Having cut the boards and chosen the order of cuts to use when making the panel, the next step was to calculate the number of boards needed and to assemble them.

The number of boards used varied widely. It depended on the size of the panel, that of the tree, the characteristics of the wood type, and also, perhaps, the custom for a given school. The panel used for The Resurrection from the 14th-century Portuguese School,1Tomar (Convent of Christ), cf. Fig. 89 (Painting index, no. 254). which is 4.1 m in height, was made from oak boards measuring about 2 m and assembled using butt joins. The vertical assembly of this panel is, as far as we know, unique.

Study of the relationship between the dimensions of a panel, the number of boards used and the nature of the wood is interesting, but it does not allow us to extract a specific rule.2Editors’ note: It should be noted that until the introduction of the metric system in 1799 an abundance of measurements was employed, each of which could vary from town to town; see Verhoef 1982 and Doursther 1940. We find, for example, that oak is cut into wide boards in Portugal and narrower ones in France – extremely so in the case of the Troyes School. This means that the construction of a wide panel requires a large number of boards when these are laid vertically, as is sometimes the case. Portugal offers the noteworthy example of a single oak board measuring 1700 × 850 × 40 mm.3Vasco Fernandez, Creation of the Animals, Lamego, cf. Fig. 88 (Painting index, no. 312). Another revealing example is mentioned in the texts. On 17 August 1534, the painter Nicolas Dipre (c.1495–1532) was commissioned to paint an altarpiece for the chapel of Notre-Dame de Carduelh in the church of Saint-Martin in Caderousse. He supplied the wood consisting of a single piece of walnut measuring 7 palms in width and 3½ in height4Chobaut 1939: 102. (1729 × 864 mm),5For the conversion of measurements see Mémoires de l'Académie de Vaucluse 1939: 87. based on the standard canne6Editors’ note: Unit of measurement used in southern France and Italy, between 1.70 and 3 m in length depending on the region (see Dictionnaire de l’Académie française). d’Avignon.7Editors’ note: In Flanders there are examples of panels made up of planks placed in different directions, for instance, the triptych for the Lamentation from the church of Watervliet, Belgium; see Goetghebeur et al. 1966.

Although this is of only secondary interest, given the alterations the panel undergoes over the centuries, we often noted significant differences between the boards used in a single support in terms of their width, thickness or height, or of all these dimensions at once.8Editors’ note: Standard sizes had already been introduced during the making of altarpieces in the 15th century; see Jacobs 1989; Van der Straelen 1855. We examined forty panels in which the boards displayed different widths. Twelve came from Spanish schools, thirteen from French, seven from German schools of the Rhine, and only four from Italian schools. The differences were substantial in some cases, such as that of the boards used to make the panel for the Adoration of the Magi from the 14th-century Aragonese School,9Barcelona (Painting index, no. 438). or those used for Eva Prima Pandora by Jean Cousin (1500–before 1593).10Paris (Musée du Louvre) (Painting index, no. 143). In both cases, all the boards differ from one another. The Triptych du Musée de Cluny, from 15th-century northern France,11Paris (Musée de Cluny), triptych, The Flagellation; Calvary; Pietà (Painting index, no. 190). Editors’ note: Goldsmith and Bucklow 1998; Massing 2003; Kempski 2003. likewise consists of five boards, each of differing width. There are fewer panels in which the boards differ in height. One example is the early 15th-century Catalan Crucifixion by the Master of the Cardona Pentecost,12Barcelona, Painting index, no. 865. the panel of which consists of two boards measuring 1090 and 1130 mm.

It is quite common for the thickness of the boards to vary, with quite pronounced differences in some cases. Portuguese,13Master of Paraíso (now attributed to the workshop of Gregório Lopes (c.1490–1550)), altarpiece, Lisbon (Painting index, nos. 277–280). Aragonese14Juan de la Abadia, St Catherine Altarpiece, Barcelona (Painting index, no. 885). and Catalan15Catalonia, 13th century, Scenes from the Life of St Andrew, Paris (Musée des Arts Décoratifs) (Painting index, no. 873). panels are perhaps more likely to display this characteristic than those from other schools. Some 15th-century German panels from the Upper Rhine region also display it, however, such as the Entombment and Incredulity of Thomas from the Passion Altarpiece by the workshop of Martin Schongauer (1448–1491),16Barcelona (Painting index, nos. 340 and 343). which seem to us the most significant.

‘It should be noted’, Cyprien Monget writes, ‘that all this joinery (cabinets, altarpieces, etc.) is gluée, that is to say, joined using fish glue.’17Monget 1898: 169 (text dating from 1388–89).

