Opleiding Conservering en Restauratie, Master, Specialisatie Schilderijen
Masterscriptie
“A study of the influence of stretching methods onto cusping patterns in canvas supports.”
Student: Mirjam Hintz 10383913
Universiteit van Amsterdam
Scriptiebegeleiders: Drs. Maartje Stols-Witlox Emilie Froment
Bill Wei
Table of Contents
ACKNOWLEDGEMENTS ...2
ABSTRACT ...3
DEFINITION OF KEY TERMS ...4
INTRODUCTION
...5
1.1 CUSPING AS ART TECHNOLOGICAL TERM...5
1.2 NECESSITY OF RESEARCH...6
1.3 RELEVANCE TO THE CONSERVATION FIELD...8
2 PREVIOUS RESEARCH...9
2.1 INDIVIDUAL CANVAS PREPARATION...10
2.2 COMMERCIALLY PREPARED CANVASES...12
2.3 SECONDARY CUSPING...13
2.4 WARP AND WEFT...14
2.5 FIBER TYPE AND WEAVE PATTERN...16
2.6 CANVAS PREPARATION...16
2.7 STUDY OBJECT ST. FRANCIS OF ASSISI...18
3 RESEARCH QUESTIONS AND RESEARCH PLAN...20
4 METHODOLOGY...22
4.1 PREPARATION...23
4.2 STRETCHING OF TEST CANVASES...23
4.3 VISUAL DOCUMENTATION...25
4.4 QUANTIFICATION & EVALUATION OF RESULTS...26
5 TEST CANVASES...30
5.1 CLAESSENS ARTISTS' CANVAS...31
5.2 DIE LEINENWEBER'S OLD LINEN...33
5.3 COMPARISON OF THE TEST CANVASES...34
6 EXPERIMENT RESULTS AND INTERPRETATION...38
6.1 EXPERIMENT I: LOW, MEDIUM AND HIGH PULLING FORCE...38
6.2 EXPERIMENT II: WEAK, MEDIUM AND STRONG SIZE...44
6.3 EXPERIMENT III: POSITION ON THE CANVAS BOLT (CENTER PIECE, END OF BOLT, SELVEDGE)...49
6.4 DISCUSSION...53
6.5 STUDY OBJECT REVIEWED...54
7 CONCLUSION...57
GLOSSARY OF TERMS...62
LIST OF IMAGES...66
APPENDIX I – IMAGES OF THE STUDY OBJECT...71
APPENDIX II – PROTOCOL OF THE CANVAS ANALYSIS...74
CLAESSENS ARTISTS’ CANVAS...74
PROTOCOL DIE LEINWEBERCANVASANALYSIS...76
PROTOCOL ST. FRANCISCANVASANALYSIS...78
APPENDIX III - EXPERIMENT CONDITIONS, TOOLS &
MATERIALS...80
APPENDIX IX – EXPERIMENT RESULTS IN DETAIL...82
I.1 High pulling force machine-wc...83
I.2 Medium pulling force machine-wc...84
I.3 Low pulling force machine-wc...85
I.4 High pulling forces hand-wc...86
I.5 Medium pulling force hand-wc...87
I.6 Low pulling force hand-wc... 88
II.1 Weak size, machine-woven canvas...89
II.2 Strong size, machine-woven canvas...90
II.3 Weak size, hand-woven canvas...91
II.4 Strong size, hand-woven canvas...92
III. 1 Selvedge on the left, hand-woven canvas...93
III.2 Selvedge on the left and bolt end at the top, hand-woven canvas...94
III.3 Selvedge on the left, machine-woven canvas...95
APPENDIX X- REPEATEDEXPERIMENTS...96
I.2* Medium pulling force machine-wc, unused version...97
II.1* Weak size, machine-woven canvas, unused version...98
II.2* Strong size, machine-woven canvas, unused version...99
II.4* Strong size, hand-woven canvas, unused version...100
APPENDIX XI – CD WITHIMAGESOFTHETESTCANVASES...101
Firstly I would like to thank Drs. Maartje Stols-Witlox for her support and supervision of this project. My grateful thanks are also extended to Dr. Bill Wei, who provided me with skilful technical and scientific support and much practical advice. I would like to offer my special thanks to Emilie Froment for her enthusiastic encouragement and useful critiques of this research work. Sincere thanks go to Dr. Robert Erdmann for inspiring this project and for his valuable and constructive suggestions during the planning of this research work. I am also grateful to Dr. Rene Peschar for his mathematical advice and supportive feedback. Lastly I would like to give my sincere thanks to Dr. Jørgen Wadum, Dr. Norman Tennent, Dr. Maarten van Bommel, and Drs. Rene Lugtigheid for the discussions and exchange during the research proposal meetings.
Keywords: cusping, linen canvas, strain, stretching techniques, canvas analysis
The objective of this paper is to improve the understanding of cusping mechanisms in the stretching of canvas supports. Valuable information about the artist’s working method during stretching and canvas preparation as well as alteration to the dimensions of a painting can be derived from the analysis of a painting’s cusping patterns. With this study a beginning is made in the development of a more systematical approach to cusping analysis. In a series of stretching experiments it is investigated how cusping is influenced by the amount of pulling force applied during stretching, the concentration of rabbit skin glue solution during sizing, and a canvas’ position on the bolt from which it was cut out off. Strain was visually recorded with before and after stretching photographs and measured digitally with the image-processing tool Adobe Photoshop CS5. The results of this study emphasize that cusp formation greatly depends on a canvas’ structural characteristic (in particular weave density and crimp) and its absorption capacity. This contribution is hoped to be of interest to the professional field of technical art history and conservation and restoration of canvas paintings.
• Pulling force: force that is exerted onto the canvases, which causes canvases to stretch. • Tension: magnitude of the pulling force that causes fibres and yarns in canvases to move further apart: stretched state. • Stress: force per area. • Strain: deformation of canvases from a reference state (initial un-stretched state) to a deformed state, caused by the pulling force. •
Stretch: lengthening, widening of elastic material. •
Tensile strength: (or ultimate tensile strength) is the amount of stress a material can
withstand before failing.
Many other specialized terms are not explained in the text and, therefore, are listed in the glossary of terms (p. 59).
Abbriviations:
avg: average
hand-wc: hand-woven canvas
machine-wc: machine-woven canvas
INTRODUCTION
1.1 CUSPING AS ART TECHNOLOGICAL TERM
In painting conservation the term ‘cusping’ refers to the condition of the canvas support at its edges. Where the canvas is attached to an auxiliary support and stretched, the canvas is put under tension. The point-like stretching of the canvas material causes displacement of
Hintz * UvA * 2014
Fig 1. Left: Magnified detail from a x-ray of Falling Leaves (F651) by Vincent van Gogh.
