The coins of Michael VIII are currently mainly assigned to the mints of Constantinople and Thessalonica, with a few given tentatively to provincial mints in Asia Minor. However these assignments cannot be taken to be settled in all cases, as they are based mainly on the anecdotal evidence of the bazaar, or a few stylistic features, rather than the hard evidence of finds. This article is concerned with the question of which of Michael’s “billon” types can reasonably be assigned to the mint of Magnesia in Asia Minor.

There are two general reasons for assuming that some of Michael VIII’s trachea were issued from Magnesia. Firstly, since Michael VIII effectively became emperor in late 1258, but did not reconquer Constantinople until 1261, it is reasonable to assume that several of the early trachea issued in his name could have been struck at Magnesia. Secondly, it is an obvious fact that that the number of Michael VIII types assigned to Constantinople in Sear (for example) exceeds the number of years in Michael’s reign, so it is reasonable to conjecture that some at least of these types may have been issued from Magnesia, or some other secondary mint in Asia Minor or elsewhere.

Up to the present time only two or three of Michael VIII’s billon coins have been suggested as issues of Magnesia in the standard references. The program of this article will be first to assess the reliability of those assignments and then to try to determine whether any other of Michael’s coins can be considered as possible issues of Magnesia.

Ideally, we might hope to solve this problem by simply surveying the data on the finds of the various Michael VIII types found in Asia Minor. Surprisingly, however, there is little relevant data to go on – relatively few Asian finds of the coins of Michael VIII have been published, and importantly, there are, as far as I know, no published catalogs of the coin holdings of the local museums in south-western Turkey. As a result, for the most part the only coins which can be definitely assigned to Asian find spots are those from officially sanctioned excavations at ancient sites, and these, with the notable exception of the work at Pergamum, have not so far revealed much data which is useful for our purposes.

At the present time, therefore, we cannot attempt a full and systematic study of the problem, but we can at least make a beginning, and although it will be seen that the results are very limited and inconclusive, I think we can at least suggest a few types of Michael VIII which could possibly be issues of Magnesia.

The available Data.

The available data to which we can turn can be summarised as follows:

1) Reports (mostly from some time ago) from official excavations in Asia Minor, namely those at Troy, Priene, Sardis and particularly, Pergamuma (although, as stated above, the number of coins of Michael VIII actually found at most of these sites is surprisingly small).

2) The Arta Hoards of 1923 and 1983. Although a long way from Magnesia, these finds are critical to any discussion of the coinage of Michael VIII. The 1923 hoard, which dates from the early 1260’s, is listed in Bendall and Donald’s surveyb of the trachea of Michael VIII; it included several early issues of Michael VIII (from both Constantinople and Thessalonica), a number of Thessalonican issues of various earlier rulers, a couple of types usually assigned to Theodore II at Magnesia, and a few other issues. The 1983 find has a very similar composition, and is probably from essentially the same date as Arta 1923 (see Note “The Arta 1983 Hoard” for details).

3) The “Capstan Navy Cut” hoard. This small hoard (if that’s what it is) was found stored in a cigarette tin in an institute in Athens, and is detailed below; like the Arta hoards, it dates apparently from the earliest years of Michael VIII’s reign.

4) The Bergama “hoard”. These coins (as Bendall has pointed out) do not comprise an actual hoard, but rather a grab bag of coins picked out of a dealer’s bowl in Bergama (Pergamum) – nonetheless, we can probably assume that most, and very likely all, of these coins were found in western Asia Minor (which doesn’t of course mean that they were all struck there). The coins involved are noted in Bendall and Donald as the “Weller” types, and will be referred to here as coming from “Bergama”, as opposed to the excavations at Pergamum.

5) Dochev’s excavations at Turnovo, the mediaeval capital of Bulgariac. These excavations yielded large numbers of coins (mainly from stray losses rather than hoards) from the time of the second Bulgarian empire, including hundreds of Byzantine coins from the Palaeologan period, 159 of which were issues of Michael VIII. Amongst other things, the figures in Dochev therefore give us a valuable, although perhaps not entirely representative, statistical sample of the coins of Michael VIII which circulated in northern Bulgaria in the later 13th century. For our purposes this data is primarily negative, in that for geographic reasons coins from Magnesia are presumably a lot less likely to feature at Turnovo than those from Constantinople – so that types which are common at Turnovo are prima facie unlikely to be issues of Magnesia, while common types not appearing at Turnovo are worth considering as possible Magnesian issues.

