Home | Biodata | Biography | Photo Gallery | Publications | Tributes
[Back to Iconometry List]


An iconometric study of Pallava sculptures
Kalakshetra Quarterly, Vol. 3 No. 2, 1980, pp. 7-15
Gift Siromoney, M. Bagavandas and S.Govindaraju

Computer methods are used for the analysis of facial proportions of South Indian carvings. The study is based on measurements made by anthropometric instruments on thirty-nine well preserved carvings in sandstone from Kailasanatha temple, Kanchipuram, one of the earliest structural temples of Tamil Nadu built around 700 A.D.

Cluster analysis is used, instead of visual judgment, for grouping the sculptures into groups that contain most similar carvings. Some of the most similar carvings would have been carved by the same sculptor. Special features of these sculptures, such as extraordinarily long nose and a small chin are not to be seen in either the earlier or the later periods.

The average values of the facial proportions are given and these may prove useful in restoration work. These values are quite different from the proportions prescribed in the Indian canonical texts. There is no reason to believe that any canon of iconometry was followed by the sculptors of Kailasanatha temple.

It is also shown that Indian canons have many general features in common. There were two systems of iconometry and they had got mingled together in the canonical texts and the existence of the two systems had been overlooked by scholars.


Art historians have been aware of the existence of canonical texts on Indian iconography and iconometry and have taken for granted the increasing influence of these Sanskrit texts in the actual making of the sculptures. Some have even gone as far as to say that the very existence and the increased use of these texts in practice had led to the decadence of Indian art during the sixteenth and seventeenth centuries (Gopinatha Rao 1914). A vast majority of art historians have assumed that the existence of iconometric canons actually led to the incorporation of canonical proportions in Indian sculpture (Khandalavala 1974). Almost all the studies done to date have not made any attempt to compare the canonical proportions with proportions of South Indian sculptures (Banerjea 1956).

The vigour and the freshness in the South Indian art of the Pallavas of the seventh and eighth centuries A.D. have been noted by several scholars and so have been the stereotyped forms of deities of the Vijayanagar period of the sixteenth century. It is generally held that the peak of artistic excellence was reached during the reign of the Imperial Cholas around 1000 A.D. The iconographic portions of the texts indicate the standard postures, gestures and weapons associated with different forms of deities of Hindu, Buddhist and Jain pantheons. Some of the canonical texts are mistakenly believed to be very ancient because of the mythical names associated with them. Some scholars claim that the canonical texts were followed in the creation of seventh century Pallava images (Nagaswamy 1973).


The canons of iconometry follow a complex system called talamana (Gopinatha Rao 1920), in which the basic units are called angula and tala and the latter stands for the length of the palm. The angula is either a fixed length (manangulam) or a proportion (dehalabdhangulam). A piece of sculpture may be made according to one of the ten main divisions of proportions. Each main division can be further subdivided into three other types. A basic type is the madyma navatala (or the standard nine face-length). The face is of length one tala, (or twelve angulas) the length from throat to navel is two tala, from navel to top of knee is three tala, from the lower knee to ankle is two tala making a total of eight tala. One tala is distributed equally between the heights of foot, knee, the neck and topknot. The nava tala thus has a total of nine tala units, in height (108 angulas). There are two other types in the nava tala division. The uttama nava tala type is four angulas taller and the adhama nava tala type is four angulas shorter. The four angulas are distributed evenly between the heights of the foot, the kneecap, the neck and the topknot. The angula unit (dehalabdhangulam) that we discuss here is a proportional unit, and figures of different sizes can be made following the same proportion. The system makes use of the fact that persons with disproportionately larger face length appear shorter and persons with disproportionately shorter face appear taller. Dwarf figures are made following a chatusra tala or a four tala system where the total height is four times the face length. Canons of iconometry describe ten divisions from the eka tala (or single tala) to dasa tala (or the ten tala).