The earliest method in general use appears to have been assembly by glued butt joins. Altar or door panels were made, Theophilus Presbyter tells us, as follows: ‘First of all, join the boards carefully, one at a time, with the joining tool used by coopers and joiners. Fix them with cheese glue ... Once they have dried, panels made in this way will adhere so solidly that they cannot be separated again by either damp or heat.’18Théophile 1843: 31. We can verify this by examining the panels themselves.

We found the most noteworthy examples in Spain (Figs 68 and 113). The Museu Nacional d’Art de Catalunya in Barcelona has numerous panels, the backs of which are wholly intact. It was in this museum that we found the most interesting documents concerning the construction of the support. The earliest evidence we discovered of assembly using glued butt joins comes from the 12th-century Catalan School.19Barcelona: Christ and the Twelve Apostles (Painting index, no. 869); Christ, the Apostles and St Martin (Painting index, no. 878). The backs of a certain number of Spanish panels from the 13th to the 16th century,20Spain, 15th century, Painting index, no. 436.—Aragon, 14th century, Painting index, no. 883; 15th century, no. 862.—Castile, 15th century, Painting index, no. 897.—Catalonia, 13th century, Painting index, nos. 870, 872, 879, 888 and 889; 14th century, no. 441; 16th century, 465, 470 and 489. and of French,21France, 15th century, Painting index, no. 132; 16th century, no. 143.—Western Provence, 15th century, Painting index, no. 505; 16th century, nos. 3 and 67.—Troyes, 15th century, Painting index, no. 216; 16th century, no. 217. Italian22Florence, 13th century, Painting index, no. 988.—Poland, 15th century, Painting index, no. 561.—Milan, 16th century, Painting index, no. 662. and German23Germany, 15th century, Painting index, no. 933A. panels from the same period also have joins of this kind.

It might be asked whether boards that do not show any trace of assembly were not joined using some other method than glued butt joins, in such a way as to be undetectable from the back or the edge of the panel. We do not think such doubts are justified in the case of the panels we studied, as these are mostly constructed from poorly joined boards – especially the panel of the Apostles and Prophets from the workshop of the Rohan Master (active 1410– 1440),24Musée de Laon (Painting index, no. 132). who belonged to the early 15th-century French School, in which any other assembly method would be visible.

Butt joins are sometimes plugged with filler – a feature detected very often, in fact, in the earliest panels. Examples are the 14th-century Aragonese panel in Fig. 9625St Ursula Altarpiece, Barcelona (Painting index, no. 883). and the series of panels by the Master of Játiva (active 1490–1515) from the Museu Nacional d’Art de Catalunya, Barcelona.26Altarpiece: Scenes from the Life of the Virgin and of Christ (Painting index, nos. 913–919). These traces of filler in joins visible from the reverse side might also come from the preparation of the front of the panel. The ground might have penetrated through the gaps on the front, filling the joins and bonding the boards together once dry.27For another example, see Fig. 105. There are plenty of Spanish texts, meanwhile, that describe this aspect of panel-making. The joins, they state, will be ‘filled ... caulked ... they will be sealed at the front and the back, so that they do not move at any time’.28Madurell Marimon 1944: 155, 166, 172, 205. (We are indebted to J. Ainaud de Lasarle for the information concerning these contracts.)

It is quite often the case that, as we noted earlier, panels do not show any trace of joins between the boards, even though we have reason to believe they were connected by butt joins. The edges of the support do not reveal anything either. This is where panels split between front and back reveal their secret. These skilfully connected boards, the joins of which are sometimes impossible to detect, are generally linked by dowels. These little wooden or iron rods are a form of tenon, which hold and fix pieces of carpentry or joinery together.29Editors’ note: The dowels do not reinforce the join, but were useful during the gluing process as they prevented the planks from slipping out off their position before the glue had coagulated. After this the dowels played no role in the maintenance of the join. They are inserted into a hole gouged into the edge of one board and then into another hole in the corresponding position on the other board.30See the Glossary.

It is very interesting to view a panel assembled in this way. One has a sense of observing the joiner’s hesitations and changes of mind. The split panel with the Tiburtine Sibyl Prophesying the Advent of Christ to Emperor Augustus, painted by Konrad Witz (c.1400/10–c.1445/46),31Dijon (Painting index, no. 49). the St Vincent Polyptych by Nuno Gonçalves32Lisbon (Museu Nacional de Arte Antiga), Painting index, nos. 298 and 303. (active c.1450–1490) (Fig. 98) and Duccio’s Maestà in Siena Cathedral (Fig. 99)33Cf. Van Marle 1924–1938, I: 365; II: 4, 7, 8,15, 20–60, 61, 64, 71, 75, 92, 131, 177, 292. are particularly revealing. Dowels that do not continue on the other side of the join indicate a change of mind on the part of the craftsman who made the panel. The latter must have been disassembled and then put back together, the traces of which can still be detected. In some cases, however, these marks could also be evidence of previous assembly.34The reader is encouraged to supplement this list by consulting the Painting index, which includes many more examples.