Right: Reversed grey scale image of 5
threads towards the edges of a canvas because of lateral forces caused by stretching.1
So called ‘primary cusping’ occurs along the edges of a canvas in that format in which it was originally stretched. The distortion is increased by the application of aqueous glue layers, which shrinks the fabric and essentially freezes the thread movement into place. Commonly canvases are prepared in larger formats and then cut down and mounted on smaller stretchers before painting. During this second stretching process ‘secondary cusping’ can occur. Cusping patterns are visible in x-rays if a paintings’ ground layer contains lead white. In that case the ground, which is pressed into the structure of the canvas acts as a mould and renders an inverse image of the textile support (fig 1).
1.2 NECESSITY OF RESEARCH
Cusping analysis can provide valuable information on the making process of a canvas painting support. However, the current ability to interpret cusping patterns has not yet reached its full capacity, and more information could be gained from its study. Moreover, cusping interpretation is largely based on assumptions, which are based on object studies, written art technical sources and simply common sense, but not on scientific evidence. This is aggravated by the fact that the reverse sides of many paintings were covered during lining treatments, whereby the original tacking margins were frequently removed.2 If no x-ray images can be taken of lined canvases, the
analysis of the original support is impossible.
In the 1970s, re-evaluation of structural treatments first triggered a wider 1 Johnson, Don H., e.a. ‘Automated Thread Counting’. In: Vellekoop, Marije, e.a. Van Gogh at Work. München: Hirmer, 2013, p. 147.
2 Van de Wetering, Ernst. Rembrandt: The Painter at Work. Revised Edition. Berkeley & Los Angeles: University of California Press, 2009, p. 93.
interest in the study of canvas supports. A new trend towards minimalism regarding structural treatments urged the need for a better understanding of the physical properties of canvases.3 Hedley, Berger, Mecklenburg and Michalski carried out
pioneering work in this field.4 Young has contributed much to our knowledge on artist
canvases by conducting research into the history, design and physical properties of fabric supports and by developing methodologies for assessing those.5 Two projects
that incorporated exceptional in-depth analysis of canvas supports are the Rembrandt Research Project (RRP), launched in 1968, and the Van Gogh project, launched in 1998. Another major effort in canvas analysis is being made within the scope of The Thread Count Automation Project (TCAP). TCAP was initiated by C. Richard Johnson, Jr., Don H. Johnson and Robert G. Erdmann in 2007. Within the project so called ‘angle maps’ were discovered to be a useful tool in analysing cusping, because they give a clear overview on the amount of cusping and its distribution over a whole painting.6 The automated analysis procedure makes use of an algorithm for counting
threads from x-rays. The technique is described in detail elsewhere.7
Although there are well-established ways of interpreting cusping patterns, a number of phenomena have been observed in angle maps, for which there are no explanations yet. For instance, it is not clear how the irregular cusping in Vermeer’s
The Art of Painting precisely comes about. Here cusping fades out from one side to
the other (fig 2). Such irregularities suggest, that the sequence with which pulling forces are applied during the stretching of a canvas influences its cusping pattern.8
3 Ackroyd, Paul. ‘The Structural Conservation of Canvas Paintings: Changes in Attitude and Practice since the Early 1970s’, Reviews in Conservation, no.3, IIC, 2002, p. 9.
4 Young, Christina. Accelerated ageing of fabric supports: is it possible? Keynote Paper. AHRB Textile Conservation Centre Conference Post Prints, October 2005, Archetype Books, pp 111-116. 5 Young, C. ‘History of fabric supports’. Stoner, Joyce Hill, Rebecca Rushfield (ed.). Conservation of
Easel Paintings. Oxon: Routledge, 2012.
6 Angle maps are false colour images, which illustrate the direction and degree of thread deviation from the average thread direction.
From: Johnson, Richard C., e.a. ‘Interpreting Canvas Weave Matches’. ArtMatters International
Journal for Technical Art History, no. 5, 2013, p. 56.
7 Erdmann, Robert, e.a. Reuniting Poussin’s Bacchanals Painted for Cardinal Richelieu through
Quantitative Canvas Weave Analysis. Presented 29 June 2013 at the AIC Annual Meeting,
Indianapolis. 24 January 2014.
<http://erg.mse.arizona.edu/Erdmann_Reuniting_Poussins_Bacchanals.pdf >.
8 Liedtke, Walter, e.a. ‘Canvas Matches in Vermeer: A Case Study in the Computer Analysis of Fabric Supports’. Metropolitan Museum Journal, no 47, 2012, p. 101.
In cases where there is no reason to assume that a paintings’ original size has been altered, and physical examinations of the painting indicates that the support has been stretched and primed individually, cusping being different on one side than on the others cannot be explained. This is for instance the case for Van Gogh’s People
strolling in a park (fig 3).9
1.3 RELEVANCE TO THE CONSERVATION FIELD
The study of cusping patterns provides information about the artist’s intent and the restoration history of a painting. Therefore, this project mainly serves the field of technical art history. However, cusping analysis is also relevant for the restoration and conservation speciality. A paintings’ cusping pattern is primarily studied if there are questions about its original format. Conservators might consider readjusting a painting’s size by adding reconstructions, for instance. Canvas analysis is also an important tool in identifying fragments of cut paintings, which can lead to their reunification.
As cusp formation is directly linked to the stress distribution in canvases, this 9 Van Gogh at work. P. 160.
Fig 2. Weft-thread angle map of Johannes Vermeer’s The Art of Painting, ca. 1666– 68. Oil on canvas, (120 x 100 cm). Kunsthistorisches Museum, Vienna (L26).
study could, in the long run, also help the condition assessment of textile supports. At present, there is no effective non-destructive method to assess the tensile strength of canvases. There are destructive tear tests, but for those, large sample-pieces are required.10 Tensile strength is particularly of interest when paintings need to be
re-stretched or keyed out. A better general understanding of textile painting supports would furthermore aid in the risk assessment of conservation treatments, which involve the use of water. If the inherent tensions in a textile support are unknown, then its reaction to aqueous treatments is unforeseeable.
Once the basic principles in the creation of cusping patterns are understood, then knowledge can be used to interpret cusping patterns observed in paintings. Thus, should there be a provable regularity in canvas cusping it would become possible to deduce more reliable information from it.
2 PREVIOUS RESEARCH
While most efforts in textile research are made to improve fabric properties for production, only a limited amount of research projects were carried out in the interest of technical art history and conservation. In this state of research, recent literature concerning canvas painting supports is reviewed. Thereby only conservation literature is considered relevant and thus here included. The issue of cusp formation is looked at from various angles, which gives an impression of the complexity of the topic. At first an overview over the research outcomes of the TCAP project will be given in which it will be explained what conclusions have been drawn from weave and angle maps so far. In order to understand these phenomena better, seventeenth-century practices with 10 Rouba, Bogumila J. ‘Die Leinwandstrukturanalyse und ihre Anwendung für die
regard to the stretching of painter’s canvases will be discussed. Aspects such as differences in warp and weft threads, fiber type, weave pattern and finishing of linen textiles will thereby be taken into consideration. This review also includes recent research into the influence of canvas preparation onto tensions in canvases.
