The description of the Tower by the Empress, in her charter, as "turris Londonie cum parvo castello quod fuit Ravengeri," and its similar description in Stephen's charter as "turris Lond[oniæ] cum castello quod ei subest," though at first sight singular and obscure, are fraught, when explained, with interest and importance in their bearing on military architecture.

It will be found, on reference to the charter granted to Aubrey de Vere (p. 180), that the Empress gives him Colchester Castle as "turrim et castellum de Colcestr[a]," a grant confirmed by her son as that of "turrim de Colcestr[a] et castellum" (p. 185 n.), and, in later days, by Henry VIII., as "Castrum et turrim de Colcestr[a]."[943] Further, in the charter to William de Beauchamp (p. 313), we find Worcester Castle described as "castellum de Wigorn[ia] cum mota," Hereford Castle being similarly described in the charter granted at the same time to Miles de Gloucester as "motam Hereford cum toto castello." Before proceeding to the inferences to be drawn from these expressions, it may be as well to strengthen them by other parallel examples. Taking first the case of Colchester, we turn to a charter of Henry I., granted to his favourite, Eudo Dapifer, at the Christmas court of 1101,[944] in which Colchester Castle is similarly described:—

"Henricus Rex Angliæ Mauricio Lond. Episcopo et Hugoni de Bochelanda et omnibus baronibus suis Anglis et Francis de Essex salutem. Sciatis me dedisse benigne et ad amorem concessisse Eudoni Dapifero meo Civitatem de Colecestrâ et turrim et castellum et omnes ejusdem civitatis firmitates Cum omnibus quæ ad illam pertinent sicut pater meus et frater et ego eam melius habuimus et cum omnibus consuetudinibus illis quas pater meus et frater et ego in eâ unquam habuimus. Et hæc concessio facta fuit apud Westmonaster in primo natali post concordiam Roberti comitis fratris mei de me et de illo.

"T. Rob. Ep. Lincoln et W. Gifardo Wintoniensi electo et Rob. Com. de Mellent. et Henr. Com. fr. ejus et Roger Bigoto et Gisleberti fil. Richard et Rob. fil. Baldwin et Ric. fratr. ejus."[945]

Turning to Hereford, we find its description as "mota cum toto castello" recurring in the confirmation by Henry II. and the recital of that confirmation by John.[946] There is another example sufficiently important to deserve separate treatment. This is that of Gloucester.

We find that, in 1137, "Milo constabularius Glocestrie" granted to the canons of "Llanthony the Second"

"Tota oblatio custodum turris et castelli et Baronum ibi commorantium."[947]

Here again the correctness of the description is fortunately confirmed by subsequent evidence; for John recites (April 28, 1200) a charter of his father, Henry II. (which is assigned by Mr. Eyton to the spring of 1155), granting to Miles's son, Roger, Earl of Hereford,

"custodiam turris Gloc' cum toto castello," etc., etc.... "per eandem firmam quam reddere solebat comes Milo pater ejus tempore H. R. avi mei;"[948]

while Robert of Torigny speaks, independently, of "discordia quæ erat inter regem Anglorum Henricum et Rogerium, filium Milonis de Gloecestria, propter turrim Gloecestrie."[949] The "tower" of Gloucester is also referred to in the Pipe-Roll of 1156,[950] and in the Cartulary of Gloucester Abbey.[951] The importance of its mention lies in the fact that it establishes the character of Gloucester Castle, and proves that what the leading authority has written on the subject is entirely erroneous. Mr. G. T. Clark, in his great work on our castles, refers thus to Gloucester:—

"The castle of Gloucester ... was the base of all extended operations in South Wales. Here the kings of England often held their court, and here their troops were mustered. Brichtric had a castle at Gloucester, but his mound has long been removed, and with it all traces of the Norman building."[952]

In another place he goes further still:—

"Gloucester, a royal castle, stood on the Severn bank, at one angle of the Roman city. It had a mound and a shell-keep, now utterly levelled, and the site partially built over. It was the muster-place and starting-point for expeditions against South Wales, and the not infrequent residence of the Norman sovereigns."[953]

It may seem rash, in the teeth of these assertions, to maintain that this mound and its shell-keep are alike imaginary, but the word "turris" proves the fact. For, as Mr. Clark himself observes with perfect truth,

