Chapter I: Nave and Aisle Vaults (2)

The second group is much larger and more widespread, and comprises all the churches employing transverse tunnel vaults over the side aisles. The examples belonging to the school of Perigord have already been discussed,[106] and mention has been made of the fact that there are possibly enough of such churches in Provence alone to constitute a fifth type in that school.[107] But the system is too widespread to be attributed to any one province. It is undoubtedly a product of Roman and very early mediaeval architecture, for it is to be seen in such buildings as the Basilica of Maxentius at Rome, and in a modified, ramping form at Aachen.[108] Its structural advantage lies in the large space which the tunnel vault affords for windows in the outer wall thus lighting both the nave and aisles. Among the many examples are the parish church of Chatillon-sur-Seine (Côte-d’Or)[109] of the twelfth century, the abbey churches of Hauterive (Savoie), Ronceray[110] (vaulted in 1115), Bénévent-l’Abbaye (Creuse),[111] and the cathedral of Lescar (Basses-Pyrénées),--in which, however, the vaults are an addition to a primitive construction.[112] In the church at Fontenay (Côte-d’Or)[113] (before the middle of the twelfth century) concealed flying buttresses appear over the transverse arches between the aisle bays, thus aiding in securing a more even abutment for the continuous thrust of the tunnel vault of the nave. A few churches like Cavaillon,[114] and the cathedral of Orange (Vaucluse),[115] have tunnel vaults over rectangular bays flanking the nave but not connected by arches to form side aisles.

The vaulting of the ambulatory gallery of Mantes cathedral, of the aisles of Fountains Abbey in England, and possibly the original vaults of the aisles of Saint Remi at Reims[116] were also transverse tunnel vaults. These latter churches differ from the ones previously mentioned, however, in that they are not tunnel vaulted in the nave and, moreover, are constructed with a clerestory so that the side aisle vaults do not serve the purpose outlined in the account of tunnel vaulted churches in the preceding paragraph.

TUNNEL VAULTS WITH CROSS RIBS

This brings the discussion of the standard methods of tunnel vaulting to a close, but there remain two curious churches in which cross-ribs were added beneath the surface of simple tunnel vaults. One of these is at Lusignan (Vienne),[117] and the other at Javarzay (Deux-Sèvres). Both date from about 1120 to 1140 though the ribs may be a later addition to give the appearance of ribbed vaulting which was introduced at about this time.

NAVES WITH GROINED VAULTS

Although usually confined to the side aisle bays, there are a few Romanesque churches in which the builders of the eleventh and twelfth centuries placed groined vaulting over the nave. The scarcity of such examples is due primarily to the difficulty of meeting the severe outward thrusts of a groined vault raised over bays of considerable span and at a point high above the ground. In the side aisles where the vaults were comparatively low, the exterior wall could be thickened by salient buttresses, and the piers strengthened by the weight of the wall above in a manner to offset the thrust, but in the nave the problem was more complicated. The builders had not yet invented the flying buttress. Hence, when they attempted groined vaults at all, they blundered along trusting that the inert mass of their walls and such timid buttresses as could be erected above the nave piers would provide sufficient offset for the thrusts even though these were now concentrated at four main points in each bay. Naturally the vaults frequently gave way and had to be reconstructed. In spite of these difficulties, the advantage of the groined vault in providing a clerestory whose windows might rise as high as the crown of the vault itself led to its occasional use.

GROINED VAULTS OVER RECTANGULAR NAVE BAYS

The vaults thus employed were of two rather distinct classes, those over rectangular nave bays which were usually but little domed up, and those over square bays which were generally distinctly domed in the Byzantine manner. Of the first type perhaps the best known example is the Burgundian church of La Madeleine at Vézelay (Yonne), (Fig. 17) dedicated

in 1104. Its nave is divided into a series of rectangular bays by transverse arches of semicircular section, and over each bay is placed a groined vault very slightly domed at the crown. To insure the stability of these vaults, the builders relied on the weight of the walls, which were carried up somewhat above the window heads, and on simple salient buttresses. To these exterior supports were added interior arches half imbedded in the walls above the clerestory windows (Fig. 17), furnishing one of the earliest examples of the use of wall ribs or formerets. The web of the vault does not, however, follow their extrados, but gradually breaks away from it toward the crown, with the apparent object of thus concentrating even more pressure upon the piers by stilting the wall line of the vault surface.[118] Even these precautions were not deemed sufficient, so iron tie-rods were employed, but these rusted and broke,[119] the vaults settled badly,[120] and if it had not been for the addition of exterior flying buttresses, which had meanwhile come into general use, the vaults would most certainly have fallen. Although not a structural success, Vézelay did prove of advantage in turning the builders away from the tunnel vault,--and this, too, in Bourgogne where it had been most highly developed,--to a new type which presented problems whose solution was to lead to Gothic architecture. Vézelay was, however, but little imitated in the Romanesque era, perhaps because of the almost contemporary development of the ribbed vault in Lombardy, Normandy, and the Ile-de-France. A few churches, such as Anzy-le-Duc (Saône-et-Loire)[121] did employ groined vaults over the nave but on a smaller scale and frequently with more pronounced doming.