2.1 INDIVIDUAL CANVAS PREPARATION
It is generally assumed that individually prepared canvases will have primary cusping on all four sides. In the seventeenth century canvases were prepared on so called working frames. Working frames (fig 4) were common since at least the end of the sixteenth century. They were significantly larger than the canvas and to mount the canvases, cords were simply passed around the strainer bars or nails in the strainer, or through pre-drilled holes in the strainer bars. 11
During canvas preparation canvases could easily be restretched in working frames, and when applying the ground with a priming knife, no strainer bars behind the canvas would disturb the process.12 After ground application canvases were
fastened in various ways onto strainers, their final auxiliary support. Usually the ends of canvases were wrapped around the edges of the strainer and fastened with hand-forged nails, wooden pins of hardwood, glue or cords. In the latter case, the cord was passed through pre-drilled holes in the slightly oversized strainer bars. The strings ran from the back to the front and back again through the same hole. With this stretching method tacking margins would be absent, and the dimensions of the painting would be just a little smaller than those of the strainer (fig 5). Sometimes leather sheets were nailed between nails or pins and the canvas in order to reduce corrosion and to distribute tension better by increasing surface contact.13
On paintings with original lacing margins, it can be seen that the edges on canvases were seamed before stretching (fig 6). The purpose of that practice must have been to strengthen the edges to prevent those threads at the edges unravel.
11 Van der Werf, Inez, 'Het spanraam: een essentieel onderdeel van het schilderij'. kM, no. 8, Winter 1993, p. 6.
12 Van de Wetering. Rembrandt: The Painter at Work: 117f. 13 Werf, Het spanraam: 7.
Fig 4. Two depictions of working frames, in which the strings are wrapped around the bars of the strainer.
Besides written sources and paintings with original stretching methods, such depictions are the only sources that tell us about stretching method of that time.
Left: A detail of A Painter in his Studio, by G. Dou. Right: A Painter in his Studio, etching by V. van
der Vinne.
Fig 6. Detail of the reverse side of Portrait of Dirck Hendricksz
van Swieten, unknown artist,
1626. Amsterdam Rijksmuseum.
The bottom edge of the canvas has been seamed. In this case the painting was stretched directly into the picture frame instead of an auxiliary frame.
Fig 5. Detail of Assumption of the
2.2 COMMERCIALLY PREPARED CANVASES
“When no primary cusping is present on one or more sides chances are that the painting has been reduced in size by later hands.”14
In saying so Van de Wetering says that the absence of cusping is an indication of a potential reduction in size. However, he acknowledges that the availability and use of commercially prepared canvases complicates things. By the seventeenth century it was already common for artists to buy their canvas supports ready made.15 It was also
not unusual for the client to provide the hired artist with the painting support.16 The
absence of cusping on one or more sides occurs when a canvases has been cut out of a larger piece of primed canvas. The preparation of whole canvas bolts (up to 7m long and about 70 cm wide) appears to have been relatively common.17 He describes this
method in the following way: “The method involves stretching the canvas strip between two poles, the short sides being attached only in the corners.”18
In Rembrandt’s Portrait of Johannes Wtenbogaert, Van de Wetering observed a wide cusp at the bottom edge that runs from one corner of the painting to the other; no cusping at the top and normal cusping at the sides. He concludes from this, that the painting comes from the end of such a primed bolt.19
The high tensions used in machine looms and the way the looms are threaded seems to influence cusping too. In Van Gogh’s painting Augustine Roulin, widely spaced cusping of 40 cm depth was noted. It is understood that these deep distortions occur because of the way the warp threads were tied together into bundles in the loom (fig 7). Johnson speculates that such strong cusping would only be present at the beginning and end of a bolt.20 Such intense cusping has not yet been noted in painting
supports made of hand-woven fabrics.
14 Van de Wetering. Rembrandt: The Painter at Work: 116.
15 Van de Weterings investigations revealed that in the years between 1632-1634 Rembrandts
workshop used canvases from 25 different bolts. From: Van de Wetering. Rembrandt: The Painter at Work: 100.
16 Van de Wetering. Rembrandt: The Painter at Work: 100.
17 Van de Wetering bases his hypothesis on a sixteenth century source that describes the use of banners, as well as Xenia Henney’s founds on the width and length of primed seventeenth century canvases in the Rotterdam Municiple Archive.
18 Van de Wetering. Rembrandt: The Painter at Work: 116. 19 Van de Wetering. Rembrandt: The Painter at Work: 116. 20 Johnson, Van Gogh at Work: 149.
2.3 SECONDARY CUSPING
Secondary cusping can be noted primarily on nineteenth century paintings. At that time it was common practice for artists to stretch their commercially pre-primed canvases with canvas pliers.21 As described earlier, secondary cusping is less
pronounced than primary cusping and the cusping pattern is often lost due to trimming during lining or relining treatments.22 Secondary cusping can also be found
in seventeenth century paintings, like in Vermeer’s The Little Street, from 1658 or Rembrandt’s Portrait of a Couple at the Gardner Museum in Boston. In the x-ray of the paintings both primary cusping and secondary cusping is visible (according to the authors Gaskell and Van de Wetering).23
Van de Wetering assumes that secondary cusping occurs when a canvas is re-stretched on a strainer shortly after it had been primed and the ground material is still soft.24 This would imply that no new cusping would form after a primed canvas had
dried sufficiently and also not if a painting was re-stretched after having become slack, for instance.
21 Buckley, Barbara A. Stretchers, tensioning, and attachments. Stoner, Joyce Hill, Rebecca Rushfield (ed.). Conservation of Easel Paintings. Oxon: Routledge, 2012, pp. 148-160.
22 Van de Wetering. Rembrandt: The Painter at Work: 116.
23 Gaskell, Ivan and Michiel Jonker, Vermeer Studies. New Haven: Yale University Press, 1998. 24 Van de Wetering. Rembrandt: The Painter at Work: 116.
Fig 7. Schematic depiction of the threading of a loom.
