# The Hanging Gardens: Babylon or Nineveh?

**Report:** https://historical-mysteries.netlify.app/hanging-gardens/report
**Case page:** https://historical-mysteries.netlify.app/hanging-gardens/

## For the site

### The mystery

The story, as Greek writers told it, is that a king of Babylon, the great city on the Euphrates
in what is now southern Iraq, built an artificial mountain beside his palace for a wife who came
from the hills of Media, in north-west Iran, and pined for them: stone terraces on brick vaults,
planted with full-grown trees, watered from the river by machinery hidden inside the structure.
By the first century BC Greek writers counted it among the seven wonders of the world. Five
ancient descriptions survive, none by an author who writes as if he had seen it. The earliest, by
Berossus, a priest of Babylon writing in Greek about 290 BC, names Nebuchadnezzar II, king of
Babylon from 605 to 562 BC, as the builder; two others say only "a Syrian king", meaning an
Assyrian one; two name no builder. The garden has never been found. Robert Koldewey, the German
archaeologist who excavated Babylon from 1899 to 1917, thought he had found it in a block of
vaulted rooms that turned out to be a storage basement. No cuneiform text from Babylon mentions
a garden, and Herodotus, the Greek historian who described the city about 430 BC, says nothing
of it.

### Why it is a mystery, and the sides

Three answers have been given. The traditional one keeps the garden at Babylon and gives it to
Nebuchadnezzar, whose own inscriptions describe a new palace, the North Palace, built on terraces
"as high as a mountain" beside his father's; on this view a planted terrace of that palace was
exaggerated by later writers. Since 1993 Stephanie Dalley, an Assyriologist at the University of
Oxford, has argued that the Greeks had the wrong city: the garden was built by Sennacherib, king
of Assyria from 705 to 681 BC, beside his palace at Nineveh, three hundred miles north on the
Tigris and a century earlier. His inscriptions describe a palace garden and a bronze
water-raising device that she reads as a water screw, four centuries before Archimedes; his
canals and aqueducts survive in part; two carved panels from Nineveh show a garden; and the
Greeks, who did confuse Assyria with Babylonia, moved the garden south. Some specialists accept
this. A third answer holds that no single garden lies behind the story, which combined a
Mesopotamian royal park, a Babylonian palace terrace and Greek engineering fantasy.

### How the case was judged

The kings' own inscriptions (Sennacherib's prisms in RINAP 3/1, Nebuchadnezzar's palace texts in
RINBE 1/1 of 2024), the five Greek and Latin descriptions in the original languages, and the
excavation record at both cities were read in the standard editions, with the arguments on both
sides as published (Dalley's 1994 article and 2013 book only in part). The evidence was sorted
into seven clusters and each was given a probability under each answer; with equal priors the
arithmetic gives Babylon 57%, no single structure 38%, Nineveh 5%. Because three clusters turn on
readings that specialists dispute, the headline is Babylon 55%, no single structure 35%, Nineveh
10%. These are judgements expressed as numbers, conditional on the framing and the weights, and
no Assyriologist has reviewed them.

### The argument (verdict and confidence)

The garden was, more likely than not, a real raised garden at **Babylon**: the terraced North
Palace that Nebuchadnezzar II says he built "as high as a mountain", exaggerated by later writers.
Every ancient witness who names a city names Babylon. Berossus places the garden in that palace, in a paragraph that follows Nebuchadnezzar's own
inscription item by item. The one cuneiform text that describes Sennacherib's garden speaks of
water drawn by bucket and says nothing of terraces. Sennacherib's garden at Nineveh is real and
well documented, and it is the closest ancestor of what the Greeks describe, but the evidence does
not support Dalley's claim that it is the garden they described.

| Hypothesis | Headline | Weighing (flat priors) |
|---|---|---|
| Babylon: a real raised palace garden behind the tradition, most plausibly Nebuchadnezzar's North Palace (a Persian-period garden on the same spot would also count) | **55%** | 57% |
| No single structure: a literary construct with no one garden behind it | **35%** | 38% |
| Nineveh: Sennacherib's garden, relocated to Babylon by confusion in the Greek tradition (Dalley) | **10%** | 5% |

The headline lifts Nineveh from 5% to 10% because the inputs are judgements and three of the
seven evidence clusters turn on readings that specialists dispute. All three figures depend on
how the hypotheses are framed and on the weights assigned; they are judgements about evidence, and the sensitivity runs in section 5 of the report show how far they move when the weights
do. Starting from a prior in Nineveh's favour still ends at Babylon 56%, no single structure 38%,
Nineveh 7%; granting Dalley her three contested readings at once puts Nineveh ahead, 55%
against Babylon's 27%.

