Spot-Checking the 2% Claim: Random Sampling Renaissance Science

We claim that only ~2% of Renaissance Latin has been translated into English. But how do we know? I took 100 random works from the USTC database and searched for translations. Here's what I found.

Porta: Natural Magick

1658 · TRANSLATED

Glauber: Furnaces

1689 · TRANSLATED

Maier: Symbola Aureae Mensae

1617 · NOT TRANSLATED

The Problem with Estimates

When we say “2% of Renaissance Latin is translated,” we're making a claim based on comparing the number of known translations (~2,000 Renaissance works) against the estimated corpus size (~100,000 unique works in USTC). But this top-down approach has a problem: maybe translations exist that we haven't catalogued.

The only way to validate the estimate is empirical: take a random sample of works from USTC and actually search for translations. If we find significantly more than 2%, our estimate is wrong.

Methodology

For each sampled work, I searched for English translations using web search, library catalogues, and my knowledge of translation series. I categorized results as:

• Translated: Complete or substantial English translation exists
• Classical/Medieval Reprint: Work by pre-Renaissance author (Euclid, Aquinas) republished in Renaissance
• Not Translated: No English translation found

Sample 1: 25 Random Science Works (seed=1234)

Notable Untranslated Works

Several works in our random sample stand out as significant gaps:

Key Findings

Comparison: General USTC Sample

Earlier random samples from the full USTC corpus (not filtered by Science) showed much lower translation rates:

The key insight: when classical reprints are removed, the Renaissance-original translation rate drops to 2-4%. Our headline estimate of “~2% translated” is validated and may even be generous.

Translated Works: Where They Come From

The translated works in our samples follow a pattern:

Recommendations for Translation Roadmap

Based on this sampling exercise, here are the highest-value targets from the Science/Natural Philosophy category:

Conclusion

The random sampling exercise validates our headline estimate: for truly Renaissance-original works, translation coverage is around 2-4%. The Science category shows higher rates (~13%) due to:

• Celebrity scientists (Descartes) attracting translators
• 17th-century English interest in alchemy
• Classical/medieval reprints inflating numbers

But even in this relatively well-served category, major works remain completely untranslated: Cardano's De subtilitate, Libavius's Alchemia, Scaliger's Exercitationes. For theology, law, pedagogy, and other genres that make up the bulk of USTC, the rate is almost certainly lower.

The 2% estimate isn't pessimistic. It's realistic.

Reproducibility

The Python code used for sampling:

import pandas as pd
import random

# Load USTC Latin editions
df = pd.read_csv('ustc_latin_editions.csv', low_memory=False)

# Filter for Science category
science = df[df['classification_1'] == 'Science']
print(f"Science works in USTC: {len(science)}")  # 3,451

# Random sample with fixed seed
random.seed(1234)
sample_indices = random.sample(range(len(science)), 25)
sample = science.iloc[sample_indices]

for _, row in sample.iterrows():
    author = row['author_name_1'] or 'Anonymous'
    title = row['std_title'] or 'No title'
    year = row['year']
    print(f"{author}: {title} ({year})")

Data source: Universal Short Title Catalogue Latin editions export (533,308 records).