2.4 WARP AND WEFT
Johnson observed that in individually prepared canvases, opposite sides show the same degree of cusping if the tension is spread evenly and the tacking spacing on the two sides are similar. Differences in warp and weft cusping result from differences in yarn properties.25
Zenker has addressed the myth, which is believed by many, that the warp threads are always straight and the weft wraps around the warp in waves. This misconception results from a lack of understanding of yarn properties and the differences between weaving machines and handlooms.26
The deformation of warp and weft threads depends on the beating-up impact of weft threads during weaving, as well as the elasticity of the threads. In most cases, in both machine- as hand-woven fabrics, warp threads are more evenly spun, while weft threads usually contain coarser fibres and are
less tightly twisted. Yarn with a greater twist is more flexible, stable and thinner. This allows the yarn to resist the constant load during the
weaving process. In handlooms weft threads were
inserted relatively loosely in order to avoid breakage. Taken that, in a fabric, the elastic properties of the weft and warp threads were similar, the pulling forces of the warp threads would be great enough to force the weft threads to lie themselves around the warp threads. If the weft threads were far more thick and stiff, the warp threads would wrap around the weft threads despite the pulling forces.27 With weaving
25 Vellekoop, Marije, e.a. Van Gogh at Work. München: Hirmer, 2013, p. 147. 26 Zenker, Evelyn, ‘Über Kett- und Schussfaden’, Zeitschrift für Kunsttechnologie
und Konservierung, no. 2, 1998, p. 348.
27 Zenker. Über Kett- und Schussfaden: 346.
Hintz * UvA * 2014
Fig 8. Illustration of a fabric with crimped warp and straight weft.
machines a greater weft density can be attained because weft threads can be beaten up tighter. Consequently warp threads have to follow a greater wrapping angle and
become wavier (fig 8).28 Johnson found
that cusping severity tends to be similar on the opposite sides of a painting and usually deeper along the warp direction. He explains this phenomenon, in relation to nineteenth century paintings, as a result of the differing tensions applied to warp and weft threads during the weaving process, as well as differences in warp and weft thread properties, which seem to cause warp threads to stretch more than weft threads.29
So called ‘weft snakes’ occur almost exclusively in hand-woven canvas supports. Weft snakes are created by a slight increase in tension across the width of a canvas sheet, due to a laid-in weft thread of insufficient length. Weft threads in the immediate vicinity of such a shorter thread go a bit slack and form subtle waves, which shows in automated weft angle maps (fig 9). Weft snakes are considered a reliable tool in determining the warp and weft direction of threads in canvases, whithout tacking edges or selvedges.30
In order to compare different canvases it is essential to determine warp and weft, as the number of threads usually differs in the two directions. The most secure way of determining the warp is by the fortunate presence of a selvedge, to which it is always parallel. In larger scale paintings the warp usually runs along the length of the strip. In small paintings it can be difficult to determine the warp and weft direction. There are some indications thatmay help: thread counts are of greater regularity in warp direction, because warp threads are held apart at regular distances by the comb.31
Especially with linen the warp density cannot be set too high in the loom, otherwise the fibrous threads would stick to each other.32 While the warp is generally very
straight and parallel, the weft can go in waves in hand-woven canvases.33
28 Zenker. Über Kett- und Schussfaden: 344. 29 Vellekoop. Van Gogh at Work: 148.
30 Johnson, Richard C., e.a. ‘Detecting Weft Snakes’. ArtMatters, no. 5, 2013, p 49. 31 Van de Wetering. Rembrandt: The Painter at Work: 98 f.
32 Zenker. Über Kett- und Schussfaden: 344.
2.5 FIBER TYPE AND WEAVE PATTERN
It appears that seventeenth century Dutch painters primarily used tightly woven plain-weave linen, although a trend towards more open plain-weave canvases was observed towards the end of the century.34 Twill weave was used only sporadically.35
As a rule, primary cusping is more pronounced in open weave fabrics because of the lower fill of threads and thus greater space between single threads.36
2.6 CANVAS PREPARATION
In Dutch seventeenth century practice canvases were stretched onto working frames either without any pre-treatment, or possibly was made wet only.37 Whether the
wetting of canvas was common practice is questionable. In historical art technological sources, such as the ‘De Mayerne Manuscript’38 or the ‘Original treatises on the arts
of painting’, compiled by Merrifield39, little is said about the actual stretching of
textile supports. In fabrics that have not been wetted and stretched before their final stretching, cusping will be more pronounced. This is because canvas shrinks when wetted.40
Regarding canvas preparation for oil paintings, many different recipes for 34 Van de Wetering. Rembrandt: The Painter at Work: 97.
35 Buckley. Conservation of Easel Paintings: 155. 36 Buckley. Conservation of Easel Paintings: 155. 37 Werf, Het spanraam: 8.
38 The ‘De Mayerne manuscript’ is a conglomeration/pilation of painting-technique formulas, which Théodore Turquet de Mayerne (1573 -1655) gathered from books, scriptures and personal contacts with artists active at the English court in the seventeenth century.
From: Bischoff, Gudrun. Das De Mayerne-Manuskript, diploma thesis. Institut für Technologie der Malerei an der Staatl. Akademie der Bildenden Künste, Stuttgart, 2003, p. 4.
39 Merrifield, Mary Philadelphia, Medieval and Renaissance Treatises on the Arts of Painting, Original Texts with English Translations, 2 vols. bound as one. New York: Dover, 1849, 1967, 1999. 40 Buckley. Conservation of Easel Paintings: 155.
grounds are listed in the De Mayerne Manuscript.41 Sizing was recommended with
fish or animal skin glue in order to reduce the absorbency of the canvas. Most artists, whom Mayerne cites in his treatise, advised against a too high concentration of size, because it was understood that this would cause tensions and cracks in overlying layers. This is why honey or sugar was sometimes added to the glue as plasticizers, which again caused blooming on paintings in humid environments.42 No
recommendations for the number of size coatings could be found.43
Dried size makes canvases much stiffer. Furthermore, the size locks the yarns up at the weave points. Therefore, yarns resist displacement and cusp deformations are retained even if the canvas is removed from the stretcher.44
Young investigated how size and ground application influence the tensions in linen canvas. During her experiments the test canvases were fixed to a biaxial tension tester. The technical details of the experiments are discussed elsewhere.45 It appears
that sizing causes first an increase in tension (as wetting), due to the swelling of fibres and yarns. Then tension decreases as the water from the size evaporates. Finally, a slight increase in tension occurs as the glue dessicates completely and shrinks.46
Oil-ground application (titanium white with linseed oil) causes a slight loss of tension, probably due to increased moisture content in the size layers. Upon drying, the tension returns to its value before ground application.47
2.7 STUDY OBJECT ST. FRANCIS OF ASSISI
41 Bischoff. Das De Mayerne-Manuskript: 400. 42 Bischoff. Das De Mayerne-Manuskript: 200.
43 In a Winsor & Newton's catalogue for 1928 seventeen different canvases were listed. They were sized with two coats of glue and smoothed with a pumice stone before different priming were applied. From: Harley, Rosamond D., ‘Artists' prepared canvases from Winsor & Newton 1928-1951’. Studies
in Conservation, vol. 32, no. 2, May 1987, p. 78.
44 Johnson, Van Gogh at Work: 147.
45 Young, Christina. Measurement of the Biaxial Tensile Properties of Paintings on Canvas. Diss. thesis. University of London, 1996. The Courtauld Institute of Art. 21 November 2013, p. 76 ff. <http://www.courtauld.ac.uk/people/young-christina.shtml>.
