‘Iron Age’ Picts and their spoken language

Okay, here’s another thing I wanted to write up before I went to Kalamazoo. You may have seen, if you are following Archaeology in Europe as you all should be, that there was a recent paper in Proceedings of the Royal Society A that apparently decodes the Pictish language or something similar. I confess to initial scepticism, not least because they inexplicably persist in using the term `Iron Age’ for a people only attested under the name ‘Picts’ from the Roman period onwards, and whose glory days are most definitely early medieval, but I am interested in the Picts, I am in favour of Science! in history and so I thought I’d better have a look. After all, I am developing a blog-tradition of critiquing scientific papers on matters historical, and I’d hate to pass up another opportunity. Now, if those instances have taught me anything, it is these things:

1. articles based on the press release usually massively exaggerate the impact, and indeed the intent, of the actual research;
2. the actual research is usually more interested in proving a method than in its applications, otherwise it would have been published in a historical forum not a scientific one; and,
3. it is unfortunately rare for the authors of that research to have read enough in the field to which they’re supposedly contributing to have an accurate sense of whether or not they really are.

And this particular case ticks all three boxes, which is to say it’s interesting, appears scientifically rigorous at first glance, but sadly isn’t going to add much to the historical or linguistic debates, even though the news coverage would have you believe it’s a revolution in the field. So first of all I’ll deal with what the paper is doing, then try very briefly to describe the debate in which it belongs, and lastly assess the former against the latter. And because these things turn out to take a while, I will do so behind a cut…

The Pictish symbols display signs of a spoken language

To have heard of the Picts is almost to have heard of their symbol stones. These enigmatic monuments exist from or in many parts of modern Scotland and, though the iconography of Christianity distinguishes some from the others, and eventually seems to have driven out local styles entirely, before that final phase the stones share a remarkably coherent and limited library of symbols, the meaning of which has been debated for a long time (though I’m getting ahead of myself there). Here they are:

The Pictish stones' symbol library

What this new paper is actually doing, as opposed to what the press coverage has said, is demonstrating that these symbols have characteristics that belong to a writing system that is representing a spoken language, lexigraphically.¹ The method is down to how random the distribution of these symbols in use is, specifically how predictable or not a second symbol is when a first symbol is known. For example, if in English you have a word, or even a syllable (or a digram, as the paper has it, a combination of two symbols), that starts with `w-‘, there are a restricted set of symbols that can then follow it. All the vowels might work: consonants such as `-r’, or `-l’ or `-s’ are possible if this is part of a large word, but `-b’, `-g’ or `-x’ are at least less likely and maybe actually impossible. A starker example is obviously `q-‘, which except in cases only found in the Scrabble dictionary is only followed by `u’ in English. For that example, the entropy they’re counting is basically zero, it’s nowhere near random. Even with letters like `e’, in actual use the sample is a lot less than random because some combinations are so much more common than others. And this is a common characteristic, say they, of lexigraphical writing systems: their symbol distribution is significantly non-random, whereas with what they insist on calling a `heraldic’ system, by which I assume that they mean pictographic—they don’t mean logographic, for reasons I’ll come to—it’s far closer to random, and the division between these two cases, they say, is pronounced enough that with a decent sample you can tell lexigraphic from `heraldic’ by application of a test for `Shannon entropy’.

The Pictish symbol stone at Gairloch, Wester Ross

OK, so far this sounds plausible: Pictish symbols do tend to occur in a certain number of repeated combinations, though you’d get no sense of that from this paper admittedly. The problem is that we’re dealing with such a tiny sample set: 23 symbols (or 35 according to a more generous classification, admittedly) repeated over about 180 stones is not a lot to conjure with. In order for the Shannon test they’re using to work, they therefore have to have an estimate of how many of the possible combinations we have. So the crux of the paper is an attempt to prove that you can do this by reasoning up from the number of single symbols versus the number of combinations of symbols. Now, I have met this kind of maths before, in dealing with estimating currency size from coin die survival, and there it sucks.² However, let’s not dismiss it out of hand just yet. They’re not actually sure if this method works but they are sure that the attempt would produce different results—because of the repetition of particular combinations being much higher in lexigraphic ones than heraldic, for example how many words in English contain the doublet `er’—in a ‘heraldic’ writing system than in a lexigraphic writing system. And for them Pictish is the latter, not the former.

