Notes on Social Neuroscience Imaging Methods and Analysis
Identity and the brain: the Gage/Broca-style question of what makes you “you.” The talk begins with Accident (Phineas Gage) and the idea that brain regions contribute to personality, feelings, and behavior; poses the question of whether selfhood is defined by brain or by brain’s interaction with the body and other systems.
Emergence of social neuroscience: from the 18th/19th century observations to late 1990s–early 2000s, modern functional noninvasive brain imaging enables the field to study how social processes map onto brain activity.
Foundings and key figures (three groups later considered to form a single field):
John Cacioppo: one of the first to coin the term social neuroscience (1994). Focused on how social world affects nervous system, peripheral nervous system, neuroendocrine function, and immune function; health/social factors influence brain and body.
Founded the Society for Social Neuroscience and the Journal of Social Neuroscience.
Elizabeth Belk, Matthew Limerick, Oksner (as named in the transcript): early proponents of social cognitive neuroscience; interested in cognitive processes in social contexts; began administering social cognition tasks in fMRI; neuropsychology background; work with split-brain patients; linked to Gazzanaga’s work (Dartmouth) and the Journal of Cognitive Neuroscience.
Gazzanaga (likely a reference to Michael Gazzaniga): founded cognitive neuroscience initiatives; edited journals; helped establish the broader field.
“Big tent” view: despite different emphases, all three strands contributed to a unified field—social neuroscience—covering social processes, cognition, and neural mechanisms.
Early mapping in the field (domain-specific, simplistic view):
Domain-specific mappings proposed a one-to-one link between a cognitive/affective domain and a single brain region:
Emotion domain (e.g., fear) linked to the amygdala.
Social cognition domain (e.g., beliefs, theory of mind, person perception) linked to dorsal medial prefrontal cortex.
This one-to-one mapping is simplistic and not strictly true; regions show specialization but also participate in multiple processes.
Transition to a more nuanced view: brain regions are parts of distributed networks; social neuroscience uses multiple imaging modalities to capture temporal and spatial dynamics.
Concept of social neuroscience as a bridge: links brain anatomy/biology to social world and experiences; integrates social psychology (group-level processes) with neuroscience (brain mechanisms).
Imaging modalities: organizing by temporal and spatial resolution; different techniques offer different trade-offs between when things happen (temporal) and where in the brain (spatial).
Common human imaging methods (noninvasive):
Functional MRI (fMRI): measures blood-oxygen-level-dependent (BOLD) signal; good spatial resolution; relatively poor temporal precision (seconds-scale), indirect measure of neuronal activity.
EEG/ERP: measures electrical activity with high temporal resolution (milliseconds); varying spatial precision depending on sensor setup.
PET (positron emission tomography): measures metabolic activity via radioactive tracer; invasive due to tracer injection; good spatial data but less common now due to fMRI.
Other techniques: Transcranial Magnetic Stimulation (TMS) and Transcranial Direct Current Stimulation (TDCS) as noninvasive brain manipulation methods; used to test causality but with limitations and safety considerations.
Temporal vs spatial resolution recap:
Single and multiunit recordings (invasive, animal models) give high spatial (cellular-level) and temporal precision but are not typically used in humans.
fMRI and EEG/ERP in humans trade off: fMRI: high spatial, lower temporal; EEG/ERP: high temporal, limited spatial specificity; PET: invasive, moderate spatial information.
Correlational vs causal logic in neuroscience methods:
Neuropsychology/lesion studies (natural lesions/ablations) provide strong causal inferences about brain regions for specific behaviors.
Functional imaging (fMRI/EEG) is largely correlational: shows regions active during tasks but does not prove that activity is causally necessary for the behavior.
TMS is a causal manipulation: disrupting neural activity in a targeted region can produce measurable changes in behavior.
TDCS is a noninvasive modulation technique with debated efficacy and safety; not as robustly established as TMS.
EEG/ERP details:
Event-Related Potentials (ERPs): voltage changes time-locked to a psychological event (stimulus or condition).
ERPs are derived from many recording sites (often ~100 electrodes) to obtain a cleaner signal with peak/valley patterns.
Pros: exceptional temporal resolution (tens to hundreds of milliseconds).
Cons: limited spatial resolution; locating sources is imprecise (rough localization by landmarks is approximate).
Example historical finding: 2002–2007 studies linked certain ERP signatures to political orientation differences, with inferences about dorsal anterior cingulate cortex involvement.
PET vs fMRI: historical and practical notes
PET: radioactive tracer, measures blood flow/metabolism; invasive; good functional data but slower and with exposure concerns.
fMRI: noninvasive; measures BOLD signal; high spatial resolution; widely used in social neuroscience.