46 Young. Biaxial Tensile Properties of Paintings: 141. 47 Young. Biaxial Tensile Properties of Paintings: 146.
The seventeenth century painting Francis of
Assisi (fig 10), attributed to the workshop of Anthony Van Eyck, demonstrates well
the complexity of interpreting cusping patterns. In the following the painting will be used to illustrate the importance of research into cusping analysis. In chapter 6.5, the cusping patterns of this painting will be discussed again in relationship with the experiment results. The painting of Francis of Assisi measures 35 x 42,5 cm (original canvas size) and it resembles a detail of a much larger altarpiece created by Anthony Van Dyck called The Crucifixion with the Virgin, Mary Magdalene, John the
Evangelist and St. Francis of Assisi, also known as Cavalry with Saint Francis. This
altarpiece is from the Church of Our Lady in Dendermonde, Belgium, and is dated around 1629-1630.48 It is believed that this painting is either a preparatory study,
possibly for the big altar, or a copy of this altarpiece. Whether the painting was made in its present format, or whether it is a fragment cut out of a larger painting is not clear. The painting has been glue-lined and there are no tacking margins present around the paintings edges. This means that the original canvas has definitely been slightly reduced in size, however whether that has happened to an extent that the painting’s composition was affected, is not clear.
The direction of threads is more clearly visible in the tracing, which was made from the x-ray of the painting (fig 11). Cusping is much more severe in the vertical direction than in the horizontal direction. There is only one subtle wide cusp in the horizontal threads, in the left bottom edge, that extends about five cm from the bottom into the painting (fig 12).
48 Lamers, Maranthe, Condition and treatment report St. Franciscus van Assisi, Antoon van Dyck. University of Amsterdam, 2014.
Fig 11. X-ray of the painting Francis of Assisi (left) and a mapping of the canvas threads.
Fig. 12. Detail of the left bottom edge of the thread mapping. The horizontal threads are pulled down towards the left side.
The vertical threads are not actually cusped but rather wavy. These subtle deformations continue over the whole painting including the centre. On the left side of the painting they are slightly stronger. Vertical thread deformations on the paintings edges does not correspond in most places with the nails, which are present in the painting now. In some areas the direction of the vertical threads changes within short distances (fig 13).
3 RESEARCH QUESTIONS AND
RESEARCH PLAN
It has been shown that the development of cusping is quite complex and influenced by many variables. There are thread deviations, formed during the weaving process of fabrics, primary cusping created during the first stretching and the canvas preparation, and secondary cusping, which results from the restretching of canvases on strainers. Keying out paintings that have become slack could cause further thread movement. Cusping is usually analysed in order to retrieve the steps a painting went through in its creation. Often the structural properties of the woven fabrics are thereby not taken into consideration. Young’s research approaches canvas analysis in a systematical manner. Her research results give us an idea of the inner tensions in canvases during the stretching, sizing and ground applications on canvas supports, but the strain development remains uninvestigated. Many open questions remain concerning the relation between cusping patterns and several stages of canvas preparation. Unknown is also the impact of many variables in the stretching process onto cusp formation.
Fig. 13. Detail of the x-ray with arrows indicating where vertical threads change direction.
The research questions, which were chosen for this project, deal with very basic links between stretching methods and the properties of cusps, which have not yet been investigated sufficiently. With this research it is intended to prepare the ground for further research, which is more relevant to the field of Conservation and restauration. The following questions are selected for investigation within this project:
I) Differences in cusping in warp and weft yarns: Zenker addressed the common and
mistaken assumption of canvases having straight warp and wavy weft yarns. Johnson et al. presume that canvases cusp more in warp direction because of the greater waviness and thus elongation of warp yarns. At present this is a common presupposition, which requires experimental corroboration.
II) The effect of sizing on cusping intensity: Young has shown that stretched sized
and dried canvases have greater tensions than stretched unsized canvases. The absorption of water in the size causes yarn swelling and permanent increases in yarn crimp, which results in shrunk canvases. Increased tensions caused by the shrinkage in the fabrics and the glue must influence cusping, however it is unclear to what extent. The relationship between size concentration and cusping has not been investigated before.
III) The position of a canvas on the bolt: Johnson et al. deduced from extreme
cusping in nineteenth century canvases that they were cut from the beginning or end of a bolt. On nineteenth century looms warp threads were strung with very high tensions and bundled together in the loom construction. Whether such strong cusping can also be found at the beginning or end of a hand woven canvas bolt is still an open question. The weave density might also alter at bolt ends or along the selvedges. The movement of threads in a canvas might therefore be influenced by the position on its bolt from which it was cut off.
Within this research project these questions will be approached in a practical manner by making 'reconstructions of issues'.49 Thus, the emphasis lies not on producing
historically accurate reconstructions of stretched canvases, but rather reconstructions, which exhibit particular phenomena.
The particular overall method of stretching which will be imitated in these experiments is that of the seventeenth century. In the seventeenth century it was still more common to prepare artist canvases individually, even if they were of small dimensions. Another practical aspect, which is of advantage, is the measurement of the applied forces during the stretching process. The measurement of the load values in each pulling point would be much more complex if the canvases were stretched on stretchers. In the seventeenth century stretching method, canvases were laced to oversized working frames. Such a construction allows the measurement of forces exerted onto each pulling point along the canvas’ edges, by attaching spring scales between the canvas and the bars of the working frame.
Linen fabrics will be used for the whole series of tests. Linen is highly sensitive to moisture; therefore preparation techniques involving water have the greatest structural impact on the fabric support. Oil ground application has only a slight and non-permanent impact on the tension within canvas supports (Young). For this reason only size will be applied to the test canvases, but no ground layers.
4 METHODOLOGY
In this chapter, the methodology is explained. Here included is the manner in which the experiments were carried out, documented and analyzed. Experiments and experiment results are discussed in chapter 6.
The experimentation consists of three principal work-phases: firstly the preparation of canvases, secondly the execution (stretching and sizing of canvases) and thirdly quantification of the severity of cusping and evaluation of results.
As mentioned in the introduction, canvas analysis is usually carried out on x-49 The term ‘reconstruction of issues’ is inspired by Norman Tennant.
rays of paintings. Because no ground layer will be applied on the test canvases, it is possible to carry out cusping analysis directly on photographs of the canvases. Images of the canvases in unstretched state serve as a reference, with which the stretched and sized canvases are later on compared.