(However, it should be noted that non-appearance at Turnovo could also indicate an origin in northern Greece. It is clear from retail sales statistics collected by me that the issues of Michael VIII at Thessalonica are much less likely to appear at Turnovo than those of Constantinople, presumably reflecting the trade patterns of the period – Thessalonican issues make up 44% of the total number of Michael’s single ruler types, but even including S.2280, they made up only 13% of the coins at Turnovo, as opposed to 36% in general retail trade. This dichotomy incidentally provides general support to most of the current assignments to both Thessalonica and Constantinople, although, as we will see below, a number of the “Constantinople” types seem to be under-represented at Turnovo).

6) Certain recent (mid 2000’s) sales on the Internet of coins of Michael VIII which were apparently sourced from Asia Minor (more of these later).

We are now in a position to try to determine which types could have come from Magnesia, and to begin with we consider three proposed Magnesian issues from the standard literature.

S.2271 & 2279.

Hendy, in his “Coinage and Money in the Byzantine Empire, 1081-1261”, assigned two of Michael VIII’s trachea to the mint of Magnesia, namely S.2271 and 2279, and this assignment was followed by Bendall and Donald. The arguments that Hendy used were essentially based on style, rather than any hard evidence – in particular, the appearance of St Tryphon (a saint of Asia Minor) on the obverse of S.2271, and the use of ligatured letters in the legends – the TP in Tryphon in the case of S.2271, and the Th in DECPOTHC on S.2279. However St Nicholas, who appears on S.2285, supposedly an issue of Constantinople, was also an Asian saint, and similar ligatures are found on other trachea of Michael VIII, such as S.2303 to 2306, which are currently assigned to Thessalonica; hence these arguments don’t really carry much weight, and we note that in Sear and PCPCd Bendall assigned both S.2271 and 2279 to Constantinople.

However, this reassignment seems to be somewhat premature, at least in the case of S.2271, since no less than four examples of this not very common type were found at Pergamum (making it the commonest type, along with S.2263 and 2265, among the 46 coins of Michael VIII found at that site), and it has also been found at both Manissa (the modern name for Magnesia), and (in a variant form) in the Bergama bowl. (Note that neither S.2271 nor S.2279 appeared at Turnovo, but by itself this is not really significant, given the relative scarcity of the types).

(It is perhaps also worth noting here that the Th ligature on S.2279 is also found on S.2286/7, a rare type probably struck in Philadelphia, not far from Magnesia, and also on S.2288, which is currently assigned to Constantinople).


On the other hand, in Sear and PCPC Bendall reassigned the common type S.2239 (= S.2283) to Magnesia. Here the evidence seems to be more compelling, at least at first sight. This type was part of the small “Capstan Navy Cut” cigarette tin find, which dated (apparently) from the earliest years of Michael’s rule at Thessalonica, that is, from the same period as the Arta hoards.

To understand this assignment we need to compare the “Capstan” find with the Arta finds. The cigarette tin contained S.2239/83 (8 copies), 2296 (2) and 2297 (4) of Michael VIII, plus a single Thessalonican coin of John III, while the Arta 1923 hoard included the Constantinople type S.2259 of Michael VIII (3 copies), the Thessalonican types S.2295 (13), 2296 (3), 2297 (6) and 2309 (9), together with 32 other pre-Michael coins of various types (mainly issues of Thessalonica).

Now, since both the Thessalonican types S.2296 and 2297 are present in both hoards, but 2295 and 2309 only appear in the Arta hoard, Bendall argued that the “Capstan” find predates the Arta hoard, and hence that it probably dates from (the indictional year) 1259-60, before the reconquest of Constantinople. But if so, then, since S.2283 is apparently not a Thessalonican type, it must be an issue of Magnesia, or at least of a mint in Asia Minor.

