LEONARDO, Vol. 34, No. 1, pp. 35–40, 2001 35© 2001 ISAST

How does a painter transform a vision of theexternal world into a picture on the canvas? Much work hasbeen reported on aspects of visual processes and on percep-tion of finished artwork [1], but so far cognitive psychologistsand art historians have had little to say about the actual pic-ture-production process [2].

An artist drawing from life shifts his or her gaze many hun-dreds of times, from the subject matter to the paper or can-vas and back again. Less often, longer periods will be spentjust looking at the sitter or at the emerging picture. Clearly,this shift of gaze is a fundamental aspect of picture produc-tion, and a direct consequence of the subject-to-picture trans-formation taking place in the artist’s brain [3].

With the help of the eyetracker, an instrument that can ac-curately measure eye movements, together with a sensor re-cording hand movement and close-up video filming, we setout to investigate how one artist produces pictures.Humphrey Ocean paints or draws realistic portraits from life,hence both his visual input and his manual output are observ-able to an outsider. This work, combined with an appreciableamount of existing data on this particular artist, comprises adetailed case study upon which we later hope to elaborate.

A PRELIMINARY STUDY: DOUBLE-PORTRAITOne of us (Tchalenko) had participated in a previous project,Double-Portrait, during which Ocean was filmed nearly con-tinuously from the model’s point of view [4] while he waspainting the camera operator and sound engineer. The re-sulting video material allowed approximate timing [5] of thepainter’s gaze and body movements, and in this way the keyfactors governing the picture-production process could beidentified. The first of these concerned the number of timesthe painter looked at his model. During a 20-×-30 cm prelimi-nary pencil sketch that he completed in 12 minutes, he fixedhis gaze 157 times on the model [6] at a rate of 13 times perminute, with an average fixation duration of 0.92 seconds.For the final 130-×-100-cm painting, completed in about 100hours over 12 days, we estimated that he made over 25,000fixations on the model at a rate of 6–12 per minute (after al-

lowing for pauses for mixingpaints, adjusting canvas, etc). Av-erage fixation duration wasslightly over 1 second. Hence, theartist was interrupting his workabout 12 times a minute to lookat the model for about 1 secondeach time.

Based on this work, we under-took a detailed study of eye andhand movement in the same artist.

TRACKING EYE MOVEMENTS: METHODSThe eyetracker (an AlphaBio Eyeputer) is a specialized video-camera system mounted on a headset (Fig. 1) and coupled toa computer. The infrared-sensitive camera records a close-upimage of the eye while the computer calculates the positionof the center of the pupil. The headset includes a second“scene” camera filming the scene in front of the head. Ini-

T E C H N I C A L N O T E

A Painter’s Eye Movements:A Study of Eye and Hand Movementduring Portrait Drawing

R.C. Mialland John Tchalenko

A B S T R A C T

The mental processes that al-low an artist to transform visualimages—e.g. those of his model—into a picture on the canvas arenot easily studied. The authors re-port work measuring the eye andhand movements of a single artist,chosen for his detailed and realis-tic portraits produced from life. Hiseye fixations when painting ordrawing were of twice the durationof those when he was not paintingand also quite different from thoseof novice artists. His eye-hand co-ordination pattern also showed dif-ferences from that of novices, be-ing more temporally consistent.This preliminary work suggeststhat detailed and quantitative analy-sis of a working artist is feasibleand will illuminate the process ofartistic creation.

R.C. Miall (neuroscientist), University Laboratory of Physiology, Oxford University,Parks Road, Oxford OX1 3PT, U.K. E-mail: <chris.miall@physiol.ox.ac.uk>.

John Tchalenko (filmmaker), Camberwell College of Arts, Peckham Road, London SE58UF, U.K. E-mail: <j-t@dircon.co.uk>.

Fig. 1. The eyetracker was mounted on a helmet and incorporatedan eye camera recording the pupil position and a scene camera re-cording the view in front of the subject. The head-position moni-tor allowed calculation of the gaze position of the eye even duringhead movement. A similar monitor was attached to the artist’spencil. (Photo: John Tchalenko)

36 Miall and Tchalenko, A Painter’s Eye Movements

tially, the eyetracker had to be calibratedby asking the artist to steadily look at aset of fixed targets. After this, the coor-dinates of each point in space that theartist viewed could be calculated in realtime, superimposed on the scene cam-era image and digitally stored on thecomputer for detailed analysis.