A more important and independent group of groined vaulted churches is to be found in Normandy. In this school, the churches were usually covered with wooden roofs though the aisles were occasionally groined. But there are three churches in which the choir also has groined vaults. These are, La Trinité or the Abbaye-aux-Dames at Caen (Calvados) (cir. 1066), Saint Nicolas at Caen (cir. 1080), and Saint Georges-de-Boscherville at Saint Martin-de-Boscherville (Seine-Inférieure) (late eleventh and early twelfth century). The choir of the third of these churches, though later in date than the others, is more primitive in type, for it is covered by interpenetrating vaults, in which, however, the deep lunettes above the windows rise so nearly to the crown that the result resembles groined rather than tunnel vaulting.

In both the other examples true groined vaulting is used, but at La Trinité it is in practically square bays, and carried by walls running down to the ground,[122] making it easier of construction than that at Saint Nicolas[123] where the bays are rectangular and the choir has true side aisles. This church is similar in structural principles to La Madeleine at Vézelay--except that the wall ribs are omitted,--and these two churches may be said to represent the highest point reached by groined vaulting with practically flat crowns during the Romanesque period.

Other examples might be cited, ranging from such an unusual church as Saint Loup-de-Naud (Seine-et-Marne) in the Ile-de-France,--which is of uncertain date,[124]--to churches as late as the thirteenth and fourteenth centuries, among which are Severac-le-Château (Aveyron) and Saint Pons-de-Mauchiens (Hérault).[125] Occasionally, also, groined vaults were used in the crypt as at Saintes (Charente-Inférieure),[126] even when tunnel vaults were used in the upper part of the church, a peculiarity explained by the fact that underground it was easy to dispose of the thrusts which could not so readily be offset in the nave.

The question of the origin of the method has frequently arisen and a number of writers, including Choisy,[127] suggest the East as a possible cradle of the style because of the numerous churches in Palestine thus vaulted, but Rivoira[128] shows rather conclusively that it was the Cluny influence which carried the method to the East rather than the reverse, a theory strengthened by the fact that the earliest example there, which is the church of Saint Anne at Jerusalem,[129] would seem to be after rather than before the beginning of the twelfth century.[130] Moreover it is quite reasonable to attribute the development of this advanced type of vault to the builders of Bourgogne themselves, for they were surely progressive enough to have taken such a step.

GROINED VAULTS OVER SQUARE NAVE BAYS

Churches with groined vaults over square nave bays are much more numerous than those with rectangular bays, just described. The most important of these belong to the school of the Rhenish Provinces, which had, perhaps, clung to Byzantine and Carolingian traditions in this respect. As a rule the large churches of this school were originally planned for vaulting only in the side aisles.[131] These were usually divided into square bays by round headed transverse arches, and then each bay covered by a more or less domed up groined vault, which, from its size and form, might be erected with comparatively little centering.[132] There was no triforium gallery, but a wall with blank arches took its place beneath the clerestory windows. In many of the churches[133] shafts were carried up on the inner face of alternate nave piers, probably to support the cross beams of the roof, or possibly to carry transverse arches, but not to carry vaulting.

By the early twelfth century, after numerous fires had played havoc with the churches, the Rhenish builders seem to have at last made an effort to replace the wooden roofs with vaults. In doing this, they sought a form of vault which would exert as little as possible of outward thrust and thus be stable at the considerable height at which it must be placed. The Lombard builders had by this time developed the domed up cross-ribbed vault, but, as has been admirably shown by Porter,[134] the ribs which they employed had for their sole purpose the saving of wooden centering, since the masonry of the vault proper was heavy enough to stand without their aid. It was natural then for the Rhenish builders, who copied their neighbors in Lombardy in many particulars,[135] to look to them for a method of vault construction, which they found in domed up vaults like those of Rivolta-d’Adda (1088-1099) or Sant’Ambrogio at Milan (cir. 1098). These the Rhenish builders chose as models, but being plentifully supplied with wood for centering, it would seem as if they purposely did not adopt the diagonal ribs, but built groined vaults of simple domed up type, placing them over square nave bays each corresponding to two aisle bays in the true Lombard manner. This system may be seen to advantage in the cathedral of Speyer[136] (probably vaulted cir. 1137-1140). With extremely heavy walls like those of the Rhenish churches, and with good masonry for their construction, such vaults proved comparatively safe even over naves of such a span as that of Speyer which is almost fifty feet in width.

This account of the Rhenish school completes the discussion of groined vaulting as applied to the naves and choirs of Romanesque churches. The heavy walls and the general excellence of masonry construction which they required, together with the necessity for large interior piers, did not render them popular or widely used.

AISLES WITH GROINED VAULTS IN LOMBARDY AND NORMANDY

That the use of groined vaults was far more extensive in the aisles than in the naves of Romanesque churches has already been shown by the examples cited from the schools of Poitou, Auvergne, Bourgogne, and elsewhere. To these should be added a number of churches, chiefly of the schools of Lombardy and Normandy, which have groined aisles in combination with rib vaulted or wooden roofed naves. In Lombardy, where the naves are ribbed, this combination has been admirably explained by Porter[137] in connection with the use of wood for centering. Thus he shows that groined vaults, provided that they were sufficiently domed up, could be built over the small bays of the aisles and triforia with almost no wooden framework, but that when such vaults were attempted in the nave the bays were so large as to require a considerable amount of centering beneath the vault, and therefore the builders substituted permanent diagonal arches of very heavy character.

