Overtone Analytics

Overtone AnalyticsOvertone AnalyticsOvertone Analytics
  • Home
  • Overtone Interviews
  • Methods
  • About
  • Writing
  • More
    • Home
    • Overtone Interviews
    • Methods
    • About
    • Writing

Overtone Analytics

Overtone AnalyticsOvertone AnalyticsOvertone Analytics
  • Home
  • Overtone Interviews
  • Methods
  • About
  • Writing

Methods

Two colorful human profiles made of letters facing each other.

The Overtone Delta: how we score public statements

Overtone studies how scientific and medical claims are worded as they move from the lab to the public. Every public statement about a therapy or test gets two scores from 0 to 6: how strong the words are (the claim), and how strong the public evidence was on the day they were said. The scales are built to match, so the right words for a stage of evidence score the same number as that stage.

The Overtone Delta is the difference between the two:

  • 0: the words fit the evidence.
  • +1, +2, +3: the words ran that many steps ahead.
  • Below 0: the words were more cautious than the evidence allowed.

The two scales

Evidence follows the standard order of clinical evidence:

  • 0: lab, genetic or computer evidence only
  • 1: animal studies
  • 2: first human data, in a handful of patients
  • 3: early trials in 10 or more people with adequate follow-up
  • 4: a pivotal trial met its main goal, or approval based on adequate trials
  • 5: level 4, plus years of follow-up showing the benefit lasts
  • 6: level 5, plus most patients need no further treatment for many years

Diagnostic tests use a parallel scale built on whether a test has been shown to measure correctly.

Claims run from cautious words ("may," "suggests") through "promising" and "demonstrated" to "durable" and "cure." Hedges ("potential," "may") and stated limits ("in mice," "in 6 patients," "investigational") lower a claim by a step.

What we collect, and how we check the method

At each milestone (first human data, mid-stage results, pivotal results, approval and later), we collect every public voice: the company, the university or hospital, the investigators, the regulator, the journal and the press. Each statement is recorded with its verbatim wording, date and source link, so anyone can check it.

We also score positive controls: cases where strong language was earned, such as hepatitis C "cure" rates and cystic fibrosis "breakthrough" results. A fair method should score them at zero, and ours does. When a control exposes a rule that's too strict, we fix the rule, write it down and apply it to every case.

Limits, and where it stands

A language score doesn't detect fraud or judge a drug's value. It shows how far the words ran ahead of what anyone could check at the time.

All scores are provisional: one coder so far, with an independent second coder next. Our current cases were chosen knowing how they ended, so the patterns are hypotheses to test, not findings. The scale follows the accepted order of clinical evidence, but its specific thresholds are our own conventions, written down and versioned so they can be challenged.

Rulebook version 0.4.1, November 2026.

Overtone Interviews: what to expect

Every Overtone Interviews episode follows the same protocol: the same questions, in the same order, whatever the guest or format. We publish this protocol for two reasons: to ensure prospective guests understand precisely what they're agreeing to before they accept an invitation, and to maintain the integrity of the series' goal — if we're exploring the drift of scientific language, it is essential that our approach remains transparent and consistent, rather than being adjusted quietly from episode to episode. Every guest reviews this page before consenting to an interview.

What you get

A Published Feature under your name that you approve before it goes live — easy to share on LinkedIn or X, with a natural link from your lab page or CV. It's a citable record of your work at this point in your career, highlighting not just what you eventually published, but also how you discuss your work. Practicing how to explain findings to those outside your subfield is a valuable skill that manifests in many areas, including job talks, grant panels, and media requests. This is low-stakes practice supported by an editor, rather than addressing a cold audience. You'll receive professional feedback on your message and, after our discussion, we're happy to continue the dialogue — whether it’s about how you’re presenting this work to funders, your department, or a broader audience if that adds value for you. Much of what we discuss may not make it into an episode, but that conversation is invaluable. You'll gain early access to a growing network, joining the founding cohort of this series and Overtone Analytics’ broader research community, with the first opportunity for a follow-up interview if your work progresses. There’s no cost or catch — it's unpaid, but the only requirement is your time and your words. You’ll review the draft before publication and can request changes to ensure your voice is accurately represented.

Intake snapshot

Published alongside every episode: 


- Name, current role, institution 

- Field / therapeutic area 

- Career stage 

- Stage of the work discussed (unpublished, preprint, under review, published) 

- The finding, in one sentence, in the guest's own words 

- Self-rated confidence in the finding at the time they first described it internally (1–5) 

- Self-rated confidence in the finding today (1–5) 


Together, these show how the way a finding is described can change over time.

The questions

Origin and first framing


Take us back to the moment you first thought "this might be something." What's the first sentence you used to describe it — to your PI, in a lab notebook, in a lab meeting? How did you convey the significance of the finding the first time you had to write it down formally? Do you still have that original language? Looking back at that early description, is there a word or phrase from that initial description you wouldn't use today? What has changed since then?


Bench to publish


What evolved between how you described this internally and how it reads in the submitted (or published) manuscript? Where did the pressure to sharpen, soften, or reframe the claim originate — a co-author, a reviewer, a PI, a funder, or your own second-guessing? Was there a specific comment, edit, or conversation that altered how you told the story? What got cut from the narrative along the way, and why?


Calibration and language


Which word or claim in the final paper are you most confident still holds up? Are there any words or claims that no longer align with your story—would you change anything in hindsight? How do you decide between hedged language ("suggests," "may," "is consistent with") and more assertive claims? If you had unlimited space, what caveat would you reintroduce that didn't make the final cut?


Translation and forward look


What needs to be true for this finding to have clinical or commercial implications? What do you think this finding could become that you wouldn't yet say in a paper? What's the most likely way someone outside your subfield could misapply or over-read this result? If someone revisits this paper in five years, what do you think will still hold up — and what might appear overstated, or understated?


Reflection and advice


What's one thing you wish someone had shared with you about scientific writing before your first submission? What advice would you provide to an early-career scientist trying to describe a finding honestly without underselling it?


Protocol updated October 2026: one question added and one widened under Translation and forward look. Guests interviewed before this date were not asked them.

For returning guests

If we talk again after the work has advanced, we'll ask: How does the finding read to you now compared with when we first talked? What's different between where you thought this would go and where it actually went? Did the language in the eventual publication, patent or trial protocol match what you expected when we first talked?

What we don't publish

We also track a short set of internal, non-public fields for each interview—funding mechanism, model system, and a few language-calibration ratings—to support our own research into how scientific language changes. These details aren't published with individual episodes; when we share findings from that research, it will be in aggregate or de-identified form. This research includes developing our own methods and tools for analyzing scientific language.

Get new case studies and interviews by email

Sign up for new case studies, interviews and methods updates by email.

Overtone Analytics

Join us in listening for the overtone 

Get Started

Copyright © 2026 Overtone Analytics - All Rights Reserved.

Powered by

This website uses cookies.

We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.

Accept