4.1 PREPARATION
In preparation for the stretching experiments, individual threads on the test canvases are marked with a black permanent pen. A grid of marked threads in 5 cm intervals in both weave directions is thereby created. The marked lines make it possible to monitor cusping during the evaluation process. The edges of the canvases are seamed with cotton yarn to avoid unravelling. Including the seam, the canvases are 40 x 40cm large. If woven fabrics were a flat, regular and symmetric network of threads, which stood in a right angle to each other, the evaluation of the test results would be much easier because one could assume that all angle variations of threads in the stretched canvases were created by the stretching process. However, canvas is uneven and unsymmetrical and usually fans out at the selvedges or buckles were it was folded. Hence the unstretched canvases can only be made as flat as possible and then photographed. For this purpose, the canvases are taped with double-sided tape onto a cutting mat. This way already a small amount of force is exerted onto the textile. This force is not measurable and considered small enough to be negligible in these experiments.
During the execution phase, the canvases are stretched and attached with cords to a wooden strainer. The force that is applied to the canvases during stretching is measured with electronical spring scales. Recording the pulling force is necessary to make the experiments reproducible and to answer some of the research questions. These scales are attached directly on the pulling points of the canvas, which has the advantage that it will be known how much force is applied at each of the pulling points. In appendix IX the amount of force applied on each of the pulling points is listed in tables. Although the scales record kilograms, pulling force will be referred to in Newtons in the following text. Kilograms are converted into Newton by multiplying the Kg-sums by 9,8.
Stretching sequence
First of all, sixteen spring scales are tied with hemp ropes to the edges of the canvases at 10 cm intervals. The canvases with the scales are attached to a working frame with ropes that are passed through eyelets at the bottom of the scales. Instead of wrapping the rope around the bars of the stretching frame, the rope is passed around screws in the wooden stretcher bars. That way the stretching procedure is more standardized and the pulling forces can be adjusted more easily.
Although, it appears that in the seventeenth century canvases were stretched onto working frames in one go with a single rope50, this is practically not achievable
in this reconstruction. The stretching method has to be modified, because the scales go on hold and then switch themselves off too quickly. When the scales are restarted they do not retain the value measured earlier but are reset to zero. The stretching and measuring procedure becomes manageable by splitting it up into three phases, starting at the four corners and then stretching the sides. In doing so the corners, marked ‘1’ in fig 14, are stretched simultaneously with one rope. This is done first. The three pulling points on each side are also always stretched in one go, using a single rope. Thus, the three points on the left and right, numbered ‘2’ are stretched simultaneously. Practically, this is done by tightening firstly the rope that connects pulling points 2.1, 2.2, and 2.3, without applying much force yet. Directly thereafter the opposite three points on the right, 2.4, 2.5 and 2.6 are stretched. Then it is necessary to go back to the left side and adjust the tension with the help of additional screws (fig 15). By 50 See chapter 2.1.
passing the rope around the extra screws, pulling forces can be increased locally by a few Newtons. Until all ‘2-points’ are stretched to the desired tension the scales are prevented from turning themselves off, by regularly pressing the on-off button. Then points 3 are stretched.
Sizing
After stretching, all the canvases are sized with warm rabbit skin glue. Just before applying the size, the scales are turned on again. At this point, the scales are automatically reset to zero, although they are still attached to the stretched canvases inside the working frame. When applying the liquid size, the canvases shrink. The tension in the canvases increases because the shrinkage is restrained by the stretching system tension. The spring scales measure the force that is thus exerted onto the canvas. Appendix IX lists these shrinkage forces.
4.3 VISUAL DOCUMENTATION
Fig 15. Additional srews aid in optimizing the pulling force.
Fig 14. Stretching sequence for the canvases on the working frame.
Photographs are taken with a Nikon D90 DSLR camera with 12.3-megapixel resolution. The canvases are photographed horizontally; placed on the ground and photographed from above (fig 16). In all photographs, a scale (a 1,5cm square of graph paper) is included for calibration, so that the conversion from pixels to mm is possible in Adobe Photoshop CS5.
Furthermore, photographs include a perfect rectangle so they can be orthorectified. This is necessary when the camera angle is not vertical during photo-graphing. An oblique camera angle would cause perspective distortion as depicted in fig. 15. A cutting mat is placed underneath the canvases so the rectangular lines on the cutting mat can serve as a reference for the orthorectification. To do so the images are cropped ‘in perspective’ in Adobe Photoshop CS5.
4.4 QUANTIFICATION &
EVALUATION OF RESULTS
Fig 16. Camera set-up.
Fig 17. Location of the measurement locations, the zero point and the central reference lines.
In the evaluation of this project, the severity of cusping in the stretched canvases, is assessed. In order to be able to compare the severity of cusping in the different canvases, cusping needs to be quantified. There are three aspects of cusping that can be measured. Those are the maximum deviation from a straight line, the maximum angular deviation and the depth of cusping. The maximum deviation from a straight line, named yspan (red marks in fig 17), represents the largest distance a thread departs from a straight line, or in other words, the largest distance between the bottom and the tip of a cusp. The maximum angular deviation (θspan) is the maximum angle that a thread is deviated
from its normal course. The deviation angle reaches its maximum just before the top of the cusp. Depth of cusping describes the distance cusping extends from the edge into the center of a canvas. It is assumed that threads with a large yspan also have a large θspan and a large depth, and vice verser. As all three
aspects of cusping are related to each other, it is considered sufficient to measure only the yspan in order to asses the amount of cusping representatively. Only biaxial forces
are taken into consideration. In the corners, the canvases are not stretched biaxially (in horizontal or vertical direction) but in a 45° angle. The pulling point in the corner introduces a third direction in which force is applied, which would complicate the analysis of the cusping. For this reason only the force applied on the pulling points 2.1-2.6 and 3.1-3.6 are included in the analysis.51 The cusping effect is studied only
along the sides of the canvases, as demonstrated in fig. 17. These measuring locations are named ‘top 1’, ‘top 2’ etc., in all canvases. The grid of marked threads serves as frame reference, which allows measuring the deviation of threads. For the calculation of the maximum deviation of threads only the outermost threads are measured, although the other lines give a good idea of how far cusping penetrates into the canvases.
As the canvases changes dimension while stretching and sizing, absolute values of the yspan of a single canvas during one state (unstretched/ stretched/ sized) cannot be compared to the canvas in another state. Therefore, a constant reference is necessary, in order to account for the change in dimensions and to quantify cusping in percentages. For this purpose a zero point (fig 17) in the center of the canvases is chosen (±2cm off center), through which a vertical and a horizontal auxiliary line are 51 The author is aware that about the possibility to calculate the vertical and horizontal component of the diagonal forces. However, due to the limited scope of this project it was decided to focus on the cusping in the center of the canvases only.
drawn. In the Adobe Photoshop CS5 program, images of the canvases in unstretched and stretched state are placed on top of each other, so, that the zero points lie exactly on top of each other. By making one image half transparent, both canvases are visible. The overall images are made clearer by lowering the threshold of the images, so they become black and white.
The length loriginal (fig. 18) describes the distance between the auxiliary line, which
runs through the zero point, and the outermost marked thread in the unstretched canvas. The absolute value for the yspan of a cusp is described by the difference in strain between the bottom and tip of a cusp. To calculate the yspan, lbottom is substracted from ltip. Thereby lbottom is the distance between the auxiliary line and the
bottom of the cusped thread. Ltip is the distance between the auxiliary line and the top
of the cusp. A line, which connects the left and right tip, helps to determine ltip.