### How far the evidence supports it

**Label: probable.** The evidence clearly favours Babylon over the other two answers but does not
prove it, and the percentage moves when the same evidence is weighed differently. The case for
Babylon rests on the unanimity of the witnesses
that name a city, on Berossus, and on what Sennacherib's own text does and does not say. Each of
those is a reading that a specialist could contest, and the report says so at each point. It is a
reading of the sources by non-specialists, and no Assyriologist has reviewed it.

### What this report adds

None of it is new evidence; all of it is new analysis, or new arrangement, of evidence already in
print. An explicit weighing that shows where the weight lies (location and Berossus, not the
screw) and what Dalley's case needs jointly. A check of Nebuchadnezzar's silence across the whole
2024 edition of his inscriptions, set beside his own texts on the North Palace terrace and the
Chicago Assyrian Dictionary's 1956 note on them. A reading across the Assyrian corpus showing that
"a replica of Mount Amanus" and "a wonder for all peoples" are stock phrases used by three kings,
and that the second is said of the palace, not the garden (on the reading of siḫirti ekalli that
the dictionary itself adopts, CAD S 236). A demonstration, from Sennacherib's own texts and from
Viggiano's 2025 study of the bricks of Nineveh, that the baked brick and bitumen of Strabo and
Diodorus do not tell Babylon from Nineveh. The bucket-tackle argument against the screw is Bagg's
(2014), not the report's.

### How it stands to what historians hold

**Agrees with a substantial existing position.** The verdict agrees with Bichler and Rollinger
(2005), Bagg (2014, who rejects the Nineveh thesis and treats the classical location at Babylon as
reliable), Potts (2013) and Pedersén (2021), who writes that "most scholars have accepted
the identity of Nebuchadnezzar's North Palace and the palace described by Berossos" and that "a
few have even opted for the alternative that the gardens were not in Babylon but elsewhere"
(2021, 125). That is his assessment; the report has not counted the field. Nineveh has had
published support besides Dalley (Renger 1998 and Veenhof 2001, as reported by Bichler and
Rollinger; Meijer 2017, hedged; Foster 2004, known by its title and opening only). The editors of Sennacherib's
inscriptions take no position and file the Nineveh identification as an "opinion" with "critical
reviews" (RINAP 3/1, n. 44). Dalley's thesis is the view the report argues against.

### Where it falls short

- Dalley's 1994 article and 2013 book could not be read in full; what was read is stated in sections 6
  and 8 of the report, and a reader with library access should check section 6 against pp. 29–41, 61–82
  and 107–126 of the book.
- Reade 2000, Bagg 2000, Foster 2004, Streck 2018 and Wiseman 1985 are known through Bichler and
  Rollinger, Potts, Pedersén and the RINAP notes. Foster 2004 is known by its title and opening only.
- The Bavian and Jerwan inscriptions were read in Luckenbill's and Jacobsen's translations, not in
  RINAP 3/2. The Philo text was read in a poor scan.
- The weighing's inputs are judgements. Sensitivity runs (priors, leave-one-out, two scenarios that
  favour Dalley) are printed in the report and reproducible from the script.
- No Assyriologist has reviewed the report.

### What would overturn the verdict

For Nineveh: a Neo-Assyrian text calling Nineveh "Babylon"; a Greek text placing a wonder-garden at
Nineveh; a witness to Berossus without the garden sentence; an Assyrian water screw found, depicted
or lexically secured; a Nineveh text or relief showing the garden on vaulted tiers. For Babylon to
become near-certain: excavation of the North Palace's northern slopes and riverside producing tree
pits, garden soil or a planted terrace with water channels, or a cuneiform text naming a garden on
the "new palace".

### Figures and tables made for this report

Both figures are drawn on the page as inline SVG and are in this folder as files. The captions
below are the page's own, word for word. The case page may carry any two to four of these.

