Discussion Guide — Chapter 36

1. "This chapter cannot rate claims, so why include it at all?"

What to listen for. The strong answer distinguishes two things the chapter can honestly do — report present-tense status, which is checkable and falsifiable, and teach a procedure, which does not expire — from the one thing it cannot. Listen for the word exists: the evidence does not exist, which is different from prediction being hard. Watch for the weak version, "because the future is uncertain," which sounds right and licenses exactly the hedged prediction the chapter refuses to make. A very good answer notices that the chapter's real subject is the reader's future reading, not the field's future.

2. Run the six questions on a claim someone in this room actually believes.

What to listen for. Have a student volunteer a real belief. Listen for which question does the work — it is almost always 6 or 1, which is the chapter's own claim. Listen also for the moment a student discovers their source never answered question 4 at all, because no comparator is the most common silent omission and the hardest to notice. Push back on any student who concludes the claim is therefore false; the correct output is usually "I know much less than I thought," not "this is wrong."

3. "A mechanism story does not move the base rate." Is that too strong?

What to listen for. The best answers grant the objection real force before answering it. Target quality genuinely does predict success across portfolios, and the chapter concedes this when it lists precedent in a mechanistic class as something that moves you. The discriminating move is separating the story from the record: a mechanism story is present in every failure as well as every success, so it cannot discriminate; a track record of approved drugs acting that way is evidence. Listen for students who reach for Chapter 5 unprompted — knowing how something would work is not evidence that it does. Weak answers either restate the rule without engaging the objection or capitulate entirely.

4. What would actually have to be measured to settle the muscle question, and why is a scan not it?

What to listen for. Two distinct arguments, and strong students give both. First, what lean mass contains — fluid, connective tissue, organ mass, glycogen and its water — so the scan is not measuring contractile muscle in the first place. Second, the inferential chain — mass to strength to mobility to falls to independence, each link an assumption. The best students also reach the wrinkle: an agent acting on the activin pathway improves the measurement by mechanism, so a positive scan is close to guaranteed and therefore a weak test. Listen for the function endpoints named concretely — grip strength, gait speed, chair-rise, falls, independence — rather than "clinical outcomes" in the abstract. Also listen for who is in the trial: an answer that never mentions older adults has missed where the concern concentrates.

5. "Acting on the brain" versus "crossing the blood-brain barrier" — and which error is worse?

What to listen for. First check that both routes are correctly described: circumventricular organs with fenestrated capillaries (the area postrema), and vagal afferents, where the information crosses and the molecule does not. Then the interesting part: most students will call the "it enters the brain" error worse because it inflates the drug. Push them toward the second error — concluding that central effects must be imagined because the peptide cannot cross — which is equally wrong and, in a clinical setting, more likely to dismiss something a patient is actually experiencing. A good answer ends at the chapter's narrow, unquotable formulation and notices why it is unquotable.

6. §36.10 says a forecast names what would prove it wrong. Turn that on this chapter, and then on this course.

What to listen for. This is the closing prompt and should not be rushed. Listen for students applying the test to the author rather than defending him: the three lists are gradeable, which is a structural virtue and says nothing about whether they are right. Then listen for the harder application — what would prove the course wrong, and what would change the student's own mind about a compound they have formed a view on. The answer that indicates the chapter landed is one where a student names a specific result that would move a rating they hold, including a rating they are attached to. Watch for the two failure modes: the student who has become uniformly cynical (nothing can be known) and the student who treats "ask what would make you wrong" as a rhetorical weapon for use on others only. Both have taken the form of the lesson and missed the content.