The following online article has been derived mechanically from an MS produced on the way towards conventional print publication. Many details are likely to deviate from the print version; figures and footnotes may even be missing altogether, and where negotiation with journal editors has led to improvements in the published wording, these will not be reflected in this online version. Shortage of time makes it impossible for me to offer a more careful rendering. I hope that placing this imperfect version online may be useful to some readers, but they should note that the print version is definitive. I shall not let myself be held to the precise wording of an online version, where this differs from the print version.
Published in The Coat of Arms n.s. vol. XIV, nominally 2001 but actually 2002, pp. 41–58.
Prof. G.R. Sampson
Heraldry creates its effects through contrasts of colour and shape. In an earlier article I analysed historical trends in choice among the colours available in the heraldic palette. The present paper looks similarly at trends relating to shape – changing preferences among ordinaries, charges, and ways of partitioning the shield surface.
It is easy to guess at some trends without doing research. The roots of heraldry lay in warfare; accordingly, early heraldry was full of lions. It would be no surprise to find that this particular beast became less frequent in later grants, as the range of professions pursued by those granted arms broadened beyond the military – and we shall see that this is so. More interesting, to my mind, are historical trends which are less predictable because they relate to abstract issues, such as choice among ordinaries. We shall see that some of the strongest historical trends are quite difficult to account for.
As in my paper on tinctures, in order to explore the topic I began by forming samples of coats granted at three periods – the middle ages, the 19th century, and recent decades. The mediaeval and 19th-century samples were selected in essentially the same way as for the tinctures paper (though, since I had kept no record of the previous samples, the individual coats chosen will be different). For recent grants, because ordinaries and charges are far more diverse than tinctures, I needed a larger sample than was used for the earlier study; as well as the sources used previously (the last few years of The Heraldry Gazette and The Coat of Arms) I also drew on The Armorial, vols. 1-3, 1959-62, and a sampling of grants from The Armorial Who Is Who, 1976-78. Thus my recent sample comprises coats granted between 1945 and about 1998, though with the earlier and later decades of that period better represented than the middle part.
To make the three samples as comparable as possible, I limited them to grants made by authorities in the British Isles (in the recent sample I included only grants made in London, Edinburgh, or Dublin – though many grantees resided overseas); and, since mediaeval heraldry was almost entirely personal, in this study (unlike the tinctures study) I excluded institutional arms from the later samples. As in the earlier study, I tried to avoid coats which were essentially just differenced versions of older arms (and I excluded coats which were identified as "matriculated", rather than "granted", at the Court of Lord Lyon). But of course a large proportion of coats involve some degree of graphic allusion to earlier arms – I excluded only the most obviously derivative cases, such as quartered arms.
To make comparisons between the samples meaningful, it is necessary to categorize the elements fairly coarsely, ignoring minor differences and incidental trappings. So for instance a lion, irrespective whether rampant, passant, with tail fourchee, etc., is for present purposes just a lion. Diminutives are not distinguished from full-size elements: fess, bar, or barrulet are all counted as "fess". If a lion is blasoned as having a crown or a collar, I count the lion but not the crown or collar – for heralds (if not for zoologists) these are normal accoutrements of lions; though if a lion holds some charge in its paw, I count both the lion and what it holds, because there is no standard thing that lions hold in their paws. When a beast stands on a mount or a rock, the beast is counted but the mount or rock is not.
Importantly, also, I did not separately count multiple instances of the same element in a coat. Some items (e.g. martlets) almost always occur in groups, other elements (e.g. the chief) occur singly, but that in itself does not make the martlet a more usual charge at any given period than the chief. What we are interested in is how often particular types of element were selected for inclusion in a coat, not how many of them were used together. Thus my counts are not counts of individual graphic elements, but rather counts of coats containing (one or more examples of) particular elements. The coats of France Ancient or France Modern would each in my system score one for the category "fleur de lys", the same as a coat containing a single fleur de lys.
To make the counts comparable, the number of coats in each sample was chosen to make the total number of graphic elements counted in each sample close to 300 (the mediaeval sample contains 302 elements, the 19th-century and recent samples each contain 304 elements). Because early coats were simpler than later coats, the numbers of coats in the samples are unequal (there are 190 mediaeval, 97 19th-century, and 94 recent coats).