- `figure1-probabilities.svg` and `figure1-probabilities.png` (verdict panel). Caption: "Figure 1. How the probability divides between the three hypotheses. Made for this report from the weighing in section 5."
- `figure2-transmission.svg` and `figure2-transmission.png` (section 4). Caption: "Figure 2. How the descriptions of the garden were passed down. Two lines are independent of each other, the Babylonian priestly tradition (Berossus, Abydenus) and the Alexander-historian tradition (Clitarchus; Strabo's unnamed source), and both put the garden in Babylon. Their boxes are outlined in blue; a solid arrow means the later writer used the earlier one, and the dotted arrow marks a likely but unproven line. Alexander Polyhistor, a Greek compiler of the first century BC, and Megasthenes, a Greek envoy to India about 300 BC, are the intermediaries through whom Berossus' words reached the later writers. Made for this report from the texts quoted in this section, Diodorus 2.7.3 on his own sources, Eusebius, Praeparatio Evangelica 9.40–41, and Bichler and Rollinger 2005."
- `table1-evidence-clusters.csv` (section 5, the table's values as a file). Caption: "Table 1. The seven evidence clusters and the probability assigned to each under the three hypotheses. Made for this report; every value is a judgement stated so that it can be changed and rerun. P(E | H) is the probability of seeing the evidence if hypothesis H is true; the last column divides the Nineveh figure by the Babylon figure."
- No image is reproduced from elsewhere. The reliefs (BM 124939; Layard, Original Drawing IV 77)
  are described, not shown.

## Files in this folder

- `report.html` — copy of the published report (same content as the link above).
- `A_nineveh_assyrian.md` — Assyrian sources: Sennacherib's garden texts, Jerwan and Bavian water
  works, the Amanus formula, the reliefs, the screw question. ~6,700 words.
- `B_babylon.md` — Babylonian sources: Nebuchadnezzar's inscriptions (RINBE 1/1), Koldewey's
  excavations, Pedersén 2021, Berossus, the North Palace terrace. ~7,400 words.
- `C_classical.md` — every Greek and Latin testimonium quoted in the original with edition and
  translation; the silences; the Berossus interpolation debate; the wonder-lists.
- `D_scholarship.md` — the published debate: Dalley 1993/1994/2003/2008/2013/2017; Reade 2000;
  Bagg 2000; Foster 2004; Bichler and Rollinger 2005; van der Spek 2008; Potts 2013; Pedersén 2021;
  Streck 2018. ~10,200 words.
- `E_viggiano.md` — Viggiano 2025 read in full, with page-referenced quotations on baked brick and
  bitumen at Sennacherib's Nineveh. ~3,600 words.
- `E_rinap3_materials.md` — all 150 attestations of agurru (baked brick), kupru/ittû (bitumen)
  and pīlu (limestone) in RINAP 3/1–2 with the editors' translations. ~13,000 words.
- `F_dalley_access_log.md` — every route tried for Dalley 1994 and 2013, with the result of each.
- `F_dalley_1994.txt` — what could be read of Dalley 1994: the extract, all 75 footnotes, the 45
  page-referenced quotations in Bichler and Rollinger, and a page-by-page reconstruction.
- `F_dalley_2013.txt` — what could be read of Dalley 2013: table of contents, Bagg's précis, about
  230 page-anchored snippets and the reconstructed passages, including the Conclusion.
- `F_dalley_other.md` — Dalley 1993, 2003 (with Oleson), 2008 and 2017 in full; Bagg 2014 and
  Potts 2013 in full; Reade 2000 extract; popular restatements of the thesis.
- `likelihood_model.py` — the seven-cluster weighing with sensitivity runs (priors, leave-one-out,
  a scenario favourable to Dalley, a scenario that puts Nineveh ahead). Python 3, no dependencies:
  `python3 likelihood_model.py` writes `likelihood_results.json`.
- `likelihood_results.json` — the weighing's output as used in the report.
- `figure1-probabilities.svg`, `figure1-probabilities.png` — Figure 1, the probability bar in the
  verdict panel (55 / 35 / 10).
- `figure2-transmission.svg`, `figure2-transmission.png` — Figure 2, the diagram of how the ancient
  descriptions were passed down (section 4).
- `table1-evidence-clusters.csv` — Table 1, the seven evidence clusters with the probability
  assigned to each under the three hypotheses and the posterior row.
- `make_figures.py` — draws Figure 1 and Figure 2 (Python 3, no dependencies): `python3 make_figures.py`
  writes the two SVGs; the PNGs are renders of them at twice the drawn size.

## Method

The dossiers A to D (Assyrian, Babylonian, classical, scholarship), E (Viggiano and the RINAP 3
materials check) and F (the Dalley readings) were compiled separately; every load-bearing primary
passage was then re-verified against the editions (RINAP 3/1 via archived ORACC pages, RINBE 1/1
open access, CAD PDFs, Perseus, Wikisource and LacusCurtius for the Greek and Latin). The live
ORACC site failed certificate checks; archived copies were used.