So, if Cézanne (and other painters) have unintentional discrepancies in their depiction of reflections on water, how might these occur?
To understand the nature of the problem, we need a functional model of the painting process, based on what we know about human vision. I apologise for the length and technical nature of this article, but I hope that it will give you insight into a fascinating topic which has been relatively little-studied.
For the purposes of this model, when a painter creates a painting consciously, they are creating a physical representation of an image which they have constructed in their mind. If a painting is intended to be a representation of some physical object(s), then the most fundamental model of painting is that the painter constructs a mental image from their observation of the object(s), which they then represent in paint on the painting surface. This is shown in the diagram below.
In reality, painting consists of an iterative process in which the painter observes both the motif which they are trying to represent, and their own painting in progress. The process of painting is constantly adjusted so as to keep the painting in close accordance with the mental image, and continuing observation of the motif adjusts that mental image.
In some circumstances, the success or otherwise of the painter in the painting process may adjust the mental image too. For example, they may realise that they need to further simplify the image because of limitations in their technique, the media, or time, or they may modify their composition in the light of deeper insights into the motif in front of them, or as a result of what they see in their developing representation of the motif.
Many painters who are thoroughly realist in their approach also modify what they paint to achieve what they consider to be an aesthetically superior painting. Few landscapes have ever been precise representations of the motif, as has been revealed by comparison of paintings with preliminary sketches and photographs.
Constable, who considered it perfectly reasonable to omit or move trees, for example, stated in a lecture at the Royal Institution in 1836: “The works of the truly great men who have shone in art were not mere copies of the productions of nature, which can never be more than servile imitations.” This is an issue which I examined in my series Truth in (landscape) painting. Some professional landscape photographers have been known to ‘garden’ and ‘improve’ a view prior to photography. Thus the mental image of a painting in progress need not correspond to the mental image of the motif itself.
How we do it
Research over the last couple of centuries has gradually detailed the body systems involved in such activities. Initial emphasis was placed on the role of the eyes and optic system, something of which Cézanne was mindful. However the optic system itself does not create any mental image: extensive processing of optic information is required for that, the function of substantial areas of the cerebral cortex known collectively as the visual cortex.
It is there that any mental image, of the motif and the intended painting, is formed. A painter then has to convert their mental image into a long and elaborate series of instructions to control the hand, arm, and other parts of the body involved in the act of painting. For this other parts of the cerebral cortex are involved, the motor cortex. These control the muscles and other parts of the motor system to hold the brush and apply paint from it to the painting surface. This is in turn observed by the optic system, analysed by the visual cortex, and provides feedback to motor control.
This is shown in the more elaborate version of the model above. Additional parts of the central nervous system involved in visual processing and motor control, such as the cerebellum for the regulation of movement, are not included here, but subsumed within the overall block structure.
So far this basic model has ignored the vital influences of other higher centres in the brain, showing what might be expected from a ‘painting automaton’. Cézanne made it clear that in constructing his mental image, he included more than just the visual sensation of the motif; this is undoubtedly true for many other painters, although they have perhaps been less explicit of these influences.
One of the bolder claims that could be made of ‘modern painting’ is this broadening of the sensations admitted to the process of creating the mental image. Although this appears to have been one of the major components of Impressionism, it has been a constant factor in many paintings since the Renaissance.
Clearly any mental image acting as a ‘blueprint’ for a painting will also be influenced by emotion, including the emotional reaction to the motif, the experience and intent of the painter, their intended style, and other conscious and sub-conscious factors. Another necessary modification is to add higher decision-making processes between the mental image and motor control.
A more surprising addition to the basic model is representation of a direct link from visual processing to motor control. This is the result of recent studies which merit elaboration.
Goodale (2014) cites the example of a woman who, as the result of an episode of carbon monoxide poisoning, developed visual form agnosia, in which she was unable to recognise objects which had previously been very familiar, or the simplest geometric forms. She was still able to identify objects by touch, though, and still identified the colour and visual texture of objects she saw. This was a failure in her visual processing for perception, but not in that for action, as certain aspects of visual processing were completely unaffected by this damage.
