Introduction to Connectomics: instructor model responses

Worked calculations, claim sorting and an exemplar study brief.

Learner worksheet · Lecture plan

These answers use the worksheet’s synthetic data. They are examples, not the only acceptable wording. This page is public and suitable for formative assessment.

Worked calculations

There are 12 synapses, 3/4 connected pairs (75%), 12/4 = 3 synapses per eligible pair, and 12/3 = 4 synapses per connected pair. The conditional mean excludes the zero pair. Neither mean is a firing rate or physiological weight.

Claim sorting

  1. A. Six reviewed synapses in the specified region is a structural observation, under the exercise’s assumption that the counts and identities are correct.
  2. B. The inference requires a relationship between count and physiological strength, with other determinants controlled. A defensible replacement is “A1 makes more observed synapses onto B1 than B2.” Testing physiological strength needs an appropriate functional measurement.
  3. C. No activity or causal evidence was supplied. “A1 has an anatomical route to B1” is supported. Task-linked recording and a suitable perturbation would be needed to test the causal claim.
  4. Unsupported extrapolation, C in this exercise’s scheme. The observed zero is region-limited. Replace it with “No A2-to-B1 synapse was found in the reviewed region.” Establishing absence outside it needs additional structural coverage and stated detection sensitivity. This is a sampling problem, not necessarily a need for physiology. Accept that distinction rather than grading the letter alone.

Example study brief

Question: Within a prespecified reconstructed region, is the observed fraction of A-to-B pairs containing at least one synapse greater than for A-to-C pairs with comparable opportunities for contact?

Measurement: Acquire synapse-resolving EM and reconstruct the relevant neurites and partner identities. The analysis unit is a directed cell pair within the region. For an illustrative pilot, plan 20 eligible A-to-B and 20 matched A-to-C pairs. This is a proposed pilot size, not a power calculation or evidence of population coverage.

Endpoint: Difference in connected-pair fractions, in percentage points. Report both numerators and eligible-pair denominators. Define eligibility before inspecting the outcomes, including boundary exclusions and reconstruction coverage.

Null: No difference in connectivity associated with target label after matching on contact opportunity and location. Permute B/C labels only within prespecified matched sets while holding the observed graph fixed. State that label exchangeability within those sets is an assumption. Pairs sharing neurons are not independent replicates.

Error check: Review uncertain contacts and identities without knowing which outcome favors the hypothesis. Repeat the endpoint with uncertain pairs included and excluded. A difference that reverses under plausible corrections weakens the claim.

Non-claim: This pilot would not establish stronger physiological transmission or behavioral causation. Those require additional functional evidence. A single region also does not establish a general rule across brains.

Feedback guide

Score four dimensions 0–2 each: correct denominators; evidence-matched claims; an explicit population/comparison; and an error check plus non-claim. Use 2 for an explicit, coherent response, 1 for a recoverable omission, and 0 for absent or contradictory reasoning. A proficient response has at least 6/8 and no zero in evidence-matched claims or non-claim/error reasoning. This is a local teaching rubric, not a validated assessment instrument.

Do not reward an elaborate statistical method without a defensible denominator. Ask “What observation would change your mind?” before suggesting a method.

Teaching material: CC BY-SA 4.0, NeuroTrailblazers.