On close scrutiny we find that the surviving texts do not follow this system uniformly. A second system must have got superimposed on this ancient canonical system and both the systems are found in the texts. We shall illustrate with some examples of the possible origins of this dual system of proportions in the texts. In Indian art the important figures in a group are often represented as taller figures and inferior beings are represented as smaller figures. To such smaller figures a lower tala is often prescribed. However, if both the larger and the smaller figures were to represent deities of equal rank (say Siva and Vishnu) then strictly speaking they should be made in the same proportion, or in other words in the same tala. To the sculptor then it would involve using different lengths to represent the angulaa larger size for the angula of the taller figures and a smaller size for the angula measure of the smaller figure and both figures may occur in the same sculptural panel. To overcome this difficulty sculptors would have recalculated the size of face, limbs etc., maintaining the same proportion for figures of different heights. Taking the larger figure as standard and assuming that it is in standard nava tala it would have a height of 108 angulas. Other figures could be made in different sizes but in the same proportion. A shorter figure could be made in the same proportion but with a height of only 96 units where the unit may be equal to the angula of a standard nava tala figure. A much larger figure also made in the same proportion could be of height 120 angulas of the standard figure. Twelve angulas (10x12) make ten talas or dasa tala. This larger figure may be reckoned to be made according to the second system in dasa tala even though its proportions would be close to the standard nava tala figures. Then its face would not be 12 angulas but more. These examples illustrate the possible origin of the second system. This second system has also got mingled in the texts with the first system. When the proportions of a figure are given in dasa tala one has to check carefully whether it belongs to the first system or the second system. If it belongs to the first system, its face length would be 12 angulas irrespective of its total height. The existence of the two systems in the texts has not been fully realised so far by scholars. We shall call the first system the pure tala mana system and the second system the derived tala mana system.

The different texts give the iconometric measurement often using both the systems. The measurements relate to height, width and sometimes the circumference of the different parts of the figure. Even though there is a lot of variety in measurements between the different texts, we claim that there are certain common general features. First the face length is equally divided between the fore-head, nose and nose-to-chin irrespective of the system. Secondly the pubis (base of the male organ) is the midpoint of the height of a nude figure. In other words the distance from the sole of the feet to the pubis is equal to the distance from the pubis to the topknot. Thirdly deities are prescribed a higher tala compared to human figures. One may interpret it as belonging to the pure talamana system or the derived system depending upon the text as well as the theme. Fourthly children will be represented in a lower tala like the chatusra tala (four tala) of the pure system. The face length will be comparatively large for children. There are a few exceptions to the general features that we have described but we are not sure of the authenticity of such texts.


Around 700 A.D. King Rajasimha Pallava built the famous Kailasanatha temple at Kanchipuram in South India and the temple is adorned with exquisite sandstone sculptural panels. The main sanctum tower is enclosed by a group of small shrines forming a rectangular enclosure. In front of the temple there are eight small shrines, also with sculptural panels. King Rajasimha also built the Shore Temple at Mahabalipuram and the hill top temple at Panamalai. There are many other Pallava monuments at Mahabalipuram but they belong to the pre-Rajasimha period.

Did Rajasimha and his architects follow any of the Silpasastras that have come down to us? So far scholars have concentrated on comparing the iconographic details of the texts with the sculptures of Rajasimha. In a study of the Somaskanda panels we found that there were enough uniformities in all of the panels of Rajasimha but these panels differed in iconographic details from the later traditions that have come down to us in the agamic literature (Lockwood, Siromoney, Dayanandan 1974). Then our main interest was in dress and ornaments and the various gestures and weapons. Now we add a new dimension to the study of Pallava sculptures by taking iconometric measurements. All the measurements reported in this paper were taken by one of the authors using anthropometric instruments.

If some canon of iconometry had been strictly followed one would expect the sanctum figures (of Somaskanda) in the front shrines of Kailasanatha complex to be of the same size since most of those shrines are of the same dimension. However, it is easy to check that the relief carvings of Siva in the back wall of the front shrines are of different sizes.

At Kailasanatha temple, if figures with tall makutas (crowns) are closely examined, the upper portion (or the head and torso) excluding the crown will be seen to be significantly shorter than the lower half (below waist). According to the canons both portions must be of equal height. Furthermore, in some figures the upper half of the leg is longer than the lower half but according to the canons the portion of the leg above the knee-cap must be equal to the length between ankle and knee-cap. If any of the canons that have come down to us had been used these would have been avoided. Canonical proportions (which would go well with Jain and Buddhist figures) and tall crowns do not go very well together. If a tall crown is added on to a well proportioned figure, that figure will appear out of proportion. This can be seen from the Manmata figure of Sundaravaradaperumal temple of Uttiramerur.

For male figures the canons propose that the face length be divided equally between the length of the fore-head, nose and nose-to-chin. This general rule is followed in the early sculptures of North India (Plate 1). It also agrees with the general anatomical features (Plate 2). The early Chola sculptures from Kodumbalur follow these proportions (Plate3). However, in Kailasanatha temple, gods, goddesses, chauri bearers (Plate 4), drummers and male worshippers are all represented with extraordinarily long noses. Compared to the distance from nose to chin, the nose is about one and half times long. The eyes are long and narrow. The lips are thick. These features differ from the features of ganas and the dwarapalakas who have bulging eyes and shorter noses.

