tions exist to these earliest forms, as we shall see, but they are all modern structures, in themselves, and render the enquiry less easy than it might have been. But if the modernity of our documents be remembered, and due allowance made for this circumstance, the difficulties will be in no sense insuperable. We must not, however, expect our modern structures to illustrate in each and every particular the precise stage of the development which the logic of the temples as a whole will put before us, at each given point. It is for considerations such as these that it is not now possible to illustrate among the temples extant in Tirhut to-day any of exactly the most primitive type I have taken theoretically as our starting point in this enquiry. But we see what is essentially the same in certain rude little shrines at Sonpur (Plate I). Here, as in the structure which we have hypothecated, we see a simple square chamber, constituting the cella, with a simple, ordinary roof, rising to a point, and with a narrow porch in front. These are the constituents of our simplest form of temple, and in the present example we find them in as simple forms as now are traceable among the existing and recorded monuments. It will be admitted that in primitiveness they are almost all that could be wished. They show, however, one feature which is regrettable namely the false arches applied decoratively to the sides of the cella wall. These are, of course, extremely modern elements, their ornamental cusps betraying Muhammadan influence, and they render the monument less suitable for our present purposes than could be wished. However, if we eliminate this feature, and consider for the moment that these arches are not there (they are of course wholly non-essential), we shall have the primitive type we postulate, with perfectly plain, undecorated walls, and an equally undecorated pointed roof, square in plan, as is the cella proper. It was this very circumstance of the plainness of these surfaces which, so far as I can judge, gave rise in course of time to the entire development which we seek to trace. Nowhere are the beauties of the play of light and shade more appreciated than in India, whether because of the brightness of the sun or because of the innate æstheticism of the Hindu heart I cannot say. At all events, the Hindu has never been unmindful of this feature, and an aesthetic utilization of shadow is a conspicuous part of the beauty of most, if not all, Indian monuments. The monotonous expanse of this plain wall, calculated in Indian conditions to become a mere intolerable glare, was not long endurable, we may be sure. The builder sought to diversify this surface accordingly, with a view, so I conceive, of introducing shadow, and to this end conceived the simple, but epoch-making expedient of advancing the central portion of his wall a little way (Plate 2). By building out a central projection of this kind, he at once broke his plain surface by distributing it in two planes, and thereby gained the desiderated shadow. But let us now observe that even in this most modern of examples, a striking characteristic of the building is its prevailing lack of eaves. There is, to be sure, a slight projection around the top of the wall, intermediate between it and the actual roof, but it is in no sense conspicuous, and, in the most primitive examples, may not have existed at all. I rather infer that it did not exist, originally, because in the case of the Shiva temple at Padram, in Sāran District, where we see the first instance of the projection on the cella wall of which I speak, we find that the contour of this projection has been followed up into the region of the roof as well, and that its outer edge here also conforms, very naturally, to the configuration of this roof, running parallel to the edge again. This result would have been not only facilitated by the absence of eaves, it would have been rendered almost a constructional necessity, for, with the side wall distributed in two planes, as it is here, the roof, if rising from the inner one of these planes, only, would have left the top portion of the second one bare, unfinished and objectionable. But so slight a rim as surrounds the top of the wall in this temple at Padram would not have prevented the carrying of the outline into the region of the roof, even if we assume that it or its counterpart did actually exist in prehistoric instances. [J.B.O.R.S. 124 TEMPLE TYPES IN TIRHUT. Once carried upwards in this way, however, the projection on the cella wall automatically, or with the minimum of deliberate invention, supplies us with what is one of the most striking peculiarities of Hindu architectural ornament, namely, the decorative miniature. Fergusson at one place remarks (Volume II, page 99), that "almost all the ornaments on the facades of Buddhist temples are repetitions of themselves; but the Hindus do not seem to have adopted this system so early, and the extent to which it is carried is generally a fair test of the age of Hindu temples". How such a form could have developed naturally in a temple of purely Orissan type I cannot myself perceive; but in the case of this temple at Padram it is clear that there is more than constructional propriety in it, it is almost a structural necessity. Were it not a fact that the Orissan temples are centuries older than any of the structures now extant in Tirhut, one would be tempted to suspect that the miniature as such originated with the Tirhut type, and, having become established in Indian architecture as a decorative device, was subsequently applied with less architectural propriety elsewhere. It is perhaps not impossible that this is really the case, despite