The Norman groined aisles are, however, of a different sort, for they either have level crowns or are but slightly domed up in type.[138] The abbey church of Jumièges (Seine-Inférieure) (1040-1067) is among the earliest examples of this construction and is the only Norman church with groined vaults in both the aisles and triforium.[139] La Trinité at Caen[140] and the abbey church of Lessay (Manche)[141] are also Norman churches with groined aisles, in both cases with level crowns. In La Trinité, as in the early churches of Poitou, the bays are not even separated by transverse arches.[142] In Saint Étienne at Caen, and in the choir of the cathedral of Gloucester, the aisles are vaulted in both stories like those of Auvergne, the lower groined, the triforia with half tunnel vaults, but it seems very probable that these latter were added only when vaulting took the place of the wooden roof in the central portions of the church.[143]

Curious instances of the persistence of groined vaulting are to be seen in the triforia of such transitional churches as Saint Germer-de-Fly (Oise)[144] and Vézelay, where the remaining portions of the church have ribbed vaults. For this persistence an explanation is later attempted.[145]

AISLES WITH SEMI-GROINED VAULTS

An unusual form of aisle vault appears at Creully (Calvados)[146] (twelfth century), where the aisles are covered with a half tunnel vault intersected toward the outer wall by lunettes, which thus convert it into a semi-groined vault. Its obvious advantage lies in the combination of inward pressure, which it exerts in support of the nave vaults, with the added window space which it affords without increasing the height of the exterior walls.

RIBBED VAULTS

The introduction of ribs beneath the diagonal intersections of groined vaulting gradually brought about a revolution in Mediaeval building, and transformed the massiveness of Romanesque construction into the light and graceful architecture of the Gothic era. Much has been written in an effort to discover the origin of the new system. It is not, however, the intention here to add to the number of theories advanced, except in an incidental manner, but rather to classify the various forms of ribbed vaulting as applied to naves, choirs, and aisles of the churches following immediately after those of the Romanesque period which have just been described. As a geographical basis is no longer practical for such a classification, because of the widespread distribution of the new method of construction, a structural basis will be substituted, and the vaults will be divided into two major groups according as they were used over square or rectangular nave bays, and then subdivided according to their minor characteristics.

RIBBED VAULTS OVER NAVES WITH SQUARE BAYS

Lombardy affords the first examples of ribbed vaults over nave bays of square plan. According to Rivoira[147] the earliest are in the church of Santa Maria e San Sigismondo at Rivolta d’Adda[148] (before 1099), though this was closely followed by the more important church of Sant Ambrogio at Milan (between 1088-1128) (Fig. 18), which furnishes an admirable example of the Lombard type. Its nave is divided into four great square bays, each corresponding to two bays in the side aisles. (Plate I-a.) Of these the eastern bay is treated as a crossing and covered by a dome above a lantern on squinches, but the remaining three have four-part domed up vaults with heavy ribs of square section, used not only transversely and along the walls but also diagonally, thus forming a complete system or skeleton of arches beneath the vault surface in the manner of true Gothic architecture. But there are many reasons to believe with Porter[149] that the builders of Lombardy employed these ribs purely as a permanent centering of masonry,--which was less expensive than a temporary centering of wood in a country where the latter material was very scarce,--and that they failed to appreciate the fact that such ribs made possible a great reduction in the weight of the panels, or web. of the vault, and in other ways could be made to aid in reducing and concentrating its pressures. The masonry of the vault is still excessively

thick,--between sixteen and twenty inches,--and would stand equally well were the ribs removed. Moreover its thrust is so great that the builders dared not raise its imposts sufficiently high to admit of a clerestory beneath the formerets, and instead of rendering possible a lighter construction as Gothic vaults were destined to do, these vaults of Saint’ Ambrogio required for their support a wall forty inches thick and ramping walls above the transverse arches of the triforium together with interior tie-rods and wooden chains in the masonry[150] to offset their severe outward thrust. All these facts show that the Lombard vaults are still fundamentally Romanesque in type. Even in San Michele at Pavia (early twelfth century), where the system was a little more developed, in that a small clerestory was introduced, the principles were still the same as in Milan. As a matter of fact, the Lombard builders never made any further advance in the handling of ribbed vaults, and even went backward rather than forward. For the builders found that groined vaults of domed up type could be built so lightly as to require but little centering, and a return to this simple form was made in such churches as San Lanfranco at Pavia.[151] Later on, in the thirteenth and fourteenth centuries, French methods of ribbed vaulting were introduced, but throughout the whole period of Lombard supremacy the tendency was to avoid vaulting entirely, and when adopted, it was of the heavy character just described.

THE SYSTEM OF ALTERNATE SUPPORTS

The Lombard churches are important in the present connection, however, because of the method in which they are divided into vaulting bays. They furnish the earliest examples of the system of alternate light and heavy supports,--employed according to Cattaneo[152] as early as 985 in the three original bays of SS. Felice e Fortunato at Vicenza. This system of piers with alternate transverse arches produces one square[153] bay in the nave to two square bays in the side aisles, and it occurs not only in vaulted churches but also in others in which a wooden roof rests upon these transverse supports.[154] Its advantage in the vaulted churches is particularly important, however, and of a two-fold character. In the first place, it renders the four enclosing arches uniform, and it makes them as nearly as possible of equal span with the diagonals.[155] And in the second, it saves a considerable amount of centering by rendering possible the construction of a vault covering a space corresponding to two rectangular bays on four instead of seven ribs.[156]

Outside of Lombardy, the four-part cross-ribbed vault over square nave bays was but seldom employed in churches with side aisles also divided into square compartments. It appears, however, in the cathedral of Le Mans, (Sarthe) (middle of the twelfth century), where it would seem to be due to the influence of the neighboring single aisled churches of Anjou,--which are later discussed,--and it was frequently used in reconstructing the vaults of the Rhenish school. In the Gothic period also, the system occasionally appears in a modified form, and naturally enough these revivals occur where Norman and Rhenish Romanesque had caused the principles of Lombard architecture to be strongly entrenched. Thus the church of Saint Legerius at Gebweiler[157] (cir. 1182-1200) furnishes a Rhenish, and the choir of Boxgrove Priory church (cir. 1235), an English application of this method. In the latter, the vaults are no longer highly domed up, and are therefore far removed from their Lombard prototypes, only the general division of the church reflecting this influence.