Thus the theory is that the Arta hoard contains the first four issues of Thessalonica, plus the first issue of Constantinople (S.2259, from 1261-2 presumably), while the cigarette tin contains the first two issues of Thessalonica, plus the first issue of Magnesia (S.2283, from 1258-9).

Now at first sight Bendall’s argument certainly looks attractive, particularly since we might expect that at least one reasonably common issue of Michael VIII would have been struck at Magnesia at the beginning of his reign. Nonetheless there are serious difficulties with this idea.

To begin with, why didn’t the cigarette tin include the second issue of Magnesia (whatever that was) as well as the first? (there would likely have been at least one more issue at Magnesia before the reconquest of Constantinople). Also, we might perhaps have expected that S.2295, as the commonest Thessalonican type in both Arta finds, would have been the earliest issue of Thessalonica, but it did not appear in the tin.

Secondly, and perhaps more importantly, why weren’t there a lot more pre-Michael VIII types from Thessalonica in the cigarette tin? (as there were in both Arta finds).

Thirdly, we note that S.2283 is one of the commonest types of Michael VIII overall. At Turnovo, for example, it was easily the commonest type of Michael, actually accounting for 20% of the total number of coins for that ruler at that site, and 24% of the Constantinople types. Now its frequency there certainly seems to be exaggerated, for some reason, but nonetheless in trade offerings its frequency is still about 7% of all of Michael’s types, making it the most common type overall. Therefore, if this type really is an issue of Magnesia then we might have expected it to feature strongly among the coins of Michael VIII which have turned up in Asian Turkey (in official excavations, for example, and in trade), but in fact it is no more prominent there than many of the other common types of Constantinople, and if anything, it is somewhat under-represented at Pergamum. Two examples of the type were indeed found at this last site, and one example appeared in the Bergama coins, but these figures are consistent with a Constantinople origin for this common type. (An example of this type “found at Manissa” was also offered on Ebay in 2004 by the “Asian source” referred to above, a seller on Ebay of coins apparently sourced in Asia Minor, but most examples of this type in the market don’t appear to derive from Turkey).

At this point we need to consider the second Arta find of 1983, which on the face of it was probably laid down soon after the first; this hoard included the Constantinople types S.2259 (3), 2260 (2), 2277 (1?) (or Const. Tikh Asen?) and 2239/83 (23), the Thessalonican types S.2295 (17), 2296 (2), 2297 (13) and 2309 (2), together with 82 coins of earlier rulers, mostly from Thessalonica. (As well, we note here the Ioannina hoard – this generally parallels the Arta 1983 hoard, and includes S.2295 (2), 2296 (1), 2297 (3), 2309 (3) and S.2239/83 (4)).

The publication of the Arta 1983 hoard postdated Bendall’s theory, and it raises a serious difficulty for his ideas. The problem is the presence in quantity of S.2283 in this hoard but not in the first Arta hoard; more specifically, given that S.2283 is supposed to be a pre-1261 issue of Magnesia, then, even though Arta 1923 is only about half the size of Arta 1983, we should still expect this type to also appear in significant numbers in the 1923 hoard. In fact of course S.2283 doesn’t appear at all in Arta 1923, clearly suggesting, prima facie at least, that it postdates that hoard. (For a possible, if at first sight unlikely, solution to this particular difficulty see below).

On the face of it, then, the evidence here is rather confusing, since we seem to have good arguments both for and against assigning S.2283 to Magnesia. However, we have to remember that we don’t actually know the history of the coins of the “Capstan” hoard before they turned up in the cigarette tin, and hence whether they constitute a real (or complete) hoard or not. And in fact, the almost total lack of pre-Michael VIII types in the tin strongly suggests that we don’t have a complete hoard here.

Now, if we are not dealing with a complete hoard, then of course the whole argument is weakened, and we could assign S.2283 to Constantinople, after S.2259 (and S.2260 to Magnesia – see below), as the second Arta hoard suggests.

But if so then how do we explain how such an odd collection of types as was found in the cigarette tin could have been put together, so as to include S.2283 but not the common Thessalonican types 2295 and 2309, both of which appeared at Arta. This obviously could involve any number of scenarios, but one idea, foreshadowed above, might be that the cigarette tin coins (which seem to have first appeared in the 1920’s) were originally part of the Arta 1923 hoard – this might seem rather unlikely, but it could explain a lot, particularly the absence of S.2283 in Arta 1923. It could also of course make S.2283 the second (or maybe even the first) issue of Michael VIII at Constantinople.

