The only large celestial globe evidently surviving from antiquity is that in the Naples Archaeological Museum, held upon the shoulders of the statue of Atlas, known as the Farnese Atlas. The globe is marble, 65 cm in diameter, covered with constellations depicted in relief (no stars). Whose astronomy inspired the artist's celestial vision is a long-debated question. A scrupulous scholarly history, covering centuries of researches into the globe, was provided in Globusfreund (1987) by Vladimiro Valerio, noting that N.Halma & astronomer J.Lalande were among the earliest experts to take an interest in the Farnese globe about 200y ago, and that Matteo Fiorini provided an able mathematician's perspective on it over a century ago.
In 1898, the eminent scholar Georg Thiele argued in detail (though undogmatically) that the sculptor followed the astronomy of Hipparchos (fl. 130 BC). (See translation of Thiele by H.Thurston.)
Valerio (op cit p.105) provided careful measurements of the key features of the Farnese globe, and suggested the possibility of a Ptolemaic origin.
On 2005/1/11, with promotion by the American Astronomical Society (AAS)
by the Journal for the History of Astronomy,
Bradley Schaefer (JHA Ed.Board) — ignoring the globe's
contra-Hipparchan features —
Thiele's Hipparchan thesis and created a “new” discovery
out of it for a world press-story,
simple on-colure stellar evidence remarkably similar to Thiele's.
[Twice dishonored? Several Nazi destroyers were sunk by Brits in a fjord near Narvik in the Spring of 1940. The rusting hull of one of them is still visible there: the “Georg Thiele”.]
Dennis Rawlins has posted criticisms of Schaefer's shaky conclusion and
noting that Schaefer had unfortunately attempted
to go beyond Thiele's properly cautious discussion
— announcing via AAS the sort of unattainable
(pop-sci newsbite-)certainty that the “science press” prefers:
 With hubristic confidence, Schaefer wades into demonstrably unreliable (sometimes dynamically) entertaining photogrammetry on many more-scattered, valuelessly-mushy “star”-positions.
 He effectively merges two statistically-inconsistent star-samples whose precession-computed dates are 300 yrs apart (each over a century before and after Hipparchos) to — essentially by averaging the two dates — arrive at his Hipparchan-epoch conclusion.
 Schaefer seriously mis-computes (sign-error) the effect of atmospheric extinction on an observer's ever-visible arctic circle, when tampering with his own potentially valuable finding that the Farnese globe's polar circles were arranged for 38°-39°N latitude, thereby (quite fallaciously) nudging the globe's latitude to come out nearer Hipparchos'.
 Still nudging, Schaefer mistakenly renders the last datum as 37° by confusing Hipparchos' 37° latitude for Athens (which is actually at 38°N) with that for Rhodes (36°N in Hipparchos & reality).
 Despite a mess of central problems with his careless paper, Schaefer has publicly retracted nothing, even continuing to publicly insist upon his exact erroneous Farnese-globe date, 125BC, which (undone by yet another Schaefer sign-error) he has mis-computed by over a decade — even if one entirely accepts his own method on his own “star”-selections.
 It is demonstrated that Schaefer is astonishingly ignorant of the numbers which form the majority of the star-data in Hipparchos' only surviving work (data familiar to every specialist in the field), which firmly delimit the bounds of Hipparchos' constellations — bounds which are frequently in gross conflict with those displayed upon the Farnese globe.
Provided further are evidences of JHA slapdash-refereeing (which led to brushing aside JHA Assoc Editor James Evans' prescient 2004 pre-publication doubts about the paper).
DR also provides his own (explicitly tentative-speculative) proposal
for a possible source of the Farnese globe's astronomy,
Krates of Mallos:
[A] The points on the globe corresponding to the most potentially useful epoch-sensitive stars (the on-colure ones) suggest the early 2nd century, the era of Krates.
[B] He was famous as a globist.
[C] Krates (evidencing a pioneer's non-geocentric universal-perspective?) emphasized the import of the Galactic Circle — a fragment of which may survive on the Farnese globe, a hitherto-mysterious line-segment at the globe's Cygnus, which DR revealed (2005/4/1) exhibits both the direction and the location of the Milky Way.
[D] Krates' Pergamon 39°N latitude was consistent with the globe's polar circles' indicated latitude.
A great deal of further expert investigation of the Farnese controversy (as well as helpful accessory material, including the 1739 M.Folkes projection of the Farnese globe) has been contributed by physicist Dennis Duke at www.csit.fsu.edu/~dduke/farnese4.pdf. [The Duke paper has been published in the 2006 Feb issue of the Journal for the History of Astronomy 37:87-100.]
[On 2005/7/27&29, Dennis Rawlins visited the Naples Archaeological Museum
to view the Farnese globe 1st-hand.
The Rawlins photographs resulting from these visits
have been transmitted to Duke (since Schaefer had earlier
to acknowledge Duke's request to use Schaefer's photos), for assistance
in checking the accuracy of the Folkes projection, which Duke is using
for his polite but devastating 2006 Feb JHA
on the Farnese-globe's numerous non-Hipparchan aspects — a paper whose
has (on the evidence of a lordly, insulting, accusatory 2005/9/18 BS email)
gone astonishingly unappreciated by Schaefer.
[We are here glad to compliment the DR-hating JHA for ultimately doing the right thing in publishing Duke's quietus to scholarly acceptance of Farnese-BS. Let's hope that this creditable event can act as a spark in initiating a long-needed apolitical truth-priority trend in the astronomical-history field.
(Note added 2010. As witness events since this pacific note was posted (e.g., HAD's DIO-shunning 2011/1/9 Neptune session), as usual nothing essential has changed.)]
Note: neither DR nor Schaefer were able to photograph the globe from above. This limitation was one contributing factor in the Schaefer photogrammetric approach's accuracy-problems. Duke was cleverly able to overcome this difficulty by using one of Thiele's 1898 photos.]