The panels of the St Vincent Polyptych by Nuno Gonçalves all bear the same dowel traces on the back. The location of the dowels can provide general assistance in reconstructing the original composition of an altarpiece. A detail from the edge of one of these panels (Fig. 97) allows us, moreover, to examine how a dowel used for assembly was cut – roughly, it would appear – using a knife.

Where a dowel has been lost, the open hole left behind offers valuable evidence of the construction method. When they are revealed on the backs of split panels, they have the appearance of elongated grooves (Fig. 75). Similar traces can be seen on the edges of panels that have not been split (Figs 88 and 97). Assembly using dowels, which is impossible to detect when the panel is well joined, means we have to exercise considerable caution in our conclusions.35Editors’ note: Marette omits mentioning radiography (see the Introduction to this publication and the Introduction to Part 1) as a technique for detecting dowels in the thickness of wood panels, along with much other valuation information about panel assembly and condition. Distinguishing between assembly with dowels and with well-glued butt joints is virtually impossible,36Editors’ note: The introduction of X-radiography of panels helped to answer this type of question, as well as providing significant new information on panel-making practice in general. although some assistance is provided by the period of the painting, and by the greater perfection of dowelled joins.

Although this panel construction method was common in Europe in the 16th century, it already appeared in the 15th and even the 14th century, if we are to judge by the aforementioned Maestà panel by Duccio in Siena (c.1255/60–c.1318/19). If dowelled joins were already being used in Duccio’s time, we might justifiably wonder whether the perfect construction of 15th and even 14th-century Italian panels is not due to most of them already being assembled with dowels.

Butterfly keys are small pieces of wood cut with dovetails at either end. They remain in common use today to join the separate boards making up a panel. They are positioned on the back and recessed across the join between the boards, which they reinforce (Fig. 89).38Editors’ note: It should be noted that the use of butterfly keys was mainly at the front of panels; finding butterfly keys that are visible on the back often indicates a later addition/treatment for purposes of reinforcement. Butterfly keys of this kind serve a different function in the panel of the 16th-century Aragonese Adoration of the Magi.39Barcelona (Painting index, no. 438). Here they play an essential part in the assembly of the panel, rather than merely reinforcing joins already made. These butterfly keys can be detected on the face of the panel (Fig. 90). They are recessed halfway into the wood and do not come through to the back.40Editors’ note: Butterfly keys inserted in the face of the panel in Flemish painting can be seen in a triptych by Master John, The Lineage of Saint Anne, c.1513–1517; see Currie 1996: 119. Viewed from that side of the panel, the boards seem held together simply by glued, butt joins (Fig. 91).

An integral tenon cut into a single dovetail works on the same principle (Fig. 91). Joins like this were used two centuries earlier to assemble the boards of the panel for the Coronation of the Virgin with Adoring Saints by Jacopo di Cione (c.1325–after 1390),41Florence (Galleria dell’Accademia) (Painting index, no. 573). and were a prototype of sorts. The particular feature of this system and the way it differs compared to the previous one lies in the shape of the tenons and their degree of movement. They are cut into dovetails on the board itself, and so form part of it, similar to the pieces of a jigsaw. They are placed in mortises cut into the board to be joined. The butterfly key, by contrast, is a separate element.

Panels assembled using either of these procedures are cut on the front, and so appear perfectly uniform from the back (Figs 91 and 93); it is in this respect that they resemble one another. This method for joining boards is interesting for both its great age and because of the unique nature of these two examples.

Tongue and groove joins are quite common. The edge of one of the two boards is cut into a continuous tenon, while that of the other board is cut into a continuous mortise.42See the Glossary. The spline is a variation on the same method. In this case, the two edges of the boards to be joined are cut into a mortise and the join is made using a wooden baton recessed into the two mortises.43See the Glossary. This method, which was rare in the Middle Ages, is very common in modern joinery. We nevertheless found examples of it in the Rhineland and in Italy in the 15th century, in panels made of poplar and fir.

