**Last updated: Version 8/9.0 (June 2026)**
---
**Scientific analysis based on the primary source:** Mildner, S. (2026). *Geodynamic Reinterpretation Model for Ptolemy's Germania Magna: General Model Description, Cartometric Foundations*, (v9.0). EarthArXiv (Preprint). [https://doi.org/10.31223/X5KB51](https://doi.org/10.31223/X5KB51)
([📥 **Download NEW-v9.0-PDF**](https://zenodo.org/records/20758648/files/Geodynamic_Model_Description_for_Ptolemys_Germania_Magna___eartharxiv__c9.0.pdf?download=1))
---
***Disclaimer***
*This article is a technical companion piece summarising what is new in Version 8/9.0 of the model. It does not repeat the cartometric, geodynamic, or narrative arguments already presented in the articles listed below — those remain current and are not superseded. Version 9.0 adds three independent robustness checks (a control-region comparison, a sequential Bayesian re-analysis, and a formal statistical power assessment) plus a new monograph part that gives the Long-Transgression hypothesis an explicit falsifiability framework. The model has not been evaluated by peer review.*
---
## Why a Fourth Layer of Robustness Checks?
The earlier articles on this site established the cartometric model (the affine baseline and the kinematic corrections), validated it with a formal out-of-sample blind test, and subjected it to a full statistical battery — AIC/BIC, leave-one-out cross-validation, bootstrap, permutation testing, and Moran's I. A subsequent update (v8.2) added an identification-multiverse analysis testing whether the result depends on which modern places are matched to which Ptolemaic names.
Version 9.0 closes three remaining gaps that careful readers could flag as worth addressing explicitly, plus it gives one existing hypothesis a sharper, falsifiable shape:
| # | Addition in v9.0 | What it answers |
| --- | --- | --- |
| 1 | **Gallia control-region test** ($T_{\rm ctrl}$) | "Maybe every Ptolemaic region has a longitude offset like this — why is the Elster Cluster special?" |
| 2 | **Sequential Bayes factor** ($\mathrm{BF}_{10}$, $n=2\to6$) | "Were the six cluster points chosen because they happened to show a large effect?" |
| 3 | **Formal power analysis** (§C.1) | "Isn't $n=6$ too small a sample to draw any statistical conclusion from?" |
| 4 | **Part VIII: Long-Transgression falsifiability framework** | Embeds the 120 km coastline-shift claim in the full Holocene Baltic Sea record and converts it into five explicit, independently testable predictions |
Each of the first three is, in essence, an answer to a specific, legitimate objection. The fourth is a structural addition: it does not introduce new statistics, but it does give the hypothesis discussed narratively in the companion review article (linked above) a formal place in the monograph's falsification programme, alongside Cohen's $d$ now reported next to every $t$-statistic throughout the volume.
---
## 1. The Gallia Control-Region Test ($T_{\rm ctrl}$)
### The Objection
A natural methodological worry is that Ptolemy's coordinate system may simply carry a general, model-wide longitude compression — in which case the Elster-Cluster (EC) displacement would be nothing special, just one more patch of an empire-wide distortion rather than evidence for a localised geodynamic block translation.
### The Test
Version 9.0 answers this with an external control region. The same Model-A affine transformation used throughout the monograph — calibrated exclusively on the three river-mouth anchors K1 (Rhine), K2 (Elbe), K3 (Vistula/Oderberg) — is applied, unmodified, to ten well-identified Ptolemaic cities in Gallia (six in Gallia Narbonensis, four in Gallia Belgica; identifications following Stückelberger & Graßhoff, 2006, Books 2.7–2.10). Gallia is geographically and tectonically unrelated to the proposed Zechstein décollement and lies far outside any plausible reach of the Elster-Cluster mechanism — exactly the kind of region where, if the "universal compression" objection were correct, a comparably large and *coherent* offset should appear.