People sometimes ask questions like "Are 300 cases are enough for a representative sample?", but statistically speaking there is no general answer to that kind of question. A particular sample size will be more than large enough to reveal a major change in the incidence of some feature between the periods compared, yet not large enough to demonstrate an equally real but lesser change in some other feature. In practice, one gathers samples as large as time and energy allow, and then checks which differences between the samples are "significant" (meaning that there would be a low probability of getting such a skewed distribution merely as a random sampling effect, if the underlying frequency of the feature in question were actually unchanged). Conventionally, a probability less than one in twenty is taken as establishing significance – in the present context, a genuine change in heraldic style between historical periods.
The tables in this paper show frequencies of various features in the three samples. For each feature I compared the frequency in each sample with the frequency of the same feature in the other two samples taken together. Where the statistical test shows a significant difference (with a probability less than one in twenty of occurring by chance) between the frequency in one sample and that in the other samples for the same feature, I mark the figure for the distinctive sample with an asterisk. Two asterisks mean a very significant difference, probability less than one in a hundred; three asterisks an extremely significant difference, probability less than one in a thousand. A figure without asterisks means that the difference between it and the other figures in the same row is not significant. (Even if that difference appears large, it might well be merely a sampling effect.)
Thus an asterisked figure in the "mediaeval" column means a change between middle ages and 19th century; an asterisked figure in the "recent" column means a change between 19th century and today; asterisked figures in both of those columns means a change that has continued in the same direction over both intervals. If a 19th-century figure is asterisked, the trend of change must have altered between the earlier and later intervals (otherwise the 19th-century figure would be intermediate between the other two figures, and hence not significantly different from their average).
Geometric v. Iconic Elements
The graphic elements used in heraldry are endlessly diverse but, for statistical purposes, we can group them into logical categories. The most fundamental classification is between geometric elements and elements which are iconic (that is, which depict the shape of some physical thing). Geometric elements include ordinaries such as the bend, geometric charges such as roundels (bezant, hurt, etc.), and also partitions (cases where a field – or an element within the field – is party per bend, party per pale, chequy, etc.).
The distribution of elements across these broad categories is as follows:
There has been a very marked and continuing reduction from the middle ages to the present day in the use of geometric elements, and a corresponding increase in use of iconic elements. Evidently mediaevals were concerned chiefly with visual distinctiveness, and found abstract geometry an adequate way to achieve it, whereas later ages have increasingly wanted to see actual things portrayed on their coats – usually, no doubt, things having symbolic significance in their lives.
Although the contrast between geometric and iconic elements seems basic, it is noticeable that in the recent sample it is beginning to break down. There have always been one or two standard charges which partook of both categories, most obviously the mullet, which is named after a physical object (spur-rowel) that happens to have a simple geometric shape. The high frequency of the mullet as a charge reflects its geometric simplicity – rowels are not very important things even in horse-borne societies; but the fact that mullets are often shown pierced is an iconic feature – rowels have spindle-holes. (In this paper I have counted the mullet as a geometric rather than iconic charge.) In the recent sample, though, there are many other one-off cases of unusual charges defined in geometric terms that are intended as stylized representations of physical objects.
Thus, Baron Nolan (created 1994) has a coat including three "bars wavy couped composed of two troughs and a wave invected of one point on the upper edge and engrailed of one point on the lower edge", and Baron Smith of Clifton (created 1997) has four "fusils … per pale argent and or per chevron counterchanged" – the former representing open books seen end-on, and the latter representing flints found at an archaeological site. Should these be counted as geometric or iconic charges? I have followed the language of the blasons and counted them as geometric, but the truth is that they are intermediate. This new trend is clearly influenced by the stylized logos of present-day graphic design.
An interesting case concerns a pattern of regularly-spaced vertical and horizontal stripes, not shown laced over and under as in square fretty but continuous – what in ordinary speech would be called a grille. This occurs in arms granted in 1990 to Anthony Mallard, where it is intended to symbolize a portcullis, but since "grille" is not a traditional heraldic term it was then blasoned as "four pallets … joined to six barrulets". By the time of the grant to Baron Lofthouse of Pontefract (created 1997), the same patterns are blasoned straightforwardly (but innovatively) as "grilles". It would be artificial to count the former coat as including pallets and barrulets and the latter not; I have counted the "grille" as a new type of partition, alongside fretty and chequy, and treated both these coats as exemplifying it.