Both systems are highly visible on the edge of the boards and are readily identifiable. The woodwork for the Studiolo of Federico di Montefeltro (1422–1482), Duke of Urbino, painted by Justus of Ghent (c.1410–c.1480)44Editors’ note: Commonly identified with Joos van Wassenhove. and Pedro Berruguete (c.1450–1503/04),45Fourteen Portraits of Illustrious Men, Musée du Louvre (Painting index, nos. 514–527). was assembled in the same way, as were the Besançon Passion Altarpiece46These panels have been transposed. We had the opportunity of examining them during that process, which enabled us to examine the fittings very closely (Painting index, no. 933). and the Bergheim Predella by the Master of the Stauffenberg Altar and the Housebook Master (active 1475–1490),47Bergheim Predella: Sermon of John the Baptist (Painting index, no. 941). now in the Museum of Colmar. We ought not, however, draw too hasty a conclusion from these observations: other schools may have used the same assembly method, the traces of which were later lost due to the destruction or restoration of the panels. Our findings nevertheless allow us to point out the rarity of similar examples in the relevant period.

Boards assembled using halved cross-lap joins are also rare. We found examples in Spain in the 14th and 15th centuries.48Barcelona (Painting index, nos. 61, 384, 387 and 389). Three panels from the Castilian School Funeral Processions cycle were assembled in this way Fig. 95). This method of construction – cut into walnut boards – differs slightly from the one used to assemble the panel for Lluís Dalmau’s Virgin of the Consellers. The oak boards of this panel have halved joins with a deeper shoulder (Fig. 94). This can be seen by examining the edge of the two panels. It is the variation in the depth of the shoulder, therefore, that distinguishes them.

It should be noted that these methods are inferior to tongue and groove joins as they increase the size of the gluing surface and so risk reducing the strength of the panel. Incidentally, the consequences of such a system can be seen in the photographs we reproduce here. Over time, the joins gradually loosen, causing the paint layer to tear all the way along.49Editors’ note: Half-lap joins are seen when two pieces of wood with perpendicular grain direction are joined. See for example, Rubens’ panels as described in Poll-Frommel and Schmidt 2001.

Once the panel had been assembled, it was painted either across or with the grain of the wood. The grain was generally oriented in the direction of the panel’s longest side. A panel wider than it is high therefore has the grain running across it,50Examples include the Syndics of the Drapers’ Guild by Pieter Pietersz (Painting index, no. 229) and the Banquet of the Crossbowmen’s Guild by Cornelis Anthonisz (Painting index, no. 227), in the Rijksmuseum in Amsterdam; the 15th-century German Calvary (Painting index, no. 933A), in the Musée de Saint-Omer and the Avignon Pietà (Painting index, no. 405) in the Louvre. while one higher than it is long – which is the case for most paintings – has the grain running vertically.51Editors’ note: Marette is here probably referring to altarpieces and neglects landscape paintings. In some cases, however, a very wide panel will also have the wood grain oriented vertically. More specifically, we noted an English School panel of this kind in the 13th century. It is 300 cm wide by 93 cm high, and is made up of twenty vertically assembled boards.52Scenes from the Life of the Virgin, Paris, Musée de Cluny (Painting index, no. 55). This altar frontal most likely originally belonged together with the Thornham Parva retable. Editors’ note: Goldsmith and Bucklow 1998; Kempski 2003; Massing 2003. Examples also exist of horizontally oriented wood grain in a very tall panel, such as the 16th-century French St Lupien in the Musée du Troyes,53Painting index, no. 219. or in the Adoration of the Shepherds by Pieter Aertsen (1508–1575).54Rijksmuseum, Amsterdam (Painting index, no. 228). Editors’ note: It is possible that Marette did not know that this was a fragment (90 × 60 cm). Strictly speaking it is not known whether it was a landscape or a portrait format before it was mutilated (unless the author tacitly considers it a fragment of the high altar of the Nieuwe Kerk, destroyed by the iconoclasts in 1566). These are, however, quite rare.

The orientation of the painting relative to the grain of the wood does not appear to follow any particular rule. This might be because vertically oriented scenes are much more frequent; and it is also possible that artists used the grain of the wood as an element in their paintings. We found, however, that works painted across or perpendicular to the grain were much rarer and were always wider than tall.

Painters may have taken the grain into account even if they did not attempt to exploit the texture of the wood systematically for aesthetic purposes. If so, however, this would only have occurred as of the period in which the painted panel had ceased to be an element of woodwork. Before then, the orientation of the wood grain would, of course, have been chiefly determined by structural considerations. Artists drew and then painted onto a previously applied coating. This explains why the image on the front of 12th- and 13th-century Catalan altars were generally painted in the smallest dimension of the panel, and hence always painted across the grain.55Editors’ note: The editors are uncertain of Marette’s argumentation in this section.