But whether or not this is the case, it seems that S.2283 is more likely an issue of Constantinople than Magnesia, since this type seems to be far too common at Turnovo and trade generally (and, in particular, in the Arta 1983 and Ioannina hoards), to be an issue of Magnesia. At Turnovo the Magnesian issues of the Nicean rulers are generally quite scarce – for example, there were only 3 examples of the Magnesian bronze coins of Theodore I (and these were all small module types designed for use in Bulgaria anyway), 14 examples of John III, and none of Theodore II, suggesting that the Magnesian issues rarely reached northern Bulgaria. (By contrast there were 21 large module coins of Theodore I from Nicea at Turnovo).

On the other hand, there were no less than 33 examples of S.2283 at Turnovo. Now it’s true that the economy of Bulgaria was in deep decline in the 1240’s and 1250’s, and that it revived in the 1260’s under Constantine Asen, but I doubt that this is sufficient to explain the large number of 2283’s at Turnovo, particularly as compared with the small numbers found in Asia Minor. Both S.2283 and the similar type S.2285 are clearly over-represented at Turnovo for some reason, making up 60 of the total of 159 coins of Michael VIII found there. One interesting possibility, as noted in the Appendix below, is that these types are in fact early (perhaps even pre-1261) issues of Michael at a mint in Thrace, such as Adrianopolis; this could also explain the presence of S.2283 at least in the Arta 1983 and Ioannina hoards. However, we note the obvious fact that S.2285 does not appear in any of the Epirote hoards, which at first site to rule out a pre-1261 date for both of these types.

Furthermore, as noted above, S.2283 is also clearly over-represented in the Arta 1983 and Ioannina hoards, and much commoner than the Constantinople type S.2259, an issue which is nearly as common as 2283 in general trade. Again, the reason for this discrepancy is unclear, but as before it would seem to raise difficulties in assigning 2283 not just to Magnesia, but also to Constantinople. On the other hand, the presence of this type in Asia Minor surely rules out Thessalonica as its source, so that, as just suggested, we seem to be led back to the possibility of another mint somewhere in northern Greece. Could the answer be that the S.2283’s were struck in north-west Turkey (in say Nicaea or Nymphaion), or perhaps a Thracian mint, in 1261 as pay for the army during Michael’s progress to Constantinople, while S.2285 was struck (somewhere) at a later date?

Two new candidates

So, if the preceding examples are not all totally convincing as issues of Magnesia, do we have any other candidates, preferably ones whose claims are based on solid facts rather than just minor stylistic features? In fact, there are quite a few, and of these two types in particular stand out.

PCPC 39, DO 104-5.

This type, which is not listed in Bendall and Donald or Sear, has St Michael standing on the obverse holding a distinctive slim patriarchal cross, and Michael enthroned on the reverse, as on S.2276, but holding a sceptre cruciger and globus.

This type was not found at Turnovo, but quite a number of examples of it have been offered on Ebay in the last few years, most apparently sourced from Asian Turkey, and of these at least two were specifically said to have been found at Manissa. Two examples were also found at Pergamum.

Combining these examples with the fact that this type was not found at Turnovo, we might reasonably conclude that PCPC 39 is very likely an issue of Magnesia.

On the other hand, the fact that the type was not included in the Bergama bowl is a little strange. It is not clear what to make of this, but perhaps we shouldn’t place too much importance in the Bergama find, given that it is really just an eclectic selection. It is true that the prime selection criterion was apparently rarity, so that PCPC 39 is exactly the kind of coin that we might have expected to have been picked out of the dealer’s bowl, but perhaps there just weren’t any (collectable) examples in the bowl at the time – after all, it’s not really a common type, and was not recognised until relatively recently (although it actually appears in Sabatier Pl. LVIII, No.8 as an issue of Isaac II!).

P.S. (Apr. 2015). For another possible but unlisted issue of Magnesia with a likely connection with PCPC 39 see Coin 10 on the Unlisted Palaeologans page.