NAVES WITHOUT SIDE AISLES

More important by far, are the churches without side aisles but with naves in square bays with four part cross-ribbed vaults. This method is to be seen in the cathedral of Fréjus (Var),[158] which is considered by Porter[159] to exhibit the earliest extant ribbed nave vaults in France. These are distinctly of Lombard type, and would seem to show a strong Lombard influence entering France from the south. It may possibly be that this same influence followed the route taken earlier by the dome on pendentives, and thus gave rise to the domed up ribbed vault so common in the churches of Anjou.[160] Of these latter, the cathedral of Saint Maurice at Angers (Maine-et-Loire) (Fig. 19), presents perhaps the best existing example. Its nave vaults which date from as early as 1150[161] are among the largest and finest in France, having a span of some fifty-six feet. As in Lombardy, the crown is highly domed up while to facilitate the construction of the web of the vault with the least possible centering, pointed diagonals and enclosing arches are employed. By this means the entire vault was constructed on the ribs with no centering at all for the lower courses, and a simple _cerce_, a device consisting of two curved boards sliding along each other, for those near the crown. At the same time the outward thrusts were greatly reduced by the pointed section of the vault.

ANJOU RIDGE RIBS

FIG. 19.--ANGERS, CATHEDRAL.

Since the Anjou churches possessed naves of wide span, it is not surprising to find that their builders soon added ridge ribs beneath the vault. That these were not mere cover-joints to conceal an irregular intersection of the masonry, as Choisy suggests,[162] would seem to be proved by the fact that the courses meet in a straight line at the ridge in by far the greater number of Anjou churches in which they are employed,--for example in La Couture at Le Mans (Fig. 20), Airaines,[163] and numerous churches with small torus ribs, as well as by the fact that such ridge ribs are sometimes omitted even when the masonry is laid up in courses of equal width and therefore interpenetrating at the ridge, as in Avesnières (Mayenne)[164] near Laval. If not, however, primarily a cover-joint, these ribs did at least possess both a structural and decorative quality. In the first place they helped to keep the keystone of the diagonals rigidly fixed during the building process, and furthermore, they gave an absolutely straight line to the vault crown which was always difficult to adjust, particularly in a vault of large size. One of the best and earliest examples of the employment of such ribs appears in the nave of Notre Dame-de-la-Couture at Le Mans (Fig. 20) which dates from about 1200, and a later example is afforded by the church of Saint-Avit-Sénieur (Dordogne),[165] where the vaults are of the thirteenth century and replace an original series of domes on pendentives of true Perigord type.

In all of the Anjou vaults thus far discussed, the ribs are of comparatively heavy section and placed entirely beneath the vault surface, but there was to be a decided change in the thirteenth century. It has already been noted that domed up vaults could be erected almost without centering and exerted little if any pressure upon the ribs beneath them. Realizing this, the builders of Anjou soon began to reduce the size of the ribs until they became little more than torus mouldings running along the groin and ridge of the vault. As an actual fact, however, these torus mouldings were carved upon a sunken rib flush with the surface of the panel, which, if it no longer furnished a support for the vault, at least formed a sort of permanent centering dividing the surface to be vaulted into distinct severies and marking the line of their intersection in an absolutely correct curve. Such vaults are closely allied to those of groined type, the ribs playing practically the same part as those of brick in Roman concrete vaulting. Since, however, in the Anjou system the ribs always were merely a permanent centering which could easily be removed without destroying the vault, a sunken centering was quite as efficient in serving the purpose of vault division while the torus afforded a certain amount of surface decoration.

Of this typical Anjou construction, there are numerous examples. At Poitiers, in the church of Sainte Radegonde the ribs are of reduced size but not quite flush with the vault surface and the same is true at Saint-Hilaire--Saint-Florent near Saumur (Marne-et-Loire),[166] while the choir and transept of Angers cathedral (Fig. 19), and the later bays of the cathedral of Poitiers furnish examples of the standard type. After a short period of experiment, the builders of Anjou became very skillful in the construction of these ribs and vaults and frequently employed them over bays of unusual plan and elevation as, for example, in the chapel north of the choir aisle in Saint Serge at Angers (Fig. 21).

An instance of the influence of Anjou construction upon the neighboring territory, as well as of the relationship between this Gothic style and the Romanesque school of Perigord, may perhaps be seen in the Old Cathedral of Salamanca in Spain.[167] Here the three western bays of the nave are covered with ordinary domes but with diagonal ribs beneath them, while the two remaining bays have regular domed up Anjou vaults. The date of this cathedral, cir. 1120-1178, may, perhaps, explain this peculiar combination as being due to an Anjou-Gothic influence displacing one of Perigord-Romanesque, in much the same manner as such an influence displaced the Perigord-Romanesque architecture of western France.