"in the convention between Stephen and Henry of Anjou (1153) the distinction is drawn between 'Turris Londinensis et Mota de Windesorâ,' London having a square keep or tower, and Windsor a shell-keep upon a mound."[954]

So the keep of Gloucester, being a "turris" and not a "mota," was clearly "a square tower" and not "a shell-keep upon a mound." The fact is that Mr. Clark's assertions would seem to be a guess based on the hypothesis, itself (as could be shown) untenable, that "Brichtric had a castle at Gloucester." Assuming from this the existence of a mound, he must further have assumed that the Normans had crowned it, as elsewhere, with a shell-keep. But the true character of this great fortress is now determined.

Two examples of the double style shall now be adduced from castles outside England. In Normandy we have an entry, in 1180, referring to expenditure "in operationibus domorum turris et castri," etc., at Caen;[955] in Ireland the grant of Dublin Castle to Hugh de Laci (1172) is thus related in the so-called poem of Matthew Regan (ll. 2713-2716):—

The phrase, it will be seen, corresponds exactly with those employed to describe the castles of Carlisle and Appleby, at the same period:—

Having thus established the use of the phrase, let us now pass to its origin.

I would urge that it possesses the peculiar value of a genuine transition form. It preserves for us, as such, the essential fact that there went to the making of the mediæval "castle" two distinct factors, two factors which coalesced so early that the original distinction between them was already being rapidly forgotten, and is only to be detected in the faint echoes of this "transition form."

The two factors to which I refer were the Roman castrum or castellum and the mediæval "motte" or "tour." The former survived in the fortified enclosure; the latter, in the central keep. The Latin word castellum (corresponding with the Welsh caer) continued to be regularly used as descriptive of a fortified enclosure, whether surrounded by walls or earthworks.[957] It is singular how much confusion has resulted from the overlooking of this simple fact and the retrospective application of the denotation of the later "castle." Thus Theodore, in the seventh century, styles the Bishop of Rochester, "Episcopus Castelli Cantuariorum, quod dicitur Hrofesceaster" (Bæda, iv. 5); and Mr. Clark gives several instances, from the eighth and ninth centuries, in which Rochester is alternatively styled a "civitas" and a "castellum."[958] So again, in the ninth century, where the chroniclers, in 876 A.D., describe how "bestæl se here into Werham," etc., Asser and Florence paraphrase the statement by saying that the host "castellum quod dicitur Werham intravit." Now, it is obvious that there could be no "castle" at Wareham in 876, and that even if there had been, an "army" could not have entered it. But when we bear in mind the true meaning of "castellum," at once all is clear. As Professor Freeman observes, "Wareham is a fortified town."[959] Its famous and ancient defences are thus described by Mr. Clark:—

"In figure the town is nearly square, the west face about 600 yards, the north face 650 yards.... The outline of this rectangular figure is an earthwork, within which the town was built."[960]

Such then was the nature of the "castellum," within which the host took shelter.[961] Passing now to a different instance, we find the Greek κώμη ("a village") represented by "castellum" in the Latin Gospels (Matt. xxi. 2), and this actually Englished as "castel" in the English Gospels of 1000 A.D.[962] Here again, confusion has resulted from a misunderstanding.

As against the castellum, the fortified enclosure, we have a new and distinct type of fortress, the outcome of a different state of society, in the single "motte" or "tour." I shall not here enter into the controversy as to the relation between these two forms, my space being too limited. For the present, we need only consider the "motte" (mota) as a mound (agger) crowned by a stronghold (whether of timber or masonry), but not, as Mr. Clark has clearly shown, "crowned with the square donjon," as so strangely imagined by Mr. Freeman.[963] In the "tour" (turris) we have, of course, the familiar keep of masonry, rectangular in form, and independent of a mound.

The process, then, that we are about to trace is that by which the "motte" or "tour" coalesced with the castellum, and by which, from this combination, there was evolved the later "castle." For my theory amounts to this: in the mediæval fortress, the keep and the castellum were elements different in origin, and, for a time, looked upon as distinct. It was impossible that the compound fortress, the result of their combination, should long retain a compound name: there must be one name for the entire fortress, either "tour" (turris) or "chastel" (castellum). Which was to prevail?