For example, when presented with objects of different widths (but identical colour and visual texture) and asked to grasp one, her hand opened to the appropriate width of the object in mid-flight, even though she was unable to discriminate between the objects visually. This and other less accessible evidence leads Goodale to conclude that there are direct connections between centres involved in visual processing with the centres for motor control, which do not enter conscious thought.
In the visual areas of the cerebral cortex, information in the ventral stream (purple in the diagram above) of the early visual areas projects in a perceptual stream to the inferior temporal cortex, where it forms the basis of vision-for-perception. Information in the dorsal stream (green in the diagram above) projects instead to the posterior parietal cortex, where it forms the basis of vision-for-action.
Solso (2001) used fMRI to compare blood flow changes in regions of the brain in a skilled portrait painter and a novice when drawing a series of faces. Both showed increased blood flow in the right posterior parietal region, which is normally associated with facial perception and processing, although the skilled painter showed lower levels than the novice.
He found evidence of the skilled painter using more of his right frontal lobe, an area normally associated with planned motor movements and skilled functions. Unfortunately this study was limited to drawing, and with just the pair of subjects. However he suggests that the skilled painter was ‘thinking’ portraits rather than seeing them, perhaps in the way suggested above.
Miall et al. (2009) performed functional imaging of untrained participants whilst they drew simple cartoon faces from memory and without being able to see their drawing hand. They observed activation of the face sensitive areas of the lateral occipital cortex and the fusiform gyrus during image memorisation, and when drawing. Drawing from memory also activated parts of the posterior parietal cortex and frontal areas, consistent with it being a visuo-motor activity utilising mainly the dorsal visual stream. Differences between drawing with vision and drawing from memory were more subtle.
This has been further refined by Yuan & Brown (2014), who have used higher resolution fMRI and careful task design to contrast mark making against fixation and blind drawing. Mark making from memory was accompanied by prominent activations related to motor control of the drawing hand and forearm, in the sensorimotor cortex and posterior cerebellum. Additional activations were observed in the frontal eye fields medial to the primary hand activations, and in motion perception areas and a form-processing area thought to be important in processing form from motion.
These reflect the dynamic visual-hand and visual-eye coupling that occurs during the making of marks and the emergence of resulting images. Copying showed more intense activations in those areas, with the addition of activations in the basal ganglia which probably relates to the imitative aspect of copying, and in the primary visual cortex and its surrounds, which probably reflected a static image model formed by repeatedly glancing at the subject being copied.
Thus image data from the primary visual cortex is divided into: a ventral stream, which feeds the colour area and the inferior temporal cortex, concerned with vision-for-perception. This links into the thalamus and prefrontal cortex, including the orbitofrontal cortex; a dorsal stream, which feeds the parietal cortex, a supplementary motor area in the pre-motor cortex, and the pre-frontal cortex, concerned with vision-for-action.
As no studies appear to have been made of related phenomena in painters, I suggest that it is likely that those whose painting skills have become well-practised, perhaps even semi-automatic, may not involve much conscious processing of visual images in order to control the motor actions in painting. This is consistent with the proposal made by James D Herbert that brushstrokes are rarely conscious actions, and much more habitual.
Anecdotal evidence to support this is the way in which some drawing and painting actions may take place whilst the artist is still looking at the motif, not the painting. This unfortunately invalidates the assumptions in many papers which attempt to extrapolate from research on the viewer role to that of the painter.
Tchalenko & Miall (2007) found that ‘keen amateur’ artists were able to complete quite accurate ‘blind’ copies of the complex curve of the outline of a head. These were ‘blind’ to the extent that the participants directed their gaze at the image that they were copying, and did not look at the copy which they were making. However, whilst the shapes of individual parts of the outline were quite accurate, relative proportions between those parts was lost, in scaling error. This occurred wherever the hand had to be lifted from the paper to move on to the next section of the curve.