On the other hand, no less than five examples of the rare joint rule type S.2318 were found in the dealer’s bowl at Bergama, easily the commonest type in this “find” of only 18 coins, and, importantly, an example was also found at Pergamum. Since this type did not appear at Turnovo it is tempting to conclude that it is definitely a local issue of Asia Minor, presumably of Magnesia. However, we have little other significant data on this type, and its scarcity means that its absence at Turnovo doesn’t necessarily mean much. And of course, it has the same St Nicholas obverse as S.2285, which is presumed to be an issue of Constantinople.

Nonetheless this type is clearly very possibly an issue of Magnesia (after all, as noted above, Nicholas is ultimately an Asian saint), and if so then its relatively late date suggests that other types could well have been produced there as well.

Other possible candidates

While the last two types stand out as possible issues of Magnesia, there are a number of other issues that might also be considered on more general grounds.

It is noticable that three relatively common types, namely S.2260, 2263 and 2275, did not appear at all at Turnovo, while only two examples of S.2265, another common type, were found there (but note that S.2263 may have been listed under S.2262 – see next section). We also note that all of these types have shown up at official excavations in Anatolia, and in fact they were amongst the commonest coins of Michael VIII found at Pergamum, the numbers being S.2260 (3 examples), 2263 (4), 2265 (4) and 2275 (2). As well S.2260, 2263 and 2275 were included in recent retail offerings from Asia, and two S.2260’s were found in the Bergama bowl. (S.2260, 2263 and 2275 also appeared among the small number of Michael VIII types found at Anaia in Anatolia).

Then we have the odd case of S.2273. The iconography of this type (the model city) is definitely Thessalonican, and it was not found at Pergamum, but of 5 examples of noted by me in the retail market in the last few years, two were apparently from Asia, with one stated as having been “found at Manissa” – almost unheard of for a Thessalonican issue. This type was also a no-show at Turnovo, although it is not particularly common to begin with.

Does all this suggest that these various types were issues of Magnesia? Possibly – although the fact that most of them are fairly common in the market might also be taken by itself to militate against a Magnesian origin (since it is noticable that, for pre-Michael types at least, the issues of Magnesia are relatively scarce in trade*, just as they are in most reported 13th century finds). Perhaps, then, there is another reason for the non-appearance of these types at Turnovo – for example, they may be issues of Thessalonica, or some other mint in northern Greece (or Asia), since, as noted earlier, it is clear that the Thessalonican issues of Michael VIII are also under-represented at Turnovo. However, it must be said that the fact that most of these types appeared at Pergamum virtually rules out their being issues of Thessalonica, since it is clear that regular Thessalonican types rarely appear in Asia (none of the recognised Thessalonican issues of Michael VIII were found at Pergamum, for example, and only two of the coins in the Bergama bowl appear to be issues of Thessalonica).

If these last types seem unlikely to be issues of Magnesia because they are too common, there are still a few scarcer types worthy of attention, namely S.2284, the seraph type S.2289, S.2277, and finally, the rare “empty throne” type, S.2272.

S.2284 is not a particularly common type, and it did not appear at Turnovo, but there were two examples at Pergamum, and of four examples noted by me in recent sales (other than the two in the “Despot” sale) two came from the Asian source. Also, this type is often characterised by jewelled decorations, a feature not normally found on the coins of Michael VIII, but common enough on many of the Magnesian issues of earlier reigns. It is certainly not the worst candidate for Magnesia.

S.2289 is also not very common, but there were no less than three examples at Pergamum, and one in the Bergama bowl. Again, this type did not appear at Turnovo, and it is hence a possible Magnesian issue. Note also this example of S.2318 apparently overstruck on a seraph type:

S.2318 overstruck on Seraph type (2.10g)
S.2318 overstruck on Seraph type (2.10g)

S.2277 is a scarce type (although the common and similar type of Constantine Asen is often mistaken for it), and its exagerated style suggests a provincial origin. It did not appear at Turnovo, but there were two examples in the Bergama bowl and another at Pergamum, again consistent with an origin at Magnesia, or at least in Asia Minor. (Note that the reported presence of this type in the Arta 1983 hoard is open to doubt, although stylistically it is similar to some of the unlisted coins in that find which are possibly issues of western mints).