A graph of the randomness of Pictish symbol combinations, with additions by Language Log

Now, this is not without its problems even in its own terms, and as you might expect, Language Log has been all over them. The biggest problem is that the paper’s random test set also comes out on the `written significance’ side of the barrier they set. Furthermore, so do many other things that are not at all linguistic: on the graph above that I’ve poached from Language Log, the red cross is the randomness of 32 throws of a six-sided die, and by the standards of this paper that’s significant, because it’s not uniform. So a lot seems to fall over if you actually mean `random’ rather than `evenly distributed’. But even then, the point I think still stands: the symbols have meaning. Problem is, I think we had more or less got that already.

Historian’s perspectives

Let’s start with the Pictish language we’ve supposedly discovered here, and then talk about the stones a bit, then bring them together for the grand finale. (Yes, this post will in fact end some time, I promise.)

The Pictish language

The language of the Picts has been a matter of debate for long enough that Walter Scott parodied the arguments (in his The Antiquary). This is mainly because we have so little of it left. (We don't have 'none', as the woman who appears to have written the article that first led me to this seems to have believed when she headlined that “New Written Language of Ancient Scotland Discovered”, but that I can dismiss because the RSPA paper makes no such claim, and in fact discusses the corpus of material quite thoroughly given the space available.³) The only definitively Pictish text we have is the king-list, which you can find, apart from anywhere else, inscribed in the walls of the relevant gallery of the National Museum of Scotland, or for those not actually in Edinburgh right now, here. Even this is mainly Latin, but the names appear to be rendered in Pictish, not Gaelic, and even some Gaelic ones have been Pictished, so we have the beginnings of a clue. There are also a bunch of place-names, and of some of these it has been particularly noted that they include the sound `P’, which {edit: I had this sentence and the next all backwards, sorry; thanks to commentators] evolved in the Brittonic and Gaulish languages out of a kw- sound that the Goidelic languages spoken in Ireland retained, for which reason these two branches are usually called P-Celtic and Q-Celtic. What exactly these names mean is a matter of debate, but this factor at least means that they cannot simply be Gaelic.⁴ Then there are 14 36 of the stones inscribed with legends in Ogam, and another 2 in Latin, so both in foreign scripts implying perhaps foreign scribes, but also apparently containing Pictish names, and of which large parts have not yet been understood.⁵ And then we have Bede, who tells us that in his day four five languages were spoken in the British Isles, English, British, Pictish and Gaelic, [edit: as well as Latin, of course] and we have Adomnán of Iona, who has stories of Columba going to Pictland, even bits of Pictland quite close to Iona, and needing an interpreter.⁶ And that’s basically it.

The 'Drosten stone' from St Vigeans, image from Wikimedia Commons

There are deep-seated issues here because since Kenneth Jackson, and probably before, there’s been an argument that the incomprehensible bits of Pictish as we have it are evidence of the survival of a non-Celtic language, possibly even a pre-Indo-European one like Basque, and of course the nationalists just love this and so it will not die. Of recent years, however, Kathleen Katherine Forsyth has all but demolished this thesis for the academic world, arguing that the stones are just hard to understand, not actually gibberish, and that all the other evidence, primarily the place-names in Pett- or Pit-, indicate that Pictish was a P-Celtic language and that it was the only one spoken across the area known as Pictland.⁷ Most people have been happy with this, but a few small villages still hold out. I have tentatively suggested that the chronology of this needs to be looked at because there are reasons to believe quite a lot of ‘Pictland’ was already Gaelic-speaking long before the Gaelic takeover, not least Atholl, whose name, attested in the eighth century in the Irish annals (though heaven only knows that’s not a simple thing to assert), is plainly derived from the Gaelic for ‘New Ireland’.⁸ More importantly, Graham Isaac, a very learned Celticist, has repeatedly stood up in public and said that there are still a lot of place-names attested from early Scotland that do not make sense in Celtic, and for which some other explanation remains necessary.⁹ What this most immediately means is that we may be somewhat off-base in asserting a single Pictish language, for all that Forsyth does, and this ties up well with the numerous other indicators I’m fond of pointing out that we have plural cultures in this single political zone.

