Stephen Walker 26 February 2006

Note: Direct examination of the St. Ninian's Isle Treasure in 2011 and subsequent research and experiments has caused me to modify my conclusions.
 The most significant changes are that I now believe that the mold material was more likely to have been a gypsum plaster and that some of the pieces,
 the sword chape, brooch number 17 and the conical mounts all show evidence of being carved in the positive version. A new paper on the subject was
 presented at the International Conference of Insular Art at the University of York on July 19, 2011. An abbreviated version of that presentation,
 including step-by-step photographs of the process reproducing brooches 24, 25 & 26 will appear on this web site in the near future.

As a craftsman, I have tried unsuccessfully for many years to duplicate the early medieval chip-carving style of cast metalwork. My results typically failed to get the background space as evenly faceted and to accomplish the even and fluid line control of the foreground. It seemed that there must be some missing secret to how this once common technique was executed. While looking at the brooches of the St. Ninian’s Isle Treasure in January 2004 I gained a clue as to what this secret might be. Three of the brooches are nearly identical in their overall layout and size, but the chip-carved detail differs between the three. These silver gilt penannular brooches, discovered on St. Ninian’s Isle in the Shetlands in 1958 are now in the National Museum of Scotland in Edinburgh and are dated to the 8th century. As I tried to imagine what steps the craftsman would have to take in order to modify the designs before they were cast, I realized that the chip-carved detail, which differed from piece to piece, was all in the recessed areas. The high ridges surrounding that detail would make access to carve in the positive version very difficult. Yet this is consistently how chip-carving is found in the Insular tradition, surrounded by high cell walls that would interfere with tool access if the model was being created in a positive version. But in the negative version of the mold, this surface would be the high spot, easily accessible to carving.

Brooch 24

Brooch 25

There are two very well documented methods used to create ceramic molds for casting complex metal objects in ancient and medieval times. Moist clay can be used to make an impression of an object to be copied. In this manner a brooch can be reproduced or a model for a casting can be made of any other hard material and used to create a clay mold that will serve to reproduce that shape. The second method is ŕ cire perdu or lost wax which has advantages for complex shapes. A wax model of the design is made that is covered in clay. When fired, the wax is eliminated; leaving a negative space that is then filled with molten metal. After cooling, the clay is broken away revealing the same shape that was modeled in wax, now in metal

Brooch 26

A third, and more direct method of making a cavity for the casting of metal is to carve the negative shape directly in the mold material. Open-faced stone molds for the casting of ingots and ax head blanks exist from the earliest metalworking sites. Morten Axboe states, “Most of the authors, who have lately been dealing with the manufacture of brooches ect., seem to consider a positive version as the first step in the working process.” Axboe then laid out a theory that negative carving was used as a matrix that created a positive wax model for lost wax casting. This correctly recognized the negative nature of chip-carving. I would like to propose a simpler version of Axboe’s process by suggesting that the chip-carving style of ornament, such as that seen on the St. Ninian’s Isle Treasure is accomplished by directly carving in the open face of the mold. Three brooches from the St. Ninian’s Isle group show similarities and differences that can be explained by a combination of impressing a positive model into clay molds and then adding chip-carved detail by working directly on the mold surface.

Brooches Numbers 24, 25 & 26 all share the same dimensions, overall design and peculiarities of certain details to assume that they all began as impressions from the same model. This model, which I will call the “frame model”, would be the entire silhouette of the brooch, with the circles, cusps and borders of each terminal, but with the space in the recesses left plain. The arcs of the brooch each have five ridges, although the two ridges on either side of the center are modified by carving into a twist pattern on Number 24 and rows of beads on Number 25. Number 26 is left as a series of plain ridges. The three chip carved lobes that extend from each terminal are identical in their shape and dimension, which would indicate that these elements were part of the original common frame model. The cartouche differs considerably between each brooch, but each has the same length and width. For this reason it can be assumed that the cartouche on the frame model was a lower border of the simpler form seen on brooch Number 24 and that this was modified by carving deeper in the molds and by carving the circles in the center of the cartouches on brooches Numbers 25 and 26. The slight differences in dimensions among the three brooches are all within measurements that can easily be accounted for by the process and finishing. The overall width of brooches Numbers 25 and 26 are 70 mm. Number 24 is 65 mm across at the same point. The gap between the terminals are closer on this brooch, but if the terminals were bent to the same gap as the others, all three would be reasonably the same.


Based on these observations I propose that these three brooches were cast by the following process:

1, The frame model was impressed between two slabs of clay. These two halves are keyed in such a way as to allow them to fit correctly together after the model is removed.

2, After the clay was dried enough to be sufficiently rigid, the model was removed.

3, The chip-carved ornament was then engraved in the open face of the mold. A funnel was also carved to give the molten metal a gate to be poured into the mold. Mold fragments and casting flaws suggest that this was typically done at the cartouche.

4, The two halves of the mold where returned together and fired in the manner of pottery.

5, Molten silver was poured into the mold, while the mold was still hot from the fire.

6, After cooling, the mold was broken away to reveal the casting.