The “empty throne” type S.2272 is very rare, but one turned up at Pergamum. Given that this is apparently the only find spot of this type known, it is possible that it may be an issue of Magnesia, or at least of a mint in Asia Minor, although given the lack of any other hard data on this type, we can hardly draw any firm conclusions about it. (This type is also found in silver of course, which is otherwise unknown at Magnesia for Michael VIII).

It should be noted that apart from the types considered so far above several other types of Michael VIII assigned to Constantinople appeared at Pergamum, but the only ones to appear in significant numbers were S.2259, 2264, 2267 and 2285 (3 examples each). Given that these are all common types, both at Turnovo and in trade, their appearance at Pergamum is not unexpected.

(For a more detailed discussion of the statistics of the various issues considered here see the Appendix below).

* In recent times (2015) this has started to change, due perhaps to increased trade of crisis coins and other artifacts from Syria through Turkey.

Some further types of interest

There are several more types worth considering for reassignment, although not necessarily to Magnesia. Firstly we have S.2262. This scarce “blessing” type is quite similar to the much commoner type S.2263, and hence could possibly be a provincial version of the latter. One copy of S.2262 was reported at Pergamum, but Dochev reported no less than 8 examples at Turnovo. However, Dochev reported no copies of S.2263, and hence may well have lumped the two types together, so it’s difficult to know what to make of these figures.

Then we have S.2294 and 2298, currently assigned to Thessalonica. These are both scarce types, but they show a close resemblance to certain of the other much commoner early types of Thessalonica, and hence give the impression that they could be provincial issues copying the original types. However, there is no hard evidence to support this idea, or to associate these types with any particular mint. Neither type appeared at Turnovo, but given their scarcity this means little. (Note also that an odd coin with the reverse of S.2298 and B. Christ on the obverse was offered on Ebay in Oct. 2010).

Finally, we have S.2278. This is a very uncommon type, yet according to Dochev there were no less than 4 examples at Turnovo. If this figure can be trusted, this must raise questions as to the source of this type.

The rationale of the Magnesian issues

Two questions that can be legitimately asked are when and why (and how) were the hypothetical Magnesian types issued. Clearly, coins from the earliest years of Michael’s reign are to be expected from Magnesia, but except for the S.2260 (presumably a coronation type) there is no clear indication that any of the types considered here date from that period.

Once the official mint had transferred to Thessalonica and Constantinople, it might seem unlikely that would there have much expertise in coin production left in Magnesia, or indeed any need for it. However, the joint rule type S.2318 evidently dates from 1272, or even later, and it seems likely that at least one issue of Michael (S.2286/7) was produced at some stage in Philadelphia (one example of S.2287 was found at Pergamum), so it appears that a local need for coinage did eventually arise again in Asia Minor. It is therefore possible that other later issues of Michael VIII were produced in the region to fill that need, and in fact there is good evidence that local production of coinage continued into the reign of Andronicus II (see Note “S.2464 and the Asian trachea of Andronicus II”).


Overall, the results of our labours might seem to be rather minimal, but I think we have two strong candidates for issues of Magnesia, namely PCPC 39 and S.2318. Furthermore, we have several other candidates worthy of further consideration for Magnesia, namely S.2260, S.2272, 2277, S.2284 and 2289. As well, there is some evidence at least that a number of other types might need to be considered for reassignment, if not to Magnesia, then at least to somewhere other than Constantinople, namely S.2265, 2273 and 2275.

As well as these new candidates, S.2271 is definitely still a likely issue of Magnesia, although statistically this type (and S.2279) fits comfortably in Constantinople. S.2239/83 is clearly an early type, but there are problems in assigning it to Magnesia, and hence for the present I am assuming it is most likely an issue of Constantinople, although there are difficulties with this allocation as well, and it may be that in the end this type, and also S.2285, may need to be assigned to some other mint or mints, as noted earlier.

One major problem still remains however, namely the identification of the earliest issues of Magnesia. While S.2260 would seem to be a likely coronation issue at Magnesia, it is unlikely to have been the first issue of Michael at that mint, but just which issues proceeded it is yet to be clarified.