For this project we combined theserecordings with those from a hand-movement sensor (a Polhemus Fastrakmotion-analysis system) to follow themovement of the artist’s pencil. This de-vice records the three-dimensional (3D)position of a lightweight marker at-tached to the artist’s hand or pencil [7].While these eye and hand movementswere being recorded, the drawing’sprogress was filmed continuously with aclose-up video camera.

EXPERIMENTAL PROCEDUREA prime consideration throughout theproject was to obtain data about an artistat work, rather than from an artist per-forming scientific tests. The methodol-ogy was therefore as follows:

1. Selecting a model: The artist, wear-ing the eyetracker, sat facing an emptychair, and was presented one by onewith four male sitters. The duration andlocation of his initial fixations duringthe first few seconds were recorded inorder to establish what he was lookingfor in a prospective model.

2. Brief sketches: The artist wished tomake brief sketches of his candidatemodels to help him in his selection. Hedrew in ink in a small sketchpad held inhis left hand, each sketch lasting 1–2minutes.

3. The main portrait: The eyetrackingsystem was set up to record the artist’sright eye as he sat with a near-verticaldrawing pad positioned on an easel about45 cm in front and to the left of him. Themodel sat just to the right of the drawingpad, at the same distance, to minimizeparallax errors in the video recordingsmade from the scene camera. Theeyetracker was calibrated and used forabout 15 minutes each hour, and the art-ist then worked normally without theeyetracker for the next 30–40 minutes,while model and artist rested for the re-maining 10–15 minutes per hour. Thus,five recordings were made, spanning theentire portrait-drawing process. For thethird and fifth session, we placed a mo-tion-tracking monitor on the back of theartist’s hand to record the position of thepencil within an accuracy [8] of about 10mm and with a resolution of about 1 mm.

This provided records of the spatial loca-tion of major hand movements and alsoof the timing of all hand movements.

4. Secondary portrait: The followingday, three short portraits were per-formed, intended to take about 10 min-utes, so that the eyetracker could be wornthroughout each. The motion tracker wasattached directly onto the shaft of theartist’s pencil, increasing the spatial accu-racy of the recording of the pencil tip toabout 2 mm, with a resolution of 1 mm.

5. Comparison studies with untrainedsubjects: Some time after the detailedstudy, the artist and three untrained sub-jects made a series of very brief sketches.Each subject sketched from a black-and-white photocopy of a face [9] and was re-stricted to 1 minute per sketch. Both eye-gaze and pencil positions were recorded.

RESULTSSelecting a ModelAs each model candidate entered thecamera’s field of view and sat down,Ocean had already fixated on theperson’s left eye. He then made a num-ber of rapid fixations, each lasting an av-erage of 0.4 seconds, until his eye cameto rest temporarily, for about 1 second.

With the last candidate, subsequently se-lected as the model for the main por-trait, these first few fixations were essen-tially confined to the person’s eyes [10].This high rate of fixations, about 140per minute, and concentration with thesubjects’ eyes, are typical of eye move-ments reported by others [11].

Brief SketchesFixation rate and duration over the foursketches were remarkably consistent(Table 1). Overall average fixation dura-tion was 1 second, and the rate just un-der 22 fixations/minute.

The Main PortraitThe eyetracker data can be usefullyviewed as a time series to emphasize thetemporal pattern of Ocean’s eye move-ments (Fig. 2). We used the spatial loca-tion of the eye’s gaze to color-code thetime series, with segments in light greydenoting the periods Ocean was lookingat the model (Nick), and in black, at thepaper. Blinks, which hide the eye’s pupilfrom the eyetracker camera, appear asvertical lines.

Ocean’s eye movements adopted aregular rhythm from the beginning ofNick’s portrait. For the first 35 seconds

Fig. 2. Nick’s portrait: the first 2.5 minutes. Horizontal position of the artist’s eye againsttime, color-coded light grey when the eye fixated the model and black at all other times.The descending vertical lines are blinks, which are commonly made at the same time aslarge saccadic eye movements. (top and bottom left) The initial visualization period isshown in panel a, when the artist explored the blank paper with a rapid series of fixations.Compare the long fixations on the model’s face with the brief fixations made to positionson the paper. (top and bottom right) Areas marked 5, 6 and 7 denote the drawing of a con-tinuous line on the paper. The fixations on the model’s face are rhythmic and of nearlyequal duration.