MRI safety concerns: the static magnetic field is very strong; risk comes from ferromagnetic objects; rare fatal incidents reported; noninvasive otherwise.
Anatomy and orientation basics for reading imaging data
Lateral view of the brain includes cortex, temporal, occipital cortex, cerebellum, parietal cortex.
Anatomical terms (common orientation):
Anterior (front) / Posterior (back)
Rostral (toward the nose) / Caudal (toward the tail) — often overlapped with anterior/posterior in humans.
Superior (toward the top) / Inferior (toward the bottom)
Lateral (toward the sides) / Medial (toward the middle)
For cross-sectional viewing, three planes exist:
Coronal, sagittal, axial.
How these terms are used to describe brain regions (e.g., anterior prefrontal cortex, posterior prefrontal cortex).
From neurons to fMRI signals: key physiological chain
Neuron basics: resting potential ~
Resting potential: .
Threshold for action potential: .
Action potential peak: .
Afterpotential and repolarization return to baseline.
Synaptic transmission: neurotransmitter release at terminals; binding to receptors on dendrites; if enough input, neuron fires an action potential.
Myelin increases conduction efficiency along axons.
The direct neuronal signal is in the form of action potentials; many neurons fire to generate a population signal.
fMRI measures BOLD signal, not direct neuronal firing: neural activity increases energy demand, leading to hemodynamic changes.
BOLD signal specifics:
(ratio of oxyhemoglobin to deoxyhemoglobin) and is a proxy for neural activity.
Neural activity leads to a cascade resulting in changes in blood flow/oxygenation that fMRI detects.
The fMRI signal is an indirect, sluggish reflection of prior neural activity due to the hemodynamic response.
The hemodynamic response function (HRF): the typical temporal profile of BOLD following neural activity is delayed and protracted, roughly peaking around several seconds after the neural event and returning to baseline over ~10–20 seconds or more.
How fMRI data are collected and analyzed
Data acquisition basics:
Structure: structural MRI provides anatomy; functional MRI provides task-related BOLD data.
The signal is collected over many time points while participants perform tasks.
Movement is a major source of noise; even small head motions distort the signal.
Preprocessing steps (overview): motion correction, smoothing, normalization, etc. Smoothing increases signal-to-noise ratio by averaging across nearby voxels.
Voxels: the brain is divided into 3D pixels called voxels; typical fMRI analyses examine voxel-by-voxel activity.
Rough scale: tens of thousands of voxels (often around 32,000 or more) depending on resolution.
Each voxel gives a time series of BOLD data.
Types of analysis approaches:
Whole-brain analysis: voxel-by-voxel tests across the entire brain to identify regions showing significant differences between conditions (contrast analysis).
Pros: exploratory, comprehensive; can reveal unexpected regions and networks.
Cons: difficult to interpret, multiple comparisons problem, may produce diffuse results; requires clear contrasts to be interpretable.
Reporting: results often include peak coordinates in a standard space (x, y, z), voxel counts, and effect sizes for each activated region.
Region of Interest (ROI) analysis: predefined spherical regions (often 3–6 mm radius) based on prior literature or hypotheses; average BOLD across voxels in the ROI.
Pros: increased statistical power, theory-driven inferences, reduced multiple comparisons burden.
Cons: may miss important signals outside ROI; assumes the relevant process is localized to the chosen ROI and may miss distributed networks.
ROI analysis is a form of univariate analysis (still voxel-by-voxel within the ROI but often summarized by ROI average).
Regression analyses: correlating brain activity (within ROI or whole-brain) with non-brain measures (behavioral data, personality scales, real-world outcomes).
Pros: connects brain data to real-world variables and behavior; can reveal brain-behavior relationships beyond in-brain contrasts.
Cons: high risk of false positives when used exploratorily (multiple comparisons); beware of “voodoo correlations.”
Key study design concepts in fMRI analysis:
Contrasts: comparing conditions (e.g., Task A vs Task B) to infer relative engagement; activity is always relative to a baseline/control.
Evoked design vs block design: different ways to structure task timing; evoked designs isolate transient responses; block designs enhance detectability through sustained stimuli.
Convolution with HRF: to model expected BOLD responses, task timing is convolved with the HRF to predict the fMRI time course.
Example study (whole-brain analysis): a nature video with embedded words used to study attention control; instructions were to ignore words and maintain gaze; the analysis identified multiple dlPFC regions involved in attentional control; results often report peak voxels with coordinates in a standard space plus voxel counts and effect sizes.
Example studies illustrating methods and interpretations
Affect labeling vs reappraisal (Brooklyn et al., 2014): ROI analysis of amygdala during affective labeling (labeling what you feel) versus observing (natural viewing) and reappraisal (reframing the emotion.
Design: emotional images (e.g., snakes, disgust) with conditions: observe, label, reappraise.
