The so-called rule of tinctures is usually stated as follows: "colour cannot be placed upon colour, nor metal upon metal." Fox-Davies adds: "This is a definite rule which must practically always be rigidly obeyed."
There are a number of standard exceptions:
In French, arms that seem to violate the "rule of tinctures" are called "armes à enquerre", because they were supposed to prompt you to inquire about their origins. French heraldry uses the term "cousu" (literally, sewn) to get around an apparent violation by an ordinary (most commonly a chief). For example, Sandberg (Netherlands) as blazoned in Rietstap is: d'argent au chevron cousu d'or accompagné de trois trèfles de sinople (Argent, a chevron or between three trefoils vert). The use of the word "cousu" is common for chiefs which are in fact augmentations of honor. Many French cities have a chief of France in this manner. But there are other examples where no augmentation is involved: Belot in Franche-Comté (Argent three lozenges azure a chief embattled or), Surville in Ile-de-France (Gules a cross bottony couped argent and a chief azure), de Maumigny (Argent a chevron sable and a mullet of five points gules in base, a chief or), Robinet de Cléry (Azure a chevron or and a rose in base argent, on a chief gules three mullets of five points argent). In each case the chief is called "cousu".
At this point, the exceptions are so numerous that one wonders whether the rule is really meaningful. It is, however: for it will apply to such blatant examples as Azure a lion gules or Argent a chevron or. Such arms should, in principle, not exist.
The rule's history has been studied by Bruno Heim (Or and Argent; Gerards Cross (Buckinghamshire), 1994; Van Duren). The first heraldic treatises date from the 14th c., and the earliest known is the Dean Tract, also called "De Heraudie". Brault dates it to around 1340 (Early Blazon, p.xxi). The text states: Le Roy de Jerusalem porte l'escu d'argent croiselee d'or a une croise potente d'or. Et si avient malement colour d'or en argent (the king of Jerusalem bears on a shield argent crusily or and a cross potent or; ...). This is not the statement of a rule, however. Furthermore, a number of 13th and 14th century rolls of arms or other heraldic documents blazon arms which violate it without any particular notice. Other 14th c. treatises like Bartolo ca. 1350 and Johannes de Baudo Aureo ca. 1390 make no mention of the rule.
The Argentaye tract (ca. 1410), the Liber Armorum of Bernard du Rosier (ca. 1440), the Blason des Couleurs of the herald Sicile (ca. 1440-50) and many subsequent treatises state the rule, always citing the arms of Jerusalem as an exception justified by the exceptional nature of the kingdom. Although important texts such as Guillim's Display of Heraldry in the 17th c. make no mention of it, it tends to become a staple of heraldic treatises. By the 19th century, it has reached the status of a "strict law" (Boutell) or "definite rule" (Fox-Davies) or "primary fundamental canon of Heraldry" (Woodward, who nevertheless documents violations).
It is far from being absolute, or even a rule. And as one moves away from France and England, the rule seems to apply less and less: Spain, Germany, and especially Hungary offer numerous instances where the rule is ignored, even in the 19th century. The Netherlands offer an egregious example in the arms of the city of Amsterdam: Gules on a pale sable three saltorels argent. Violations usually occur with color-on-color, though examples of metal-on-metal, albeit rare, can be found in any country.
Arms of van Geen (Netherlands), granted 1832: Azure on a bend sinister vert three bugle-horns in bend sinister or and in dexter chief a sword paleways of the last.
Pastoureau studied a database of about 10,000 arms from 13th-15th c. armorials, which covered only Western Europe. He claims that only in Castille and the kingdom of Grenada does the frequency of violations rise above 2%. But his database covers the traditional areas of heraldry, and not, for example, Hungary.
Bruno Heim, whose personal arms violate the metal-on-metal rule, has had a life-long interest in the matter, and collected many examples of violation over time. They are presented in his book Or and Argent. He claims to have found 1500 violations of the metal-on-metal rule in Rietstap's General Armory (which contains 120,000 arms). That's a rate of about 1.2%. Heim's violations, however, can often be classified as standard exceptions to the rule. Moreover, it is often hard to say if the examples he collected are not the result of errors.
The Jerusalem arms are the best-known violation, but there are many others. In fact, Rietstap cites 6 families bearing Azure a lion gules (Avila in Castile, Bemdorf in Saxony, Betti in Florence, Christol in Toulouse, Molnet in Lorraine and Strodl in Erlangen) and one bearing Argent a chevron or (Morelle in Holland).
Here are some of Woodward's examples, and a few more culled from Rietstap:
Writers such as Woodward recognize the weakness of the rule in Continental heraldry, but insist that it is better observed in England: "in our own country, distinct violations are of great rarity". But violations can also be documented in England.
John Gibbon's Introductio ad Latinam Blasonam (1682, p. 152) lists a few British exceptions to the tincture rule. I was able to check them against Burke's (1844 edition). Some are confirmed as such (with the violation), some are "corrected", some are absent. I give the list, with Burke's variations between [brackets]. The only metal/metal case does not appear in Burke (Grendon), all others are color/color. Gibbon argues that the rule shouldn't be emphasized as much, since it is already violated for chiefs, labels, bordures (and France's Anjou arms, Ancient and Modern, each give an example of the latter). Gibbon's sources are either Rolls or windows in churches.
Bruno Heim cites a number of British coats violating the metal-on-metal rule. Some are from medieval rolls, others from more recent sources; some appear in Burke but corrected. In some cases, I suspect that an error was made at some point, and that no violation occurs, but some seem to be bona fide examples:
Heraldry is a European-wide phenomenon that emerged rather spontaneously over a wide area, and which has generated hundreds of thousands of coats of arms over several centuries. With respect to that mass of coats of arms, our concern in studying heraldry is obviously not to edict rules which cannot be obeyed by so many dead people, but to see if there is some kind of consistent pattern or order that one observes. The word "rule" should be understood, not as a prescriptive code of conduct, but as in the phrase: "as a rule, things happen a certain way".
One such consistent pattern is the "rule of tinctures": call it an empirical regularity if you want. It suffers precious few exceptions, and a number of those exceptions follow a predictable pattern. For example, you will not be surprised to see a chief of color on a field of color in French heraldry, but you will be very surprised to see a bend of color on color, though it may happen. Pastoureau's figure is limited to Western European heraldry, but the man did sort through 10,000 coats of arms from the Middle Ages. Anybody got a larger database? When 99% of arms conform to a pattern, I am tempted to call it a rule.
What's the point of deciding whether there is a "rule" of tinctures? One that I see is this: suppose you come across a coat of arms, on a seal or a silver spoon or whatever, that violates the rule. Should you accept the testimony of the object at face value, orf should you allow for the likelihood of a mistake? This will guide you in your identification research. If it is a rule, albeit with recognized exceptions, then you will tend toward the second.
If you accept that usefulness, then it may be of use also to categorize the "exceptions" to the "rule" (deviations from the pattern, if you prefer). For that reason, the listing of exceptions is not an exercise in futility or semantics, but a genuine tool in understanding heraldry.