Miall and Tchalenko, A Painter’s Eye Movements 37

lowed precisely by Ocean’s eyes, in asmooth movement [14]. Occasionally,the eyes would rapidly look away to fix-ate upon the corresponding detail onthe model, or upon another part of thedrawing, before returning to follow thepencil tip. Practice movements are seenin many tasks and sports requiringskilled movement and serve to refresh ashort term “motor memory” of how thebody moves [15]. In other instances,the “practice” resulted in very faint pen-cil marks on the paper, perhaps to aidin deciding on the exact form of the fi-nal line. The difference in pencil posi-tion between these two movementtypes, practice and drawing, were soslight (probably less than 1 mm fromthe paper surface) that our trackingtechnique could not separate them,and close-up video was required fortheir study.

Fig. 3. Ocean’s eye and hand movements while drawing Nick’s lips. (top and middle left)The synchronization of eye and hand movement, with a fixation on the model just as thepencil approaches the paper. (top and middle right) A period of rehearsal, with rapid cy-clic pencil motion and corresponding smooth movements of the eye. (bottom right) Eightseparate pencil strokes (a to h) were used, most of which were preceded by practice move-ments of the hand as shown on the graph.

Table 1. Ocean’s eye fixation on the four candidate models during brief pensketches.

Total Duration No. of AverageCandidate of Drawing Fixations on Fixation FixationsNo. (Minutes) Model Duration (Sec) per Min

1 1.35 29 1.05 21.52 1.35 31 1.10 22.63 1.98 41 0.98 20.74 2.16 47 0.82 21.8Average 1–4 0.99 21.6

(Fig. 2a), he scanned the blank paper,occasionally referring back to themodel, suggesting a visualization pro-cess. He then started drawing Nick’sright eye [12], falling into a pattern ofregular fixations on the model’s face,each lasting 0.6–1.0 seconds, at a rate ofabout 12 per minute. One of the firstcontinuous lines he drew (at 55 sec-onds) was a 4.5-cm contour of themodel’s eyebrow (segments marked 5, 6,7 in Fig. 2b). While drawing this line, hestopped twice to glance back at themodel, suggesting that his eye was cap-turing about 1.5 cm of detail per fixa-tion. The drawing of the right eye pro-ceeded for about 1 hour with theaccumulation of such small marks.

Some subtle variations of this basicpattern appeared. For example, therewere long fixations on the model at thestart, during the first minute. An hourlater, as Ocean drew the hair, therewere rapid sequences of alternation be-tween the subject and the paper. In thethird hour, as he drew the lips, therewere regular and spaced sequences.These differences suggest that theremay be a consistent relationship be-tween the complexity of the visual ob-ject being viewed and the viewing pat-tern. In particular, when capturing thesweep of the hair, Ocean made rapidcomparisons of the model and the pa-per. As he added detail to the alreadypartly drawn lips, a regular process ofvisual capture followed by drawing tookplace. The eye frequently returned tothe same location on the model [13], ata rate that would indicate visualmemory was refreshed about every 5seconds.

We analyzed in detail Ocean’s handmovements while he drew the model’slips and observed two interleaved move-ment cycles (Fig. 3). The first had a pe-riod of about 20 seconds: Ocean raisedhis hand from his lap close to the paperand drew and/or practiced, then low-ered his hand again to his lap. In the sec-ond, with a period of about 5 seconds,he fixated on one detail on the model

and then on one or more positions onthe paper as he drew that detail. Theprecise timing of these movements re-quires more exploration, but this strat-egy likely provided the most recent visualinput prior to hand movement.