SQUARE NAVE BAYS OUTSIDE OF LOMBARDY AND ANJOU

Besides its use in Lombardy and Anjou, the square nave bay with four part cross-ribbed vaults, was employed to some extent in other parts of Europe throughout the Gothic period.[168] Some of these are churches without side aisles, but aisles are more commonly found, divided into rectangular bays corresponding in number to those of the nave. Of the single naved churches, San Francesco at Assisi,[169] is a good example. Although dating from 1236-1259, its vault ribs are still heavy and almost square in section, as if derived from Lombard prototypes. But they differ in being of pointed section and in not giving to the vaults a domed up crown. In this they would seem to be examples of French influence upon Lombard tradition.

SQUARE NAVE AND RECTANGULAR AISLE BAYS

An early church with square nave bays and ribbed vaults over rectangular bays in the side aisles (Plate I-b), is to be found at Bury (Oise) (Fig.

22). It probably dates from about 1125, and is an important monument of the Transitional period. Its nave vaults are quite highly domed and in this respect seem somewhat Lombard, but their pointed arches and awkward construction indicate an effort on the part of the builders toward reducing this doming and a dawning consciousness of the value of the pointed arch in the construction of ribbed vaults. This is further shown in the side aisles. Because of the rectangular shape of the bays, the problem was presented of getting three sets of ribs of different span to rise to the same or practically the same height. Not being thoroughly familiar with the flexibility of the pointed rib, the builders at Bury were naturally somewhat clumsy in its use. Thus, the diagonals were made segmental in elevation to lower them to the level of the pier arches, while masonry was piled on the crown of the transverse ribs, or their voussoirs widened, to bring them up to the level of the vault panel.[170] A few such experimental steps as these at Bury, were all that were necessary to give the builders a mastery of the use of the pointed arch in ribbed vaulting.

RIBS WITH CARYATID SUPPORTS

But there is another feature of the side aisle vaults which is worthy of note before turning to the more developed churches which resemble Bury in their arrangement of vaulting bays. This is the use of small caryatid figures which appear at the springing of the diagonal ribs (Fig. 23).[171] These would seem to serve a purely decorative purpose, perhaps to distract attention from the great size of the ribs behind them, or to give an apparent lightness to the vault itself by seemingly placing its burden upon such insignificant shoulders, or more probably still, the figures served to break the transition from shaft to rib by concealing the impost of the latter. Whatever their explanation, other examples besides those at Bury are to be seen. Of these, the angels--now badly mutilated--at the base of the ribs in the narthex of Saint Ours at Loches (Indre-et-Loire) (Fig. 24)[172] are especially interesting, and perhaps account for the tiny figures employed at the springing of the ridge ribs in a number of churches in Anjou, such as Angers, Saint Serge (Fig. 21), as well as for the larger figures in the apse of Notre Dame-de-la-Couture at Le Mans (Fig. 20).[173] It may even be through the influence of such figures as these that grotesques were used to support the small shafts in the arcade of the triforium passage in the cathedral of Nevers (Nièvre) (Fig. 25).

SQUARE NAVE AND RECTANGULAR AISLE BAYS CONTINUED

Returning to the churches later in date than Bury but vaulted on the same plan, it will be found that there are but few examples in France, an interesting fact for which an explanation will later be attempted.[174] The lower story of the Sainte Chapelle at Paris (cir. 1250) furnishes one of the rare examples, but here the nave and aisles are of the same height and so do not exactly resemble the system at Bury. Because of their narrowness, the side aisle vaults of the Sainte Chapelle did not furnish proper abutment for those of the nave, and the builders found it necessary to add tie-rods and even transverse half arches forming veritable interior flying buttresses at about half the height of the transverse ribs. This is, however, a most unusual arrangement.

It was in Italy more than elsewhere that the method of square nave and rectangular aisle bays was adopted. Many of the largest churches of the Gothic period in that country were thus constructed. Among these, Santa Maria Novella at Florence (end of the thirteenth and beginning of the fourteenth centuries) has nave bays which are practically square, while the cathedral of Santa Maria del Fiore (fourteenth century) in the same city is a much larger church more strictly following the type.[175] This vast edifice presented such a vaulting problem that the builders did not hesitate to resort to the use of iron tie-rods to counteract the thrusts,--a subterfuge common enough in Italian architecture, of which the church of the Frari at Venice (after 1250) presents an exaggerated example.