Plate 1. Buddha figure from Sarnath 500 A.D. This pre-Pallava figure is in Madras Museum. The face length is divided into three equal parts. This is in accordance with the Silpa Sastras and human anatomical proportions as far as nose length is concerned.

Plate 2. South Indian girl. The facial proportions are quite different from the carving of the Chauri bearer.

These special proportions of the stapati/s of Kailasanatha temple have not come down to us in any text. We have made an attempt in this paper to get at these special proportions. These will be of value in restoration work. We give here the average values of these proportions of facial measurements taking the face length to be twelve angulas (Table 2). Ten different facial measurements were taken from each sculpture (Table 1). The total face length was taken from the base of hairline (or head-dress) to the chin. The morphological face length was taken from the top of the nose to chin. In the canons the nose was supposed to start from a point in line with eyes. At Kailasanatha temple the nose starts from the brow.


No. Variables Description
1. Nose LengthStraight distance between the root of the nose and subnasale.
2. Nose breadthStraight distance between the most laterally placed points on the nasal wings (alaria).
3. Lip LengthStraight distance between the two corners of the mouth (Chelion).
4. Lip breadthStraight distance between labrale superior and labrale inferior.
5. Face breadthStraight distance between the two tragia.
6. Eye lengthStraight distance between the internal and the external corners of the eye.
7. Eye breadthMaximum distance between the eye lids.
8. Nose to ChinStraight distance between subnasale and chin (gnathion)
9. Face lengthStraight distance between trichion (hair line or crown base) and chin (gnathion).
10. Morphological Face lengthStraight distance between the root of the nose and the chin (gnathion).



Variables A B C D Overall average
1. Nose length   5.28   5.30   5.34   5.51   5.37
2. Nose breadth   3.67   3.41   3.72   3.39   3.57
3. Lip length   4.61   4.25  4.44   4.06   4.36
4. Lip breadth   1.79   1.50  1.74   1.78   1.75
5. Face breadth 11.21 10.51 11.19 11.39 11.21
6. Eye length   3.28   3.24   3.46   3.50   3.40
7. Eye breadth   0.76   0.66   0.88   0.76   0.79
8. Nose to chin   3.64   3.53   3.64   3.41   3.56
9. Morphological face length   8.97   8.85  9.00   8.92   8.95
10. Nose/Nose to chin   1.45   1.50  1.47   1.62   1.51
11. Eye ln./Eye br.   4.32   4.91  3.93   4.61   4.35
12. No. of carvings 13   3 11 1340
          A: Gods;     B: Goddesses;     C: Human Males;     D. Human Females


When statistical techniques and computer facilities were not available it was the practice to rely entirely on visual comparison of figures. Recently in the study of proportions of Greek sculptures computer aided cluster analytic techniques have been made use of (Guralnick 1976). In this study we make use of similar techniques to compare South Indian sculptures.

First the ten facial measurements (Singh, Bhasin 1968) are taken and standardised by dividing throughout by face length. Then the figures are taken two at a time and compared. The dissimilarity between two figures is measured in terms of the Euclidean distance between them. In other words, each figure is imagined as a point in nine-dimensional space (since there are nine measurements for each figure) and the distance between the points calculated. If the distance between two figures is small then the similarity between them is high. This distance function is a common measure used in cluster analysis. On the basis of the distance measure the figures are grouped using a cluster analysis algorithm (Hartigan 1975). We have chosen the single linkage method since it is the simplest. In the single linkage method figures A, B, C will belong to a single group if A and B have a small distance between them and (say) B and C. Even if A and C are not very close they will belong to the same cluster (but A and C cannot be very much apart). An artist may draw two similar figures A and B and some features of B may be incorporated in the third figure C so that B and C may be very similar. Instead of the single linkage method, one would use other methods of clustering, if necessary.. We give here only the main findings of our study. Even though we have given the average values for four different categories of figures (Table 2) a cluster analysis shows that gods, goddesses, chauri-bearers and male worshippers have close resemblance. The figures that are most similar to each other are two human male figures and a Bikshadana. One male figure and the Bikshadana are on the western end of the southern side opposite to prakara shrines no. 22 and no. 23. This close facial similarity would support the theory that the same artist was responsible for both these figures which are in fairly close proximity. Furthermore the similarity with another male figure which is on the northern side opposite to prakara shrine no. 38 would support the theory that the same artist worked on the northern side also. Other evidences make one believe that the same artist was responsible for a large number of figures on the main sanctum walls. The slight differences in facial resemblance between three human figures opposite to shrine no. 22 are brought out. These three figures form a part of single panel and most likely to be the work of a single artist.

