1. Prologue: encounter with a complex linguistic scene

This article, which falls under the larger domain called History of Descriptive Linguistics and, inside that domain, concerns what Auroux (1994) calls grammatisation ‘grammatization’,² concentrates mostly on some of the traces left by speakers of Portuguese who were trying in the 16^th and in the 17^th centuries, by writing grammars and compiling dictionaries —see the extracts from the Arte composed by Henrique Henriques [1520–1600] (henceforth HH) and from the Vocabulario Tamulico Com a Significaçam Portugueza (henceforth VTCSP) inside Figures 1 and 2— in order to fullfill their Missionary activities, to master a South Indian Language, which they referred to as malauar (see Figure 1) or as Tamul (see entries 342_L_j & 342_L_k in Figure 2, and see the transcription-translation provided in (1) & (2) below).

Figure 1

Beginning of the 16th cent. Arte attributed to Henrique Henriques (HH) (cod. 3141, Biblioteca Nacional de Portugal, Lisbon).

Figure 2

Five entries (342_L_h to 342_L_l) from the 1679 VTCSP (Vocabulario Tamulico Com a Significaçam Portugueza).

As those missionaries would progressively discover, the tamul language which they were trying to master —usually referred to in English as ‘Tamil’— turned out to be a symbiotic combination of three languages, difficult to separate in practice, because each component of the Tamil Triglossia (see Figure 3) had its own role to play in the global picture of the everyday life in Tamil Nadu, linguistically punctuated by solemn occasions, often religious. In this linguistically complex environment those missionaries also wanted to play a significant role, an ambition which was, to some extent, fullfilled for some of them.

Figure 3

Three languages called ‘Tamil’ (the Tamil triglossia).

(1)	[tamiḻ]. l, [‘or’] [tamuḻ], a lingoa, Ta- // =mul ‘the language, Tamil’.

(2)	[tamiḷaṉ]. H. Tamul. ‘Tamil man’.

As can be seen in the above citations (1) and (2), which contain as additions standardized transcriptions of the Tamil characters and English translations of the Portuguese glosses found in the VTCSP entries 342_L_j & 342_L_k, two variants of the name of the Tamil language are mentioned by Antam de Proença [1625–1666] (henceforth AP), who had compiled the data which would be printed after his death in the form of the 1679 VTCSP book. Those two variants, which are tamiḻ and tamuḻ, are a clear illustration of what is normally referred to as the Tamil Diglossia (B vs. A), expanded by me into a Triglossia, inside Figure 3 (C vs. B vs. A), which I shall also symbolize occasionally by means of .³ The first form mentionned in (1), namely « », transliterated as [tamiḻ], is the standard form, representative of Variety B (formal Tamil), and is also found in Variety C (Classical Tamil). The second form mentioned, namely « », transliterated as [tamuḻ], is one of the possible variants found in some dialects of spontaneously spoken vernacular Tamil (Variety A). Since it is the form which is the closest to the Portuguese form tamul, to which the French form tamoul is comparable, we can probably conclude that this dialectal variant⁴ is the one with which the Portuguese travelers first came into contact, when Vasco da Gama reached India. As for the form seen in (2), namely « », transliterated as [tamiḷaṉ], it looks like a kind of compromise between the A and B varieties, because the use of the i vowel in the second syllable seems to indicate an attempt at using formal Tamil (i.e. the B-variety), whereas the use of the consonant ḷ instead of the standard ḻ seems to indicate that the word was uttered by a speaker of a dialect in which ḷ and ḻ had merged, as is the case for instance in Madurai nowadays, although a school teacher in Madurai would now, if seeing (2) in the notebook of a student, immediately correct it and criticize it as a serious mistake.

As can be seen by a rapid look at Figure 5, the temporal span covered by all the sources mentioned in this article is extremely vast. It starts with the Tolkāppiyam, an ancient grammatical treatise, which belongs to the most ancient period for Classical Tamil literature (the Variety C of Figure 3). It goes up to the modern period, where historians of linguistics are trying to evaluate what took place during this long time span, and where the linguistic situation has also changed. We can indeed see that if there exist nowadays many human beings who have a simultaneous mastery of English and of Tamil (in several of its varieties), the number of human beings who have a simultaneous mastery of Portuguese and Tamil is probably very small. It is even smaller if we restrict ourselves to those who are simultaneously capable of deciphering the Portuguese sentences contained in a 16^th or 17^th century MS or book, along with the Tamil passages, in ambiguous Tamil script or in approximate Roman transcription, which are found in the same book or MS. The same remarks can be made concerning the grammars and dictionaries composed in Latin, such as the three grammars which are represented by the letters Z, B and W inside Figure 5. Those letters stand for the works of B. Ziegenbalg [1682–1719], C. J. Beschi [1680–1747]⁵ and C.T. Walther [1699–1741]. If two of these grammars had not been translated into English, very few people nowadays would have even an approximate idea of their content. It has been my intention, in writing this article, to rediscover myself and to help others to rediscover many forgotten topics, by exploring what is in fact ancient primary linguistic field work. That field work —or initial terrain— is especially important, from the point of view of general linguistics, because it represents a first contact between languages which were not part of the same linguistic area, and were thus in a situation of maximum alterity.⁶