Of course, these conclusions can only be provisional at this stage, due largely to the limited amount of solid statistical information on the distribution of Michael’s coins in Asian Turkey (and for that matter in Greece). One thing that would be very helpful would be a census of the coins of Michael VIII in the local museums, or even just the local markets, in Anatolia, particularly in the south west. This information would undoubtedly be very informative – perhaps some readers could help us out here.

Appendix: The statistics of the issues of Michael VIII.

The numbers of the various types of Michael VIII at Turnovo and in trade generally constitute two separate samples of the coinage of Michael as a whole. As we have seen, in some cases there appears to be a considerable difference between the frequencies of occurrence of certain types in the two samples, and the question is, how significant are these differences?

The differences between the frequencies of occurrence (measured as a fraction of the total number of coins in each sample) can be assessed by statistical means. For any given type we can assess the significance of the difference by expressing it in terms of the standard deviation (s.d.) of the difference, which is the square root of the sum of the squares of the individual s.d’s for the type in the two samples (Weatherburne, Sec. 46).

Doing this we find that for the nominally Constantinople types there seem to be rather too many types for which the difference in the frequencies in the two samples is more than 2 standard deviations. In fact, there are in all 7 types in this category from the 28 separate types present in at least one of the samples (excluding S.2280, which is probably an issue of Thessalonica). Since we would normally expect perhaps only 1 or 2 types (5%) with differences exceeding two s.d’s for this number of different types, perhaps some at least of the 7 “deviant” types belong elsewhere. (Note that for purposes of calculation I have combined S.2262 and 2263, since they are presumably combined at Turnovo)

We should note here that the calculations are somewhat biassed by the obviously excessive numbers of two particular types at Turnovo, namely S.2283 and 2285 (for which the frequency differences are 3.1 and 4.5 s.d’s respectively, which surely indicates an over-representation of these types, for some reason). Omitting these two types from the calculations reduces the spread of figures somewhat, but we still seem to have too many deviant types from Constantinople, and hence a need, prima facie at least, for some reassignments.

The most obvious candidates for reassignment are those relatively common types which did not appear at Turnovo, namely S.2260, 2273 and 2275, plus S.2265, only two examples of which appeared at Turnovo. For all these types the difference in frequency of occurrence was somewhat more than two standard deviations of the difference (except for S.2273, where the figure was 1.6 s.d’s). If we transfer these four types to (say) Thessalonica, then we find that we have only two deviant Constantinople types left at Turnovo, out of 22 types present (apart from the two anomalous types S.2283 and 2285), a much more acceptable figure.

At Thessalonica, even including the extra types, we have only one deviant type out of 23 types present (including S.2280), showing that the extra types could, from a statistical view at least, fit in quite easily. (The deviant here is the type DOC V Table 9B, No. 61, which appears in Dochev as Pl. XV, 9. This type has B. St Demetrius on the obverse as on S.2303, with a standing emperor holding an akakia and a trilobate sceptre on the reverse – three examples of this scarce type turned up in the excavations at Turnovo, but I know of only two examples in commercial sales).

However, from a statistical point of view it must be said that the numbers involved in these calculations are generally rather too small to inspire great confidence, and in any case all of the four types in question here (except for S.2273) appeared in numbers at Pergamum, which, as stated earlier, surely rules out Thessalonica as their source. Hence there may well be other explanations for the no show of these types at Turnovo. In general, it has to be said that the Turnovo coins are probably not an entirely representative sample of the coins of Michael VIII, so that the Turnovo figures have to be used with caution.

One thing that is clear however is that, as compared with general trade, the Thessalonican types of Michael as a whole are under-represented at Turnovo against those of Constantinople (by a factor of about two), and the question that needs to be asked is why? Probably the main reason is that Bulgaria is basically divided into separate northern and southern regions by the central Balkan mountain range, and that whereas most of the recent trade offerings presumably come from Bulgaria generally, and to some extent Turkey and (perhaps) Greece, the Turnovo figures reflect only the coins which circulated in northern Bulgaria, where coins from Thessalonica were apparently less likely to appear*.

