Cracks and the Perception of Paintings

The Effect of Cracks on the Perception of Paintings

by Spike Bucklow

Introduction

The illusion of pictorial depth can be compromised by visual features that emphasise the surface of a painting. The effect of varnish upon the perception of the pictorial image has been widely acknowledged, but the perceptual effect of craquelure – the pattern of cracks that naturally develops across the face of a painting – has been neglected.

Varnish 

A discoloured and textured varnish layer tends to interfere with the reading of depth in a painting. Any illusion of depth that is created by geometric perspective will continue to function when seen through the yellow filter of an aged varnish. However, the perspective of envelopment – warm colours in the foreground going to cool colours in the background and high contrasts in the foreground going to low contrasts in the background – may be severely disrupted by an aged varnish. Aged varnishes therefore visually ‘flatten’ paintings.

Craquelure

This investigation into the perceptual effects of craquelure was restricted to those cracks that formed slowly by brittle failure of dry paint. It was assumed that these cracks had no specific colour, but were just darker than the paint they disrupted. It was also assumed that they were not associated with any out-of-plane distortion of the paint.

The fine network of cracks usually goes un-noticed, except in light passages such as flesh or sky. Even when visually obvious, the network is often overlooked because it is perceived to have little to do with what the artist intended to depict – we subconsciously filter it as irrelevant. Yet, as the detail of light transmitted through a portrait on canvas shows, the crack network can be a surprisingly high proportion of the painted area.

A detail taken from a portrait showing a distinct pattern of craquelure

 A network of cracks was digitally extracted from a Flemish panel painting and imposed onto black and white printed images. Pairs of images – with and without imposed crack patterns – were shown to 160 students and their responses to the images were recorded.

Two of these tests examined the degree to which the crack network influenced the ability to discriminate between passages in the underlying images. In one case, the contours between passages were defined by very low differences in contrast. In the other case, the contours were defined by a boundary that had to be mentally constructed from very few clues. In both cases, the presence of a crack network was shown to significantly impair recognition of the underlying figure. Another test examined the extent to which a crack network could influence the perceived depth in a black and white photograph. The response to these images suggests that the crack network – like a discoloured yellow varnish – visually flattens the pictorial image.

	Photograph of Jesus Green, Cambridge	Same photograph with imposed crack pattern

In studies of visual perception, it is known that visual textures interact with each other and it is possible that the high contrast between the cracks and the underlying image interfered with the perception of more subtle contrasts in the underlying image. This – like a discoloured yellow varnish – would impair the function of any perspective of envelopment intended by the painter.

The project produced one other finding. The crack pattern from a Flemish panel painting had been digitised, so the area covered by the network could be accurately measured. The area of paint constituted around 90% of the visible image and the crack network was approximately 10% of the image. The crack network was printed onto white paper with black ink, and – from the known reflectivity of the ink (3% or Munsell value 2.0) and paper (90% or Munsell value 9.5) – the overall reflectivity of the printed image was calculated at about 82%. The image was shown to subjects and who compared its tonality with Munsell value neutral greys. The black and white crack network was matched with the Munsell value neutral grey 7.5, which has a reflectance of 50%.

	Bar chart reflecting Munsell values perception		Detail of the digitally extracted crack pattern

The crack network has a disproportionately large effect on the apparent tonality of the image. A physical reflection of 82% was perceived as a reflection of 50% of the light falling on the image. Of course, cracks in real paintings are not necessarily black, and the paint passages they traverse are not necessarily white. Nonetheless, the colour of the cracks is assimilated into the colour of the neighbouring paint (this is the von Bezold spreading effect that enables us to reconstruct shades of grey from black and white in wood-cuts and etchings, etc.).

If we assume that the cracks across a painting are a uniform colour then they will have a large effect where they cross contrasting paint passages and a negligible effect where they cross paint passages of a similar colour. The presence of crack networks therefore influences the tonal organisation of paintings, effectively reducing their dynamic range.

The fact that this does not interfere with interpretation shows how much high-level visual processing is undertaken when making and looking at paintings. For example, a sun-lit cloud is tens of thousands times brighter than the shadowy foliage under a tree, yet when the artist paints a landscape his brightest clouds can only be thirty times brighter than his darkest shadows (assuming that they – like the white paper and black ink – reflect about 90% and 3% of the light falling on the painting, respectively). The artist is able to re-construct a dynamic range of 10,000s to 1, with paint that reflects 30 to 1 of ambient light. If the painting develops a crack network that reduces the perceived reflectivity of the bright clouds, then the dynamic range is further reduced. White paint that reflects c.90% of the light becomes cracked paint which may reflect c.80% of the light, but the ‘spreading effect’ means that it is perceived as if it reflects only c.50%. The painting’s dynamic range therefore shrinks from 30 to 1 when new, to 17 to 1 when heavily cracked, yet it still adequately represents a scene of 10,000s to 1.

Publication

S Bucklow Paradigms and Pigment Recipes; Vermilion, Synthetic Yellows & the Nature of Egg, Zeitschrift für Kunsttechnologie und Konservierung, 13, 1 1999: pp. 140-149