Categories of Geometric Element
Geometric elements can be divided into ordinaries, geometrical charges, and partitions. I shall not draw arbitrary distinctions between ordinaries and "sub-ordinaries": I count an ordinary as any element which is defined in purely geometric terms, one or more of whose edges coincide with the edges of the field, and which occupies a minor proportion of the field, creating a "figure versus ground" appearance. A partition is a geometric division of the field into roughly equal areas. A geometric charge, in my terms, is a geometric element normally placed within the field, surrounded by it on all sides. (Thus I count a plain cross, with arms extending to the edges of the field, as an ordinary, but a cross crosslet, cross moline, cross flory, etc. as a geometric charge.) In most cases this classification is straightforward.
A problem does arise in the case of fields divided into stripes – barry, bendy, etc. Should a bendy coat be counted as a field containing multiple bends, which are ordinaries, or as a field with a particular type of partition, like a chequy field? Modern English heraldry makes a distinction between cases where the number of stripes is odd, which are treated as multiple ordinaries on a field, and cases where they are even, which are treated as partitions. But this distinction is not recognized by the heraldic traditions of other nations, and it is post-mediaeval: mediaeval heraldry was often unspecific about numbers of stripes, and the same coat can sometimes be found blasoned alternatively as, say, "barry" or "three bars".
It is vital in the present exercise not to make graphic distinctions which are more refined than mediaeval heraldry itself made, otherwise numerical comparisons between mediaeval and later periods would be meaningless. Consequently I have counted all these cases, irrespective of numbers of stripes, as cases of multiple examples of an ordinary, rather than as partitions. Counts for "fess" in what follows include cases of barry fields, for instance.
The breakdown is then as follows:
The greater use of geometry by mediaeval heraldry was, specifically, a far greater use of ordinaries; and in recent decades geometric charges have also declined in favour.
The decrease in use of ordinaries after the middle ages did not affect all ordinaries alike. The figures for specific ordinaries are:
The simple ordinaries consisting of one straight stripe crossing the shield in different directions, pale, fess, and bend, occur in the 19th century with roughly equal frequency; but in the middle ages, the fess in particular, and also the bend, occurred at much higher frequencies than in the 19th century, and far higher frequencies than the pale.
The steep decline in frequency of the fess between middle ages and 19th century is astonishing. The significance statistic for the difference between these two periods is massive, much larger than any other test statistic in this research and well off the scale of the statistical tables available to me. In other words, it is sure beyond a peradventure that heraldic taste really did swing strongly against the fess in the 19th century; and this is all the more surprising since the pendulum has swung much of the way back in the subsequent century. The bend, on the other hand, has continued to decline significantly in popularity.
If I had thought about the matter before doing the research, I might have surmised that pale would be commoner than fess in the middle ages, with their emphasis on the chivalrous virtues: an upright posture surely has nobler connotations than a supine posture. In fact, the pale itself was even rarer in the mediaeval sample than the above figures show: the total of three represents one pale and two paly fields (whereas in the corresponding 19th-century figure, and in the fess and bend figures for both periods, the large majorities of cases represent the individual ordinaries rather than the varied fields – paly, barry, bendy – produced by multiplying the ordinaries.)
Establishing the numerical facts is one thing; it is another to explain them. One possibility might be to explain the decline in popularity of the fess in the same way as that of the lion. A horizontal stripe is a psychological barrier; like the modern road sign for a toll booth, it says to an advancing enemy "you stop here" – whereas a vertical pole is no hindrance. But while this could perhaps account for the special frequency of the fess in the middle ages, it does not explain why the bend too is far more frequent than the pale; a diagonal surely has no relevant psychological connotations. If one compares the mediaeval figures with each other rather than with the corresponding 19th-century figures, it is really the low figure for the pale that calls for explanation more than the high figure for the fess.
The explanation I tentatively lean towards has to do with the physical shape of shields. These were often rounded from side to side, to encourage blows to glance off, and sometimes even had a crease down the centre. This perhaps meant that a pale would not stand out visually unless the shield was seen square on; seen from an angle, a pale might blend into the curvature of the surface, whereas fess or bend would be plain for what they are at any angle. In the modern period of paper heraldry, on the other hand, fess and pale are the most straightforward orientations for stripes, and a diagonal is a less obvious choice.