the absence of quotable instances today of this Tirhut form in really ancient examples. But whatever the history of the miniature in the Orissan type, it will be, I think, obvious as we proceed that in the Tirhut type it developed in this simple, unimaginative way. And why? Because, having at last conceived this principle of achieving light and shade by the advancement of one portion of his wall, with this resultant single miniature as decoration for the sikhara the architect very naturally next proceeded to repeat the process, when, by advancing yet a further portion of the wall symmetrically with the first and thereby distributing the surface in three planes instead of two, he attained not only added play of shade, but also a second miniature, as in the Har-Mandir at Ghataru in Muzaffarpur (Plate No. 3). That this is a great step forward all observers will admit, for this Har-Mandir is as chaste and beautiful as it is simple in its every part. Nor is it surprising that the success of this venture should have encouraged the builder to continue in the former course, and advance yet a third portion in the same way, with the result observable in the Mahadeva temple in Mahalla Garhi Tir in Chapra (Plate No. 4). It speaks equally for his good taste that he recognized this as the proper limit of this style, for no temple recorded in Tirhut shows any attempt to advance further along these lines. The Chapra example is the culmination of this style, and will appeal to Europeans as to Indians in its every line. It is indeed remarkable how closely it approximates in feeling to many spires on Christian churches in the Western world, although it is, I take it, perfectly obvious from what we have already seen that both in origin and development the whole is absolutely local and This fourth type, wherein the cella wall is broken by three indigenous. projections and the tower decorated with three miniatures, forms, as I have just remarked, the culmination of the development in this direction. But if we now turn back to Type 2, the Shiva temple at Padram, where there is only one such projection, we observe that in distributing the surface of the wall in two planes, this projection also divides it into three vertical panels, one formed by the projection itself, the other two by the parts of the wall at either side remaining unadvanced. At some stage of the architectural history this threefold division appears to have come prominently into notice, and the architect conceived the idea of balancing this triplicity rythmically by a corresponding threefold division of his tower in horizontal storeys (Plate 5). The result is seen in the Shiva temple at Bagaha in Champaran District. Here we have a three-storeyed tower corresponding rythmically to the three panels on the cella wall, each of the three storeys being formed of rows of miniature sikkaras, individually evolved, so I conceive it, in the simple way that we have seen, but become stereotyped as an accepted architectural ornament before the creation of Type 5 was possible.* A form transitional between Types 4 and 5 is afforded us by temples of Type 2 where minor miniatures are placed at the four corners rising to half the height of the main or central miniature. This results in a threefold horizontal division of the tower, also, and is presumably the origin of the rythm between these horizontal storeys and the vertical panelling. 126 TEMPLE TYPES IN TIRHUT. [J.B.O.R.S. At first sight one may be inclined to doubt whether there is really any connexion between these three storeys here in the tower and the three unobtrusive panels in the wall, but so far as my present survey of the Tirhut temples enables me to judge (and I have photographs of almost every building of any interest at all), this rythm between the vertical panelling and the horizontal banding is a constant one. A few insignificant exceptions do exist, but these are apparently spurious modern forms, built by humble modern masons ignorant of the true principles of their art. They form no integral part of the development and as mere architectural mistakes may be left out of consideration in this paper. Besides, these subtypes are the great exception. In the vast majority of cases the relationship between the divisions of the cella wall and those of the sikhara or tower is faithfully maintained. This results in the interesting fact that the number of such horizontal storeys is regularly an uneven one. The projections on the cella wall naturally divide it into either three or five or seven vertical panels, according to the number of these projections; they cannot divide it into four or six or eight. In consequence we find that the horizontal storeys similarly advance in odd numbers, the next step in the development being illustrated by the Kamaleswarnath temple at Triveni in Champaran District (Plate 6). Here we see that the side of the cella wall has been broken into five vertical panels by means of the two projections; but instead of decorating the tower vertically by two resultant miniatures, as we saw in the case of the Har-Mandir at Ghataru where the cella was similarly constructed, we here find that the sikhara has been banded horizontally by five rows of miniatures. The temple is a very modern one, of course, and the asthetic variation in the size of these miniatures makes the counting of the storeys in the photograph less easy than we could desire, perhaps, although there actually are five storeys in reality; but in the case of the next example (Plate 7), the Mahadeva asthan at Saurath in Darbhanga, and in the other examples we shall see, the counting is obvious enough. Here the projections on the cella wall are three, which divide the |