Visit the blog
announces a new imprint

Search Articles

Search Authors

Advanced Search

Join our Mailing List
Friday, March 01, 2002

Arts & Letters

Cavafy’s Cavafy versus Savidis’s Cavafy: The Need to De-edit the “Acknowledged” Poems

Cavafy was his own best editor. Few poets can ever have been so concerned with the fine points of orthography, punctuation, and layout, or exercised such close control over the organization and printing of their work. Cavafy left very little for subsequent editors to do. If only they had been content to do very little instead of undoing the poet’s meticulous work!

In the course of my research into Cavafy’s own printings of his poems, I have been dismayed to discover that the late George Savidis was a far more interventionist editor than he ever admitted. Savidis’s 1963 edition of the “acknowledged” poems, entitled Poiêmata [Poems], remains the best edition we have of the Cavafy “canon,” if by “best” is meant closest to the author’s latest discernible intentions; but it still does not give us the poems exactly as Cavafy printed them. In the preface to the 1963 edition, Savidis does acknowledge imposing the standard distinction between pou with a grave accent (meaning “that” or “who/whom/which”) and pou with a circumflex (meaning “where”) – Cavafy tended to use the latter indiscriminately – and between other comparable pairs where Cavafy’s practice was inconsistent (Cavafy, Poiêmata [Poems], 1963, I, 12). Other, more significant, changes are not acknowledged, however. Although Savidis states (ibid.) that he used Cavafy’s latest printings of the poems as the basis for his edition, he also says that he consulted earlier printings and, in the case of the pre-1911 poems, other sources. Consequently, it is not entirely clear what Savidis’s edited versions of the poems actually represent.

The 1963 edition is, in any case, no longer available, having been superseded by Savidis’s “new edition” of 1991, which departs more radically from the orthography of Cavafy’s own printed texts, but without any explicit acknowledgement. This “new edition” has the title, Ta poiêmata [The Poems], and even the addition of the definite article is a departure from Cavafy’s practice. Cavafy’s own collections of his poems were all simply entitled Poiêmata, with the addition of a date or dates.

Cavafy’s quirks of punctuation and accentuation, and his occasional surprising spellings, require no editorial “correction”; his inconsistencies do not need to be standardized. We can be confident that they were in most cases deliberate and well considered – and sometimes they are poetically significant. It is regrettable that Savidis’s 1991 edition removes iota subscript from most of the many datives and subjunctives in which Cavafy employed it. In his 1963 edition, Savidis informs us (I, 12) that he actually “added a few subscripts” where Cavafy had omitted them! Equally regrettable is that Cavafy’s frequent contractions of particles or definite articles with verbs – such as thatan (= tha êtan, “would be”), narthoun (= na erthoun, “that they might come”) tathelen (= ta êthelen, “wanted them”) – have been systematically replaced by the unsightly and irrational (more modern and supposedly more demotic) forms, tha ’tan, na ’rthoun, and ta ’thelen.

Still more regrettable, however, are Savidis’s “corrections,” in both the 1963 and 1991 editions, of Cavafy’s creative misspellings. Take, for example, the first two lines and the penultimate line of Epestrefe (“Return”). As printed (many times) by Cavafy, they read:

Epestrefe suchna kai perne me,
agapêmenê aisthêsis epestrefe kai perne me –
. . . . . . . . . . . . .
Epestrefe suchna kai perne me tên nuchta

Return often and take me,
beloved sensation return and take me –
. . . . . . . . . . .
Return often and take me in the night

In both editions, Savidis “corrects” perne to pairne (there is no difference in pronunciation). Fortunately, he did not also “correct” epestrefe to epistrefe, but epestrefe was an established variant associated especially with Constantinopolitan speech, whereas perne is an unprecedented spelling of the imperfective imperative of pairnô (stress on the first syllable). Cavafy must have known perfectly well how the word is spelled, and he misspelled it deliberately – in part, I think, to draw attention by visual means to the short “e” sound that dominates these lines, but, more important, to create a semantic ambiguity. Perne is a hybrid of pairne and perna, the corresponding imperative form of pernô (stress on the last syllable), meaning “pass” (intransitive) or “pass by” or “pass through,” etc. The use of this hybrid suggests that the speaker is calling on the erotic sensation not only to seize or take possession of him, but also to pass through him, thus forging a link with line 4, which speaks of the moments when “old desire passes again through the blood” (epithumia palêa xanaperna sto aima). In this case, Savidis’s corrections are not merely unnecessary, they actually diminish the poem.