This question may have been decided by either of two considerations. On the one hand, the relative importance of the two factors in the fortress may have determined the ultimate form of its style; on the other—and this, perhaps, is the more probable explanation—the older of the two factors may have given its name to the whole. For sometimes the keep was added to the "castle," and sometimes the "castle" to the keep. The former development is the more familiar, and three striking instances in point will occur below. For the present I will only quote a passage from Robert de Torigny, to whom we are specially indebted for evidence on military architecture:—

(1123) "Henricus rex ... turrem nihilominus excelsam fecit in castello Cadomensi, et murum ipsius castelli, quem pater suus fecerat, in altum crevit.... Item castellum quod vocatur Archas, turre et mœnibus mirabiliter firmavit.... Turrem Vernonis similiter fecit."[964]

More interesting for us is the other case, that in which the "castle" was added to the keep, because it is that of the respective strongholds in the capitals of Normandy and of England. The "Tower of Rouen" and the "Tower of London"—for such were their well-known names—were both older than their surrounding wards (castra or castella). William Rufus built a wall "circa turrim Londoniæ" (Henry of Huntingdon):[965] his brother and successor built a wall "circa turrim Rothomagi."[966] The former enclosed what is now known as "the Inner Ward" of the Tower,[967] the "parvum castellum" of Maud's charter.[968]

Of "the Tower of Rouen" I could say much. Perhaps its earliest undoubted mention is in or about 1078 (the exact date is doubtful), when Robert "Courthose," revolting from his father "Rotomagum expetiit, et arcem regiam furtim præoccupare sategit. Verum Rogerius de Iberico ... qui turrim custodiebat ... diligenter arcem præmunivit," Ordericus here, as often, using turris and arx interchangeably.[969] Passing over other notices of this stronghold, we come in 1090 to one of those tragic deeds by which its history was destined to be stained.[970] Mr. Freeman has told the tale of Conan's attempt and doom.[971] The duke, who was occupying the Tower, left it at the height of the struggle,[972] but on the triumph of his party, and the capture of Conan, the prisoner was claimed by Henry for his prey and was led by him to an upper story of the Tower.[973] At this point I pause to discuss the actual scene of the tragedy. Mr. Freeman writes as follows:—

"Conan himself was led into the castle, and there Henry took him.... The Ætheling led his victim up through the several stages of the loftiest tower of the castle," etc., etc.[974]

Here the writer misses the whole point of the topography. The scene of Conan's death was no mere "tower of the castle," but "the Tower," the Tower of Rouen—Rotomagensis turris, as William here terms it. He fails to realize that the Tower of Rouen held a similar position to the Tower of London. Thus, in 1098, when Helias of Le Mans was taken prisoner, we read that "Rotomagum usque productus, in arce ipsius civitatis in vincula conjectus est" (Vetera Analecta), which Wace renders:—

Again, even in the next reign, a royal charter, assigned by Mr. Eyton to 1114-15, is tested, not at the "castle" of Rouen, but "in turre Rothomagensi."[975] And so, two reigns after that, a century later than Conan's death, we find the custodes of "the Tower of Rouen" entered in the Exchequer Rolls, where it is repeatedly styled "turris."

Thus at Rouen, as at London, the "Tower" not only preserved its name, but ultimately imposed it on the whole fortress. And precisely as the Tower of London is mentioned in 1141 by the transition style of "turris Londoniæ cum castello," so in 1146 we find Duke Geoffrey repairing "sartatecta turris Rothomagensis et castelli," after it fell into his hands.[976]

Here then we have at length the explanation of a difficulty often raised. Why is "the Tower of London" so styled?[977] And although, in England, the style may now be unique, men spoke in the days of which I write of the "Tower" of Bristol or of Rochester as of the Tower of Gloucester.[978] Abroad, the form was more persistent, and special attention may be drawn to the Tower of Le Mans ("Turris Cenomannica),"[979] because the expression "regia turris" which Ordericus applies to it is precisely that which Florence of Worcester applies, in 1114, to the Tower of London, to which it bore an affinity in its relation to the Roman Wall.[980]

All that I have said of the "turris" keep is applicable to the "mota" also, mutatis mutandis, for the motte, though its name was occasionally extended to the whole fortress, was essentially the actual keep, the crowned mound, as is well brought out in the passages quoted by Mr. Clark from French charters:—