It does not.
| Region | $n$ | Mean $\Delta\lambda$ | $\sigma(\Delta\lambda)$ | $t$ | $p$ |
| --- | --- | --- | --- | --- | --- |
| **Elster Cluster (model-primary)** | 6 | $-88.9$ km | $14.5$ km | $-15.04$ | $2.4\times10^{-5}$ |
| Gallia Narbonensis | 6 | $-52.2$ km | $37.8$ km | $-3.39$ | $0.020$ |
| Gallia (all 10) | 10 | $-45.5$ km | $55.6$ km | $-2.59$ | $0.029$ |
Two diagnostics separate the two patterns cleanly:
- **Spatial coherence.** The EC cluster's scatter ($\sigma = 14.5$ km) is **2.6× tighter** than the most comparable Gallia subset. A tight $\sigma$ is the signature of a uniform physical translation — every point moving by nearly the same amount; a wide $\sigma$ is the signature of diffuse measurement noise, exactly what one would expect from generic ancient surveying error accumulating differently from city to city.
- **Statistical extremity.** The EC's $|t| = 15.0$ is **4.4× larger** than Gallia Narbonensis's $|t| = 3.4$. The modest Gallia result is fully consistent with the well-documented general westward longitude compression that affects much of Ptolemy's map (a known artefact of his adopted circumference for the Earth); the EC signal sits on top of that background and is categorically stronger.
In short: yes, Ptolemy's map has a general longitude distortion — but that distortion, measured directly in an unrelated control region using the identical method, is markedly weaker and markedly less coherent than the Elster-Cluster signal. The control test does not weaken the case; it sharpens exactly what makes the Elster Cluster distinctive.
---
## 2. Sequential Bayes Factor: Ruling Out Cherry-Picking
### The Objection
Six points were selected to form the Elster Cluster. Were they selected, even unconsciously, *because* their displacements looked unusually large — in which case the headline statistic would be an artefact of selection rather than a genuine signal?
### The Test
Version 8/9.0 introduces a *sequential* Bayes-factor analysis: starting from $n=2$ and adding one EC point at a time, the model recomputes the Jeffreys–Zellner–Siow default Bayes factor ($\mathrm{BF}_{10}$, Cauchy prior scale $r=\sqrt2/2$; Rouder et al., 2009) at every step, under three different point-entry orderings — ascending $|\Delta\lambda|$ (the most conservative: weakest signal enters first), descending $|\Delta\lambda|$ (the most aggressive), and the actual historical order in which the six identifications entered the model (S3 → S7 → S6 → S5 → S-L → S-C).
| $n$ | Mean $\Delta\lambda$ | $t$ | $\mathrm{BF}_{10}$ (ordering A) | Jeffreys' classification |
| --- | --- | --- | --- | --- |
| 2 | $-74.1$ km | $-18.07$ | $2.7$ | Anecdotal |
| 3 | $-78.4$ km | $-15.88$ | $10.5$ | Strong |
| 4 | $-80.7$ km | $-19.44$ | $72.2$ | Very strong |
| 5 | $-84.7$ km | $-16.35$ | $238.6$ | Extreme |
| 6 | $-88.9$ km | $-15.04$ | $751.9$ | Extreme |
The result: **all three orderings are strictly monotone — no ordering produces a reversal.** There is no subset of EC points whose removal would *strengthen* the case, and no single point that "carries" the result on its own. Evidence accumulates steadily regardless of which point is considered first, last, or in between. This is the quantitative answer to the cherry-picking objection: a post-hoc-selected set chosen to maximise $t$ would be expected to show exactly the kind of fragility — a point or two propping up an otherwise weak signal — that the monotone curve rules out. At $n=6$, the posterior odds against a zero displacement stand at 752:1 (independently cross-verified in the Wolfram Language to five significant figures).
---
## 3. Is n = 6 Too Small? A Formal Power Analysis
### The Objection
A sample of six points seems small for any one-sample $t$-test, and small samples are conventionally associated with unreliable or underpowered statistics.
### The Test
Version 9.0 makes the power calculation explicit (new §C.1 of Appendix C) rather than leaving it implicit. The Elster-Cluster effect size is Cohen's $d = -6.14$ — more than twelve times the conventional "medium" effect ($d \approx 0.5$) and nearly eight times the conventional "large" effect ($d \approx 0.8$) used as benchmarks in the social and life sciences.