If this is the correct explanation for the rarity of pales in the middle ages, it seems to imply that a partition per pale on the other hand would work well – a central crease in the shield would divide it into two wings whether or not they were coloured differently. We shall see below that partition per pale is indeed commoner than per fess in the mediaeval sample, though the reverse is true of the 19th-century sample. This perhaps offers some support for the shield-shape theory. I must admit, though, that this theory does not seem wholly satisfactory – for instance, it does not explain why 19th-century heralds moved so decisively away from the fess while continuing to use the visually more complex chevron at mediaeval levels.
Another development, the post-mediaeval decline in the use of crosses, is less marked than the cases of fess and bend; but it is easier to explain, in terms of the fading after the middle ages of the relationship between heraldry and the Crusades.
Two significant developments in recent heraldry are the decline of the saltire, and the rise of flaunches. The changing fate of the saltire is perhaps to be explained in the same way as that of the bend (and the Scotch connexion has evidently not been sufficient to shore up its popularity – or perhaps the saltire is nowadays seen so much as a national emblem that it would be cheeky for private individuals, even Scots, to use it). The only explanation I can suggest for the sudden trend in favour of flaunches is that, because the recent sample represents a relatively short historical period, its figures could be influenced by the individual tastes of particular heralds.
The figures for partitions are:
In this and later tables, rather than giving separate figures for every possible category (including those too infrequent to show any definite historical trends), I lump together as "others" all categories which are not represented at least three times in one of the samples. This will become particularly important in connexion with iconic charges, where the tables would otherwise grow very cumbersome. In the case of partitions, "others" covers cases of partition per bend, per saltire, gyronny, per pall, compony, and "grille" discussed earlier.
The only significant mediaeval/19th-century contrasts are the great increase of partition per fess, and matching decline in quarterly coats. I have attempted, I believe successfully, to exclude from the samples quarterly coats deriving from marriages, and to include only cases where a unitary coat is party per cross; perhaps the decline in numbers of the latter happened because people were afraid that such coats would be mistaken for the former. I cannot guess why partition per fess should have grown significantly in popularity in the 19th century (and fallen back again since). There is a link with the increased popularity of the pale – three of the 19th-century coats have the pattern "per fess a pale counterchanged", to give a six-piece partition. But even without these cases the increase in partitions per fess is large.
The most significant point in this table is that partition per chevron is far commoner now than before. Although the change is striking, it does not seem to be explainable as an individual herald’s preference. The coats in question were granted at different times over the last fifty years and in several cases are known to have been designed by different people.
Non-Field Partitions and Fancy Lines
Most types of partition can apply either to the field, or to individual ordinaries and charges placed on the field. (One exception is compony, which applies only to stripe-shaped ordinaries.) Partitioning of the field is much the commonest case, and the mediaeval sample has only such partitions (if we ignore two compony elements, for the reason just given). But the 19th-century sample has eight cases, and the recent sample four cases, of partitions applied other than to the field: for instance one 19th-century coat, for Pidcock, includes "a pied cock per fess or and argent". The nil mediaeval figure, and the high 19th-century figure, are both significant.
I also examined whether the outlines of ordinaries and partition lines were plain or "fancy" (for want of a better word). The great variety of distinctions now recognized among lines, e.g. wavy, engrailed, invected, indented, etc, were not observed in the middle ages; but there seems always to have been at least a contrast between plain and fancy lines. So I classified outlines in that simple two-way fashion. (Cotises by the side of an ordinary were counted as one kind of fancy outline, and a saltire "parted and fretty" was counted as a fancy saltire. I counted things like "fusils conjoined in fess", which in the middle ages were often interchangeable with an indented fess, as a fancy fess, rather than as a set of fusils.)
I expected that fancy lines might have become commoner at the later period, with the need to differentiate new grants from an ever-growing body of existing coats. However, that was only marginally true. The "fancy" cases numbered 31, 28, and 35 in the successive periods. These figures do imply a lower proportion of fancy outlines in the mediaeval sample, because that contained quite a lot more geometric elements capable of taking fancy outlines. But the proportional differences are small. Fancy outlines as a proportion of "fanciable" elements are 22% (mediaeval), 27% (19c), and 31% (recent); although this looks like a steady rise, the differences are not statistically significant.