Figure 4

Tamil Triglossia (simplified form of Figure 3).

Figure 5

Brief chronology of the early stages in the description of Tamil HH = Henrique Henriques; Z = Ziegenbalg; B = Beschi; W = Walther; P = Portuguese; L = Latin; E = English; t = Tamil as object-language; {T\t} stands for a description of Tamil (=t) written in Tamil (=T).

In order to illustrate in a compact manner the linguistic diversity, a number of symbols, such as {T\t}, {P\t}, {L\t} & {E\t},⁷ have been placed on a vertical axis found on the right side of Figure 5, and complement the temporal information of the horizontal axis. Inside those four symbols, the capital letters T, P, L & E stand respectively for the Tamil, Portuguese, Latin & English languages, suitably extended in order to be usable as metalanguages for describing Tamil, noted here by the lower case letter t, in order to signify that it is the object-language in a description, which can be either a grammar, as in the case of the Arte attributed to HH, or a dictionary, as in the case of AP’s 1679 VTCSP.

2. A preliminary question: the infinitive

The original inner incentive for the writing of this specific article has been my attempt at finding an answer to a question which I would now, with hindsight, reformulate in the following manner:

That question (3b) had first in its background, my personal opinion that the traditional Tamil terminology, which has been developed by Tamil grammarians over the course of many centuries, is in fact preferable and more natural. It was also reinforced by a second observation, formulated in what follows as a question, namely:

If we were to stay within the {T\t} framework, where the types discussed in (3a) and in (3b) are both called viṉai-y-eccam (henceforth VE), we could refer to them compactly as VE2 (whose model is ceya) and as VE1 (whose model is ceytu), adopting the order in which Tamil grammarians enumerate them, which corresponds in fact to their frequency, because VE1 forms are more frequent than VE2 forms, as revealed by the observation of Tamil textual corpora.

However, since we are in this article exploring the roots of {E\t} type grammars, and since all English-speaking grammarians and linguists refer to the VE2 form as ‘infinitive’, we shall also make frequent use of that expression, although we shall see, for instance in section 6, that the current unanimity has not always existed.

My attempt at answering question (3a), by moving back in time, from the 21^st to the 18^th century, had first taken me, when making the initial conference presentation, to an exploration of the two most visible {L\t} types of grammars —where L stands for Latin— namely the 1716 Grammatica Damulica, composed by B. Ziegenbalg and printed in Halle, and the 1728/1738 Grammatica Latino-Tamulica, written in 1728 by C.J. Beschi and printed in Tranquebar in 1738. To those two should be added a third {L\t} type grammar, written by C. T. Walther and printed in 1739 in Tranquebar, as an (unasked for)¹² supplement to Beschi’s grammar. Placed in the wider chronology, those three sources are represented by the three letters Z, B and W inside Figure 5, as has already been said in the previous section. In the same Figure 5, one can also see, just below, inside the {E\t} line, two more names, with a date, Horst (1806), and Mahon (1848). These are the names of the authors of the two English translations¹³ of Beschi (1738) and it is certainly thanks to the existence of these two translations that Beschi’s grammar has undoubtedly remained directly influential for the longest duration of time. Additionally, it has been reprinted (both in the original and in translation) several times, whereas Ziegenbalg’s grammar has had to wait until 2010 for finally becoming accessible to the English-speaking world, thanks to the efforts of Daniel Jeyaraj. As for Walther’s grammar, which has never been translated, it has probably had very few readers, especially in the modern period. However, because Ziegenbalg’s grammar has been the (veiled) target of Beschi’s criticism, who for instance did not agree with his use of the label infinitivus substantivus,¹⁴ his positions were somehow present in some of the implicit debates. As for Walther, who is indebted, in different ways, to both Ziegenbalg and Beschi, but was also a kind of general linguist, interested in many languages, including Hebrew, his distinct formulations also throw some new light on the debate between his two immediate predecessors, but we must reserve their detailed examination for a future study.¹⁵

However, as I have tried to suggest in a preliminary manner in the introductory section, the nature of the linguistic configuration which the European missionaries encountered when they reached South India was much more complex than is usually reported in some simplified accounts of the discovery of Tamil. For that reason, it has progressively appeared to me that an answer to the limited question asked in (3a), above, could really be informative only if it started from the very beginning of the external description of Tamil, or at least from what is reachable of that very beginning, namely the Arte composed by HH, which falls under the formula {P\t}, where P stands for Portuguese.

