Applied Research Methods
Week 1 – Introduction
Importance of faces
= can produce radically different images on our retina when it changes expression and/or orientation
Central role in human interactions
Contribute to speech perception
Communicate a wealth of social information:
- Age, gender, personal identity (physical structure)
- Mood and emotional state (facial expression)
- Interest / attentional focus (direction of gaze)
Faces as visual stimuli
Faces as a category highly homogenous (similar)
- Share basic component parts in a fixed configuration (2 eyes, a nose, a mouth)
Individual faces are highly different
- Vary in many dimensions (head shape, individual features, relative feature placement, color, texture)
Dynamic and changeable due to movable parts that change shape and relative position
- Facial expressions (a smile versus an angry frown)
Behavioral measures of face processing
Object recognition operates via feature recognition (based on the shape of its parts)
Faces however all share the same features
One solution could involve recognizing the relations between the parts of the face rather than the parts
- Configurational – computation of the spatial configuration of the components
- Holistic – integration over the whole face
Inverted faces
= provide an ideal case, because they are identical in terms of within-class
similarity, complexity, and features.
Upright faces are much easier to recognize than inverted faces
Thatcher’s illusion, faces are differently processed when inverted
- Yin’s studies (the face inversion effect)
o For houses and planes, participants were slightly worse
when inverted.
o For faces, participants made a lot more mistakes.
We recognize faces differently than other objects
Gaze cueing
Do we move our attention with the direction of the gaze of an individual?
Participants following the gaze of the person will be faster to detect the colored patch on top of the cued face
compared to the uncued face. Social faces are faster recognized than scrambled faces.
People that score high on the ASQ do not differentiate in response time between faces and scrambled face parts.
People that score low on the ASQ score higher for faces than scrambled faces.
There is a link between autism and the degree to which you are cued into a social situation
Developmental measures of face processing
Specialized processing of social stimuli in newborns
Infants track face-like stimuli longer than scrambled faces
The brain is innately sensitive to faces
fMRI measures of face processing
Structural images Functional images
High spatial resolution Lower spatial resolution
Takes ± 5 minutes Takes ± 1 second
Not sensitive to neural activity Blood-oxygen-level-
dependent (BOLD)
Poor temporal resolution
Blood-oxygen-level-dependent
Blood is slow to arrive after stimulus presentation
fMRI is a slow and indirect measure of neural activity
The Fusiform face area (FFA)
Responds more to faces than other objects
1. Identify FFA in participant
2. Measure FFA response to multiple conditions
The FFA is driven by interpretation (vase-face phenomenon)
1
,Applied Research Methods
Magneto/electro-encephalographic measures of face processing
EEG MEG ECoG or iEEG
Directly measures neuroactivity Uses magnetic fields Electrodes put under scalp
High temporal resolution Expensive because of helium cooling Big negative response to faces is
found compared to other categories
Poor spatial resolution Better resolution than EEG High spatial resolution
Weak signals Signals travel through tissue without High temporal resolution
being interrupted by the scalp
Averaged ERP results in meaningful The activity of face-selective Invasive
waveforms by filtering out noise component N170
Voltage is plotted the other way Facial response in FFA N200
around (negative-up/positive-down)
Single neuron electrophysiological measures of face processing
Great temporal and spatial resolution
Only recording a single cell
Cells cued to a specific orientation
Only in animal studies, neurons in monkeys can be selective to the face
Golden standard
Neuropsychological measures of face processing
Prosopagnosia = ‘face blindness’
Due to brain injury in the right temporal lobe
Socially crippling
- Cannot recognize faces that are familiar or own face
- Can recognize faces as category vs other objects
- Can recognize familiar people by voices and non-facial cues
- Vision otherwise is ok
Group analysis approach
1. Take a large group of patients with lesions in particular brain regions
2. Measure structural MRI to map out the lesions
3. Have patients perform behavioral tasks and see who is impaired in specific tasks
4. For each brain region, compare performance with and without lesions
Tells you which brain areas are casually involved
Non-invasive brain stimulation measures of face processing
TMS
Generating electrical current above the
skull to disrupt activity temporarily
Good spatial and temporal resolution
Can only measure brain regions right
underneath the skull.
Variation among patients
tFUS can access deeper areas in the brain
The FFA and TMS
The FFA is too deep to reach
- Right occipital face area (rOFA)
- Right lateral occipital area (rLO)
- Right extrastriate body area (rEBA)
The Rofa and rEBA are causally and selectively involved in processing their preferred category
Conclusion
Different methods offer both advantages and disadvantages in their use
Each is associated with assumptions, techniques, designs, and analysis approaches
The combination of methods allows for converging evidence
2
,Applied Research Methods
Literature - The methods of social neuroscience
Dimensions of cognitive neuroscience
1. Temporal resolution = accuracy of when an event occurs
EEG/ERP, MEG, TMS, and single-cell recording all have millisecond
resolution
PET and fMRI have minute and second resolution
2. Spatial resolution = accuracy of where an event occurs
Lesion and functional methods imaging have millimetre resolution
Single-cell recording is analyzed on the level of the neuron
3. Invasiveness = equipment located internally or externally
PET is invasive because the equipment is internally located
EEG is non-invasive because the equipment is located externally
Measuring behavior and cognition
1. Performance-based measures = measuring response times and accuracy
Mental chronometry, studying the time-course of information processing.
e.g. 4 + 2 = 6 is calculated faster than 4 + 3 = 7
Measuring accuracy, using error rates, percentages correct, or percentile of performance.