LIGHTING PROBLEMS IN NAVES WITH SQUARE BAYS

Several factors enter into the lack of popularity of the vaulting system just described especially in the more northern countries, but the fundamental one would seem to be the difficulty of properly lighting churches thus covered. If an examination be made of the churches with a single broad nave and no aisles it will be seen that in Italy, where a comparatively small proportion of window space was necessary, the builders were content with a single window in each nave bay as for example, in San Francesco at Assisi. In France, on the other hand, the light thus admitted would have proved inadequate, and in such churches as the cathedral of Angers (Fig. 19) and Sainte Radegonde at Poitiers two windows were introduced under each wall rib. This is, however, an awkward arrangement because these windows do not properly fill the wall space, and though this is better accomplished by adding a circular window above the upright pair as was done in La Couture at Le Mans (Fig. 20), still the effect even then is not satisfactory and much solid wall which might be utilized for windows is wasted. Moreover, in a church with side aisles, the clerestory arrangement was still more troublesome since important structural difficulties were involved. To raise a great four part vault high above the aisles in order to obtain a large clerestory was no easy task because of the excessive thrust which such a vault exercised at its four points of support. In Italy, where the amount of light required was not great, a very low clerestory with small, circular windows, one to each bay, was all that was essential, and so in such churches as Santa Maria Novella and the cathedral at Florence the nave vault was placed at a point only slightly above the vaults of the aisles, and its thrusts offset by simple ramping walls beneath the side aisle roofs. Such a church in France would have been inadequately lighted, and even if a greater structural skill permitted the French to erect loftier clerestories than those in Italy, there remained the difficulty of arranging the windows to get the maximum of light and the best appearance. A single opening occupying the entire space beneath the wall rib would have been all head and no jamb. One upright window would have admitted too little light for a large nave, and two windows near together not only left a great deal of wall space unused but were most awkwardly placed in churches where one nave bay corresponded to two bays in the aisles as in Le Mans cathedral,[176] because they were not on an axis with the arches of the nave arcade. On the other hand, if placed on this axis, the resulting windows were necessarily of small size like those in such Rhenish churches as the cathedral of Speyer where a second stage of windows has been added one in the center above each lower pair in a far from satisfactory manner since it brings a window above the intermediate pier.

ORIGIN OF SEXPARTITE VAULTING

In view of these facts it is at least a reasonable assumption that the lighting problem had much to do with the discarding by the French builders of the simple square four-part nave vault. As a matter of fact, however, they did not exactly discard it, but evolved from it a vault in six cells, which, while it still retained the old division of the nave into square bays, each corresponding to two bays in the aisle, at the same time permitted the uniform treatment of these in elevation and made possible larger windows,--one to each aisle bay,--symmetrically placed and, in the course of time filling the entire space beneath the wall ribs. This six-part ribbed vaulting would seem to have originated early in the twelfth century, in the French province of Normandy. This province has already been mentioned as the center of a Romanesque school, which extended over the greater part of England after the conquest of 1066, and reached its height during the reign of Duke William, the Conqueror (1035-1087), when a vast number of churches were constructed, many of them of large size. These were in general wooden roofed throughout, though, occasionally, as has been shown,[177] groined vaults were used in the choir or aisles, or both. Toward the beginning of the twelfth century, however, the Norman builders determined to vault the naves of a number of these churches, among them the two abbeys at Caen, and the result of this determination was the evolution of the true and false six-part vault.

Like the Rhine provinces, Normandy had always been strongly influenced by the methods of building developed in Lombardy. Whether this was due to the presence in Normandy of such men as Lanfranc,--who was born in Pavia in 1005 and became successively prior of Bec (1045-1066), abbot of Saint Étienne at Caen (1066), and archbishop of Canterbury (1070-1089), and who may have kept Normandy closely in touch with Lombardy,--or whether there were other more powerful influences, it is impossible to state, but in any event the architectural analogies between the two schools are striking. This is especially true of the type of shafted pier most frequently found in Normandy, and of the alternate system of light and heavy supports, which, while it does not characterize all the churches of the school, is found in many of them. Thus when the Norman builders determined to vault their great churches at Caen, one would naturally expect to find them turning to Lombardy for a method of vault construction, especially since Sant’ Ambrogio at Milan had been successfully completed at least a quarter of a century before their determination was made. And in fact this is probably what they did. But there were certain differences in structure between the churches of the two schools which made it impossible for the Norman builders to adopt unchanged, the heavy square, domed-up, cross-ribbed vaults of Lombardy. The first of these differences lay in the fact that the Norman churches were originally built for wooden roofs,--which may even have been in place, in many cases, when the vaults were begun,--while the Lombard churches were planned from the ground for their vaulting. The second difference was, that the Norman interior system possessed a clerestory window of considerable size centered above each of the arches opening into the side aisles,--that is two in each wall of what would be a square nave bay,--while the Lombard churches either had no clerestory at all, as at Sant’ Ambrogio, or one in which the windows were small and there was no attempt to center them as in San Michele at Pavia.

It was natural that the Norman builders should have preferred to preserve their interior and exterior elevations as nearly as possible as they were when only a wooden roof was used, both to avoid the expense which would be involved in reconstruction and to preserve the large clerestory so essential in a northern country. To vault these churches and at the same time save this clerestory would seem to have been the problem, therefore, which the builders set themselves to solve. That they attempted to use the four-part vault in its solution will be seen from an examination of the seven vaulted churches[178] still remaining in which the old system of square nave bays is found, for in four of these a variant of four-part ribbed vaulting was employed while in the other three a new method was developed out of the four-part type.