Geometric charges break down as follows:
|cross (as charge)||25*||12||14|
We have already seen a possible reason for declining use of crosses. However, the significant post-mediaeval decrease in crosses as charges relates entirely to crusilly fields, rather than to single crosses. The mediaeval sample contains eleven crusilly fields, the two later periods none – a very significant difference. Crosses as individual charges on the other hand occur at essentially at the same frequency in each sample. (Could this be because many types of cross, e.g. cross moline or cross flory, were never identified with the Christian cross in the first place, so that their frequency was not affected by changes in the role of religion in society?)
Since lozenge and mascle (which were not always distinguished in mediaeval blason) are treated in the table as a single category, it may seem illogical to count the annulet separately rather than as a (voided) roundel; but it seems to me that the annulet is usually thought of as a quite separate charge, unrelated to the roundel. It is notable (but mysterious to me) that both crescent and annulet were far more popular in the 19th century than before or after. (I believe that none of the items in my counts were cadency marks.)
Iconic charges fall naturally into the following main classes:
|inanimate natural objects||0||2||6*|
Having already seen that later heraldic periods have increasingly favoured things over abstract shapes, we expect the figures to increase from left to right; but for what was originally by far the highest-frequency iconic category, "fauna" (animals, birds, fish, etc.) the frequency has remained stable. It is among the less-frequent categories, flora and (particularly) inanimate artefacts and natural objects, that increases have been proportionately large.
The figures for "fauna" can be analysed in two different ways.
One is according to whether the whole animal, or just a head (or other body part) is used:
(I classed an occasional demi-lion as a whole lion, since it seems closer in spirit to that than to a head alone.) In the mediaeval sample, heads are the only body parts used. Five each of the 19th-century and recent cases are other body parts, e.g. lion’s or bear’s gamb, feather, ermine spot; but the mediaeval figure would still be significantly low even if those elements were excluded from the figures for the later samples.
The rarity of animal heads in mediaeval heraldry was unexpected to me. It is also interesting that although (as we shall see) lions account for more than half of all beasts in the mediaeval sample, none of the eight mediaeval heads is a lion’s head. One, in the arms of a 14th-century William Gascoigne, is a lucy’s head: evidently mediaeval heraldry was not restricted to "noble" charges, if it was prepared to use a coarse fish-head.
The other way to analyse the fauna figures is by species:
|martlet, other birds||6||12||9|
|stag, buck, hart||1||4||2|
As surmised earlier, the shift away from the single-minded mediaeval focus on warfare is reflected in a significant and continuing decline in use of the lion.
The representation of human beings has become significantly commoner in the recent period (though none of the samples include any whole humans – the total of six in this row comprises three cases of hands, and one each of arm, heart, and Moor’s head).
A further significant change, to my surprise, was that mythical beasts were completely absent from the mediaeval sample. I had imagined that they might be relatively frequent at the early period, considering that the boundary between myth and reality was blurred in the middle ages. Again, though, this can be explained in terms of changing concerns of society between the respective periods. Except for one unicorn, the 19th-century mythical beasts were all griffins, which in classical mythology were guardians of treasure; thus a griffin makes a suitable charge for a banker (cf. the logo of the late Midland Bank). Clearly, financiers were prime candidates for grants of arms in Victorian times but not in the feudal period; I have not attempted to check the professions of the six griffin owners in my 19th-century sample, but I surmise that most of them were money men. (Some of the recent mythical beasts are dragons chosen for their Oriental associations.)
The other significant change has been a much greater use of diverse species after the middle ages. This is not, as one might suppose, because exploration and Empire had created an awareness of species that were unknown in the middle ages; only one of the 19th-century "other fauna" (flying fish), and very few in the recent sample, appear to come under that heading. Post-mediaeval heralds just seem to have been more inclined to draw on a wider range of well-known fauna, such as fox, horse, ram, bull.