I am not yet sure how extensive Cavafy’s practice of creative misspelling is, but let me give one further example. Addressing, in his imagination, Caesarion, the doomed son of Cleopatra VII, in the poem simply called “Caesarion,” Cavafy writes, “My art gives to your face a dreamlike sympathetic beauty.” In Savidis’s editions, the Greek work for “gives” is spelled dinei in the normal way; in Cavafy’s printings, however, it is deinei. Again, this spelling is, as far as I know, unprecedented (unlike Cavafy’s taxeidi for taxidi, “journey”); and, again, it may been seen as a hybrid, evoking both deichnei (“shows”) and dinei (“gives”), suggesting that the “beauty” that the poet’s art “gives to your face” (sto prosôpo sou deinei) is something that that face in any case possessed, so that we might translate the hidden, but hinted at, subtext as “shows in your face” (sto prosôpo sou dei[ch]nei). Again, something has been lost in the editing.

Now that the best available edition is Savidis’s “new” edition of 1991, which is new in a number of regrettable ways, there is a compelling need for an edition that gives us Cavafy’s Cavafy, the poems of the canon as he actually printed them. I will return later to the difficulties of identifying Cavafy’s final printings of each poem, which would, in most cases, provide the texts for such an edition. Now, however, I want to discuss another problem facing the would-be editor of Cavafy: the arrangement of the 154 poems of the canon.

As is well known, no commercial edition of Cavafy’s poetry was published in his lifetime. The first such edition appeared in 1935, edited by Rika Sengopoulou, printed in Athens and published in Alexandria. In the last three years of his life (more precisely from June1930 until April 1933), Cavafy had three privately printed collections of his poems in circulation. Two of these, Poems (1905-1915) and Poems (1916-1918), were bound booklets, in which the poems were arranged thematically (for an analysis of Cavafy’s thematic sequences, see my “Philosophical, Historical and Sensual: Cavafy’s Thematic Arrangements,” Byzantine and Modern Greek Studies). The dates refer to the years of first publication of the poems that they contained. In the third collection, Poems 1919–1930 (or –1931 or –1932, as the second date was changed each year when the first new poem of that year was added), the poems were arranged chronologically in order of first publication. (After 1918, first publication invariably meant first printing at the poet’s expense.) This chronological collection was unbound. The poems (not the pages) were numbered, and the leaves were crudely pierced through the top left corner and held together by a brass split-pin. This bundle of leaves was presented in inner and outer wrappers, of slightly different dimensions but both larger than the sheets on which the poems were printed. The first page of each wrapper bore the title details (usually printed, except for the last two digits of the second date), while the list of contents occupied the third and fourth pages of the inner wrapper.

Poems (1905-1915) and Poems (1916-1918) contained 40 and 28 poems, respectively; the chronological collection in its latest form, Poems 1919-1932, contained 69 poems. The total, 137, leaves us 17 short of the 154 poems of the Cavafy canon. The last poem of the canon, “On the Outskirts of Antioch,” was not published until after Cavafy’s death, although he revised it for printing during the last weeks of his life – and to this it owes its place in the canon. The 16 pre-1905 poems present a real challenge to an editor, and none of the solutions so far essayed is satisfactory. To understand the difficulty, we must go back to the early stages of Cavafy’s self-publication.

Between 1891 and 1904, Cavafy had occasionally experimented with the printing of his own poems as broadsheets or pamphlets (four single poems and one pair of poems), but late in 1904 he commissioned a small, perfect-bound booklet in an edition of 100 copies. It was called Poems 1904 and contained 14 poems, 13 selected from the 40-odd poems he had published to date, and one, “Desires,” appearing for the first time. In 1910, he produced a second, enlarged edition, with the title Poems 1910, in which he added a further seven poems first published, or republished in revised form, in the years 1905-1909 to the original 14 poems. (At this stage in Cavafy’s career, first publication meant in most cases publication in a newspaper, literary journal, or almanac.) In both of these editions, the poems were arranged thematically. Although Poems 1910 was never republished, and its distribution ceased in 1915, there is an important late manuscript version of it, known as The Sengopoulos Notebook, which dates from around 1927. It was in this version that Cavafy inserted “Walls,” a poem of 1897, into the thematic series. Although the poem had never been included in any printed collection, it had been republished a number of times in periodicals – most recently in the December 1926 issue of Alexandrine Techne – and was always regarded as part of the canon. “Walls” brings the total of the pre-1905 canonical poems to 15. The sixteenth is “The funeral of Sarpedon.” Although first published in 1898, this poem was not included in Poems 1904. A revised version was published in 1908 and included in Poems 1910, but it seems that Cavafy regarded it as a poem of 1898 (see below).