Behavioral data: self-reported unpleasantness lower during labeling and reappraisal compared to observation; labeling also reduced unpleasantness similarly to reappraisal.
Neural data (amygdala, left and right): activity decreased during labeling and reappraisal relative to observed trials.
Interpretation: both labeling and reappraisal dampen amygdala activity and reduce subjective unpleasantness; ROI focus on amygdala shows clear results; convergent evidence between brain data and self-report.
Takeaway about ROI analyses: supports the idea that ROI-focused analyses can yield interpretable results when prior literature supports the region’s involvement; note potential for other regions to show different patterns in broader data.
Regression example: Wagner et al. (2011) on smoking cues
Design: smokers vs nonsmokers watched movie scenes with smoking vs non-smoking scenes; ROI analysis first tested for differential activation to smoking cues; then regression linked dorsal anterior cingulate cortex (dACC) activity to post-scan craving within smokers.
Finding: greater dACC activation during smoking cues predicted greater craving after the scan among smokers.
Interpretation: illustrates brain-behavior correlations; supports the idea that brain activity in a specific region during cue exposure can relate to subsequent craving; a clear example of brain-behavior regression analysis.
Practical and conceptual notes
The field emphasizes the need for clear contrasts and appropriate baselines because brain activity is always reported relative to something (contrast-based interpretation).
Movement control and preprocessing choices significantly influence results; motion is one of the most challenging confounds in fMRI data.
The “salmon study” reference (mentioned as a reading) is an example used in class to illustrate whole-brain analysis and the interpretation of voxel-level results in a broader brain-map context.
Reporting conventions: in whole-brain analyses, researchers often present coordinates (x, y, z) in standardized space, voxel counts, and effect sizes for peak activations.
Conceptual caveats and critical thinking for interpreting imaging data
No method provides perfect measurement; each technique has trade-offs between spatial and temporal resolution, invasiveness, and causal inference.
“One region, one function” is a simplification; most cognitive and social processes recruit distributed networks across multiple brain regions.
Reverse inference caution: inferring a specific cognitive process from activation in a brain region can be misleading because regions contribute to multiple processes.
Statistical issues: multiple comparisons across voxels throughout the brain; need corrections to control false positives; beware of inflated correlations in exploratory brain-behavior studies (voodoo correlations).
Ethical and safety considerations mentioned
IRB approvals are necessary for experiments involving emotionally or morally risky stimuli (e.g., aversive images).
MRI safety concerns include the risk of metal objects in the scanner; real-world incidents emphasize strict adherence to safety.
Practical lab/learning notes from the lecture
The speaker emphasizes the balance between theoretical (hypothesis-driven) and exploratory work; planned ROI/regression analyses are generally more defensible when there is prior evidence.
There is ongoing discussion about the practical use of noninvasive brain stimulation techniques (TMS, TDCS) in social neuroscience, including limitations, cost, and safety concerns.
The lecture stresses the importance of understanding when to use whole-brain analyses versus ROI analyses depending on the research question and prior knowledge.
Quick glossary of terms used
BOLD: Blood Oxygen Level Dependent signal; the primary fMRI signal reflecting changes in blood oxygenation.
HRF: Hemodynamic Response Function; the brain’s blood-oxygenation response to neural activity over time.
VOI/ROI: Region of Interest; a predefined brain region for targeted analysis.
Voxel: a three-dimensional pixel, the smallest distinct unit in a 3D brain image.
Dorsal Anterior Cingulate Cortex (dACC): a brain region implicated in conflict monitoring and emotion regulation in several studies.
ERPs: Event-Related Potentials; time-locked EEG responses to stimuli or events.
TMS: Transcranial Magnetic Stimulation; a causal manipulation technique to disrupt neural activity.
TDCS: Transcranial Direct Current Stimulation; a noninvasive, electrical brain stimulation technique with safety and efficacy debates.
Voodoo correlations: inflated brain-behavior correlations due to circular analysis or multiple testing without proper controls.
Summary takeaway
Social neuroscience integrates social/psychological questions with brain mechanisms using diverse imaging and manipulation techniques.
Early domain-specific mappings gave way to network-based understandings of brain function in social contexts.
A robust research program uses method-appropriate designs (whole-brain and ROI analyses), clear contrasts, and, when possible, converging behavioral data to interpret neural results; always mindful of correlational versus causal inferences and methodological pitfalls.
Final thought for exam prep
Be able to differentiate and describe: domain-specific vs network-based brain mappings; correlational vs causal methods (fMRI/EEG vs lesion studies vs TMS); ROI vs whole-brain analyses; the meaning and limitations of BOLD signals; how HRF affects data modeling; and how to interpret contrasts, baselines, and regression brain-behavior results.