Ocean’s drawing was frequently ac-companied by repeated practice strokes(Fig. 3). The pencil would move severaltimes just above the paper’s surface, fol-

38 Miall and Tchalenko, A Painter’s Eye Movements

Secondary PortraitsThe intermittent recorded test sessionsimposed an unnatural break to theartist’s normal routine. We thereforeagreed that he would produce a portraitwithin a shorter period, allowing us tomonitor eye and hand movementthroughout. Figure 4 shows the com-plete record of eye and hand movementduring an 11.5-minute pencil drawing.One can see the very rhythmic sequenceof eye movements between the paperand the model, while the upper recordshows a pattern of hand movement simi-lar to that seen in Fig. 3. The only signifi-cant difference was that the rest periodsof the hand were less frequent, Ocean’shand dropping about once per minute.This may have reflected increased timepressure compared with the previous,longer portrait. In the eye-movementtrace, the steady pattern of fixations onthe model is again clear, with an averagerate of 13 per minute, although for briefperiods, e.g. at 5.3 minutes and 8 min-utes, the rate increased. As before, itseems likely that this very rapid compari-son of model and drawing was made inorder to assess accurately the spatial ar-rangement, as each period preceded thedrawing of a new feature (the right eyeat 5.3 minutes, the lips at 8 minutes).The slower and steadier sequences thatwere seen occurred while Ocean drewthe finer details of each feature.

Figure 5 shows the completed drawingand the corresponding spatial record ofthe pencil-tip movements. The differ-ences between the black lines and the fi-nal drawing reflect practice hand move-ments that were not committed to paper.

Comparison with UntrainedSubjectsPreliminary comparisons were madebetween Ocean and three novices. We

imposed a time limit of 1 minute persketch while they copied photographedfaces. Ocean’s fixation duration re-mained at 0.6–1.0 seconds, typical ofthat recorded previously. The novices’durations were about half as long. Fur-thermore, Ocean’s fixations were al-ways single, whereas the novices’ weregenerally multiple. Ocean locked hisgaze onto one position, apparently tak-ing in a single detail, while the novicesfixated on two or more positions, some-times quite separate. The briefer andless consistent fixation durations of thenovices were more typical of everydayeye movements and more typical ofOcean’s fixation pattern when notdrawing or painting.

It seems possible according to thisdata that a graded pattern of eye move-ment correlates with drawing skill. Themost fluid and accurate drawing by ourthree novices was made with the mostconsistent, longest fixations on themodel; the least accurate, with the leastconsistent eye movements. However,more tests are needed to confirm this.

DISCUSSION

Our study of Ocean at work has shownthat his eye movements while drawing aportrait are different from his normaleye movements. While drawing, hemade a sequence of regular single fixa-

tions on selected details of the model’sface (Table 2).

Between fixations on his model, Oceanwould look at his drawing with shorter,more rapid fixations or, alternatively, withsmooth movements that followed thepencil tip across the paper. Such eye-hand coordination was most dramaticallyillustrated by the practice sequences,when the pencil stroke was preceded by arehearsing action just off the paper’s sur-face. When changing his gaze from pic-ture to model, or model to picture,Ocean’s fixations were precisely targeted.

In contrast, untrained subjects didnot show clear changes in eye move-ment when drawing; nor did they showprecise fixation on individual details ofthe model.

How then does a skilled painter trans-form a vision of the external world intoa picture on the canvas? The followingfive remarks help situate our observa-tions of Ocean in the more general con-text of his way of working.

1. The capture of visual informationdetail by detail, rather than in a moreholistic manner, is reflected in the waythe drawing or painting is built up. Eachdetail and each element is of intrinsicimportance. In Ocean’s words: “Theshape you are putting down is always ab-stract. . . . That next rhomboid is theside of the nose—or it is an abstractshape. Each bit of the picture has to beable to exist in its own right” [16].

Table 2. Ocean’s average fixation timings while drawing a portrait.

Portrait Fixation FixationsDuration Duration (Sec) per Min

Double-Portrait Sketch (pencil) 12 min 0.9 13Brief Sketches (pen) 2 min 1.0 22Nick—Main Portrait (pencil) 2 min 1.0 22Luke 2—Secondary Portraits (pencil) 11.5 min 0.9 (st. dev. 0.47) 13From Photo Sketches (Pencil) 1 min 0.86 26

Fig. 4. The complete recording of eye and hand movement when drawing Luke 2. The upper graph is the pencil position (actually a calcu-lation of motion through 3D space), color-coded black when the pencil was within 1 cm of the paper and grey otherwise. The eye move-ments made during this 2-minute interval are shown as well (lower graph). The great regularity of Ocean’s eye movements and theperiodic pauses in his drawing are apparent.