In 1912, Cavafy began to circulate his post-1909 poems, alongside Poems 1910, as loose sheets arranged in chronological order. At first, these chronological collections lacked wrappers and the poems were unnumbered, but they soon evolved to resemble the last chronological collection described above. In 1918, Cavafy withdrew the earliest poems from this collection and combined them with the later poems of Poems 1910 to produce a new thematic collection, Poems (1909-1911), which over the years, through the addition of and earlier later poems, became Poems (1908–1914), then Poems (1907-1915), and eventually the collection already discussed, Poems (1905–1915). One consequence of this process is that there are six poems (poems of 1905-1909) which are found in both Poems 1910 and Poems (1905-1915). These do not include “The Funeral of Sarpedon,” which Cavafy chose not to incorporate into Poems (1908-1914) or its successors, suggesting that it remained for a him a poem of 1898. In terms of subject matter and treatment, it would certainly have been out of place among the later poems. Besides, Cavafy seems to have remained dissatisfied with it, until he revised it again in 1924.

In view of the interpretive significance of Cavafy’s thematic arrangements – and there can be little doubt that Cavafy intended the poems to complement and comment on each other through their juxtapositions – this overlap between these two thematic collections presents the editor of Cavafy’s poems with a serious problem. In the first commercial edition of Cavafy’s poems published in 1935 – which defined the contents of the canon – Rika Sengopoulou simply avoided the problem by arranging all the poems chronologically (but inaccurately) by date of first publication. It is the great merit of Savidis’s 1963 edition that it restored Cavafy’s thematic sequences (though only imperfectly in the case of Poems 1910) and corrected the chronological sequence of the post-1918 poems.

The first volume of the 1963 edition begins with Poems (1905-1915). This is followed by Poems (1916-1918), while the earliest canonical poems, headed “1896-1904,” come at the end of the volume (“1896,” the supposed date of “Walls,” was a mistake, as Savidis later recognized). This last group of 16 pre-1905 poems consists of Poems 1910, plus “Walls” in the position determined by The Sengopoulos Notebook, but minus the six poems included in Poems (1905-1915). The result is an unwarranted hybrid of Poems 1904, the original Poems 1910 and the later manuscript version – in other words, a sequence that does not correspond exactly to any thematic sequence Cavafy himself envisaged.

In the first volume of the 1991 edition, Savidis tries a different solution, which is, in one sense, preferable. Again, he begins with the 1905-15 and 1916-18 collections. These are followed by “Appendix 1,” which reproduces the sequence of poems in Poems 1904, although it does not reproduce the texts of that edition. The two remaining pre-1905 poems, “Walls” and “The Funeral of Sarpedon,” are relegated to “Appendix 2.” Thus, with the exception of Savidis’s facsimile edition of The Sengopoulos Notebook and my own English translation of Poems 1910, incorrectly retitled “Poems 1897–1910” by the editors of Greek Letters, Poems 1910 has never been republished in its entirety.

The second volume of Savidis’s 1963 edition reproduced the chronological sequence of poems from Poems 1919-1932 with the addition of the posthumous poem, “On the Outskirts of Antioch.” In the 1991 edition, Savidis relegated this last poem to an appendix. Since Cavafy had prepared the poem’s text for the printers shortly before his death, it would have been more in keeping with his general practice, and evident intentions with respect to this poem, to simply add, “On the Outskirts of Antioch” as the seventieth poem in the final chronological collection, and to rename that collection, Poems 1919-1933.

This brings me to my proposal for a new edition of Cavafy (for which S. Ekdawi and I have argued in greater detail elsewhere). The proposal is simply that a future edition of the Cavafy canon consist of the following four collections in this order:

  1. Poems 1910 entire, with the addition of “Walls” (22 poems in all).
  2. Poems (1905-1915), including the six poems already printed in Poems 1910.
  3. Poems (1916-1918).
  4. Poems 1919-1933, including “On the Outskirts of Antioch.”.

It is not easy to decide which texts should be used for Poems 1910, and the answer might depend on whether a simple or a critical edition was planned – and both are needed. In a simple, or popular edition, it might be best to use the last published versions (most of the pre-1905 poems were republished, with Cavafy’s authority, in journals in the mid-1920s), or the manuscript versions of The Sengopoulos Notebook, which contain important late revisions, but also perhaps some minor errors. With the six poems common to Poems 1910 and Poems (1905-1915), there would be an opportunity to include two different versions. In a critical edition, there would be an argument for reproducing the texts as they appeared in Poems 1910, and using the text of the 1897 pamphlet, Teichi – My Walls (which included an English translation by Cavafy’s brother, John), for “Walls,” as the only earlier publication of the poem over which Cavafy had complete control. Although Cavafy authorized most of the republications of his pre-1905 poems, and in many cases probably furnished the texts that were used, publications in periodicals do not have quite the same authority as the printings that Cavafy himself controlled. In a critical edition based on the texts of Poems 1910, the later texts could be reconstructed from the critical apparatus.

