which of the two times the clock hands form the greater angle
Block design
spatial imagery task used in two experiments
Experiment 1:
• Spatial imagery task + control condition requiring only a verbal-semantic judgement (that does not rely on spatial or imagery processes)
Experiment 2:
• Compared activity during mental clock test (imagery) with that evoked by the same operation performed on visually presented material (perception) and by a different non-spatial control task (syllable counting, the attentional load of which was comparable to that of the imagery task)
These are examples of Cognitive subtraction
- Hypothesized results taken from lesion-based evidence: increase of neuronal activity, as indicated by BOLD signal, in parietal lobes bilaterally during mental clock test
- Analysis included the whole cortex to assess the coactivation of other brain areas
What it a good comparison task and what not is subjective. Both tasks involve a judgement but only the first task involves imagery:
Imagery condition: imagery + judgement
Control condition: judgement
exp 1 data analysis
preprocessing
Identification of cerebral regions responding to the stimulus
- By means of a cross-correlation analysis
- Reference vector used was a ‘box-car’ ideal vector (value 0 during control, value 1 during imagery period)
- Activated brain areas were selected in the cross-correlation maps by imposing a conservative intensity threshold for the cross-correlation coefficients
Tailarach transformation = Each subject’s structural 3D data sets were transformed into Talairach space
Guassion smoothing = Prior to averaging, the functional 3D maps were smoothed with a Gaussian kernel of 5mm FWHM
Averaging and Normalization
= Individual Talairach 3D maps were averaged across subjects and superimposed on a normalized anatomical 3D data set
- Activated brain areas were selected in the average map by imposing lower overall correlation values (r > 0.25) to take into account the interindividual spatial variability of activated clusters
exp 2
Experimental Paradigm
Compared the activity during the imagery condition with that evoked by the same cognitive operation on visually presented material (perception) and by a different non-spatial control task (syllable counting)
- Imagery Condition (same as in experiment 1)
Imagery + judgement - Perception Condition
o Clock faces generated on computer screen
o Subjects had to decide at which of the two times the clock hands formed a greater angle
Judgement - Syllable Counting Condition
o Subjects had to count the syllabled of the auditorily presented pairs of times and report whether the total syllable number was even or odd
Judgement
data analysis exp 2
preprocessing #gerey and white matter boundary restruction
Wagner
The event-related design permitted identification of regions that demonstrate differential activation during the encoding of words subsequently remembered and those subsequently forgotten.
We cannot categorize our words into remembered and forgotten beforehand (so we cannot form blocks with those categories). Which word is remembered depends on each participant, so we have to look at the activity for this specific word separately.