"Le motte et les fossez d'entour ... le motte de Maiex ... le motte de mon manoir de Caieux et les fossez d'entour."[981]

Here the "fossez d'entour" represent the surrounding works, the "castellum" referred to in the charters of the Empress. But between "the right to hold a moot there," "the moat (sic) and castle" as Mr. Hallam rendered it, "the moat (sic) probably the motte" of Mr. Clark (ii. 112), and the clever evasion "mote" in the Reports on the Dignity of a Peer (Third Report, p. 163), the unfortunate "mota" of Hereford has had a singular fate.

And now for the results of those conclusions that I have here endeavoured to set forth. The three castles to which I shall apply them are those of Rochester, of Newcastle, and of Arques.

In an elaborate article on the keep of Rochester, Mr. Hartshorne showed that it was erected, not as was believed by Gundulf, but by Archbishop William of Corbeuil,[982] between 1126 and 1139. But he did not attempt to explain what was the "castle of stone" which Gundulf is recorded to have there constructed. As everything turns on the exact wording, I here give the relevant portions of the document in point: —

"Quomodo Willelmus Rex filius Willelmi Regis rogatu Lanfranci Archiepiscopi concessit et confirmavit Rofensi ecclesiæ S. Andreæ Apostoli ad victum Monachorum manerium nomine Hedenham; quare Gundulfus Episcopus Castrum Rofense lapideum totum de suo proprio Regi construxit.

"Gundulfus ... illis contulit beneficium ... castrum etenim, quod situm est in pulchriore parte Hrovecestræ.... Regi consuluerunt [duo amici] quatinus ... Gundulfus, quia in opere cæmentarii plurimum sciens et efficax erat, castrum sibi Hrofense lapideum de suo construeret.... Dixerunt [Archiepiscopus et Episcopus] ... quotiescunque quidlibet ex infortunio aliquo casu in castro illo contingeret aut infractione muri aut fissura maceriei, id protinus ... exigeretur.... Hoc pacto coram Rege inito fecit castrum Gundulfus Episcopus de suo ex integro totum, costamine, ut reor, lx librarum."[983]

Though castrum is the term used throughout, Mr. Parker in his essay on The Buildings of Gundulph, 1863, assumed that a tower must be meant, and wrote of "Gundulf's tower" in the Cathedral: "This is probably the tower which Gundulph is recorded to have built at the cost of £60."[984] So too, Mr. Clark wrote:—

"As to his architectural skill and his work at Rochester Castle, ... the bishop [was] to employ his skill, and spend £60 in building a castle, that is, a tower of some sort. What Gundulf certainly built is the tower which still bears his name.... It may be that Gundulf's tower was removed to make way for the new keep, but in this case its materials would have been made use of, and some trace of them would be almost certain to be detected. But there is no such trace, so that probably the new keep did not supersede the other tower."[985]

Mr. Freeman guardedly observes:—

"The noble tower raised in the next age by Archbishop Walter (sic) of Corbeuil ... had perhaps not even a forerunner of its own class.

"Mr. Hartshorne showed distinctly that the present tower of Rochester was not built by Gundulf, but by William of Corbeuil.... But we have seen (see N. C., vol. iv. p. 366) that Gundulf did build a stone castle at Rochester for William Rufus ('castrum Hrofense lapidum' [sic]), and we should most naturally look for it on the site of the later one. On the other hand, there is a tower seemingly of Gundulf's building and of a military rather than an ecclesiastical look, which is now almost swallowed up between the transepts of the cathedral. But it would be strange if a tower built for the king stood in the middle of the monastic precinct."[986]

Thus the problem is left unsolved by all four writers. But the true interpretation of castrum, as established by me above, solves it at once. For just as William of Corbeuil is recorded to have built the "turris" or rectangular keep,[987] so Gundulf is described as constructing the castrum or fortified enclosure.[988] We must look, therefore, for his work in the wall that girt it round. And there we find it. Mr. Clark himself is witness to the fact:—

"Part of the curtain of the enceinte of Rochester Castle may also be Gundulph's work. The south wall looks very early, as does the east wall."[989]

But Mr. Irvine had already, in 1874, pointed out, in a brief but valuable communication, that a distinctive peculiarity of Gundulf's work—the absence of plinth to his buttresses—is found "in the castle wall at Rochester (also his)."[990] Thus, it will be seen, the character of the work independently confirms my own conclusion.