When the facial features of Siva of the east-facing Shore Temple are compared with the Kailasanatha figures, there is very close similarity between that figure and that of Kirata on the prakara shrine no. 16 and other figures. This would support the theory that the artist of the Shore-Temple Sanctum also worked on the sculptures at Kanchipuram. Or one may argue that the craft school or guild had reached such a level of perfection that they produced very similar figures.

In addition to the method of cluster analysis, another method called factor analysis was applied to the analysis of facial proportions of the sculptures of Kailasanatha temple (Siromoney, Bagavandas, Govindaraju, 1979). Factor analysis would reduce the number of variables from nine proportions to a smaller number of variables or factors. Each factor would be a linear combination of the proportions. If the sculptors had strictly adhered to any canon based on the talamana system, one would expect to find one major factor in terms of which the other variables can be expressed. This factor would represent the angula or a multiple of it as a basic unit used in the sculptures. A computer analysis, however does not reveal any single major factor. In other words there is no evidence to support the theory that the sculptors followed closely a canon of iconometry based on the angula or a tala.

The unusually straight long noses and long narrow eyes are the main characteristics of the Kailasanatha sculptures. These features are found in the Shore-Temple and at Panamalai. We shall proceed with the assumptions that these are Rajasimha characteristics. These features are not: found at Krishna Mandapa of the pre-Rajasimha period and other monuments of Mahabalipuram. These special features are not found at Muktesvara, Mathangesvara and the Vaikuntha Perumal (Srinivasan 1971) temples of Kanchipuram. These temples belong to the period of Nandivarman Pallavamalla of the post-Rajasimha period. The Mukundanayanar temple at Mahabalipuram that does not have any inscriptions is likely to belong to the post-Rajasimha period, since the figures on the Somaskanda panel do not exhibit the typical iconometric characteristics of Rajasimha period (Soundararajan 1969). However, some of the sculptures on Airavatesvara temple exhibit typical Rajasimha features and this temple is generally attributed to the Rajasimha period.

We have examined the facial features and found that the proportions are quite different from the canonical prescriptions. Because of the strictures against non-conformity with canonical proportions, Stapatis may often redefine the features in such a manner that the proportions may appear to fit in with the canonical prescriptions.

Plate 4. Chauri bearer from Kailasanatha temple Kanchipuram. This beautiful Pallava piece of Rajasimha's period has an unusually long nose. Early eighth century. This is not in accordance with any of the Silpa texts nor with common human anatomical proportions.

Plate 3. Tripurasundari from Kodumbalur. Early Chola sculpture from Madras Museum. The nose is not unusually long as in Rajasimha's sculpture.

We have shown that sculptures of the Kailasanatha temple have peculiar stylized characteristics that are different from features prescribed in canons of iconometry and different from the natural face. Many of the features prescribed in the canons follow the natural human face to some extent. Only when the characteristics of sculptures of a particular period differ clearly from the canonical proportions we conclude that either there were no canons or that the canonical proportions were not strictly adhered to. If there is not much divergence between the canonical proportions and the proportions of a sculpture then one may conclude that the canons were followed or alternatively that the sculptor was inspired by a particular human face.

We wish to examine more closely whether the Rajasimha characteristics are found in sculptures of the period preceding the Rajasimha period and also during the period following it.


The Krishna Mandapa is a well-known rock-cut panel of Mamallapuram. Judging from the costumes and jewellery depicted on the sculptures, it is closer in style to the period of Mahendra of the early part of the seventh century than to the style of Rajasimha. We shall treat Krishna Mandapa as a monument of pre-Rajasimha period. We give the facial measurements taking the face length to be 12 angulas (Table3).

The average values are quite different from the values of Kailasanathan sculptures. The nose is not unusually long. The ratio between average nose length and average nose-to-chin length is less than one (0.89) for human male figures and just a little over one (1.06) for female figures. The eyes are not long and narrow as in the Kailasanatha sculptures. The eye-length to eye-breadth ratio is 1.33 for the male figures and 2.13 for the female figures. The size of eyes prescribed in the canons is two angula long and one angula broad. The faces are quite elongated.

In Krishna Mandapa there are some figures of children depicted on the panel. The problem of depicting a child figure was solved by the authors of the canons who proposed that a four-face length proportion should be used for such figures. If the sculptors who created the Krishna Mandapa were aware of the canonical solution they would have used it to depict the standing child figures. Instead we find a standing child figure represented almost in adult proportions but on a smaller scale and shown as holding the hand of an adult. Its total height is a little over seven times the face length! In this case it is unconvincing to argue that the sculptors knew the canons but did not make use of the knowledge to depict a child with longer face length.