Flora break down as follows (I group trefoils, quatrefoils, and cinquefoils together as "N-foils"):
|fleur de lys||6||7||10|
The marked recent increase in use of roses is perhaps a surprise. (A number of the coats in my recent sample were grants to New Labour peers; in at least one case a red rose was used for that reason.) Apart from that, after the middle ages heralds drew on a greater diversity of flora as they did of fauna. In the recent (though not the 19th-century) sample, "other flora" do include a number of introduced or exotic species (e.g. narcissus, New Zealand cabbage tree) alongside familiar traditional items such as oaktree or bunch of grapes. But the main difference between mediaeval and later use of botanical charges seems to be a move away from generic towards specific items. A quatrefoil or cinquefoil, and probably even a fleur de lys, represents a generalized flower rather than the flower of a particular plant species; but then, a mediaeval warrior might not have known one flower from another – the great burgeoning of interest in natural history lay far in the future. By the time of the recent sample, it has become commoner to show identifiable species than generic flowers.
The figures for artefacts are:
|wreath, chaplet, civic crown||2||4||2|
It is ironic that the two most frequent objects in the mediaeval sample, maunch and water-bouget, were things whose outlines became unrecognizable in later periods; so it is natural that their frequency fell. Apart from them, later periods saw a much greater propensity to use artefacts as charges, with a greater variety of objects used.
The difference between mediaeval and post-mediaeval figures for weapons (a category I use to cover charges such as sword, seax, pheon, battleaxe, arrow) are large enough to suggest a specific trend. (If we had set the arbitrary threshold of statistical significance as one in ten, as some do, rather than one in twenty, then the mediaeval weapons figure would have been asterisked.) At first blush it might seem strange that mediaeval armigers, whose careers were predominantly martial, should have made relatively little use of weapons as charges. But consider that a present-day delivery company does not make its fleet of vehicles look distinctive by painting pictures of delivery vans on the side; it chooses some quite different emblem, perhaps a tulip or a sundial. The very fact that mediaeval coats of arms were part of real fighting equipment perhaps made weapons too undistinctive to use as charges. In later centuries, only a minority of grantees will have had military backgrounds, and that minority will often have wanted their coats to reflect this.
By far the most significant change under the "artefacts" heading has been the marked and continuing increase in the variety of artefacts found in the samples. Of course, by the 19th century many things had been invented which the mediaevals knew nothing of, and one might imagine that the great growth under "other artefacts" reflected modern heralds using possibilities that were unknown to their mediaeval predecessors.
Overwhelmingly, that is not the case. One coat in the recent sample does include a "computer ferrite core", but this is very unusual. Almost all the 19th-century and recent "other artefacts" are objects such as buglehorn, cup, lymphad, horseshoe, which would have been as familiar (or more familiar) in the middle ages as at the time of the grant. For the 19th-century sample, the sole possible exception I find is a fasces (I am not sure how well known this classical symbol would have been to mediaeval Englishmen).
Nor is it the case that the increase occurred because later periods used "trivial" objects whereas mediaeval heralds limited themselves to artefacts that played lofty roles in life. There is nothing very noble about sleeves, or cushions (which account for two of the mediaeval "other artefacts"). Looking through the nineteen "other artefacts" in the 19th-century sample, I recognize at least half of them, probably more, as occurring in mediaeval coats I have encountered outside the limited random sample used in this study.
The large numerical differences in the "other artefacts" row are not really explainable in terms of later heraldry using new things, or different sorts of thing. The best explanation for these figures, I believe – and indeed for the "other" figures for fauna, and to some extent those for flora, although the move from generic to specific charges complicates the latter case – is that earlier heraldry was more strongly swayed by fashion than later heraldry.
We saw something like this in my paper on choice of tinctures. All the heraldic colours were already available in the middle ages: but, predominantly, mediaevals went for gules – that highest-frequency colour was very frequent, lower-frequency colours had much lower frequencies. At later periods, although frequency differences remained, they diminished, and gules became just one colour among the others.
As with tinctures, so with charges. Most (of course not all) the charges that have ever been used were already used, at least occasionally, in the middle ages; but at that period a few beasts and artefacts – lion, eagle, maunch, water-bouget – were very frequent, less common charges were very much less common so that most of them are not represented in my sample at all. At later periods, frequency-differences became narrower, so that a longer "tail" of charges were each frequent enough to score one or two instances in my samples. Over the centuries, heraldry became a less conformist and more diverse art.
