Part I · Chapter 1 · Lecture 1

Why study brain & behavior

The five reasons the course gives, the three disciplines it draws on, and the reading model that shows how a cognitive model turns into a testable brain prediction.

1. The brain constructs what you experience

Start with something that should bother you. You feel like you see the whole scene in front of you, in full detail, all the time. You don't. Try the figure below: two versions of the same picture alternate with a blank flash between them. One thing differs. Most people take a surprisingly long time to find it — even though, once found, it's obvious.

Change blindness — find the difference

—

What to notice: the change is large, but the blank frame wipes out the motion signal your visual system normally uses to spot changes. Without it you have to search object by object — which shows how little of the scene you were actually holding onto.

Definition Change blindness — the surprising failure to notice even large changes in our visual environment. The lecture's takeaway: what we perceive and remember is far less complete than it feels.

The same lesson holds for hearing. Shepard's tone is an auditory illusion in which a tone seems to rise (or fall) forever, even though the overall pitch never actually changes. Press play:

Shepard's tone — a staircase that never gets higher

Volume

What to notice: every bar slides right (every component rises by an octave), but the loudness envelope never moves. A component fades in at the low end as another fades out at the high end, so after one full cycle the sound is identical to where it started — yet every step sounded like "up".

Key idea The brain doesn't just record sound or light as it is — it constructs what we hear and see. Depth in a 2D painting, the endless rise of Shepard's tone ("things keep building but never resolve → tension"), the scene you think you see in full — all built, not recorded. This theme returns in every vision chapter.

2. Five reasons to study the brain

#	Reason	The lecture's example
1	Clinical implications	Parkinson's disease (PD) — a progressive neurological disorder with tremor, stiffness and slowness of movement, caused mainly by degeneration of dopamine-producing neurons. Treated with drugs and deep brain stimulation (Ch. 7).
2	To know thyself and others	Change blindness — your perception is less complete than it feels.
3	To appreciate culture (art & music)	Depth cues in paintings (Ch. 11); Shepard's tone creating tension in music.
4	To design better AI	NeuroAI; CAPTCHA = Completely Automated Public Turing test to tell Computers and Humans Apart — it works exactly because some things are easy for brains and (were) hard for machines.
5	To design better robots	Listed in the study guide alongside AI.

Common mistake Treating these as trivia. On an exam they usually appear as "give an example of why…" — have one concrete example ready per reason (PD, change blindness, Shepard's tone, CAPTCHA).

3. Three disciplines, one question

Mind

Cognitive psychology

Studies the mental processes that shape how we think and behave. Its models can be built without considering biology — boxes and arrows of representations and processes.

Mind ↔ brain

Cognitive neuroscience

Studies how the brain gives rise to those mental processes. Takes the psychologist's model and asks where and how each box is implemented.

Brain ↔ machine

NeuroAI

Brings neuroscience and AI together to understand and build intelligent systems. (Plus many others: systems, molecular, cellular neuroscience.)

Analogy Cognitive psychology writes the flowchart for a program; cognitive neuroscience finds which chips run each step. (This is exactly Marr's algorithmic vs implementation levels — Ch. 4.) The analogy leaks in one place: the brain has no clean separation between software and hardware — plasticity means running the program physically rewires the chips (Ch. 2).

4. The reading model: from box-and-arrow to brain prediction

Read these aloud: blick, plim, sproke. Now: yacht, sew, colonel. Nobody taught you "blick", yet you pronounced it instantly. And nobody can sound out "colonel" from its letters — you just know it. That suggests two different routes from print to sound.

Definition Phonological decoding — sounding out a word using letter-to-sound mappings. Needed for non-words (blick).

Definition Orthographic recognition — recognizing a familiar written word and retrieving its learned pronunciation. Needed for irregular words (colonel).

Now the cognitive-neuroscience move the lecture makes: suppose a brain region converts letters to sounds, and it's damaged — what pattern would you predict? Click a route to damage it.

Dual-route reading model — lesion a route

The lecture's answer: damage the phonological route and irregular words stay relatively preserved (colonel can be recognized from stored orthographic knowledge) while non-words are impaired (blick must be sounded out). And then the loop closes: experimental results are used to refine the cognitive model.

Intuition A cognitive model is valuable to a neuroscientist because it makes predictions. Without the two-route model you'd just see "reading is worse". With it, you know to test non-words and irregular words separately — and the pattern tells you which route broke. This is the seed of the dissociation logic in Chapter 5.

going deeper What about regular words like "mint"? Either route can read them, so they survive damage to either route — that's why they're useless as a test. The two real clinical patterns match the widget: damage to phonology ≈ "phonological dyslexia" (non-words fail), damage to orthography ≈ "surface dyslexia" (irregular words get regularized: colonel → "coll-oh-nel"). Finding both patterns in different patients is a double dissociation.

5. Why active learning feels worse and works better

Deslauriers et al. (2019): 149 Harvard students in intro calculus. Group A watched problems solved in a standard lecture (passive); Group B solved them in small groups (active). They measured feeling of learning (self-report 1–5) and a test of learning.

The dissociation between feeling and learning (schematic)

Schematic of the direction of the result only — bar heights are illustrative, not the paper's values. What to notice: the two measures point in opposite directions.

Key idea Active learning (generating: explaining, discussing, applying, testing yourself) leads to more learning — but students feel they learn less and enjoy it less. Why? Generating predictions takes effort and invites errors, and effort and errors make you feel less smart even while they promote learning. Passive learning (listening, watching, rereading) is easy, and that ease creates an illusion of learning.

Definition Dunning–Kruger effect — when you know less, you are less able to recognize what you don't know. It's why rereading notes feels like mastery: you can't see the gaps you'd need to know the material to see.

Practical upshot for this midterm: every chapter here has hidden-step problems and quizzes. Commit to an answer before opening the reveal. If it feels harder than rereading, that's the point.

6. Check yourself

A patient can read "yacht" and "colonel" aloud perfectly but can't read "plim". Which route is most likely damaged?

Why: Irregular words (yacht, colonel) can only be read by the orthographic route, so it's working. Non-words (plim) can only be read by sounding out — so the phonological route is the broken one. Visual analysis damage would hurt everything, including yacht.

Which statement best describes the difference between cognitive psychology and cognitive neuroscience?

Why: That's the lecture's exit-ticket answer, word for word. The second option is backwards — and note cognitive models can be built without any biology at all.

Students in an active-learning section rate their learning lower than a lecture section, but score higher on the exam. The best explanation from lecture is:

Why: The study randomized students, so the "smarter" option is the thing the design rules out. Self-reports aren't always wrong — they're specifically fooled by fluency. Dunning–Kruger adds that knowing less makes you worse at spotting what you don't know.

Try itDesign the reading experiment

You suspect region R in a patient's brain converts letters to sounds. You have one hour with the patient and a list of words. What do you test, and what result supports your hypothesis?

Step 1 — What's actually being asked?

Not "is reading impaired" — you need a pattern that is specific to letter-to-sound conversion. So you need at least two word types that load on different routes.

Step 2 — Pick the stimuli

Non-words (blick, plim, sproke) require phonological decoding. Irregular words (yacht, sew, colonel) require orthographic recognition.

Answer & the takeaway

Support = non-words impaired, irregular words relatively preserved. If both were impaired you couldn't tell a specific phonology deficit from general damage — which is exactly the weakness of single measures you'll meet again in Chapter 5. The transferable skill: a cognitive model tells you which contrasting conditions to test.