Discourse graphs — drawn from published papers, and from pilot records
Papers as discourse graphs, with their open issues made actionable
A published paper, drawn in the grammar of Discourse Graphs: the question it asked, the hypothesis it held, the experiments it ran, the results each produced and the issues it left open. Every result carries the place in the paper it came from and a short quote. Every open issue is checked against the work that has cited the paper since, and then given a proposed experiment, with the control and the sensitivity statement a researcher would ask for first.
Four sets drawn from papers
- Six NEJM papers, each open issue checked and designed
Two null trials and four mechanistic papers, drawn from their CC BY author manuscripts in PubMed Central. Every open issue carries a follow-up check against the citing literature, a query of the registry, a proposed experiment and a mocked hand-off to a laboratory. - Eight open-access SAGE papers
One page per paper, from psilocybin to welded joints. Each open issue carries a proposed experiment and a panel for the researcher’s objections; two pairs of papers are drawn together. - Seven physics-informed neural network papers
CC BY papers by Karniadakis and co-authors, drawn from their full text. - Eleven ACM papers, positive and negative
CC BY papers from ACM venues, chosen to show that the grammar draws a confirmed result as readily as a failed one.
How to read these graphs
A discourse graph breaks a piece of research into typed one-sentence nodes and draws how they bear on one another. The grammar is that of Discourse Graphs (Chan and Akamatsu). Read each graph left to right: a question, the experiments that addressed it, the results they produced, and the hypothesis every result is drawn against.
- QUE Question
- What the work set out to answer.
- HYP Hypothesis
- The answer the authors expected. Every result earns one edge against it.
- EXP Experiment
- One attempt at the question: a trial arm, an assay, a survey, an analysis.
- RES Result
- One finding from an experiment, with where in the source it comes from.
- ISS Issue
- An open question or untried branch the work leaves for someone else.
- addresses
- An experiment addresses the question.
- produces
- An experiment produces a result.
- supports
- A positive result that backs the hypothesis.
- opposes
- A negative result that counts as evidence against the hypothesis, or a refutation of an earlier claim.
- informs
- A partial or mixed result that bears on the hypothesis without settling it.
- consistent with
- A negative or inconclusive result that does not count either way: no positive control passed, or there is no sensitivity or power statement, so the absence of an effect cannot be told from a measurement that failed.
- relation
- A thin dotted line: a link the source itself states between two experiments, such as one being derived from another.
A dashed box marks a result the rule above counts as uninformative, or an issue. In the drawing, click a node to jump to its text below.
8 graphs drawn from records: the pilot studies
These graphs are drawn not from the papers but from the Negative-Results Records the pilot encoded from them. Every node and edge comes from a field the record already holds, and a negative is drawn as opposing the hypothesis only when the record’s informativeness rule counts it as evidence of absence; otherwise it is merely consistent with it. How the two models line up.
- Laccase mediator systems do not detectably degrade PFOA or PFOS; the reported loss is sorption to the enzyme
An Artifact of Perfluoroalkyl Acid (PFAA) Removal Attributed to Sorption Processes in a Laccase Mediator System · 4 attempts, 11 findings · headlinenegative-not-replicated - Is TMEDA inhibition a reliable test for hidden BH3 catalysis, and is any amine a better trap?
Hidden Boron Catalysis: A Cautionary Tale on TMEDA Inhibition · 4 attempts, 20 findings · headlinepartial - RuF6 does not oxidise XeF6 to xenon(VIII); F2 is evolved and only xenon(VI) salts form, one of them new
Failed Attempts at Oxidation of XeF6: Synthesis and Characterization of [Xe2F11][RuF6] · 3 attempts, 13 findings · headlinenegative-not-achievable - Computationally guided search for OSDAs to crystallise aluminosilicate STF and IFR in hydroxide media
Lessons from Failed Attempts of Computationally Guided Synthesis of Aluminosilicate STF and IFR Zeolites in Hydroxide Media · 4 attempts, 16 findings · headlinenegative-not-achievable - Alcalase pretreatment for CRISPR-Cas9 HDR knock-in in cell-walled Chlamydomonas reinhardtii (iced project)
Evaluation of Alcalase pretreatment for Chlamydomonas reinhardtii CRISPR knock-in · 6 attempts, 11 findings · headlineabandoned - Phylogenetic trait mapping to find the function of a divergent fungal actin (iced)
A structurally divergent actin conserved in fungi has no association with specific traits · 3 attempts, 11 findings · headlinepartial - Label-free flavoprotein imaging of the mouse cortex as a behaviour-independent assay for touch, pain and itch (iced)
Label-free neuroimaging in mice captures sensory activity in response to tactile stimuli and acute pain · 3 attempts, 9 findings · headlinepartial - Can Raman spectra distinguish yeast strains and species, and do batch effects confound the answer?
Leave-one-batch-out cross-validation reveals strong batch effects in Raman spectroscopy of yeast cultures · 3 attempts, 8 findings · headlinepartial
How the two models line up
Discourse Graphs is a grammar for arguing: typed one-sentence notes, linked by whether they support or oppose one another. The Negative-Results Record is a format for recording an attempt: what was tried, under what conditions, with which controls, and whether the negative counts. The graphs on this page are the first drawn over the second. Every node and edge comes from a field the record already holds, and nothing was added to the schema to draw them.
| Discourse Graph | Negative-Results Record |
|---|---|
| Question (QUE) | question on the path; attempts and findings carry their own |
| Hypothesis (HYP) | hypothesis on the path |
| Experiment (EXP) | an attempt record, with stages[] and the performer of each stage |
| Result or Evidence (RES, EVD) | a finding, with effect, controls, sensitivity and evidence[] pointing into the source |
| Claim or Conclusion (CLM, CON) | outcomeClass plus outcomeSummary; there is no separate claim node |
| Source | sources[], each with a role, a licence and its version status at the time of ingest |
| Issue (ISS) | openQuestions, untriedBranches, nextSteps |
| supports, opposes | confirms, refutes, contradicts between records; between a finding and the hypothesis, derived from outcomeClass and informativeness |
| informs | informs, informedBy |
| produces | wasGeneratedBy |
| consistent with | no field; it is what the informativeness rule yields when a negative does not qualify as evidence of absence |
The last row is the interesting one. The Discourse Graphs lab conventions say a claim is supported by a result only if certain conditions hold, and otherwise the result is merely consistent with it, and the conditions are left as placeholders for each lab to fill in. The record fills them in with code: a negative supports the absence of an effect only if a positive control passed, the negative control stayed clean, and a sensitivity or power statement exists. That rule is what decides, on every graph here, whether a red edge is drawn or a dashed grey one.
The Discourse Graphs grammar is described at discoursegraphs.com and in Chan and Akamatsu’s 2024 preprint, arXiv:2407.20666. The schema is at portolansoft.com/nrr/schema.