Speed-accuracy trade-off = the faster you are forced to respond, the less accurate you are
Advantages Disadvantages
Reflects real behavior Hard to link directly to neural substrates
Simple to analyze and interpret Low ecological validity
2. Observational measures = code ‘what’ or ‘how often’ behavior is displayed
Infancy research
- Preferential-looking paradigm
- Habituation paradigm
Non-human research
Advantages Disadvantages
Nature of task unknown to participant Difficulty in scoring behavior
Can be used in naturalistic settings Observer bias
3. Subliminal perception measures = unconscious of stimuli presented
- Verbal reports
- Wagering (bet on performance)
- Bodily measures (skin conductance)
4. Survey measures = questionnaires and measures
- Fixed or open-ended questions
- Reliability check = repeat questions at different points in time and tapping the same knowledge that requires a
different response from the participant
- Factor analysis = can be used to explore if concepts can be fractioned in underlying variables
Advantages Disadvantages
Can be used when experimental manipulation is not Self-report may not reflect the real behavior
possible or unethical
Measures thoughts and attitudes Social cognition may occur unconsciously
5. Bodily measures
Autonomic nervous system (control bodily functions)
- Sympathetic system = arousal
- Parasympathetic = rest
Somatic nervous system (coordinate muscle activity)
- Skin conductance response = sweat
- Electrical muscle activity = eyeblink startle response
Advantages Disadvantages
Present when unaware of the stimulus Not straightforward to link bodily response to brain
and cognition
Easy to record and analyze
6. Single-cell recording = direct measurement of action potential
Implant of a small electrode into the axon (intracellular) or outside the membrane (extracellular)
The number of times that an action potential is produced (spikes/second) is measured
- Rate coding = stimulus is associated with an increase in the rate of neural firing
- Temporal coding = stimulus is associated with greater synchronization of firing across different neurons
Advantages Disadvantages
Directly related to neural activity Invasive
Excellent spatial and temporal resolution Information limited to probed regions
3
, Applied Research Methods
7. Electroencephalography and event-related potential measures
EEG is sensitive to post-synaptic dendritic electrical activity
Records electrical signals generated by the brain, through electrodes placed at different of the scalp
- A whole population of neurons must be active in synchrony to generate a large enough electrical field
- Neurons must be aligned in parallel orientation to summate the response
- EEG compares the voltage between two or more different sites (mastoid bone or nasal reference)
ERPs are electrophysical changes elicited by stimuli and cognitive tasks
EEG waveforms refer to the current task but also spontaneous activity (low signal-to-noise ratio)
To increase the ratio one can average the EEG signal over many trials
- Positive and negative peaks are labeled with P or N and their corresponding number
- Timing and amplitude of the peaks are of interest in ERP data
Advantages Disadvantages
Excellent temporal resolution Poor spatial resolution
Direct measure of neural activity Impossible to reach some subcortical areas
The organization and structure of the brain
- Anterior / rostral = front of the brain
- Posterior / caudal = back of the brain
- superior
White matter Axons and support cells Gray matter Neuronal cell bodies
Anterior / rostral Front of the brain Posterior / caudal Back of the brain
Superior Top of the brain Inferior Bottom of the brain
Lateral Outer surface of the brain Medial Center of the brain
Raised surface of the cortex Gyrus Dipped surface of the Sulci
cortex
Four lobes of the brain Frontal, parietal, temporal, Subcortex Basal ganglia, limbic system,
occipital diencephalon (thalamus and
hypothalamus)
Functional imaging
Functional imaging methods like PET and fMRI measure dynamic physiological changes associated with patterns
of thought and behavior via changes in blood flow/blood oxygen (hemodynamic measures)
Structural imaging methods like CT and MRI measure the stable properties of the brain
- When metabolic activity of neurons increases, blood supply to that regions increases to meet the demand
- PET measures changes in blood flow directly, fMRI measures concentration of oxygen in the blood
- The brain is always physiologically active! Active means that the physiological response in one condition is greater
relative to another condition.
The BOLD response
Blood oxygen-level dependent is the amount of deoxyhaemoglobin in the blood in different regions of the brain
When neurons consume oxygen, they convert oxyhaemoglobin to deoxyhaemoglobin. This has strong
paramagnetic properties and distorts the local magnetic field. This distortion indicates the concentration of
deoxyhaemoglobin in the blood.
The hemodynamic response function (HRF) is the way the BOLD signal evolves over time in response to an
increase in neural activity that has three phases:
1. Initial dip: neurons consume oxygen which causes a small rise in deoxyhaemoglobin and thus a decreased
BOLD signal.
2. Overcompensation: increased oxygen consumption increases the blood flow to these regions. This is initially
greater than the increased consumption, so the BOLD signal increases substantially. This is measured by
fMRI.
3. Undershoot: blood flow and oxygen consumption dip before returning to their original levels
- HRF is stable across sessions with same participant in same region, but varies across individuals and regions
Constraints on experimental design
Different HRFs can be superimposed on each other, so one does not have to wait for the BOLD response to return
to baseline before presenting another trial.
During fMRI, fewer trials that are spaced out over time and ‘null events’ are common to allow the HRF to dip
toward baseline
The amount of data is related to the number of brain volumes acquired, not the number of trials
- Block design, trials belonging together are grouped together during stimulus presentation (block designs are the
only option in PET but fMRI can use either)
- Event-related design, trials are randomly interspersed during stimulus presentation but are then treated separately
at the analysis stage
4