For the later three collections, the text of a popular edition and the base text of a critical edition should, in each case, be – I think obviously – that of Cavafy’s last printing of a poem, incorporating, in one or two cases, manuscript amendments in his hand that appear in copies of those final printings. This sounds simple enough, but the identification of the final printings presents the would-be editor of Cavafy with a formidable problem. Copies of all three collections can be found together in a number of libraries – including the Gennadius Library in Athens and the university libraries of Princeton, Harvard, and Oxford. However, one cannot simply pick a set of these collections and use the texts they contain, since the printings of many of the poems are not final printings. This requires some explanation.

The only true editions that Cavafy ever produced were Poems 1904 and Poems 1910 (by “true edition,” I mean one in which all copies were printed and bound at the same time and are, consequently, textually identical). With every other collection that Cavafy produced, each copy is a potentially unique combination of printings; and this applies as much to the bound thematic collections from Poems (1909-1911) onward as it does to the pinned-together bundles of sheets in the chronological collections. Cavafy printed his poems in small runs and reprinted them frequently (every two to three years on average). Sometimes a reprinting of a poem involved textual amendments; sometimes the only difference from the previous printing was in the page or poem number, or in the presence or absence of a signature of a printer’s colophon, or in the date in the colophon. Sometimes there was no difference. Poems intended for the thematic collections were often (but not always) printed on double leaves. Cavafy was economical in the way that he used up his stocks of printed poems. Those intended for a collection that had been superseded had their page or poem numbers changed (usually by hand, occasionally by overprinting) for incorporation into a new collection. Even when a thematic collection is made up of signatures of stitched double-leaves, most copies contain a number of reused single leaves tipped in on stubs.

Savidis published a schematic list of all the printings of the post-1904 poems made for collections other than Poems 1904 or Poems 1910. This list (Oi kabafikes ekdôseis [1891-1932], pp. 301-23) can be used as a basis for identifying printings in actual copies of Cavafy’s privately printed collections. I have constructed a database to do this. Certain bibliographical data from each page of a collection being examined are entered into the database, generating a search string that, when compared with data from Savidis’s list of printings, produces either an exact match or a number of possible matches. In rare cases, it produces no match; and I have, in fact, discovered a small number of printings that were not found by Savidis (or, conceivably, found but incorrectly described by him). I have examined copies of collections in libraries and archives in Egypt, Greece, the UK, and the US, and have so far identified at least one copy of more than 90 percent of the vital final printings, and at least one copy of 65 percent of all printings. (The number of printings per poem ranges from one to 13, and the total number of all printings, excluding Poems 1904 and Poems 1910, is in excess of 700.)

At one time, Savidis contemplated the production of a critical edition of Cavafy’s acknowledged poems – he refers to it in his introduction to his 1968 edition of Cavafy’s Unpublished Poems – but the very nature of his 1991 edition of the canon suggests that he had by then abandoned the idea. The idea needs to be revived, however, not least because Savidis’s retrogressive1991 edition is now the standard edition of Cavafy’s most important poems. Ideally, perhaps, a critical edition should also include all the manuscript variants and all the unpublished earlier versions of the poems. With many poets, manuscripts may be used to reestablish authorial intentions that were overridden by editors, but, in Cavafy’s case, this potentially crucial function of manuscripts is scarcely relevant. His close control over the printing process meant that his printings provided the best indications of his intentions. When printer’s errors did occur, Cavafy made corrections by hand before distributing the poems. It is an indication of his thoroughness in this regard, that, when a letter was omitted in the date in the printers’ colophon in one printing of “Anna Comnena,” Cavafy (in every copy I have seen) corrected DEKEBRIOS to DEKEMBRIOS. Unusually, then, in Cavafy’s case, publications (in the form of the author’s own printings) clearly override the authority of the manuscripts; and it is the printings rather than the manuscripts that can provide a corrective to versions of the poems published in periodicals, anthologies, and critical articles.