The simple truth is that one cannot really understand what Ziegenbalg, Beschi and Walther were trying to do if one does not study the work of their predecessors, who wrote in Portuguese. Ziegenbalg, Beschi and Walther had of course their own sources of information in the Tamil-speaking world but they also had at their disposal a relatively large body of already compiled knowledge, which took the form of grammars and dictionaries, written in Portuguese (and sometimes in French), some of which are still available to us, as MSS or as printed books, although much remains to be done for the information which they contain to be really accessible to a modern reader. When they make statements about Tamil, we should not believe that everything they say is based on independent direct experience. What they tell us is partly based on the sources which they had studied and which they reformulate, trying to improve upon them, but, by so doing, they tend in fact to replace the initial discovery of the A-variety of Tamil, by an exposition targeted towards the B-variety of Tamil. Such an improvement may not necessarily be a progress, from the point of view of descriptive linguistics.

Additionally, another parameter which is also important in the evaluation of the linguistic survey made by Ziegenbalg, Beschi, Walther or AP, is what we could call their virtual ‘Tamil Library environment’. What this expression indicates is that we should try to evaluate their own estimation of the complexity which they faced in trying to have a global view of Tamil literary culture, which was, and still is to some extent, totally dominated by the C-variety of Tamil. This environment is partly accessible to us thanks to the documents which they collected. Among these:

• some are known to us by catalogues which they left although the artefacts catalogued may have disappeared (as is the case with Ziegenbalg),¹⁶

• some are still available thanks to actual collections which have been preserved (such as the BnF)

As a general observation, I shall remark that all those collections, virtual or still in existence, are dominated by the C-variety of Tamil, which is the only variety which was really admired and treasured by the inhabitants of Tamil Nadu, who would never have considered it a useful task to attempt to describe the language (Variety A) which they used in their everyday life. This is the reason why HH composed his Flos Sanctorum in the B-variety of Tamil. This is also the reason why Beschi went much further and composed a long Christian poem, the Tēmpāvaṇi,¹⁷ in the C-variety of Tamil. And, from the point of view of descriptive linguistics in general and Tamil dialectology in particular, this is why the Arte prepared by HH and the VTCSP prepared by AP are such invaluable sources of information, without equivalent, as far as the first component of the triglossia is concerned.

3. A brief overview of HH’s 16^th century Arte

Regarding the Arte attributed to HH, we are lucky that a number of scholars have worked on it. It was rediscovered and identified in Lisbon as a unique MS, the codex 3141 (see Figures 1 and 6), in the BNP (Biblioteca Nacional de Portugal) by X. Thani Nayagam. It has been critically edited by H.J. Vermeer in 1982. An English translation by J. Hein & V.S. Rajam is available since 2013. Nevertheless, it remains a very difficult task to navigate that text and to locate elements of information in it.¹⁸

Figure 6

Extract from top of folio 37 r in Codex 3141 (BNP, Lisbon).

Our starting point will be a brief presentation of the section dedicated to verbs, because it is inside that section that some elements of an answer to the preliminary question (3a) concerning the beginnings of the use of the English technical term ‘infinitive’ for Tamil can be found, as can be verified inside Figure 7, where a line reads: “15 o ẽfinitiuo¹⁹ [emphasis mine] ẽ tħca ou rca”. It is also there that the seeds for answering our enlarged question can be located. That enlarged question addresses the challenge of understanding the implicit organizational practices which govern a non- lemmatized dictionary such as the VTCSP, which is our central target in this article, and which stands between a grammar and a dictionary.

Figure 7

Vermeer 1982 transcription of folio 37 v in Codex 3141 (BNP, Lisbon): Preliminary presentation of the 3^rd conjugation.