A study of the two abbeys at Caen will illustrate this. Of the two, Saint Étienne or the Abbaye-aux-Hommes (cir. 1064-1066) would seem to be the earlier as far as its vaulting is concerned and this would seem to date from about 1135. In its nave (Fig. 26) the alternate system of supports is employed, though all the piers are of almost the same section with a single shaft carried up the inner face. The aisles are in two stories and there is a clerestory with a single window in each bay. The nave was originally covered with a wooden roof. With this elevation existing before the church was vaulted it is quite possible to account for the form which this vaulting assumed. The first step must have been to divide the nave into square bays by transverse arches,--assuming that these were not already in place. The springing of these arches must naturally have been governed by that of those which opened into the crossing, and the level of their crowns, by the wooden timbering of the roofs,--which may well have been in place when the vaults were built. The result was that these transverse arches had to rise from a point as low as the clerestory string-course and could only be a slightly stilted semicircle in elevation. If the

bays thus constituted were to be covered by four-part vaults of Lombard type, the next step would have been to erect diagonals of semicircular section thus doming up the vault at the crown, but at Caen such diagonals would have rendered necessary an entire change in the timbering of the roof because their intersection would have risen above the level of the trusses. Hence segmental diagonals were substituted. Upon this skeleton of ribs, it would have been quite possible to place a four-part vault, but the wall intersection of its panels would have cut off the heads of the clerestory windows. Several methods could have been used to avoid this. In the first place the severies could have been so shaped as to cut the walls in a curve above the window, but this would have given a flattened form to the panel and rendered it most difficult both to construct and to support when in place. A second expedient would have been to reduce the size of the windows but this, besides cutting off most necessary light would have utterly destroyed the splendid proportions between the horizontal divisions of the Norman interior. A third method would have been to move the windows toward the intermediate pier, but this would have destroyed the axis line of the aisle, triforium, and window arches, and was wisely rejected. Lastly the imposts of the ribs could have been raised, but even this would have introduced enormous structural changes: first, because it would have rendered necessary a change in the timbering, or else raising the entire roof of the church; second, because it would have placed the new impost out of level with the crossing arches; third, because it would have greatly increased the thrust of the vault, already most difficult to meet because of the segmental form of the diagonals and the lack of extensive knowledge of buttressing principles on the part of the Norman builders.

To avoid all these difficulties and still retain the windows, a new method of vaulting was evolved. An intermediate transverse arch was added meeting the diagonals at their intersection, and above the triangular window cells thus formed, separate vault panels were constructed (Fig. 26). The line of the window heads was thus left undisturbed and the six-part vault created (Plate I-c).

FALSE OR PSEUDO-SEXPARTITE VAULTING

Of course, the foregoing suggestion that the six-part vault was evolved from four-part vaulting is largely conjectural, but an examination of other churches in Normandy would seem to show that the Norman builders almost always preferred to use the simple four-part vault in a slightly modified form whenever it was possible to do so and still retain the clerestory windows, rather than to employ the developed six-part type. This modified four-part vault may properly be termed false or pseudo-sex-partite. That it was not a mere prototype of the more developed six-part form would seem to be shown by the fact that it was built in churches both contemporary with, and subsequent to those with true six-part vaults.

A good example of pseudo-sexpartite vaulting, for comparison with that of Saint Étienne (Fig. 26), is afforded by La Trinité or the Abbaye-aux-Dames at Caen (Fig. 27). It would seem probable that the upper portions of this church were extensively rebuilt at the time when vaulting was added. In this rebuilding, concealed flying-buttresses were constructed beneath the side-aisle roofs, and these, together with the solid wall which

replaces the open triforium gallery of Saint Étienne, made it possible to raise the level of the transverse arches of the vaulting to a point considerably above the clerestory string-course. Furthermore, since the wooden outer roof was probably built after the vaults, it was possible to use diagonals whose crowns were higher than those of the transverse arches, and still place them beneath the roof trusses. With such a skeleton of ribs as a basis, the builders proceeded to erect a four-part vault over each nave bay, or, in other words, enclosing two side aisle arches. Because of the higher impost of the vault ribs, the wall intersection of the vault cells easily cleared the window heads.[179] Curiously enough, however, the builders connected the intermediate piers with a transverse arch having a flat wall built upon it to the level of the crown of the longitudinal vault cells (Fig. 27). There would seem to be several explanations of this innovation. In the first place the pier system of La Trinité is regular, not alternate, and a greater symmetry was obtained by having corresponding transverse arches connecting each pair of opposite piers. Moreover such arches had been used before 1114 in the church of Saint Georges at Boscherville, and quite possibly elsewhere as well,[180] beneath a simple wooden roof, thus tying together the lofty clerestory walls. In the second place, such arches had already been introduced at Saint Étienne, though for a different reason, as has been shown, and must have proved of value in keeping the keystone of the diagonals rigidly fixed, besides having become a characteristic of what was perhaps the major church of the school; and in the third place, such an arch with its wall above aided materially both in carrying a portion of the weight of the vault to the alternate piers and in affording permanent centering, which was needed in Normandy even more than in Lombardy because the Norman vault crown was never more than slightly domed up.

Once introduced, this pseudo-sexpartite vault was not restricted to La Trinité but was, as has been said, employed in no less than four of the seven square-bayed Norman churches. At Ouistreham (Calvados)[181] (vaulted cir. 1160), the impost was raised as in La Trinité and pointed transverse arches were used, thus increasing the curve of the diagonals and improving the stability of the vault. More interesting still, however, are the two churches of Bernières-sur-Mer,[182] and Saint Gabriel (Calvados)[183] (both vaulted cir. 1150), for in them the builders have clung so tenaciously to the pseudo form in preference to the true that they have actually moved the windows of each bay toward the intermediate pier in order to use this method without raising the imposts. The latter is particularly interesting because of the extreme flatness of its diagonals for which the intermediate transverse arches must certainly have proved an added support.