The symbols on the stones

It used, I think, a long time ago, to be thought that the symbols were no more than illustrations for the stones, and this seems to be what the article here is mainly kicking against, but actually attempts have been being made for a long time to give them meaning, and this scholarship is largely missing from the RSPA authors’ reading sad to say. I was going to try and type up an account of it but then I found I already owned a better one, so I’ll just use that:

Archaeologist Charles Thomas saw the symbols as memories of late Iron Age weapons and equipment; perhaps the symbol stones were used instead of placing these things in a grave: a broken spear, a sword, a chariot. By the time the images came to be used by the Picts, in the fifth to seventh centuries, they were symbols of rank, and indicated who was remembered and who set up the memorial. So the Dunrobin stone could mean `Erected to a Warrior of the salmon-people by his wife’.

The stone depicted is known as Dunrobin I and it shows a fish, a 'tuning fork' (perhaps a sword?) and a mirror and comb. Most Pictish symbol-fish are thought to be salmon.

The anthropologist Anthony Jackson has suggested that the symbols are records of marriage treaties, typically made on boundaries of united lands, and especially necessary where descent is through the mother. The symbols refer to families or kinship groups which, occurring in pairs and triplets, record the interest that each family retains. The Dunrobin stone might be interpreted as recording a marriage between the salmon and sword families. The mirror and comb indicates that the bridewealth was paid by the senior family.
By contrast, the Glasgow scholar Ross Samson looked on the symbols as representing the elements of names; each symbol gives the sound of a Pictish word or syllable, which joined together make up a name. An equivalent in Anglo-Saxon would be Aethel-wulf – loosely ‘Noble Wolf’ – or Sigeberht ‘victory bright’. So in this interpretation the Dunrobin stone might mean ‘in memory of a woman called Salmon-sword’ or more prosaically ‘Here lies Mrs Swordfish’.

Other scholars have shown how the stones are likely to mark out territory or land and most agree that the probable function of the symbols is to celebrate a named individual, in line with practice in the neighbouring Celtic and German lands.¹⁰

Contribution to our Knowledge

That rather useful breakdown of the debate over the symbols lets us see fairly easily where this paper may have an impact. Which of these theories does it support? At first sight, it would seem to be Samson’s, not Thomas’s, and maybe Jackson’s, because Samson’s understanding of the symbols’ meaning is in fact as parts of words that would be said, Thomas’s is however purely symbolic, or as our authors would have it `heraldic’, and Jackson’s could go either way. But why is there this insistence on the term `heraldic’, rather than more usual ones like logographic? Well, that seems to be the second weak point in the argument, the first being the question of statistical distinction spotted by the people at Language Log. On analysing their results to try and answer the question of whether what they are picking up is combinations that are meaningful in themselves (that is, words, or at least particles) or phonemes that only have a sense when used together (and it seems to me that monosyllabic words would twit that distinction anyway), the RSPA authors conclude:

The larger symbol categorization proposed by Allen and Anderson in Early Christian Monuments of Scotland implies that the Pictish symbols are very constrained words, similar in constraint to the genealogical name lists.¹¹

To which, my first reaction was, yes, so they’re gravestones then, with names on, same as we already thought. And then I also thought: in that case, if each symbol is a word, how is this system definitionally different from one of pictograms or hieroglyphs anyway? and I guess that’s why they’ve used the word `heraldic’, because they are actually looking at what they think are logographs, but I’m still not convinced there’s a meaningful distinction they can draw. And if so, well, doesn’t the whole thing come to bits, if there’s no real distinction between the two cases their method is supposed to distinguish? This is the point, I guess, where a real linguist should take over (I’d page Michael Drout but he seems to be a bit busy at the moment…) but it smells of Pictish salmon to me. On the other hand, if we can trust that they are right about words, then it would seem to cripple the Samson theory and leave us only with Jackson or Thomas once more, and I have to admit that given the stones’ very largely funerary context—many were found in graves, even, suggesting that they weren’t even intended to be seen—I’d prefer Thomas, but against that, a lot of the stones have been moved and Jackson had no problem finding some on supposed boundary zones, although I suspect that that is probably like finding ancient monumental alignments in many ways.

