"Ancient Greek" is not one dialect. It is a research umbrella covering Mycenaean administrative documents, local alphabetic inscriptions, texts in the Cypriot syllabary, literary traditions and their manuscript transmission, contact between dialects and local written standards, and finally the convergence of the Hellenistic and Roman period together with the learned responses to it.
Who groups what, and how
Horrocks places Attic and Ionic and Arcado-Cypriot in an East grouping and Peloponnesian Doric with Northwest Greek in a West grouping. He discusses Aeolic through historical contact and mixed inheritance B03 §1.3 pp.14–18. His tree is expressly illustrative, and he notes alternative genetic and areal models and a limited factual base. The Atlas keeps the grouping as Horrocks's proposal, not as its own authority.
The publisher's summary of Woodard's chapter describes a first millennium classification into Attic and Ionic, Arcado-Cypriot, Aeolic, Doric and Northwest Greek B21 Publisher Summary. It is useful as an independent statement of the descriptive grouping. The full chapter has not been examined.
Porzig, Risch, García-Ramón and Garrett, who work on prehistoric subgrouping and the dating of isoglosses, are for now bibliographic leads. Their originals have not been read, and the Atlas neither attributes a detailed tree to them nor places reconstructed speakers on migration maps on the strength of those leads.
A descriptive grouping of the first millennium and a hypothesis about ancestry in the second millennium are different claims. That a dialect is attested in a city proves neither that all its inhabitants spoke it nor the history of its settlement. Pamphylian and ancient Macedonian need explicitly uncertain treatment and independent specialist evidence. No genealogy of present day Macedonian Greek is inferred from the ancient geographic name.
Types of evidence and what each cannot prove
| Evidence | What must be recorded | What it does not establish alone |
|---|---|---|
| Linear B tablet | Tablet number, face, line and signs, archive, findspot, edition, archaeological date, reading uncertainty | Full phonemic form, all contemporary spoken varieties, an alphabetic spelling |
| Alphabetic inscription | Corpus number, edition, lines, findspot separate from proposed origin, dating authority, restorations | Unmediated speech or a homogeneous local population |
| Cypriot syllabic text | Syllabic reading, editorial transcription, edition, date | Identity with modern Cypriot Greek |
| Literary text | Author, work, passage, edition and manuscript, genre, dialect mixture, transmission | A birthplace dialect, a single speaker community, an audio attestation |
| Documentary papyrus | Catalogue number, provenance separate from findspot, hand and date, editorial corrections, clause context | General frequency across the Hellenistic world |
| Scholarly reconstruction | Author, work, locator, reconstruction convention, alternative analyses | A directly attested form |
An edition's brackets, uncertainty marks and restorations must survive on screen and in export. A printed and a digital edition of the same text are linked by dependency. They do not count as two witnesses B03 §§1.2–1.4.
One evidence chain as an example
Horrocks discusses the Damonon dedication, IG V.1.123, of the fifth century BCE B03 §1.4.2(a) p.28. The Atlas has examined the scholarly discussion, not the stone, the squeezes or an independent IG edition. The example shows how every claim will be presented: claim, discussion, inscription edition, physical witness. No linguistic form has entered the Atlas as a new verified attestation.
Epic language, lyric conventions and Attic dramatic representations need records of genre and register. There is no branch called "spoken Homeric dialect". Written practice is valuable evidence, but the step to the grammar of a speaking community has to be argued.
The transition to Koine
Research looks at local persistence and convergence together, at the genre and date of each document, and compares only like grammatical environments. There is no single switch in 323 BCE for every place. Political dates are retrieval anchors, not simultaneous linguistic changes. A translation inside a Koine corpus keeps its translation status and the language of its original.
The Centre for the Greek Language's Θετίμα already offers ancient dialect studies, inscriptions and proper names B08 Εισαγωγή. Another general ancient Greek portal would duplicate existing work. The Atlas contributes a small number of evidence linked comparison explanations, connected to the original resources, with attestation and analysis clearly separated.
What needs checking
At least two competing original works on subgrouping, Mycenaean reading conventions and chronology, the status of Aeolic and Pamphylian, the treatment of fragmentary ancient Macedonian evidence, literary register terminology, and the scholarship on Koine and documentary papyri. Until then the pillar is methodologically usable but is not a final reference taxonomy.