Given the problems with the figures from Turnovo, it is interesting to compare the statistics of the coins from general trade with those of two other samples of Michael VIII’s coins, namely the overall holdings of his types in the museums surveyed by Bendall and Donald, and also the coins of Michael VIII found at Pergamum.

Although we might perhaps doubt the representative nature of the museum holdings, in statistical terms the agreement between this sample and the trade figures is actually quite good, with an overall correlation of c. 0.75, despite the seemingly high numbers recorded for certain types in Bendall and Donald. Unlike the Turnovo coins, the museum holdings presumably came in the first instance primarily from Greece and (all) Bulgaria, and to a lesser extent Turkey. The only serious discrepancy is S.2280, which is relatively scarce both at Turnovo and in Bendall and Donald, but remarkably common in recent trade, where, rather oddly, it has turned up in quantity in several hoards of the Thessalonican coins of Andronicus II (see “Some Palaeologan Finds“). The lower B. & D. figures for S.2280 are in fact more consistent with Constantinople rather than Thessalonica as the source of this issue, so perhaps we still have a problem with the allocation of this type. (But note that S.2280 did not appear at Pergamum).

The common type S.2264 is also underepresented in B. & D. for some reason, but this seems to be of no real significance, since it appears elsewhere in expected numbers.

At Pergamum the Thessalonican types are of course absent, but for the non-Thessalonican types the compatibilty between the trade figures and the Pergamum coins is quite high – the only types showing any significant degree of over-representation at Pergamum are S.2271, 2272 and 2289, which as we have seen are all very possibly Asian issues anyway, the Philadelphian types S.2286/7, and perhaps S.2262 (see discussion above for this last type).

A few common types, namely S.2261, 2269, 2270 and 2274, don’t appear at Pergamum, so it’s possible that these are later issues of Michael VIII which did not reach the city (due to increasing Turkish incursions into Asia Minor) – but the degree of significance here is not high (c.1.3 s.d’s), so too much faith should not be placed in this idea.

It is also worth noting that the figures for S.2283 and 2285 in Bendall and Donald, and also at Pergamum, are reasonably compatible with those in general trade, which makes the excessive number of these types found at Turnovo even more puzzling – can we perhaps surmise (as suggested elsewhere) that these two types are the product of yet another mint, one which had significantly greater access to Turnovo than any of the others – Adrianopolis in Thrace perhaps? It’s not impossible, but maybe there is another simpler explanation for this curious, but clearly significant, discrepancy.

* Note that, as we have seen, the types S.2283 and 2285 constitute an anomolously large fraction of the coins found at Turnovo, thus reducing the relative numbers of the Thessalonican types – but even if we drop these two types from the calculations altogether, the percentages of Thessalonican coins at Turnovo and in trade generally are still significantly different, at 20% and 39%, respectively. Reassigning the five types mentioned above to Thessalonica only increases this difference, as we would expect. No matter how we analyse the figures, the Thessalonican types are significantly under-represented at Turnovo.


a: H. Voegtli, “Die Fundmuenzen aus der Stadtgrabung von Pergamon”, Berlin 1993.

b: S. Bendall & P.J. Donald, “The Billon Trachea of Michael VIII Palaeologos, 1258-1282”, 1974.

c: K. Dochev, “Coins and Coin Usage at Turnovo (XII-XIV c.)”, 1992.

d: Simon Bendall, “A Private Collection of Palaeologan Coins”, 1988.

e: C.E. Weatherburn, “A First Course in Mathematical Statistics”, Cambridge U.P, 1962.

Ross Glanfield

March 2007

Last modified:

24 Apr. ’07: Appendix on statistics added.
31 July ’07: Coins from excavations at Pergamum included in discussion.
20 Sep. ’07: Appendix on statistics revised, incorporating Pergamum.
26 Nov. ’07: Arta 1983 hoard included.
6 Feb. ’08: S.2263 reassessed.
15 Mar. ’08: S.2277 added as a possible Magnesian issue.
24 Mar. ’08: Discussion concerning S.2283 refined.
14 Feb. ’10: S.2318 overstruck on Seraph? type noted.
23 Apr. ’15: Another possible issue of Magnesia noted (See discussion of PCPC 39).
4 Mar. ’16: S.2260 confirmed as likely coronation issue of Michael at Magnesia.