Cusping (ε cusp) is expressed in percentage by calculating the following:
ε cusp
=
Dl loriginal = ltip- lbottom loriginalLtip, lbottom and loriginal are measured in
Adobe Photoshop CS5 with the ruler tool and the numbers are then entered into Microsoft Exel tables. ε cusp, as well as averages and standard deviations, are then calculated in Microsoft Exel with the function tool.
Additional to the cusping strain, the strain of warp and weft threads over the entire canvas is calculated. Because the canvases are cusped around the edges, changes in dimension cannot be
measured representatively at neither the tips nor the bottom of cusps. Therefore another auxiliary line is drawn in Adobe Photoshop CS5, inbetween ltip and lbottom. This auxiliary line represents the theoretical edge of the canvas and is calculated in the same measuring locations as εcusp (fig.17). The distance between loriginal and this
auxiliary line is derived by calculating:
ledge =
ltip- lbottom
2 + lbottom
The change in dimension/strain is therefore: ε strain =
ledge- loriginal
loriginal
By multiplying εcusp or εstrain with 100, one gets the percentage of the unitless amount.
In the following text averages of εcusp or εstrain are used. By calculating the standard deviation of these averages one can see how much scatter from the average exists. If the SD of two average values overlap, then there is the possibility that the values are, in fact, not different from each other.
5 TEST CANVASES
The characteristics that are essential for a textile painting support are mainly determined by the structure of the woven fabric. Linen canvas is a cellulosic textile fibre. The fibre is extracted from the stems of flax plants and the original properties of linen fibers are dependent on factors such as the growth condition of flax plants, storage conditions and processing. Differences in the chemical properties within such a small substance group, as that of linen fibers, is however considered so small that it can be omitted from canvas analysis.52 If the aging process of linen canvases were
here investigated, then the chemical properties of linen would be of importance. The mechanical properties of woven fabrics are influenced by numerous parameters. The yarn properties depend on the fiber properties and the type of spinning. The fabric properties again are influenced by the weave and weave density. 52 Rouba. Die Leinwandstrukturanalyse: 80.
Also the weaving conditions play a role. This includes factors such as the weaving speed, insertion rate of weft threads, weft beat-up force, warp preparation for weaving and other technical details involved in the weaving process.53
In this chapter the canvases used for the experiments are analysed and compared to each other. Two different canvases were selected for the experiments; a machine-woven plain weave linen canvas from Claessens and a hand-woven plain weave linen canvas from Die Leinenweber. Those canvas’ characteristics that will influence the movements of threads during stretching are discussed in detail in sub-chapter 4.4. Being aware of these characteristics is essential in order to interpret the test results meaningfully.
5.1 CLAESSENS ARTISTS' CANVAS
The Claessens Artists’ Canvas from the Claessens company in Waregem, Belgium, is an unprimed linen canvas of medium structure. The weaving mill is unknown to the author. Two meters of the 2.10 m roll are purchased and rolled up to avoid folds. The canvas is allowed to acclimatize to the conditions in the testing room for several days. In the table below the structural characteristics are summarized. The specific features of the canvas were determined on the basis of at least 15 samples from the canvas. Structural properties
Article number 066
Bolt width 2.10m
Bolt length unknown
Density warp 17 (threads/cm)
Density weft 14,5 (threads/cm)
Warp yarn thickness 0.5 mm (varies from 0.3 mm to 0.6 mm) Weft yarn thickness 0.4 mm (varies from 0.2 mm to 1 mm) Warp twist direction Z- direction
Weft twist direction Z- direction
Warp twist angle 21,4°
Weft twist angle 12,4°
53 Zupin Ziva and Krste Dimitrovski, ‘Mechanical Properties from Cotton and Biodegradable Yarns Bamboo, SPF, PLA in Weft’, Dubrovski, Polona Dobnik (ed.), Woven Fabric Engenering, INTECH, 2010, 3 June 2014. <http://bit.ly/1pAWURx>, p. 28.
Warp wrap angle 25,6°
Weft wrap angle 16,3°
Canvas fill (total) 94.4 %
Warp canvas fill 86.4 %
Weft Canvas fill 58.8 %
Weave Plain linen weave
Selvage type Fringe type with leno weave Weight grams/m2 275 (Claessens’ data)
The fringe type selvage indicates that the fabric was woven on a shuttleless loom.54 A leno
weave (white yarn in fig. 19) at the edge, made with a different yarn, prevents unraveling.
The Claessens canvas is of high density. Its warp threads are thicker and more twisted than weft threads. Both, weft and warp threads exhibit abnormal fluffy parts and local thread thickenings of about 3 cm to 12 cm in length. This irregularity results from the spinning process, whereby the thread thickness varies in short intervals.55 Warp threads exhibit greater
regularity than the weft threads. The waviness in weft threads is very little, whereas warp threads are very wavy (fig 20).
54 Price, Arthur and Allen C. Cohen, J.J. Pizzuto's fabric science. 6th edition. New York: Fairchild Publications, 1994, p. 154.
55 Rouba. Die Leinwandstrukturanalyse: 81. Fig 19. Fringe type selvedge of the Claessens canvas.
Table 1: characteristics of the Claessens canvas.
Fig 20. (left) Thick and thin warp threads.
Note: warp threads are always depicted vertically in microscope images.
Fig 21. (above) The upper yarn is a weft thread, the lower yarn a wavy warp thread.
Surface properties
Claessens uses for their artists’ canvas production what is called “loom state” fabrics. This means the fabrics are used as they fall off the loom. Claessens does not apply any further additives on the fabric.
The weaving mill, however, adds a lubricant, a liquid synthetic wax (ENSIWAX CPA-CV) to the warp threads. The lubrication product is applied on the warp threads in order to avoid breakage of the yarn and to enable a better glide of the weft threads. Furthermore, it is meant to bind the fibers to the yarn and create a constant friction rate. This wax treatment certainly has an impact on the movement of threads during stretching, which makes the canvas less representative for seventeenth century stretching techniques. The possibility to wash the fabric has been considered. The disadvantage of washing is that it would be difficult to keep the washing process controlled, regular and accurately reproducible. The washed canvas would furthermore crimp, wrinkle, shrink irregularly, and possibly fan out as the edges. Considering that, in this project the quantification and evaluation of the test results are based on ‘before and after’ photographs (comparative study), the extra step of washing is conceived undesirable. As only the warp threads are lubricated with wax, it is necessary to keep the orientation of weft and warp constant in all experiments to avoid misinterpretation.