But then they go on with this, and I lose my rag somewhat:

However, demonstrating that the Pictish symbols are writing, with the symbols probably corresponding to words, opens a unique line of further research for historians and linguists investigating the Picts and how they viewed themselves.¹²

No, look, it really doesn’t. This “unique line” has been open a long time already and this paper has entirely failed to address that debate, which as said above it might have been able to do. And even if it were addressed to the scholarship, there are so many basic problems, not just with the maths but with the assumptions about a language’s nature, most of which are also punctured at Language Log. The single best comment there comes from one Pavel Iosad, who wisely says:

I have not yet seen the paper, but the interview quoted is sometimes hilarious. Apparently the Picts did not only have written language, but (as if someone thought the contrary) also a complex spoken language. You don’t say!

In any case, this all seems like a case of looking for something in a lighted place, as opposed to where you lost it…. That is, we know that some “Picts” (where “Picts” is of course pretty much a cover term for “any tribes living north of the Brigantes and Votadini, the northernmost tribes that Romans knew relatively much about”) spoke a Celtic, in all probability a Brythonic language, and some “Picts” spoke a different language, often assumed to be non-IE. This latter is recoverable from placenames and personal names, but the tantalizing issue is a number of Ogham inscriptions which are not in Goidelic Celtic and have yet defied convincing decipherment. To look for a language in, uh, pictures, when there are possibly two perfectly “normal” languages in the same place strikes me as, let’s put it mildly, fruitless.

And he is right. At its most basic the impact of this paper is “new research shows that Picts… spoke!”; at its most developed, it is still only, “Pictish symbols probably have linguistic value”, which was, as I say, pretty much agreed though a new means of showing these things is always welcome. But whether that then means we’re looking at Pictish is another question. Some part of Pictish, possibly: after all, this has a meaning that refers to spoken words, but is perhaps not really English:

British fire hydrant marker

The meaning is, I believe, “there is a hydrant located 3 yards from here at a bearing of 75 degrees” but I could easily be wrong and please check before you next start a fire in Britain. The point is, what if these symbols are, as we might suspect, only a very very limited part of a potential Pictish vocabulary?

Distribution map of Scottish place-names in Pit-

And then there is the problem of what language the ‘Picts’ actually spoke, as said above. The map above shows the spread of place-names of the form Pit- in Scotland, which as also said above is a P-Celtic signifier (and probably just means ‘settlement’). The one below, shows the distribution of the stones. Now, I know that I’m fond of arguing against the use of distribution maps as interpretative tools, because of one not knowing that one’s mapped the right things, but this is a negative correlation I’m trying to show, so I think it’s OK. Because they don’t match, do they? I mean, they overlap, quite a lot, but the stones’ heartland is further north than the ‘Pit-‘ heartland, and there are plenty of stones out further west where these place-names don’t exist. I think it is clear that the stones and at least one of the languages used by people who were called Picts by outsiders were not confined to matching areas.

Map of distribution of Class I and Class II symbol stones, from Martin Carver's Surviving in Symbols, p. 19

So okay. You’ve heard me on this before, but really, the only thing that identifies a culture group as Pictish in this area is the stones and this symbol library. Since the label Pictish appears to cover multiple material culture groups, and also multiple and intercut linguistic strata, I think the meanings the stones carried must have been readily translatable. We don’t have the kind of evidence for bilinguality that would allow it to be a question of an élite imposing its language over the top of the local ones, as far as I know, but I could easily be wrong about that. In neither of these cases would I be surprised that the Pictish symbols could be explained in linguistic terms. I would just be rather surprised if it was always the same language, and I don’t think this paper gets us any closer to knowing.