Inanimate Natural Objects
Finally, we have already seen that there has been a significant recent increase in use of inanimate natural objects. The two cases in the 19th-century sample are both goutty coats; the six recent cases also include fountain, sun, and flame. But the most notable fact about this category is perhaps how little use has been made of it at any period. It is easy to think of inanimate objects – rainbow, waterfall, snowflake – which might seem made for heraldic purposes, but which have rarely or never been used.
Overwhelmingly, it seems, heraldry throughout its history has combined
the pure abstraction of geometry with representations of things which are
either alive or closely connected with human life (as artefacts) –
almost wholly ignoring the middle ground. This combination of opposites,
I believe, is a factor that gives heraldry much of its charm.
 "Historical trends in the deployment of tinctures", The Coat of Arms, N.S. vol. XIII, pp. 271ó7.
 Ed. by Lt. Col. Gayre of Gayre and Nigg; published by The Armorial, Edinburgh.
 The Editor has pointed out to me that if my sampling for the tinctures study had happened to include many arms of local authorities, the statistics for Vert would have been inflated, since these bodies almost always want coats including green for its rural associations.
 On the other hand, in a case such as the coat of Miller of Acre Valley, Stirlingshire (Burke’s General Armory, 1884 ed., Supplement p. cxv), which contains both a cross moline in the field and a chief bearing a set of charges including a cross patee, I count two crosses, because the crosses are different and represent independent choices of charge.
 I used the chi-squared test with Yates's correction for continuity; see e.g. S. Siegel & N.J. Castellan, Jr., Nonparametric Statistics for the Behavioral Sciences, 2nd ed., McGraw-Hill, 1988, p. 116.
 Note that I applied significance tests only "horizontally", to check for changes between different periods. I did not apply them "vertically", to differences between frequencies of various ordinaries and charges at a given period. The latter would have been mathematically more complicated, and it would have been less meaningful because of interactions between the elements; for instance, if a field contains a plain cross, it may also contain a chief but is unlikely to contain a chevron.
 See p. 94 of Peter Gwynn-Jones, "Heraldry of new life peers", The Coat of Arms, vol. xiii, 1999-2000, pp. 45-57, 92-6, 186-9.
 Gwynn-Jones, op. cit., p. 56.
 Heraldry Gazette, June 1994, p. 2.
 Gwynn-Jones, op. cit., p. 51.
 Boutell’s Heraldry, revised by C.W. Scott-Giles, Frederick Warne, 1950, p. 20.
 One of the six cases of flaunches is a grant for which Peter Gwynn-Jones, now Garter King of Arms, was heraldic agent, and three others are taken from his article cited in note 7 – it is a reasonable guess that he chose to write about coats which he helped to design.
 I do not distinguish fret as charge from fretty as a method of partitioning a field; the distinction was not, I believe, made consistently in mediaeval heraldry.
 That is, a vair pattern in tinctures other than argent and azure. The incidence of furs, including vair, was discussed in my paper on tinctures.
 One of these two cases is the famous coat of Mortimer, which I have counted as a bordure compony – though it is standardly blasoned quite differently, as "barry of six or and azure, on a chief of the first two pallets between two base esquires of the second, over all an inescutcheon argent". That may be the best that the language of English blason can do, but it seems unsatisfactory as a definition of the Mortimer coat: first, because someone who had not seen the coat would scarcely be able to reconstruct it recognizably from this wording, and also because it fails to express the visual coherence of the design, treating it as an unrelated collection of ordinaries. The spirit of this coat is better captured, I believe, by treating it as a bordure compony, though one in which the pieces are disposed in a particular pattern for which blason has developed no specific descriptive term.
 Not all geometric elements can have fancy outlines. For instance, chequy is not "fanciable" – one is not likely to see a chequy wavy field (However, it is rash to assume that anything is impossible in heraldry. A few days after doing the calculations presented here, which assumed that the fret is not a "fanciable" ordinary, I discovered that various members of the M’Culloch family bear a fret engrailed.)
 Mediaeval heraldry did not consistently distinguish these.
 I count only continuous stripes as orles. An "orle of martlets" is counted under "martlet", below, and not under "orle".
 Namely inescutcheon, billet(y), and (in the recent sample) ellipse and pentagon.
 Gascoigne's coat also contains the only pale in the mediaeval sample. Perhaps Gascoigne was one of life's non-conformists.