5.2 DIE LEINENWEBER'S OLD LINEN
The canvas from the Die Leinenweber Company in Achern, Germany, is a handwoven, unprimed linen canvas of medium structure, approximately 100 years of age. The weaving mill is unknown. The canvas material is acquired as a whole bolt of 4,55 m of length, rolled up.
Article number LwAlt 115
Bolt width 71-74 cm
Bolt length 4,55 m
Density warp 16,2 (bolt middle) 15,4 (bolt ends) (threads/cm) Density weft 13,7 (threads/cm)
Warp yarn thickness 0,5 mm Weft yarn thickness 0,5 mm Warp twist direction Z Weft twist direction Z
Warp twist angle 18°
Weft twist angle 14°
Warp wrap angle 35,5°
Weft wrap angle 33,7°
Canvas fill (total) 94%
Warp canvas fill 81%
Weft Canvas fill 68,5%
Weave plain weave
Selvage type Plain selvage /tape
The warp and weft threads of this canvas are very similar to each other. Both are of an average thickness of 0,5 mm, with abnormal thickenings of over 1mm width. Both, warp and weft threads are twisted in the ‘Z’ direction. The weft zwist angle is with an average of 14,2° smaller than that of the warp, with 18,8°.
Surface properties
Table 2: characteristics of the Die Leinweber canvas.
Fig 22. An abnormally thick warp thread of 1,3 mm diameter.
The canvas has so called ‘couch marks’, that are brown-yellow spots, which occur in old linen if it was not stored in a dry and sealed environment. No size or lubricant is recognizable on the canvas by visual examination.
5.3 COMPARISON OF THE TEST CANVASES
In this section, the two test canvases and the canvas of the study object are compared in terms of their structural properties. This will enable the author to estimate how the canvases will cusp relatively to each other, during the stretching experiments. The focus in this chapter lies on those canvas properties that are relevant to cusping.
Y
In cusped canvases, yarns are forced to bend. An important property in the bending of an object is flexural stiffness. The flexural stiffness of an object depends on the Young’s modulus of elasticity56 of the material, its shape and its length.
When unravelling the twisted yarns of both test canvases, it is observed, that their fibre fineness is equal. The only visible difference is in their luster (fig 23). The fibres
56 The Young’s modulus of elasticity (E), is a measure of a material property. The greater the Young’s modulus of a material, the stiffer it is going to be.
Fig 23. Warp thread fibres of the hand-wc (top) and the machine-wc (bottom).
of both weave directions of the machine-wc have more lustre, presumably due to yarn preparation before weaving.
It is assumed that the differences in the linen fibres are minor enough and, therefore, have similar flexibility.
The thickness of yarns of both test canvases is similar (between 4 – 5 mm).
The yarns of both canvases are twisted in Z direction. In the hand-wc the weft and warp yarns have a similar amount of twist (warp 18°/ weft 14°). In the machine-wc is a greater difference in the amount of warp and weft twist (warp 21°/weft 12°).
Increasing twist enhances not only the strength but also the elasticity of the yarns. Therefore, the higher twist in the machine-wc favours more severe cusping in the latter, as well as more severe cusping in the warp direction in both canvases.
Weave
Besides the bending rigidity of fibres and yarns, the mobility of the warp and weft yarns within the fabric influences their ability to move within the system. Young observed in her accelerated aging experiments that it was primarily the woven features of most canvases that dominate the textile’s physical behaviour:
‘In most tests on new canvas supports, the woven characteristics of the canvas dominate the physical response of the whole composite, in particular,
its orthotropic behaviour.57 At low loads, typical of those to which a painting
is subjected, the warp and weft load-extension characteristics of new woven linen and cotton-duck are evident. This is also observed in archival samples but to a lesser extent.’58
The orthotropic behaviour (different in warp and weft direction) was thus greater in new canvases than in nineteenth century archival canvases. This phenomenon is also to be expected in the two test canvases used in this project, due to their different weave characteristics. The greatest difference between the hand-wc and the machine-wc, in this respect (weave characteristics- not age of the fabric), is the difference in 57 Materials with orthotropic behaviour have different material properties in different orthogonal directions or axes.
their crimp. In contrast to the weave characteristics, differences in weave density and canvas fill are minor in comparison. Both canvases, however, exhibit a lower density in their weft than in their warp directions: the machine-wc has a density of 17 threads per cm in warp, and 14.5 in weft direction. In the hand-wc the density in warp direction is 16.2 threads per cm and in weft direction 13.7 threads per cm. This also influences the canvas fill, which is in total the same in both canvases (94%), and respectively lower in weft direction: machine-wc warp 86%/ weft 59% and hand-wc warp 81%/ weft 69%.
The lower weave density and canvas fill in weft direction already suggest that cusping will be more severe along the weft direction. Crimp (waviness of threads) is, as mentioned before, the factor in which the test canvases differ the most. Linen fibres do not stretch easily. The elongation of threads depends thus on the twist of yarns and their waviness. Stretching experiments of individual yarns have shown that the hand-wc is considerably more flexible in both weave directions.59 While the yarns of the
hand-wc stretch 16.3% in warp and 10% in weft direction, the machine-wc stretches only 7% in warp and 1% in weft direction. The elastic recovery of the hand-wc is two times greater than that of the machine-wc.
The wrap angle is a further indication for the crimp and the flexibility of yarns. The wrap angle in the hand-wc is in both weave directions higher than in the machine-wc, and in both canvases the weft threads are less crimped than the warp threads: machine-wc weft/warp (16°/26°) and hand-wc weft/warp 34°/36°. Based on this, it can be assumed that cusping will be more severe in the hand-wc than in the machine-wc, and in both canvases cusping will be more severe in the weft direction than in the warp.
Water absorption
The canvases’ ability to absorb water plays an important role during the application of liquid rabbit skin glue size. Water absorption causes swelling in fibres and yarns, which results in increased fibre and yarn diameter, crimp and weave density. As a general rule, high twist angles, high weave density and canvas fill, reduce a fabric’s 59 See Appendix II.
capacity to absorb moisture. 60
The wettability of the test canvases is tested by means of a water-drop test, whereby single drops of water are placed on the surface of the canvases. The time from when a water drop is placed on the canvas, until its disappearance in the fabric is recorded (ten readings from different locations). The machine-wc absorbs a drop of water on average within one second, whereas water drops on the hand-wc are only fully absorbed after an average of 21 seconds. The wickability61 of the test canvases is
measurable by means of strip-tests (fig 24), which reveals that water can travel along the fibres of the machine-wc more than twice as fast than along the hand-wc. In the direction of the warp threads water is absorbed quicker in both canvases. From this follows, that the impact of the size will be greater in the machine-wc and also greater in the warp than the weft direction, in both canvases.
60 Morton, W.E. and W.S.E. Hearle, Physical Properties of Textile Fibres. Butterworth & Co: Manchester, 1997, p. 159.
61 The property of a fiber that allows moisture to move rapidly along the fiber surface and pass quickly through the fabric.