Another panel of the pre-Rajasimha period is the large rock-cut Penance Panel of Mamallapuram. The central yogic figure (commonly identified as Bagiratha or Arjuna) is certainly not made according to canonical proportions. The upper half of the body is much shorter than the lower half. According to the canon both halves must be equal. However the figure looks proportionate because the hands are held above the head. Furthermore according to the canons, the upper portion of the leg above the knee must be equal to the lower portion of the leg from ankle to knee. In the sculpture, the upper part of the leg is longer than the lower part. If the sculptor was aware of the canonical prescriptions he would have followed the canonical proportions at least for the leg, if not for the whole body.

As we have mentioned earlier, the Rajasimha characteristics are not found in the Pallava Sculptures of the post Rajasimha period. The sculptures of Vaikuntha Perumal temple of Kanchipuram look quite different from the Kailasanatha sculptures. This sudden change of nose length from the Rajasimha period to Nandivarman Pallavamalla's period could be attributed to the forcible or free migration of sculptors to the Chalukya country, death or mutilation of the sculptors or some other cause. Some temples in Pattadakkal in Chalukya territory have sculptures with unusually long noses too, and this might support the migration theory. On the other hand, the Pallava ideal of a long nose could have been inspired by the face of a particular Pallava queen who had an unusually long nose and a pretty face, and this later went out of fashion.

The Rajasimha characteristics are also absent in the sculptures of the Early Chola period in which the nose length is almost equal to nose-to-chin. This tendency of depicting the nose not too long and the eyes not too large continues till around 1000 A.D. When we examine the Chola bronzes of the period following 1000 A.D, we find that the noses are longer than prescribed by the canons and that the eyes are also longer. During the Vijayanagar period the eyes are depicted unnaturally large. They are both long as well as broad. We took measurements of Vishnu figures on the hundred pillared          mandapa of Varadaraja temple of Kanchipuram built in the sixteenth century. The canons had been well-established by then. Even here many of the prescriptions of the canons are not fully made use of. For example, baby Krishna is represented with a head larger than one for adult figures but proportionately not as large as prescribed by the canons. Furthermore canons prescribe that deities must be made in a higher tala than human figures. However there are many female figures in that mandapa which are represented with proportionately smaller heads than the standing Vishnu figures.


In conclusion we have shown that there exist two systems of proportions which have got intermingled in Indian canonical texts. Computer methods are very useful in analysing sculptures and for finding those pieces which resemble each other closely. The sculptures of Rajasimha Pallava were not made according to any of the canonical proportions that have come down to us. The Rajasimha characteristics of facial proportions are not found in the pre-Rajasimha or post-Rajasimha periods. Even during the sixteenth century when canons were well-established the sculptors did not follow closely the canonical prescriptions.

It looks as though the South Indian texts on iconometry were generally treated purely as theoretical exercises in proportions and seldom as practical guides by the sculptor for depicting in detail ideally proportioned figures.


Banerjea, J.H., 1956, The Development of Hindu Iconography, University of Calcutta, Calcutta.
Gopinatha Rao, T.A., 1914, Elements of Hindu Iconography, Vol. I, Part I, pp. 31, Madras.
Gopinatha Rao, T.A., 1920, Talamana or Iconometry, Memoirs of the Archaeological Survey of India, No. 3, Calcutta.
Guralnick, E., 1976, The proportions of some Archaic Greek sculptured figures: A Computer Analysis, Computers and Humanities, Vol.10, pp. 153-169.
Hartigan, J.A., 1975, Clustering Algorithms, John Wiley, New York.
Khandalavala, K., 1974, The Development of Style in Indian Painting, Macmillan, Madras.
Lockwood, M., Siromoney, G., and Dayanandan, P., 1974, Mahabalipuram Studies, Christian Literature Society, Madras.
Nagaswamy, R., 1973, Archaeology and Epigraphy in Tamil Nadu, Proceedings of the Third International Conference Seminar, Paris, Pondichery.
Singh, I.P., and Bhasin, M.K., 1968, Anthropometry, New Delhi.
Siromoney, G., Bagavandas, M. and Govindaraju, S., 1980 An application of Component Analysis to the study of South Indian Sculptures, Computers and the Humanities, Vol.14, (in press).
Soundararajan, K.V., 1969, Rajasimha's Temples, Transactions of the Archaeological Society of South India, 1962-65.
Srinivasan, K.R., 1971, Temples of South India, New Delhi.

Go to the top of the page

Home | Biodata | Biography | Photo Gallery | Publications | Tributes