Some confusion, it may be well to add, has been caused by such forms as "castellum Hrofi" and "castrum quod nominatur Hrofesceaster." In these early forms (as in some other cases), "castrum" denotes the whole of Rochester, girt by its Roman wall, and not (as Mr. Hartshorne assumed throughout) the castle enclosure. Mr. Clark leaves the point in doubt.[991]

Before leaving Rochester, I would point out that, unlike the rest of Gundulf's work, this castrum can be closely dated. The conjunction of Lanfranc and William Rufus, in the story of its building, limits it to September, 1087-March, 1089, while Odo's rebellion would probably postpone its construction till his surrender. It is most unfortunate, therefore, that Mr. Clark should write, "This transaction between the bishop and the king occurred about 1076,"[992] when neither Gundulf was bishop nor William king.

To the case of Newcastle and its keep, I invite special attention, because we have here the tacit admission of Mr. Clark himself that he has antedated, incredible though it may seem, by more than ninety years the erection of this famous keep. To prove this, it is only necessary to print his own conclusions side by side:—

The origin, of course, of the astounding error by which "the great master of military architecture" misdated this keep by nearly a century,[993] and took an essentially late work for one of the earliest in existence, was the same fatal delusion that castrum or castellum meant precisely what it did not mean, namely, a tower. "Castellum novum super flumen Tyne condidit" is the expression applied to Robert's work in 1080, and the absence of a "tower" explains the fact that Fantosme makes no mention of a "tur" when describing "Le Noef Chasteau sur Tyne," the existing keep not being available at the time of which he wrote.

We now come to our last case, that of the Château d'Arques.

"Arques," writes Mr. Clark, "is one of the earliest examples of a Norman castle."[994] It is, Mr. Freeman holds, "a fortress which is undoubtedly one of the earliest and most important in the history of Norman military architecture."[995] No apology, therefore, is needed for discussing the date of this celebrated structure, so long a subject of interest and of study both to English and to French archæologists.

As at Colchester and in other places, the very wildest theories have been generally advanced, and archæologists have only gradually sobered down till they have virtually agreed upon a date for this keep which is actually, I venture to think, less than a century wrong.

In his noble monograph upon the fortress, the basis of all subsequent accounts,[996] M. Deville enumerates, with contemptuous amusement (pp. 49, 268-272), the rival theories that it was built (1) by the Romans; (2) by "Clotaire I." in 553—the date 1553 on one of the additions for the structure having actually been so read; (3) by "Charles Martel" in 745, 747, or 749 (on the strength of another reading of the same date, confirmed by a carving of his coat-of-arms)—these being the dates given by Houard and Toussaint-Duplessis. At the time when Deville himself wrote the study of castles was still in its infancy, and of the two sources of evidence now open to us, the internal (that of the structure itself) and the external (that of chronicles and records), the latter alone was ripe for use. Now, at Arques, precisely as at our own Rochester, the written evidence has hitherto appeared conflicting to archæologists, but only because the language employed has never yet been rightly understood. On the one hand we read in William of Jumièges, an excellent authority in the matter, that "Hic Willelmus [the Conqueror's uncle] castrum Archarum in cacumine ipsius montis condidit;" and in the Chronicle of Fontenelle, that this same William "Arcas castrum in pago Tellau primus statuit;" also, in William of Poitiers, that "id munimentum ... ipse primus fundavit:" on the other, we read in Robert du Mont, a first-rate and contemporary authority, who may indeed be termed a specialist on the subject, that "Anno MCXXIII. castellum quod vocatur Archas turre et mœnibus mirabiliter firmavit [Rex Henricus]."[997]

