Can Paleography Date Inscriptions

In her 1949 book, The Scythian Period, Lohuizen used a variety of methods (stylistic, cross-reference, and paleographic) to separate the inscriptions of the Kushan period into two groups, corresponding to a first and second century of the era of Kanishka. Here the paleographic element of Lohuizen's analysis (to which Chapter 5 of The Scythian Period is dedicated) will be examined, in order to answer a broader question: can we date inscriptions by paleography? That Lohuizen's analysis is chosen is a testament to the thoroughness of her work,  in which she has provided sufficient details for the method she used to be checked. Importantly, it is the method that is at issue here, not the results, which are not accepted by any experts.

Obviously inscriptions can be dated by paleography. Epigraphers studying all periods of history depend upon dating written material by the forms of the letters, but such dating are usually quite broad, with inscriptions being separated by several centuries, and the margin of error often stretching to century or two. In Kushan studies what needs to be separated are inscriptions which are within a hundred years of each other.

In Lohuizens analysis she is not consistent in the letter forms she uses to analyze inscriptions, sometimes one criteria and sometimes another. Dani (1986) heavily criticizes her method for this reason. He believes, and I agree, that any paleographic dating would need to be systematic and use consistent criteria. Though she mixes up criteria this list indicates the criteria she uses in order of most often to least often:

The approach taken here, in analyzing Lohuizen's method, is exactly the same as that which would be taken if dating the inscriptions from scratch. Firstly, a group of inscriptions that can be securely dated (independent of paleography) is found, then this can be used as a reference point, showing how the letter form changes over time. If changes in letter forms are clear cut, with a sharp break between early and later froms then the sequence is a useful tool for dating. Such a sequence would be referred to as diagnostic. If there is no sharp break, if early forms occur in late inscriptions, or late forms in early inscriptions, then this indicates that the aksara in question cannot be used for dating.

Fortunately, at Mathura, a hundred year plus sequence of inscriptions exists which can be securely dated. The first is an inscription of Kanishka from the year 23.(226), which has to be part of the first sequence of Kushan dates because Kanishka II rules only to the year 19/20. There are also several inscriptions mentioning Huvishka and a year, and several mentioning Vasudeva and a year. Lastly there are three inscriptions mentioning Vasishka which must belong after Vasudeva. Due to regional variations in scripts (which are not well understood) it is important to take all the examples from a single site, so a number of inscriptions not from Mathura (notably those of Bala and Buddhamitra known to belong to the reign of the first Kanishka from Eastern India) have been excluded. This leaves in total 21 inscriptions over a period of about 103 years (exactly if the second century hypothesis is correct).* 

For comparison purposes, aksaras  from the Ashoka inscriptions at Delhi, The Ksatrap inscriptions from Mathura of the 1st century, Gupta inscriptions at Mathura of the fourth and fifth centuries (from Dani), have been included on the charts. So the charts full extent is a period of about 700 years with the 120 of the Kushan inscriptions highlighted.

Form of Ku

The forms of Ku occurring in the inscriptions are presented in the chart above. They are colour coded from light to dark on the basis of Lohuizen's distinction of early to late. The first thing that should be noted is that 'ku' is not a common character, and there are too few examples to allow any conclusions to be drawn. This is somewhat surprising as 'ku' is the strongest argument Lohuizen puts forward. She even includes an appendix which clearly shows a development from the horizontal bar to the curved Gupta character. So why does this chart not support her conclusions? The reason is that Lohuizen's takes examples mostly from Jain inscriptions at Kankali Tila (the Jain word for school is 'Kula'), which for the most part do not name the king so could belong to the first or second sequence of Kushan dates. She then decides whether to date an inscription late or early on the basis of how the 'ku' appears. Once she has done that she is able to present a table of those inscriptions showing a clear development of form. Of course, this is a completely circular argument.

Form of Ya

'Ya' is a very common letter and so occurs in most of our inscriptions. It highlights two things: firstly that more securely dated inscriptions cluster in the reign of Huvishka than in the rest of the sequence. Secondly, it shows how marked a distinction can exist between a character of Asokan date and one of Gupta date, to the extent that they are almost completely different characters.