Miall and Tchalenko, A Painter’s Eye Movements 39

2. The artist’s actions are essentiallydriven by the picture’s progress—theyare goal oriented rather than, as in hisfirst encounter with his candidate mod-els, stimulus controlled. Each glance atthe model is meant to develop whateverhe is drawing at the time. Such behavioris quite different from the normal way ofviewing a face or a painting. Hence, per-ceptual and cognitive studies of finishedartwork will not necessarily be relevantto the production of pictures.

3. The artist’s eye and eye-hand skillsare definable in terms of physiologicalparameters: fixation stability, fixationduration, targeting efficiency, etc. Asthese patterns were not found in ouruntrained subjects, one may assume, atthis stage at least, that they are acquiredthrough training and practice. Further-more, with Ocean, they reflect his preci-sion: “If the line lands a millimeter tothe right or a millimeter to the left, itchanges the weight, in some way, of theshape that it is describing. So when thatline lands, you just want it to land in theright position, whatever that is!”

4. Very few lines in Ocean’s portraitsrepresent actual lines on the face. Mostare subtle demarcations between areas ofdiffering light intensity, texture andcolor. Not only are their precise locationson the model’s face subjective to eachviewer, but for the artist, they also de-pend on what he wants to do with them.Ocean remarked: “I’m sure of what I amseeing, I’m not quite sure what I am go-

ing to do about it. So I make a decision.The final result is made up of a greatmany decisions.” The artist’s skills allowhim to draw with precision whatever linehe chooses; the choice of line is deter-mined by other factors.

5. The last observation centers on thischoice: undoubtedly Ocean’s reaction isdominated by the visual input to hisretina. “At any given moment I will startfrom what I can see from where I am. Itry to achieve a likeness. But what I wantis a likeness to the reaction I have tosomething I can see.” When he is doinga portrait, his vision concentrates nearlyexclusively on the model and the canvas.It is this vision that we have investigatedwith the eyetracker.

Acknowledgment

The Wellcome Trust provided support for Miall’sresearch, and a Sci-Art grant funded the project.

References and Notes

1. Reviews of existing literature may be found inP.J. Locher, “The Contribution of Eye-MovementResearch to an Understanding of the Nature ofPictorial Balance Perception: A Review of the Lit-erature,” Journal of the International Association ofEmpirical Studies 14, No. 2, 143–163 (1996); C.F.Nodine, P.J. Locher and E.A. Krupinski, “The Roleof Formal Art Training on Perception and Aes-thetic Judgment of Art Compositions,” Leonardo26, No. 3, 219–227 (1993); R.L. Solso, Cognitionand the Visual Arts (Cambridge, MA: MIT Press,1994) p. 294.

2. Recent contributions include J. Fish and S. Scriv-ener, “Amplifying the Mind’s Eye,” Leonardo 23, No.1, 117–126 (1990); V.J. Konecni, “Portraiture: AnExperimental Study of the Creative Process,”

Leonardo 24, No. 3, 325–328 (1991); C. Frith and J.Law, “Cognitive and Physiological Processes Under-lying Drawing Skills,” Leonardo 28, No. 3, 203–205(1995); D.J. Cohen and S. Bennett, “Why Can’tMost People Draw What They See?” Journal of Ex-perimental Psychology: Human Perception and Perfor-mance 23, No. 3, 609–621 (1997).

3. See accompanying article by R.L. Solso, “Cere-bral Activities in an Expert versus a Novice Artist:An fMRI Study,” Leonardo 34, No. 1, 29–32 (2001).

4. J. Tchalenko, “The Making of Double-Portrait,”(London: Dulwich Picture Gallery, 1991) p. 4.

5. To an accuracy of 1⁄25 second, although the startand end of each eye movement were difficult to de-termine with precision.

6. Strictly speaking, we could only be sure at thisstage that the artist was looking at his model, an ac-tion that we termed “glance.” However, subsequenteyetracking showed that, virtually without excep-tion, each of Ocean’s glances was made up of asingle fixation.

7. With Ocean, pencil, hand and often also his armmove as a single unit. Thus, the spatial location ofthe pencil tip can be determined from a markerplaced on the back of his hand, as there is littlemovement between pencil and hand.