At this effect size:
- Statistical power at $n=6$, $\alpha = 0.001$, is $\mathrm{Power} = 0.9995$ — a random sample of six points drawn from a population with this true effect would produce a significant result at the strict $\alpha=0.001$ threshold in **99.95 %** of repeated experiments.
- Power already exceeds 99 % at $n = 3$.
- The minimum effect size needed for conventional 80 % power at $n=6$ is $d \approx 1.44$; the observed $d = 6.14$ exceeds that threshold by a factor of $\approx 4.3$.
The conventional "$n$ is too small" heuristic is calibrated for effect sizes in the $d \approx 0.2$–$0.8$ range, where genuinely large samples are needed to detect subtle effects reliably. It does not transfer to a case where the effect itself is an order of magnitude larger than that range. As the manuscript puts it: at $d = 6.14$, even $n = 2$ would suffice to reject the null at $\alpha = 0.05$ with probability $> 99\,\%$. The actual limiting factor in this study is not statistical power but the number of Ptolemaic sites that can be confidently identified at all — a constraint fixed by the surviving ancient source, not by the analysis.
---
## 4. Part VIII: Giving the Long-Transgression Hypothesis a Falsifiability Framework
The companion review article on this site (linked above) already laid out the *Long-Transgression* hypothesis in narrative form: the proposition that the Littorina Transgression in the southern Baltic did not stabilise in the Middle Holocene as the classical model holds, but continued — episodically amplified by ongoing forebulge subsidence, the Roman Climate Optimum, and hydrographic feedbacks — reaching its actual Holocene maximum only during the Roman Imperial Period, exactly when Ptolemy's data were compiled. That continuation is what makes the model's cartometric claim — that the *Oceanus Germanicus* coastline lay roughly 120 km south of its modern position, near the latitude of Eberswalde — physically coherent rather than ad hoc.
What Version 8/9.0 adds is not a new narrative but a new monograph part (Part VIII) that situates this claim systematically within the full Holocene Baltic Sea record — reviewing the classical five-stage transgression sequence (Baltic Ice Lake → Yoldia Sea → Ancylus Lake → Mastogloia Sea → Littorina Sea) in detail, identifying specific empirical anomalies the classical model leaves under-explained (abrupt allostratigraphic boundaries, erratic accumulation rates at the Littorina/Post-Littorina transition, coherent 6th-century AD palaeoecological disruption signals), and then converting the hypothesis into five explicit, independently testable predictions:
1. **Chronostratigraphy of the amphibious zone** — targeted coring at the postulated ancient shoreline (≈52°50′N, e.g. the Oderbruch and Rhinluch peatlands) should show a rapid, datable 6th-century transition from marine/brackish to purely terrestrial deposits; a gradual, multi-millennial transition would weaken the hypothesis.
2. **Geochemical signatures in dark-earth horizons** — marine biomarkers (Cl/Br ratios), thermal-shock anomalies, and cosmochemical tracers correlated with known 6th-century chondritic horizons.
3. **Geochronology of the proposed impact structures** — U-Pb/SHRIMP dating to separate inherited Bohemian-Massif ages from any late-antique shock-event age.
4. **North Sea seismostratigraphy** — the previously unexplained ≈180 m Quaternary subsidence residual reported by Arfai et al. (2018) should show abrupt compressional or turbiditic signatures at the relevant horizon.
5. **Moran's I spatial autocorrelation** (the T39 test already part of the core statistical suite) — if the cartometric residuals were spatially random, the claimed block-displacement structure would not be statistically recoverable at all.
This is the kind of step that does not change a single existing number in the model but materially changes how falsifiable the broader palaeoceanographic argument is. Readers who want the full narrative case for the Long-Transgression hypothesis — the Blinkerwall, the Tollense Valley conflict, Jordanes' *vagina nationum*, the Vineta tradition — should read the companion review article; this update simply reports that the hypothesis now carries an explicit, citable falsification programme alongside it.
---
## What This Changes, and What It Doesn't
None of the four additions changes a single headline number from the earlier articles on this site. The Elster-Cluster mean translation (−88.9 km), the $t \approx -15.0$ ($d = -6.14$) significance, the Sudete rotation (+35°), and the 120 km coastline-shift estimate are unchanged. What Version 9.0 changes is the density of independent angles from which those numbers have been tested:
- An **external control region** (Gallia) shows the EC signal is not a generic Ptolemaic map artefact.