St. Ninian's Isle Brooch 17


Bone trial pieces are frequently seen as evidence that these sorts of designs were carved in a positive version to serve as models for impressing molds. Close examination of the quality of these pieces shows that they consistently lack the fluidity of line and crisp, regular angles of the negative space that is typical of the cast metal material. Chip-carving in the negative version in a mold for casting metal differs from the method of chip-carving wood or bone. In wood or bone the space between the raised ridges must be removed. This is done with specialized knives, making several cuts for each space, being very careful to keep the angles consistent from each negative space to the next. When carved in the negative, a simple V gouge can be used to carve grooves, which when cast, become ridges. The angled point of the gouge makes it much easier to maintain consistent background angles than carving in the positive, especially on a small scale. The consistency of the faceting on bone trial pieces is not nearly as controlled as even smaller scale cast chip-carving normally is. By careful spacing of these cuts, a very regular effect of faceting can be achieved. Carved to a depth where each groove touches its neighbors, any flat space is eliminated from what becomes the background in the casting. The regular over and under alternation of interlace is very easily accomplished by simply laying out the path of the strand, carving it successively deeper on several passes and finally establishing the over and under sequence by simply carving the grooves slightly deeper where they are meant to go over in the final cast. A double pointed gouge can be used to make a double line as seen on the larger brooch Number 17. Double line chip-carving gives a greater impression of interlace.

The ridges, borders, bezels and cells partitions in Pictish and other Insular cast metalwork are also shapes which tend to be much more easily crafted by carving the mold than by sculpting a positive model. Carving grooves in the mold material makes ridges and cusps that rise from the body of piece. For this reason it seems very likely that the frame model from which each of these three brooches progressed was itself sculpted in the negative and cast. The structures as well as the detail of chip-carved brooches are all such that negative carving of the mold rather than a positive modeling or carving in wax, wood or some other material could more easily make them. The only reason to suppose that a positive model was involved in any stage of the manufacture of these three brooches is the fact that all three are identical in size and peculiar detail of their overall shape. The chip-carved terminal lobes are so similar for each brooch that these give the strongest suggestion that the brooches have a common pattern. Since these elements themselves are themselves of a nature that suggests negative carving, I would argue that these brooches were sculpted almost entirely by negative carving, but in two stages. The first carved mold being used to make the overall frame of the brooch, which would have been cast in metal, possibly lead, and would then be used as a shortcut to manufacture a second generation of clay molds to cast modified versions of the brooch in silver.

Bone trial piece from Irish National Museum

Complex objects such as the cone-shaped mountings in the St. Ninian’s Isle Treasure, Numbers 12 – 14 and the chapes for sword scabbards, Numbers 15 & 16, all exhibit chip-carving of a nature begun as negative carving. These objects would require a more complex technique than the simple two part molds used to cast flat brooches. Axboe’s proposed lost wax intermediate step offers one possibility of how these more difficult shapes may have been crafted or some combination of intermediate positive models and multiple part molds may have been used. Chip-carving on tubular objects such as the stem of the Ardagh Chalice and some crosier ferrules from the same period exhibit characteristics of negative carving but would require a more complex technique than flat objects such as brooches. The eleventh century Irish crosiers described by Michelli as “mock kerbschnitt” appear to have been executed in the positive model, likely wax. These lack the sharp detail and controlled background observed in negatively carved material. By the eleventh century the negative carving technique seems to have been lost.

Sword cape

Three cast silver cones from St. Ninin's Isle Treasure

My experiments have used both modern materials, such as carving wax, rubber and casting investments, as well as more traditional materials such as plaster and clay. These modern materials have been very useful to understand the physical relationship between the positive and negative versions of the designs and to learn the layout and carving moves to accomplish the technique.

The actual materials originally used present a different problem. The very high degree of precision seen on many examples of this technique indicates that the carving was done in a material with excellent properties. This material must carve cleanly, leaving a smooth surface and also have the strength not to chip out between cuts. We know from Theophilus that the clay used for ŕ cire perdu was tempered with dung. In the negative version, chip carving is a series of small pyramids and sharp ridges. These tend to break out in untempered clay and in clay that is tempered with dung that is carved before it is fired. Clay tempered with dung will carve after is it fired. I believe that the porosity created when the dung burns out, makes this fired material more easily carved than untempered clay. Gypsum plaster works very well, however plaster contains small bubbles that leave a distinctive texture. That texture is not observed on historical examples of chip-carving. Clay tempered with gypsum carves successfully before it is fired. Various glues also show promise as binders for the clay. Although none of the successful clay temper combinations that I have tried yields the same quality as the best historical chip-carving, they work well enough to suggest that some form of tempered clay was very likely the material used for carved molds. Further experiments will undoubtedly produce more refined clay formulas.

Sculpting both positive and negative versions can be demonstrated as viable techniques for making clay molds for metal casting. Chip-carved ornament as well as much of the structural features of Insular brooches can be explained by negative carving in two part molds. The variation on a theme shown in the St. Ninian’s Isle brooches suggests that early medieval mold makers could develop their patterns in several stages and not rely solely on either a positive model or negative carving. Knowledge of the negative nature of this process aids our understanding of medieval molding technique. It also enables modern craftsmen the possibility to reproduce this style of ornament as an efficient and authentic process.

(C) Copyright Stephen Walker, 2006

Link to MacLeod Memorial Buckle chip-carving project

update 2013