8. See glossary for explanation.

9. P. Ekman, Emotion in the Human Face: Guidelinesfor Research and an Integration of Findings (Oxfordand New York: Pergamon, 1972).

10. A.L. Yarbus, Eye Movements and Vision (NewYork: Plenum Press, 1967).

11. Yarbus [10]; R.H.S. Carpenter, Movements of theEyes (London: Pion Press, 1988); J.M. Hendersonand A. Hollingworth, “High-Level Scene Percep-tion,” Annual Review of Psychology 50 (1999) pp.243–271.

12. Ocean nearly always starts a portrait with draw-ing one of his subject’s eyes, working outward tothe rest of the face.

13. The eye’s fixation point can be estimated toabout 2° under these conditions. The fovea, the re-gion of highest spatial acuity on the retina, is about0.5° in diameter. Hence we cannot be certain thatthe eye was actually fixating the exact detail twice,but this seems likely, given drawing of the same partoften followed each fixation.

14. The human eye makes five main types of eyemovement. We mainly considered the rapid gazeshifts known as saccades typical of voluntary eyemovement across a scene. The eye can alsosmoothly track a moving target, to maintain the tar-get image steady on the retina. See R. Dodge, Ameri-can Journal of Physiology 8 (1903) pp. 307–329. In thecases observed here, we believe that the smoothtracking may be a combination of smooth pursuit ofthe moving pencil and smooth eye movement com-pensating for small head oscillations (less than 0.6°maximum) synchronous with the hand movements.

15. J.A.S. Kelso, ed., Human Motor Behavior(Hillsdale, NJ, and London: Lawrence Erlbaum,1982); R.A. Schmidt, Motor Learning and Perfor-mance: From Principles to Practice (Champaign, IL:Human Kinetics Books, 1991).

16. These quotations are from interviews withOcean to appear in The Way of a Painter by JohnTchalenko (in preparation).

Glossary

eyetracker—a head-mounted device used to recordeye movement by providing a close-up video imageof the pupil of one eye.

Fastrak—a motion analysis system, comprising a veryweak electromagnetic transmitter mounted on atable top and a small lightweight receiver mounted

Fig. 5. (left) The portrait Luke 2. (right) A computer reconstruction of the pencil motionperformed during drawing. Black lines indicate when the pencil was within 1 cm of the pa-per; grey lines indicate transfer of the pencil to and from the paper surface.

40 Miall and Tchalenko, A Painter’s Eye Movements

on a moving object, capable of recording the posi-tion and angle of the object within a sphere with aradius of about 90 cm.

fixation—the eye moves most commonly in briefjumps (saccades), each lasting about 100 millisec-onds. In between, the eye steadily views a singlegaze position (or if the target is moving, a singleobject). This act of viewing is referred to here asfixation.

gaze position—the position on the visual sceneviewed during a fixation. In these experiments, weconsider gaze position to include a single visual de-tail, although we cannot assess by eye-movementmeasurement alone how much peripheral visual in-put is also captured.

resolution—the smallest observable movement ofthe eye; this is smaller than the spatial accuracy, asonly the relative movement of the gaze position,and not its true position, is measured.

saccade—rapid motion of the eye from one posi-tion to another, usually lasting between 50 and 150milliseconds and typically occurring 2–3 times persecond.

spatial accuracy—the accuracy with which gaze po-sition can be estimated, with respect to the actualgaze position.

Chris Miall studied zoology as an under-graduate and graduate student at ImperialCollege, London. He has since worked as aresearch scientist in the study of neural con-trol at Oxford University and various otherinstitutions. Since 1994 he has been a SeniorResearch Fellow for the Wellcome Trust, andhe is a University Reader in Neuroscience.His research focuses on the role of the cerebel-lum in the guidance of movement.

John Tchalenko studied geology and civil engi-neering and has lectured in seismotectonics atImperial College of Science in London, wherehis research specialized in Near Eastern earth-quakes. He came to films in 1975 via the Na-tional Film and TV school and directed anumber of broadcast films on science and artsubjects. His collaboration with HumphreyOcean started in 1991 with “Double Por-trait,” which preceded the present “Painter’sEye” project. At present he is head of the Draw-ing and Cognition research section atCamberwell College of Arts, London.

Manuscript received 26 October 1999.