At the beginning of the section dedicated to verbs the upper part of fol. 37r in that MS announces that nine distinct verbal ‘conjugations’ (i.e. comjugaçoes) are distinguished (see Figure 6).²⁰ This announcement is followed by a prologue (fol. 37r to fol 39v), where the general characteristics of each of the nine conjugations are briefly explained. See for example Figure 7 which contains a transcription, extracted from Vermeer (1982), of the brief characterization, given in that prologue, of the various stems seen in the two classes distinguished by the Arte in the 3^rd conjugation.

After the prologue, the verbal paradigm as it appears in each of the nine conjugations is detailed over circa 100 folios (from fol. 39v upto fol. 139r). This long section, which corresponds to pp. 33–113 in Vermeer (1982) and to pp. 81–220 inside Hein & Rajam (2013), is then followed by a shorter one, from fol. 120r upto fol. 128r, where the nine conjugations are revisited with respect to the specific morphology of what would now be referred to as neuter forms.²¹ Then comes another subsection, from fol. 129 r upto fol. 139 r, for which the title is Dos verbos passiuos.²² Then comes a coda, from fol. 140 r upto fol. 142 v, which is not part of the nine conjugation system and for which the title is ‘sum, es, fui’. The goal of that coda is to give a separate presentation of a special verb, namely aguiren ‘eu sou’ [‘I am’], which plays mostly the role of a copula.²³ All this is of course a huge amount of information, and I shall draw in what follows only from the initial long subsection. While doing so, I shall take advantage of the fact that Hein & Rajam have introduced some special notations, making use of what they call sub-tenses, which have to be used as additional specifications, along with the tenses (i.e. tempos) mentioned in the Arte attributed to HH. The use of those subtenses, indicated by capital letters, from A to R as maximum value, greatly simplify the navigation of this ocean of data, while at the same time clearly showing why it is so difficult to really master it.

In a chart which is found below (see Table 1), I have provided in compact form implicit pointers towards all the elements which are discussed inside that larger sub-section of the section dedicated to the verbal paradigm in HH’s Arte. The pointers are implicit because they have to be constructed on the basis of two or three components:

1. The row information, which tells us which conjugation (among the nine, with their eventual sub-classes) is concerned

2. The column information, which tells us which tense is intended

3. The sub-tense information, which consists in choosing a capital letter, from the ranges provided at the top of the column

Combining the conjugation and the tense information from the rows and the columns, it can be seen that the chart provided as Table 1 gives access to 66 items, falling under 6 types, in the description of the verbal paradigm. If we add to that the third element which Hein & Rajam call a sub-tense, we have access to 693 additional items, falling under 63 types, which gives a total of 759 items, falling under 69 types. For each of these 759 items, the chart tells us inside which page range of Hein & Rajam (2013) the information (in English) provided by those two scholars can be located. I must add to this that some of those 759 items are paradigms, others are list of forms and still others are individual items. The chart is as follows:

A possible way of making use of this huge amount of data, is to select a given type, and to go through the successive rows, comparing what is said about that type, in the successive rows. This will be done here first for the type which Hein & Rajam (2013) refer to as ‘present tense, subtense J’, making use of that designation for the first time on p. 87, while translating folio 44 r from the Arte. I shall refer to this type in a more compact manner as the type α-J. This type does not receive any Portuguese technical designation in the Arte, where we simply see (info-1) the form named, to which is added (info-2) an explanation of its morphology —How to derive it from the present tense?— followed by (info-3) an explanation of its meaning —How to translate it?— and then by (info-4) some examples of its use.²⁴ Such a detailed treatment is however not available for all the conjugations, and it is mostly the first conjugation which serves as the occasion for providing the most detailed set of information features.²⁵ As a first example, I shall provide here in chart form the occurrences of the type α-J in the Arte. I could also make use of Ziegenbalg’s designation for that form, namely Infinitivus Subſtantivus, but that would be going too fast in the unfolding of the chronology.²⁶ That type α-J has been chosen by me as the initial item to be examined because of the immense importance which it has in the organization of the 1679 VTCSP, where it acts as the head for what I shall call a ‘family of entries’, as we shall see in the following section. However, at this early stage, because the type α-J is, in the organization of the 16th cent. Arte, secondary to the α type (i.e. the present), from which it is derived in the description, we have no choice (except in the first row C1) but to provide its meaning on the basis of the meaning given for the primary entry α. The chart is as follows:

Finally, this chart can play the role of a transition between, on the one hand, the classification system of the Arte attributed to HH, in which each conjugation is presented on the basis of several main stem forms, starting with the 1st person present (see for instance Figure 7 for the case of the 3^rd conjugation) and, on the other hand, the classification system of the VTCSP, which will be our object of investigation in the next section.