M. de Caumont, that industrious pioneer, whose work appeared four years before that of M. Deville, boldly followed Robert du Mont, and confidently assigned the existing keep to 1123.[998] Guided, however, by M. Le Prévost (1824), he held that the original structure was raised by the Conqueror's uncle, and that Henry I. merely "fit reconstruire en entier le donjon et une partie des murs d'enceinte." M. Deville, on the contrary, in his eager zeal for the honour and glory of the castle, stoutly maintained that, keep and all, it was clearly Count William's work. He admitted that his Norman brother-antiquaries assigned it to Henry I., but urged that they had overlooked the evidence of the structure, and its resemblance to English keeps assigned (but, as we now know, wrongly) to the eleventh century, or earlier;[999] and that they had misunderstood the passage in Robert du Mont, which must have referred to mere alterations. In order thus to explain it away, he contends (and this contention Mr. Clark strangely accepts) that Robert says the same—which he does not—of "Gisors, Falaise, and other castles known"—which they are not[1000]—"to be of earlier date" (M. M. A., i. 194). Lastly, he appeals, though with an apology for doing so ("s'il nous était permis d'invoquer à l'appui de notre opinion"), to the far later "Chronique de Normandie" for actual evidence, elsewhere wanting, that the keep itself (turris) was built by William of Arques,[1001] that is, in 1039-1043.[1002]

"I went over the castle minutely," Professor Freeman writes, "in May, 1868, with M. Deville's book in hand, and can bear witness to the accuracy of his description, though I cannot always accept his inferences" (N. C., iii. 124, note). He accordingly doubts M. Deville's date for the gateway and walls of the inner ward, but sees "no reason to doubt that the ruined keep is part of the original work" (ibid.). We must remember, however, that the Professor is at direct variance with Mr. Clark on the Norman rectangular keeps, for which he claims an earlier origin than the latter can concede.

Turning now to Mr. Clark himself, we learn from him that—

"it seems probable that the keep is the oldest part of the masonry, and the work of the Conqueror's uncle, Guillaume d'Arques, and it is supposed to be one of the earliest, if not the earliest, of the rectangular keeps known" (M. M. A., i. 194).

He adds that the passage in Robert du Mont

"has been held to show that the whole structure was the work of Henry, who reigned from 1105 (sic) to 1135, and the extreme boldness of the buttresses and superincumbent constructions of the keep no doubt favour this view; but, as M. Deville remarks in the same passage, similar reference is made to Gisors, Falaise, and other castles, known to be of earlier date" (ibid.).

To resume. The external or written evidence is as follows. On the one hand, we have the clear and positive statement of a contemporary writer, Robert du Mont, that Henry I. built this keep in 1123. On the other, we have no statement from any contemporary that it was built by William of Arques (in 1039-1043). He is merely credited with founding the castellum, and in none of the contemporary accounts of its blockade and capture by his nephew is there any mention of a turris. The distinction between a castellum and a turris, with their respective independence, has not, as I have shown, hitherto been realized, and it is quite in the spirit of older students that M. Deville confidently exclaims—

"Or, conçoit-on un château-fort sans murailles? Un château-fort sans donjon, dans le cours du XIVᵉ siècle, en Normandie, n'est guère plus rationnel" (p. 310).

As to the "murailles," Mr. Clark has taught us that palisades were not replaced by walls till a good deal later than has been usually supposed; and as to the "donjon," if, as I have established, so important a fortress as Rochester was without a keep in the eleventh, and indeed well into the twelfth century, other castella must have been similarly destitute—probably, for instance, Newcastle, as we have seen, and certainly Exeter, of which Mr. Clark writes: "There is no evidence of a keep, nor, at so great a height, was any needed" (M. M. A., ii. 47). The same argument from strength of position would à fortiori apply to Arques, and there is, in short, no reason for doubting that the castrum of William of Arques need not have included a turris.[1003]

On what, then, rests the assertion that the keep was the work of the Conqueror's uncle? Strange as it may seem, it rests solely on the so-called Chronique de Normandie, an anonymous production, not of the eleventh, but of the fourteenth century! "Si fist faire une tour moult forte audessus du chastel d'Arques," runs the passage, which is quoted by Mr. Clark (i. 194), from Deville (pp. 311, 312), who, however, apologized for appealing to that authority. This "Chronique" is admitted to have been based on the poetical histories of Wace and Benoit de St. More, themselves written several generations later than the alleged erection of this keep. Of the former, Mr. Freeman holds that, except where repeating contemporary authorities, "his statements need to be very carefully weighed" (N. C., ii. 162); and of the latter, that he is "of much smaller historical authority" (ibid.). To this I may add that, in my opinion, Wace, writing as he did in the reign of Henry II., at the close of the great tower-building epoch, spoke loosely of towers, when mentioning castles, as if they had been equally common in the reign of the Conqueror. A careful inspection of his poem will be found to verify this statement. "La tur d'Arques" was standing when he wrote: consequently he talks of "La tur d'Arques" when describing the Conqueror's blockade of the castle in 1053. There is no contemporary authority for its existence at that date.[1004]