I have already made suggestions for the sources of the base texts for a critical edition. I will conclude with some suggestions concerning the scope of the critical apparatus in such an edition, bearing in mind what has just been said about the peculiar relation of manuscripts, printings, and publications in Cavafy’s case. I propose that the critical apparatus should record all variants in all of Cavafy’s own printings, and in all versions of the poems published in periodicals and anthologies during the poet’s lifetime. In addition, it should record variants from at least some of the publications in critical works that appeared during Cavafy’s lifetime, and especially those – such as Xenopoulos’s article “A poet” (1903) and Vrisimitzakis book The work of C.P. Cavafy (1917) – for which Cavafy himself supplied texts to the authors. Manuscript variants would be excluded with two exceptions: the variants in The Sengopoulos Notebook and autograph amendments appearing in printed copies of the poems. Finally, for comparison, and in part to justify the production of a new edition, the apparatus should include the editorial variants from Sengopoulou’s 1935 edition and Savidis’s1963 and 1991 editions. These are the only editions based (however problematically) on Cavafy’s own printings, every other edition of the Cavafy canon being derived from one of these three.

Only with a new and critical edition along the lines outlined above will the record be set straight, and Cavafy’s 154 most important poems be made available once more – to his now vastly increased readership – in the form in which the poet himself circulated them to a few hundred carefully chosen recipients. Curiously, readers are already better served in the case of Cavafy’s non-canonical poetry. Savidis’s notes to his editions of the Unpublished Poems and Repudiated Poems provide some (but not comprehensive) details (and not in an easily usable form) of manuscript variants (and, in the latter case, published variants as well). Renata Lavagnini’s exemplary edition of the Unfinished Poems includes full diplomatic transcriptions of all the relevant manuscripts and facsimiles of some of them.

For a poet of Cavafy’s stature – and given the relatively small size of his poetic oeuvre – it is rather extraordinary that we still do not have a full facsimile edition, or even full transcriptions, of all the extant manuscripts of all of the poems. It is to be hoped that Manolis Savidis, the present owner of the Cavafy Archive, will find the resources to make such a project possible. In the meantime, a critical edition of the Cavafy canon would be a major step forward.


  • Cavafy, C.P., 1935, Poiêmata [Poems], ed. R. Sengopoulou, Alexandrinê Technê, Alexandria.

  • Cavafy, C.P., 1963, Poiêmata [Poems], ed. G.P. Savidis, 2 volumes, Ikaros, Athens.

  • Cavafy, C.P., 1968a, Anekdota poiêmata 1882-1923 [Unpublished Poems], ed. G.P. Savidis, Ikaros, Athens.

  • Cavafy, C.P., 1968b, Autografa poiêmata (1896–1910): To Tetradio Sengopoulou se panomoiotupê ekdosê [Autograph Poems: The Sengopopoulos Notebook in a Facsimile Edition], ed. G.P. Savidis, Technografikê, Athens.

  • Cavafy, C.P., 1983, Apokêrugmena poiêmata kai metafraseis (1886-1898) [Repudiated Poems and Translations], ed. G.P. Savidis, Ikaros, Athens.

  • Cavafy, C.P., 1991, Ta poiêmata: nea ekdosê [The Poems: A New Edition], 2 volumes, Ikaros, Athens.

  • Cavafy, C.P., 1994, Atelê poiêmata 1918-1932 [Unfinished Poems], ed. R. Lavagnini, Ikaros, Athens.

  • Cavafy, C.P., 1999, “Poems 1897-1910” [= Poems 1910], tr. A. Hirst, Greek Letters, 12, pp. 55-88.

  • Ekdawi, S. & A. Hirst, 1999, “Left Out, Crossed Out and Pasted Over: The Editorial Implications of Cavafy’s own Evaluations of his Uncollected and Unpublished Poems,” Modern Greek Studies (Australia and New Zealand), 5-7,1997-99, pp. 79-132.

  • Hirst, A., 1995,“Philosophical, Historical and Sensual: Cavafy’s Thematic Arrangements,” Byzantine and Modern Greek Studies, 19, pp. 33-93.

  • Savidis, G., 1966, Oi kabafikes ekdôseis (1891-1932) [The Cavafian Editions], Tachudromos, Athens.

  • Vrisimitzakis, G., 1917, To érgo tou K.P. Kabafê [The Work of C.P. Cavafy], Grammata, Alexandria.

  • Xenopoulos, G., 1903, “Enas poiêtês” [“A Poet”], Panathênaia, 7, pp. 97-102.

Anthony Hirst is a research fellow in Byzantine and Modern Greek Literature at Queen’s University Belfast. He is currently completing a book entitled Poetry Versus Empire: Cavafy, Byzantium and Disaster.
Page 1 of 1 pages