The preference of the Norman builders for this pseudo-sexpartite vault, even to the extent of moving the windows out of center to make its use possible, may find a further explanation than any yet given in the simplicity of its construction. A comparison of one window severy of Saint Étienne (Fig. 26) with one at La Trinité (Fig. 27) will illustrate this point. In the former the surface of the vault is warped on either side of the window, while in the latter, the stone courses run almost directly back to the wall, so that the line of intersection is approximately the projection of one-half of the diagonal rib. Of course this second surface was far easier to calculate geometrically and could be put in place by less skillful builders than the warped surface required. It had, however, the fault of being in ill accord with the curve of the window head, but, on the other hand, it possessed the structural advantage of distributing the thrust of the vault over a large amount of exterior wall. This might seem a fault rather than an advantage, were it not that in such a primitive system as that of Normandy, thickness of wall was the greatest factor in abutment and thrusts which were widely distributed were thus more easily met than those which were concentrated within narrow perpendicular limits.[184] The advantage of the warped system in thus concentrating the thrusts was, in fact, realized only when inert stability which forms the keynote of Norman work gave way to the carefully balanced thrusts and counter-thrusts of Gothic architecture.

The little church of Le Petit Quévilly (Seine-Inférieure)[185] (cir. 1156) would seem at first to disprove this Norman preference for pseudo-sexpartite vaults. The imposts of its arches are sufficiently high to permit of such a type, yet the real six-part vault was employed. The explanation of this would seem to lie in the geographical situation of the church, for it is not in Calvados, like the other examples, but in Seine Inférieure near Rouen, or in other words on the border of the Ile-de-France, where the six-part vault had been adopted with enthusiasm and used as early as 1140, or some fifteen years previous to the building of Petit-Quévilly, in the large abbey church of Saint Denis.

It is also difficult to explain the use of the true form in the seventh of the vaulted churches, which is that of Creully (Calvados),[186] but the fact that it has the same low imposts as Saint Étienne at Caen combined with the evident purpose of the builders to keep the windows in the center of the bays may perhaps furnish an explanation of its appearance here.

DEVELOPMENT OF SEXPARTITE VAULTING

The true six-part vault, as used in Saint Étienne, was far from being perfect. In the first place, it possessed a number of inherent structural faults. These lie chiefly in the unequal distribution of thrusts, and the unequal size of the panels into which the vault is divided. From an aesthetic point of view, two other faults might be added: first, the decrease in the apparent length of the nave, due to the fact that it was divided into a few large bays, instead of twice as many smaller ones; and second, the fact that the crowns of the vault cells above the windows do not run out perpendicularly from the clerestory wall but at an awkward angle, thus greatly injuring the symmetry of the bays. Yet in spite of these drawbacks, which were common to all six-part vaulting, this system had a long period of popularity. There are, however, certain structural weaknesses in these early Norman vaults which were largely due to lack of experience on the part of the builders, and not to the form of the vaults themselves. Wall ribs were, for example, omitted, and the diagonals were made of segmental section, thus rendering unnecessarily severe the thrusts of the vaults. Moreover, such a church as St. Étienne was not planned from the ground for vaulting and the piers had not the proper arrangement of shafts. Last of all, the intermediate arches were of a rather ugly, stilted character, possibly so constructed with an eye to a better distribution of light, but in any event presenting an awkward appearance. All these faults were gradually overcome in the Transitional and Early Gothic churches of the Ile-de-France.

SEXPARTITE VAULTING IN THE ILE-DE-FRANCE

That it should have been this province which favored the six-part system is most curious, for at a date almost contemporary with St. Étienne at Caen, ribbed vaults of rectangular plan had probably been constructed over the naves of Saint Étienne at Beauvais and the abbey church at Saint Germer-de-Fly (Oise) (cir. 1130-40). That this method was abandoned in most of the remaining Transitional churches would seem to have been due to the fact that the vaults of Saint Étienne at Beauvais fell in, and those of Saint Germer did not prove very secure.[187] Such builders as the Abbot Suger of Saint Denis, therefore, may very naturally have looked to Normandy for a method of vaulting, since the vaults of Saint Étienne at Caen had at least remained in place.

Whatever the cause of its introduction into the Ile-de-France may have been, the six-part system was used at Saint Denis (Seine) (1140-1144) and soon became the favorite method throughout the neighboring region. Unfortunately Saint Denis and two other important churches of the Transition, the cathedrals of Senlis (Oise) (cir. 1150) and Noyon (Oise) (cir. 1140), which would undoubtedly have illustrated the progress in six-part vaulting, no longer have their original vaults, and the cathedral of Sens (Yonne) (1140-1168) (Fig. 28) remains as perhaps the most important example of the early developed type.[188] Its vaults show the great advance made in construction since the completion of Saint Étienne at Caen. The diagonals are semicircular instead of segmental arches, and the transverse ribs are pointed and all of similar curve, giving a more symmetrical appearance and greatly reducing the thrusts. Furthermore the piers are profiled from the ground according to the load which they are to carry, and, last of all, a highly stilted wall rib is added over each clerestory window, completing the skeleton of the vault and making possible a larger expanse of glass and more satisfactory illumination for the interior. Of course, the use of the flying buttress, which had been introduced a short time before Sens was built, contributed enormously to the advancement of vault construction and in large measure explains such an improved form of vaulting as this is. In fact, a heavy clerestory wall was no longer essential to the support of the vault and it was only the fact that a large expanse of glass was not safe from the pressure of the wind, which prevented the clerestory windows from occupying the entire