Accordingly Tchalenko & Miall arrived at a Drawing Hypothesis, in which the drawing of a complex shape is the result of visuomotor mapping that can be executed directly while perceiving the original and without vision of the drawing surface. However correct spatial positioning on the paper requires vision of the drawing surface.
Taking these into account, my more detailed working model is shown above.
This becomes relevant to the problem of Cézanne’s reflections when I add various conditions or circumstances in which this model could result in paintings which are unintentionally different from the motif, as in the discrepant reflections. This is shown in the diagram below, in which those are indicated in red.
For example, in late life, Monet suffered from cataracts which disrupted his ‘looking’, and resulted in quite different appearances in his paintings. Nicolas Poussin suffered from increasingly severe tremor in his hands as he grew older, which affected his motor control, and made it hard for him to paint. However because his vision was still good, he was able to correct and compensate very well for that.
Why problems with reflections?
Cavanagh et al. (2008) have explored the perception of reflections, in pursuit of understanding what appears to work despite deviations from optical principles. They suggest that “there may be something special about horizontal, reflecting surfaces” such as bodies of water.
Symmetry of the reflected image around the horizontal axis is proposed to be a basic optical cue, as is vertical alignment of objects in the real image with their reflection in the virtual image of the reflection, which are considered to be inherent in the visual system of many animals, not just humans. They consider that violation of the latter by lateral displacement (as in many of Cézanne’s paintings) degrades the perception of the reflection. Although no experimental evidence is given, they assert that in such circumstances “many viewers note that something is wrong here, often without a clear understanding of what the error is”.
They accept that more subtle consequences of optical principles, such as whether an object appears in a reflection or is blocked by intervening objects, are often not noticed, though. There is experimental evidence that mirror symmetry is more easily detected around the vertical axis than the horizontal (reflections on water having symmetry around the horizontal axis).
Symmetry judgement has been experimentally associated with activity in the parietocentral cortex, or in the lateral occipital complex (LOC, posterior to the parietal cortex). Right parietal lesions can result in mirror agnosia, in which there are difficulties discriminating reflections from real images, although this has mainly been assessed using mirrors placed vertically, rather than on the surface of bodies of water.
Although mirrors have been popular items in recent centuries, for most people over most of human existence the only situation in which someone is likely to encounter a reflection in nature is when it occurs of the surface of a body of water. Such reflections are therefore the only natural situation in which we see a picture within a picture, furthermore the interior image a likeness of the exterior, but shown upside down.
Experiments by Cornelis et al. (2009) have shown that 3D interpretation (the shape percept) of reflections about a horizontal axis (e.g. on water) appeared to show greatest change, whilst those about a vertical axis showed less. Rotation also affected it, with 180˚ rotation being similar in effect to rotation about a horizontal axis. They used a figure depicting a female torso without arms or legs in their experiments.
A series of papers (Croucher et al., 2002) has revealed that even those who are well-versed in physics, and aware of the law of reflection, have very little insight into the optics of reflections in mirrors. This has been dubbed naïve optics. Examples of this naïvety include being unable to judge when it was optically possible to see a reflection in a mirror (Croucher et al. 2002), and more.
However Jones & Bertamini (2007) found that the additional visual cue provided by reflections improved the accuracy of size and distance measurements. They also noted that other studies had shown a similar role for cast shadows. Lawson (2010) has shown using simple experiments that observers are unable to locate the projection of reflected objects on the surface of vertically placed mirrors; even though repeated testing with feedback helped reduce errors, they still remained.
Depth order problems have been surprisingly little studied. However Gillam (2011) makes its importance clear: “Occlusion is rarely discussed as a major issue in art, yet it could be regarded as the major issue in depicting a three-dimensional scene on a picture plane.” Previous work by Kanizsa & Massironi (1989) had expressed the view that completion, a more subtle expression of depth order, is “the result of a cognitive integration and, therefore, does not require any additional analysis.”