- A **sequential Bayesian re-analysis** shows the cluster membership was not selected to inflate the result.
- A **formal power calculation** shows the small sample size is not, in fact, a statistical weakness given the size of the effect.
- A **new monograph part** gives the palaeoceanographic backdrop an explicit falsification programme rather than leaving it as a plausibility argument.
Taken together with the identification-multiverse analysis from v8.2 (significant at $\alpha=0.05$ across 100 % of a $5.6\times10^{10}$-combination space of alternative place identifications) and the formal blind test from v7.2 (28–49 % RMSE improvement on held-out points), the cumulative picture is one of a result that has been attacked from several independent statistical directions — discrete identification uncertainty, selection bias, sample-size adequacy, and regional specificity — without breaking.
---
## References (APA)
Arfai, J., Franke, D., Lutz, R., Reinhardt, L., Kley, J., & Gaedicke, C. (2018). Rapid Quaternary subsidence in the northwestern German North Sea. *Scientific Reports, 8*, 11524. [https://doi.org/10.1038/s41598-018-29638-6](https://doi.org/10.1038/s41598-018-29638-6)
Geersen, J., Bradtmöller, M., Schneider von Deimling, J., et al. (2024). A submerged Stone Age hunting architecture from the Western Baltic Sea. *Proceedings of the National Academy of Sciences of the United States of America, 121*, e2312008121. [https://doi.org/10.1073/pnas.2312008121](https://doi.org/10.1073/pnas.2312008121)
Kass, R. E., & Raftery, A. E. (1995). Bayes factors. *Journal of the American Statistical Association, 90*(430), 773–795. [https://doi.org/10.1080/01621459.1995.10476572](https://doi.org/10.1080/01621459.1995.10476572)
Lampe, R., Lorenz, S., Janke, W., & Arz, H. W. (2022). Rapid sea-level rise during the first phase of the Littorina transgression in the western Baltic Sea. *Baltic Earth Reports* (ISSN 2701-7184), Report No. 15. [https://www.baltic.earth/publications/publication/110072](https://www.baltic.earth/publications/publication/110072)
Mildner, S. (2026). *Geodynamic Reinterpretation Model for Ptolemy's Germania Magna: General Model Description, Cartometric Foundations* (Version 9.0). EarthArXiv. [https://doi.org/10.31223/X5KB51](https://doi.org/10.31223/X5KB51)
Rouder, J. N., Speckman, P. L., Sun, D., Morey, R. D., & Iverson, G. (2009). Bayesian t tests for accepting and rejecting the null hypothesis. *Psychonomic Bulletin & Review, 16*(2), 225–237. [https://doi.org/10.3758/PBR.16.2.225](https://doi.org/10.3758/PBR.16.2.225)
Stückelberger, A., & Graßhoff, G. (Eds.). (2006). *Klaudios Ptolemaios: Handbuch der Geographie* (2 vols.). Schwabe Verlag, Basel.
---
## About
AncientMaps - Geography is a research-oriented platform dedicated to the study of historical cartography, ancient geography, and the reconstruction of past landscapes. The site focuses on the critical analysis of ancient and Renaissance maps, with particular emphasis on Claudius Ptolemy's Geographike Hyphegesis and the coordinate catalogue of Germania Magna.
Central to the project is the kinematic block-deformation model developed by Sven Mildner, which integrates GIS-based affine transformations, residual analysis, and physically motivated kinematic corrections to reduce spatial discrepancies between Ptolemaic coordinates and modern geography. The platform also examines the cartographic work of Gerhard Mercator and other key figures in the history of mapmaking.
Articles combine traditional historical geography with modern methods including digital cartometry, geodetic rectification, and insights from structural geology and geodynamics. Topics include the re-identification of ancient toponyms and river courses, post-Ptolemaic landscape changes, and the methodological challenges of interpreting historical geographic data.
The site aims to provide transparent, evidence-based analyses supported by statistical validation, falsification criteria, and interdisciplinary perspectives.
