Session Kit: Module 16: Scientific Visualization for Connectomics
Everything needed to run Module 16 as a taught session: prep, timing, materials, misconceptions, rubric.
Generated from modules/module16.md. Edit the module page, not this file.
At a glance
| Duration | 4 hours |
| Capability target | Produce a figure set that communicates connectomics findings accurately, including uncertainty and data-quality context, for both expert and mixed audiences. Students will leave this module able to choose the right visualization form for a given scientific claim, build publication-quality figures using standard tools, and defend every design choice in terms of clarity and honesty. |
| Learners leave with | Three-figure set exported at publication resolution with complete captions |
Before you walk in
- You can state the capability target in one sentence without reading it.
- You have one worked example you will narrate, including where you are unsure.
- Data access works — accounts, viewer, notebook — verified today, not last week.
- The rubric is visible to learners before they start, not after.
- You have decided what “uncertain” earns, and you will say so out loud.
Learners should arrive having covered:
- Basic plotting library familiarity
- Understanding of analysis outputs
Materials
Run of show
| Time | Segment | Your note |
|---|---|---|
| “Excitatory neurons in layer 4 receive more synaptic input than those in layer 2/3.” | ||
| “Reciprocal connections are enriched between Martinotti cells.” | ||
| “Axonal arbors of chandelier cells are spatially restricted to a 100-micron radius.” |
The activity
Scenario: You are preparing a three-figure package for a short connectomics paper reporting cell-type-specific connectivity patterns in a cortical volume. Your dataset includes a 50x50 cell-type adjacency matrix, morphological reconstructions for three example neurons, and synapse count distributions across layers.
- Map each claim to required visual evidence. For every result sentence, identify what figure panel and what visual encoding will support it.
- Select the appropriate plot type. Use the decision framework: topology questions get node-link diagrams or matrices; quantity questions get heatmaps or bar charts; spatial questions get renderings; distribution questions get histograms or violins.
- Draft candidate visuals with uncertainty layers. Include error bars, confidence bands, or explicit missing-data indicators from the start — do not plan to “add them later.”
- Run critique for misinterpretation risk. Show the draft to someone unfamiliar with the analysis and ask them what they conclude. If their conclusion differs from your intent, revise.
- Check accessibility. Run the figure through a colorblind simulator (e.g., Coblis or the Matplotlib colorblind check). Verify grayscale legibility.
- Revise for clarity, accessibility, and reproducibility. Add scale bars, axis labels, panel letters, and complete captions.
- Export figure package with caption metadata. Include figure files at publication resolution (300+ DPI for raster, vector preferred), caption text, and a note on the dataset version and code used to generate each panel.
What learners hand in
- Three-figure set exported at publication resolution with complete captions
- Uncertainty annotation strategy document (one paragraph per figure explaining what uncertainty is shown and why)
- Revision log from peer critique (at least two specific changes made in response to feedback)
- Accessibility check report (colorblind simulation screenshot for each figure)
Misconceptions to target
These are the errors this session exists to prevent. Surface them in the debrief rather than pre-empting them in the lecture — a misconception a learner has voiced is far easier to correct than one they are holding silently.
- They may believe: Making a figure “look good” is not the same as making it truthful. A beautiful 3D rendering with no scale bar and no uncertainty indicators is worse than an ugly but complete 2D plot.
- Surface it by asking: “What would have to be true for that to hold? What would change your mind?”
- They may believe: There is no single “best” visualization. The best choice depends on the claim.
- Surface it by asking: “What would have to be true for that to hold? What would change your mind?”
- They may believe: Complexity in a figure does not equal rigor. Simplicity with completeness is the standard.
- Surface it by asking: “What would have to be true for that to hold? What would change your mind?”
- They may believe: Cleaner-looking plots are not always better. A plot that hides uncertainty is less honest than one that shows it.
- Surface it by asking: “What would have to be true for that to hold? What would change your mind?”
- They may believe: Aesthetics cannot replace methodological clarity. A beautiful figure that only some people can read is not a good figure.
- Surface it by asking: “What would have to be true for that to hold? What would change your mind?”
Naming the norm
Every session is a chance to make one piece of the hidden curriculum explicit. Pick a moment where you would normally just do the professional thing, and say out loud why you are doing it — then ask whether anyone was taught that.
For this session, the candidate is whichever norm the activity most depends on: stating an assumption in the same sentence as the claim, recording the version a number came from, or saying “uncertain” and having it count as a real answer. See the hidden curriculum for the collected set and why naming them is a fairness intervention rather than etiquette.
Assessment
- Minimum pass: visuals map clearly to claims, include uncertainty context, use perceptually uniform colormaps, and have complete axis labels and scale bars.
- Strong performance: high clarity across expert and non-expert audiences, minimal misinterpretation risk, colorblind-safe design, explicit documentation of dataset version and code used for each panel, and thoughtful caption language that narrows interpretation bounds.
- Failure modes: overloaded figures with too many overlapping elements, missing scale context, hidden uncertainty, rainbow colormaps, gratuitous 3D renderings, captions that do not mention data quality or limitations.
Grade the reasoning, not the answer. A correct call with no evidence chain should not outscore a well-reasoned incorrect one — and saying so publicly changes behaviour within one session.
Exit prompt
Take one existing connectomics figure (from a paper, a classmate, or your own work) and perform a full audit:
- Identify the claim the figure is supposed to support.
- Add one uncertainty cue (error bar, confidence band, or missing-data indicator).
- Replace the colormap with a perceptually uniform alternative if needed.
- Write a two-sentence caption that narrows interpretation bounds and specifies the dataset version.
- Run the figure through a colorblind simulator and note any issues.
If this session goes wrong
- Nobody talks in the debrief. You asked “any questions?” Ask instead: “Which cue would you drop first if the data got worse?”
- Everyone finishes early. They are pattern-matching, not judging. Give an ambiguous case where the answer is “uncertain” and see what happens.
- Nobody finishes. The scaffolding came off too fast. Work the next case together rather than pressing on.
- A learner is silently lost. The most likely cause is unstated vocabulary. Point them at the dictionary and check back.