Brain Scanning PDF

Reminders: Sign in to AttendanceRadar Quiz: Clarifications on Quiz policies: Quiz answers should be your own work – do not discuss answers with other students...

Reminders: Sign in to AttendanceRadar Quiz: Clarifications on Quiz policies: Quiz answers should be your own work – do not discuss answers with other students until after class You can look at the readings during the quiz if necessary You can take the quiz even if not in class (though you will not be marked present for attendance purposes) Lowest 4 quiz grades are dropped Midterm Wednesday: optional review session in class Monday: midterm in class 40-50 multiple-choice questions Scantron (bring #2 pencil if you have one) If you cannot take the midterm on Monday, email Head TA Abby Wood Studying the Mind by Measuring the Brain How do we measure the brain? What can brain measurements tell us about cognition? How can brain measurements go beyond behavior? Implanted electrodes Advantages: Direct measurement of firing rates in neurons Disadvantages: Limited coverage of the brain Requires surgically removing the skull!! Surbeck et al., 2011 EEG (electroencephalography) Advantages: Non-invasive Cheap + portable Disadvantages: Electrical signals at the scalp are very blurred Limited access to deep brain structures MEG (magnetoencephalography) Advantages: Non-invasive Magnetic signals better pass through skin/skull Disadvantages: Magnetic signals very weak Requires huge, expensive, special-purpose machine (may be changing!)  fMRI (Functional MRI) Astrocytes Neurons B lo o d vesse l When neurons fire, extra oxygen flows in from nearby blood vessels Oxygenated vs. deoxygenated blood cells are different magnetically Idea: measure this blood-oxygenation-level-dependent (BOLD) signal fMRI (Functional MRI) Advantages: Non-invasive Spatial resolution of ~mm (measure ~80,000 “voxels”) Access to deep brain regions Disadvantages: Terrible temporal resolution (BOLD changes over seconds) Unclear what causes BOLD Diffusion MRI Low Fractional Anisotropy (FA) High FA Mukherjee et al., 2008 Diffusion MRI Measures direction and integrity of axon bundles Cannot measure circuits below the ~mm scale Skeptics: Jerry Fodor “If the mind happens in space at all, it happens somewhere north of the neck. What exactly turns on knowing how far north?” But the fact that a cognitive process has a specific location is a useful thing to learn from neuroimaging Informs debate between Empiricists: brain is “equipotential” (not specialized into regions) and implements general intelligence Rationalists: brain implements distinct types of cognition, supported by different regions Skeptics: Max Coltheart Cognitive theories can be tested with neuroimaging, but the theories must make specific predictions about brain states Consistency fallacy: the false argument that data supports a theory because the theory can be made consistent with that data Need the theory to be falsifiable with neuroimaging data – i.e. there is some pattern of that wouldn’t be consistent with the theory What can we measure with neuroimaging? Region(s) involved in a cognitive process Timing of different operations during cognition  Study House – Bus Camera – Market Fan-1 Fan-2 Math – River River – House … Test Math – Bus ? Camera – Market ? Fan-1: 950 ms River – House ? Fan-2: 1250 ms … Borst & Anderson, 2021 ACT-R cognitive model predicts about how Fan-1 vs Fan-2 should differ specifically during Stage 4 = “Associative Retrieval" Use EEG to measure the timing of each stage to test this theory Borst & Anderson, 2024 What can we measure with neuroimaging? Region(s) involved in a cognitive process Timing of different operations during cognition Testing whether a cognitive process is occurring Gruger et al., 2018  3 image pairs: AB, CD, EF Randomly pick which pair to show next Control: randomly switch between all 6 images Ø Hippocampus responds more to structured sequence (after some learning) Ø Effect present even in 3 month olds Yates et al., 2021 Ellis et al., 2021 Prediction: Episodic memory system (hippocampus) will “replay” weak memories in the background during rest Schapiro et al., 2018 t=1 t=2 t=3 t=4 t=5 t=6 Timepoints during rest Found that worse- remembered satellites were replayed more during rest Schapiro et al., 2018 What can we measure with neuroimaging? Region(s) involved in a cognitive process Timing of different operations during cognition Testing whether a cognitive process is occurring Measuring individual differences in cognition Falk et al., 2011 Ø Showed 16 ads to heavy smokers who wanted to quit Ø Group dropped from 21 cigarettes/day to 5! Ø Confirmed with expired CO Ø But some changed behavior more than others – can we predict how much each person will change? Measured: 1. Self-report: did they think the ads were effective? 2. Neural activity: response in medial frontal cortex Neural measure improved behavior prediction over self-report alone Falk et al., 2011 Larger size of visual region V1 -> illusions less strong Suggests that these illusions arise from interactions between low-level visual features Schwarzkopf et al., 2011 Ø Studied reading skills of 39 children (7-12 yo) over 3 years Ø Used diffusion MRI to measure FA in two critical axon bundles: Ø Arcuate fasciculus: connects language perception and production Ø Inferior longitudinal fasciculus: connects vision and language Good readers showed faster development of FA in both pathways Good readers started with lower initial FA Yeatman et al., 2012 What can we measure with neuroimaging? Region(s) involved in a cognitive process Timing of different operations during cognition Testing whether a cognitive process is occurring Measuring individual differences in cognition Measuring representations of different inputs Measuring representations with neuroimaging Ø If representations are implemented as a pattern of firing rates across neurons, Ø And firing rates influence BOLD, Ø Then BOLD patterns across voxels will capture some information about the representations currently active in the mind Measuring representations with neuroimaging i ul t im S Ø If representational maps in a region are spatially similar across people, Ø Then we can measure the extent to which people share the same representation stimuli by correlating spatial patterns Can we track learning of representations for Computer Science concepts? Meshulam et al., 2021 fMRI allows us to (imperfectly) compare representations across people Meshulam et al., 2021 Summary Neuroimaging gives us (partial) access to representations and algorithms of cognition In some cases, can give us data that is hard/impossible to get from behavior alone Not all neuroimaging connects in a useful way to cognitive theories Testing Computer Science concepts Question 1 Question 2 Correlate Meshulam et al., 2021 3 image pairs: AB, CD, EF Randomly pick which pair to show next Control: randomly switch between all 6 images Ø Hippocampus responds more to structured sequence (after some learning) Ø Effect present even in 3 month olds Ellis et al., 2021