What it does not show is the development that Lohuizen argued for.Late forms are found very early, and early forms at the end of the sequence, even a classical Kushan form in the 5th century Gupta inscriptions! It is very clear that 'ya' is not diagnostic, for the date of inscriptions. This is not (like 'ku') above simply a lack of evidence, this is a point of absolute principle, because variations in 'ya' in the Kushan period are down to the predilections of individual scribes and are not related to the date.

Form of Palatal 'Na'

Palatal 'na' is a very promising sequence. It shows the sort of development that could well be considered diagnostic. Firstly, there is a huge difference between the Early and Gupta forms, showing that the aksara underwent rapid development. Secondly, the cursive form occurs fairly consistently in the reign of Vasudeva, or later, while the earliest from (similar to the pre-Kushan form) occurs consistently in the first 50 years of our sequence. They are separated by a brief period of an intermediary form. There is however one example of a late form in an early inscription so it would be unsafe to base the dating of an inscription on the form of this character alone.

Form of 'Ma'

The sequence for 'ma' does not support Lohuizen's argument at all. So consistent are the Kushan forms that I have included only a sample from each decade. What this indicates is that the discovery of an 'open' form in a Kushan inscription should be taken as blunder on the part of the engraver not as an indication of a late date.

Form of 'Sa'

Again, this is a very disappointing sequence if it is to be used as a dating tool. Lohuizen's supposed early and late forms are jumbled together in a manner that indicates no particular pattern. The one example of a closed loop occurs earlier than a very open form, and given that it is the only example it should be seen like open 'ma' as an engravers blunder, not a dating tool. The rest of the sequence should be viewed much like that for 'ya', as demonstrating that this particular aksara will never be suitable for a dating on this short scale. 

Form of 'Ha'

It is also worth noting that there are no forms of 'ha' which show the full development Lohuizen uses as a dating tool, so it is quite impossible to tell what that form represents, either early or late. In fact, the Gupta forms at Mathura do not differ from the Kushan forms significantly. There is not enough evidence to indicate that the distinction being made here is even meaningful. 

Head marks, and Sloping

The shape of head-marks or sloping of characters has not been directly considered. The second seems to be characteristic of poorly executed inscriptions and therefore of little use in dating, and a glance at the characters above will show that well developed head marks occur quite early in the sequence, rendering them of little use in separating early and late Kushan inscriptions.


Six sequences of letters, using 21 inscriptions from Mathura of known date, have been examined. Of these one sequence provides some evidence of change over the period in question, but could not support a dating on its own; one sequence has too few examples to tell us anything; and the other four sequences show that the characters cannot be used to date inscriptions in this refined a space, either because the aksara (as 'ma' or 'ha') is fairly constant throughout the period, or because the variations are clearly a result of individual scribes, not differences in date..

Frustrating as this, it indicates that the inscriptions that belong to the second Kushan sequence will not be uncovered on the basis of paleography. Though a number of scholars have claimed high degrees of accuracy in the dating of inscriptions the majority of skilled epigraphers strongly disagree. Salamon, and Dani are both highly critical of the accuracy of paleographic dating, and the following extract from Bhadarkar's discussion of the year 61 inscription of Chandragupta is typical of the sort of comments found in the specialist writing of epigraphers:

"The characters belong to the early Gupta period when they were practically identical with those of the Kushana records. This is particularly significant inasmuch as our inscription is found at Mathura, from where a number of Kushana epigrahs have already come to light. In fact, if would have been well-nigh impossible to say that ours was a Gupta and not a Kushan record, had it not contained the name of a Gupta King"

Bhandarkar isn't talking about inscriptions seperated by just a 100 years. This inscription is dated to the year 380 AD, more than 250 years after the first inscriptions of Kanishka found at Mathura. If the changes are that small over such a lengthy period it should hardly come as a great surprise that those attempting a paleographic separation of the two Kushan dating sequences have failed.


The inscriptions used are 226, 227, 228, 231, 234, 240, 242, 245, 254, 256, 263, 275, 280, 282, 290, 291, 299, 320, 771

Contents Page Introduction



Inscription List


Brahmi Alphabet




Robert Bracey.