And now let us pass from documentary evidence to that of the structure itself. We may call Mr. Clark himself to witness that the presumption is against so early a date as 1039-1043. He tells us, of the rectangular keep in general, that—

"not above half a dozen examples can be shown with certainty to have been constructed in Normandy before the latter part of the eleventh century, and but very few, if any, before the English conquest" (i. 35).

Therefore, on Mr. Clark's own showing, we ought to ask for conclusive evidence before admitting that any rectangular keep is as old as 1039-1043. But what was the impression produced on him by an inspection of the structure itself? This is a most significant fact. While rejecting, apparently on what he believed to be documentary evidence, the theory that the keep (turris) was the work of Henry I., he confessed that the features of the building "no doubt favour this view" (i. 194, ut supra).

But leaving, for the present, Mr. Clark's views, to which I shall return below, I take my stand without hesitation on certain features in this keep. It is not needful to visit Arques—I have myself never done so—to appreciate their true significance and their bearing on the question of the date. The first of these is the forebuilding. Mr. Clark tells us that Arques possesses "the usual square appendage or forebuilding common in these keeps" (M. M. A., i. 198). But this unscientific treatment of the forebuilding, ignoring so completely its origin and development, cannot too strongly be resisted. Restricting ourselves to the case before us, we at once observe the peculiarity of an external staircase, not only leading up to a forebuilding, through which the keep is entered, but actually carried, through a massive buttress, round an angle of the keep.[1005] Rochester being believed to be the work of Gundulf, in the days when M. Deville wrote, it was natural that he should have supposed "cette savante combinaison" to have been familiar to Gundulf (p. 299). But now that, on these points, we are better informed, let us ask where can Mr. Clark produce an instance of this elaborate and striking device as old even as the days of Gundulf, to say nothing of those of Count William (1039-1043)? Where we do find it is in such keeps as Dover, the work of Henry II., or Rochester, where the resemblance is even more remarkable. Now, Rochester, as we know, was actually built within a few years of the date given by Robert du Mont, and upheld by me, as that of the construction of Arques. Oddly enough, it is Mr. Clark himself who thus points out another resemblance:—

"In the basement of the forebuilding ... was a vaulted chamber, opening into the basement of the keep, as at Rochester, either a store or prison" (M. M. A., p. 188).

Lastly, both at Arques and at Rochester, we find on the first floor, near the entrance, the very peculiar feature of a smaller doorway communicating with the rampart of the curtain.[1006] This parallel, which is not alluded to by Mr. Clark, is the more remarkable, as such a device is foreign to the earlier rectangular keeps, and also implies that the keep must have been built certainly no earlier, and possibly later, than the curtain, which curtain, Mr. Clark, as we shall find, admits, cannot be so old as the days of Count William.

No one, in short, unbiassed by supposed documentary evidence, could study this keep, with its "petites galeries avec d'autres petites chambres ou prisons pratiquées dans l'épaisseur des murs"[1007] (as at Rochester), with the elaborate defences of its entrance, and with those other special features which made even Mr. Clark uneasy, without rejecting as incredible the accepted view that it was built by Count William of Arques (1039-1043). And this being so, there is, admittedly, no alternative left but to assign it to Henry I. (1123), the date specifically given by Robert du Mont himself.

But, it may be urged, though there is nothing improbable in Mr. Freeman being wrong, is it conceivable that so unrivalled an expert as Mr. Clark himself can have mistaken a keep of 1123 for one of 1039-1043, when we remember the wonderful development of these structures in the course of those eighty years? To this objection, I fear, there is a singularly complete answer in the case of Newcastle, where, as we have seen, he was led by the same misconception into no less amazing an error.[1008]

In short, the view I have brought forward as to the separate existence of "tower" and "castle" may be said, from these examples, to revolutionize the study of Norman military architecture.