---

1. Rob Lee, Philip Jonathan and Pauline Ziman, “Pictish symbols revealed as a written language through the application of Shannon entropy” in Proceedings of the Royal Society A: Mathematical and Physical Sciences, published online ahead of print March 31 2010, not yet in print but available for free here, cited here from separately paginated PDF download.

2. Critical testing by Warren Esty, “Estimation of the size of a coinage: a survey and comparison of methods” in Numismatic Chronicle Vol. 146 (London 1986), pp. 185–215; polemic deconstruction in Theodore V. Buttrey, “Calculating ancient coin production: facts and fantasies”, ibid. 153 (1993), pp. 335–51. Accessible discussion, I hope, in J. Jarrett, “Digitizing Numismatics: getting the Fitzwilliam Museum’s coins to the world-wide web” in The Heroic Age Vol. 12 (2009), online here, §§8-12.

3. Lee, Jonathan & Ziman, “Pictish symbols”, pp. 11-13.

4. Best discussed, even now, in W. J. Watson, The History of the Celtic Place-Names of Scotland, being the Rhind Lectures on Archaeology (expanded) delivered in 1916 (Edinburgh 1926), but see now and more briefly Katherine Forsyth, Language in Pictland: the case against non-Indo-European Pictish, Studia Hameliana 2 (Utrecht 1997), online despite fearsome copyright infringement warning here, pp. 24-25, who argues following Simon Taylor that these names are in fact post-Gaelic. This may be clearer in Taylor’s paper that she cites but would take me some brain-bending to accommodate meanwhile. Where are the Ps coming from? Why do they appear to match the distribution of Class II stones and long-cist burials, both east-centred, so nicely if they’re coming from the west? and so on.

5. It’s terribly old in terms of interpretation, but the best place to see these inscriptions drawn out is John Rhys, “The Inscriptions and Language of the Northern Picts” in Proceedings of the Society of Antiquaries of Northern Scotland Vol. 26 (Edinburgh 1892), pp. 263-351, online here. For interpretation, see Forsyth, Language in Pictland, pp. 31-36 and Pll. 1-4.

6. Bede, Historia Ecclesiastica Gentis Anglorum, I.1; Adomnán, Vita Columbae, I.26. On the latter see David N. Dumville, “Primarius Cohortis in Adomnán’s Life of Columba” in Scottish Gaelic Studies Vol. 13 (Glasgow 1978), pp. 130-131; it is important to emphasise that Adomnán does not say the man to whom Columba could not speak was a Pict.

7. Forsyth, Language in Pictland, is one long argument against Jackson’s theory, which was presented in “Kenneth H. Jackson, “The Pictish Language” in Frederick T. Wainwright (ed.), The Problem of the Picts (Edinburgh 1955), pp. 129-166.

8. J. Jarrett, “The Political Range of Áedán mac Gabráin, King of Dál Riata” in Pictish Arts Society Journal Vol. 17 (2008), pp. 3-24 at pp. 14-16, largely on the basis of Watson, Celtic Place-Names, pp. 108-113.

9. As far as I can tell Isaac has not published this material, which is frustrating; I heard the pitch as “Some observations on the linguistic geography of ancient and medieval Scotland”, paper presented to EMERGE 2003 Conference, University of St Andrews, 11th September 2003.

10. Martin Carver, Surviving in Symbols: a visit to the Pictish Nation, The Making of Scotland 5 (London 1999, repr. Edinburgh 2005), p. 20, presumably referring to: A. Charles Thomas, “The Interpretation of the Pictish Symbol Stones” in Archaeological Journal Vol. 120 (London 1963), pp. 31-97; Anthony Jackson, The Symbol Stones of Scotland (Orkney 1989); and Ross Samson, “Power to the Pictish ladies” in British Archaeology Vol. 3 (York 1995), online here without pagination.

11. Lee, Jonathan & Ziman, “Pictish Symbols”, p. 11.

12. Ibid.