Examining a selection of paintings in the National Gallery, London, created between about 1460 and 1785, Gillam (2011) points out incoherence in occlusion in paintings by Duccio (particularly in his Maesta series), although these invariably require close scrutiny to discover them.
Mamassian et al. (1998) found that cast shadows were perceptually most relevant to aid recovery of information about spatial arrangement, as observed by Leonardo da Vinci. In static layouts, cast shadows aid the understanding of depth relations, and can have even greater roles as dynamic cues when objects are in motion. However sometimes complex cast shadows can confuse.
So we have a complex system of different parts of the brain which are involved in the process of painting, which is then challenged – sometimes to its limits – when trying to represent reflections on water, being a picture within a picture, which has been reflected and not rotated. Is it surprising that some of us then find it very hard, sometimes impossible, to depict those reflections as accurately as we would wish?
M Bertamini, R Latto & A Spooner (2003) The Venus effect: people’s understanding of mirror reflections in paintings, Perception 32: 593-599.
P Cavanagh, J Chao & D Wang (2008) Reflections in art, Spatial Vision 21(3-5): 261-270.
D J Cohen (2005) Look little, look often: the influence of gaze frequency on drawing accuracy, Perception and Psychophysics 67(6): 997-1009.
E V K Cornelis, A J van Doorn & J Wagemans (2009) The effects of mirror reflections and planar rotations of pictures on the shape percept of the depicted object, Perception 38: 1439-1466.
C J Croucher, M Bertamini & H Hecht (2002) Naive optics: understanding the geometry of mirror reflections, J Exp Psychol: Human Perception and Performance 28(3): 546-562.
H Y Eng, D Chen & Y Jiang (2005) Visual working memory for simple and complex visual stimuli, Psychonomic Bulletin and Review 12(6): 1127-1133.
B Gillam (2011) Occlusion issues in early Renaissance art, i-Perception 2: 1076-1097.
M A Goodale (2014) How (and why) the visual control of action differs from visual perception, Proc R Soc B 281(20140337): 1-9.
M Johnson (1993) A cognitive model for the perception and translation of a three-dimensional object/array, Visual Arts Research 19(1): 85-99.
L A Jones & M Bertamini (2007) Through the looking glass: how the relationship between an object and its reflection affects the perception of distance and size, Perception 36: 1572-1594.
R Lawson (2010) People cannot locate the projection of an object on the surface of a mirror, Cognition 115: 336-342.
P Mamassian, D C Knill & D Kersten (1998) The perception of cast shadows, Trends in Cognitive Sciences 2(8): 288-295.
R C Miall, E Gowen & J Tchalenko (2009) Drawing cartoon faces – a functional imaging study of the cognitive neuroscience of drawing, Cortex 45(3): 394-406.
J F O’Brien & H Farid (2012) Exposing photo manipulation with inconsistent reflections, ACM Transactions of Graphics 31(1): Article 4.
T Osugi & Y Takeda (2013) The precision of visual memory for a complex contour shape measured by a freehand drawing task, Vision Research (79): 17-26.
V S Ramachandran, E L Altschuler & S Hillyer (1997) Mirror agnosia, Proc R Soc Lond B 264: 645-647.
R L Solso (2001) Brain activities in a skilled versus a novice artist: an fMRI study, Leonardo 34(1): 31-34.
J Tchalenko & R C Miall (2009) Eye-hand strategies in copying complex lines, Cortex 45(3): 368-376.
S Vicari, S Bellucci & G A Carlesimo (2006) Evidence from two genetic syndromes for the independence of spatial and visual working memory, Developmental Medicine and Child Neurology 48: 126-131.
Y Yuan & S Brown (2014) The neural basis of mark making: a functional MRI study of drawing, PLoS One 9(10): e108628.