Neuronal Ultrastructure

Instructor script: "Two facts alone resolve a large share of early-annotator confusion: ribosomes rule out axon, and glycogen granules indicate astrocyte." Put the organelle table from unit §1 on screen or in hand for the rest of the session — sizes are approximate, but relative sizes and presence/absence patterns are what you actually use.

Instructor move: for each compartment on the diagram, ask the room which organelles they would expect present and — more usefully — which would be absent. Then require two independent cues before anyone names a compartment in the EM that follows.

This anchors the soma end of the compartment table: rough ER and a nucleus put you in a cell body, the one place where the ribosome cue is unambiguous rather than a judgement call. Point out the mitochondrion's double membrane and cristae — mitochondria are everywhere except thin spine necks, so they are consistent with every compartment and discriminate none.

Instructor script: "Same cue, real human tissue, rendered from the public H01 volume. Find the nucleolus first, then trace the nuclear envelope, then look outward into the neuropil." The contrast between the pale soma and the dense neuropil around it is what lets you find cell bodies before you can resolve any synapse.

The label here comes from a model, not an expert — treat it as a claim with evidence, like any other label. Ask the room for a second, independent cue before accepting it: microtubules in loose parallel arrays, ribosome rosettes, or a spine emerging. Remember the rule from unit §4: three microtubule observations are one piece of evidence; a ribosome plus a spine plus a taper is three.

The three criteria, all required: a presynaptic vesicle cluster at the apposition itself; a cleft of uniform width (20–30 nm asymmetric, ~15–20 nm symmetric); a postsynaptic density. Plus persistence: a real synapse 200–500 nm wide should appear on two to five 40 nm sections.
Instructor script: "The single most common beginner error is calling a synapse from dark contrast alone. A tangentially cut membrane, precipitate, a glial apposition or an adherens junction can all look dark. No vesicles, no synapse."

Check size against the table before naming anything: clear synaptic vesicles run 35–50 nm, dense-core vesicles 80–120 nm, microtubules ~25 nm across. This is a measurement rather than an impression. Vesicle shape is partly a fixation artifact, and it is not independent of vesicle visibility — if you cannot see vesicles clearly, you cannot use their shape either.

The schematic names the presynaptic terminal, presynaptic and postsynaptic membranes, cleft, vesicles and the postsynaptic dendrite. The translation from idealized diagram to real, noisy tissue is the step the three synapse criteria have to survive. Teach an explicit alternative-hypothesis statement for any structure a learner cannot locate in the micrograph.

Build a full evidence chain in the form of the unit's worked example — geometry, organelle content, synaptic role, neighbourhood — and stop at the point where the chain would need continuity across sections that a single still cannot supply. Stopping there, and saying so, is the correct output.

Define the tiers operationally (unit §3). High: two or more independent cues agree and continuity across three or more sections confirms. Medium: one strong cue, or two non-independent cues. Uncertain: cues conflict, or the decisive cue is not visible.
"Uncertain" is a valid and valuable output. A dataset with 8% of calls flagged uncertain with reasons is more useful than one with 100% forced — the uncertain set is exactly the review queue.
Independent cue families: geometry and caliber; organelle content; synaptic role; neighbourhood; long-range continuity. Two cues from different families is the checkable rule for high confidence.

A good patch from the unit's check-yourself: a bouton with round clear vesicles making a thick-PSD contact onto a spine head, and apparently also a thin symmetric contact onto a nearby shaft. Most likely the second contact is not a synapse from this bouton — check for vesicles at that apposition. If it genuinely is, one terminal making both types would be unusual: suspect a merge error that fused two axons. The biology looks wrong, so suspect the segmentation.

Use this to connect the cue work back to acquisition: block-face imaging gives near-perfect z-alignment and no lost sections, which is exactly what makes cross-slice continuity — step 3 of the protocol — practical. The trade-off is z-resolution that limits thin-process tracing (Unit 02's modality chart).