4. A brief overview of AP’s 1679 VTCSP

After this long contextualizing introduction, we now turn to what might be the more original part of the present study, containing elements of information which have not been available for a long time. We shall now examine more in detail the 1679 VTCSP, which is the posthumous work of AP, printed in book form in Ambalacatta (Kerala) by his colleagues. The VTCSP was rediscovered in the Vatican Library by several scholars, although one can also argue that the single available Vatican Library copy of the VTCSP was never really lost (see Vinson 1910). It has been published as a Facsimilé by X. Thani Nayagam in 1966, and has been studied several times by G. James (See James 2000, 2007 & 2009). It is however not yet as accessible as HH’s Arte. I have been engaged since 2013 in the preparation of an electronic edition²⁷ of that text, which contains 16,215 entries, as per my current estimate,²⁸ on 508 unnumbered pages. The odd pages in the VTCSP contain folio numbers in the Upper right corner but unfortunately, because of some mistake in the printing the great majority of these numbers are faulty and there are even duplicate numbers. As a clear example, the last folio in the VTCSP, whose recto is (logical) page 507 and whose verso is page 508, bears the folio number 247, when it should had been numbered as 254.²⁹ The consequence of this state of affairs is that the VTCSP folio numbers are unusable for all practical purposes. For that reason, I have decided to make use of the logical page numbers in my references to the pages: entries are referred to by means of a coordinate system indicating:

• the logical page number (from 1 to 508)

• the column: L for ‘left’ or R for ‘right’

• the rank: indicated by a lower case letter (from a to y)³⁰

This coordinate system is the one which is used, for instance, inside the seventh column of Table 2, inside the previous section.

We are now going to compare, from a point of view as global as possible, the coverage of the Tamil verbal system which is performed in HH’s Arte on the one hand and in AP’s VTCSP on the other hand, although these two works do not have the same nature, because, from the point of view of the Tamil triglossia, we can indicate, in a nutshell, that:

• HH’s Arte is the incomplete draft of a preliminary grammar, possibly summarizing the linguistic observations made during half a century, but mostly centered on the A-variety of Tamil, although containing a few clear indications of the fact that its author knew that the B-variety existed and was important.³⁶

• AP’s VTCSP is an incompletely lemmatized Tamil dictionary, with its entries ordered in the Portuguese alphabetical order. It is a mature work, based on 150 years of linguistic observations, which has in part moved away from being based exclusively on the A-variety, as is for instance visible on the rows C4, C5a, C5b & C6 of Table 2, where we see that the A-variety forms etiratu, coliRadu, puduradu and paýRadu, all found in column 2, have been replaced, respectively, by the B-variety forms ēttukiṟatu/ēṟṟukiṟatu,³⁷ koḷḷukiṟatu, pūtukiṟatu and pāykiṟatu. However, the number of A-variety forms remains very important, and there are many cross-references between the two varieties.³⁸ Additionally, as we shall see, there are many entries inside the VTCSP which are exclusively C-variety forms. Several of those entries are explicitely said to belong to the language used by Tamil poets.

5. Taking the measure of the incompletely lemmatized VTCSP

We now return to the challenge of (virtually) lemmatizing the non-lemmatized VTCSP, a preoccupation which was my initial reason for moving away from the question asked in (3a) to a more general group of questions. Before however entering the lemmatizing challenge, I must provide in this section more information concerning the organization of the VTCSP, taking as a starting point, for the sake of continuity in the exposition, the treatment seen in the VTCSP of some of the verbs which were used as models in HH’s Arte. I shall take as first object of my examination the VTCSP item which is referred to as 213_R_a in the second half of row C3a inside Table 2.

The corresponding entry, in image form, is made available here as Figure 8. That image also contains another VTCSP entry, namely 213_R_b.

Figure 8

Entries 213_R_a & 213_R_b in the 1679 VTCSP.

We shall also refer in the discussion which follows, to still another VTCSP entry, namely 214_L_a, visible here in image form, inside Figure 9. Inside my current draft XML electronic edition of the VTCSP, those three entries are represented by three entry elements inside a file, which appear, in a simplified form, as follows.

Figure 9

entry 214_L_q in the 1679 VTCSP.

(4a)	<entry coordinate = “213_R_a”>
	<line_length>3</line_length>
	<entry_head></entry_head>
	<disambiguation>pāṟkiṟatu</disambiguation>
	<CATEGORY>V</CATEGORY>
	<entry_body>Olhar.
	<p></p>.
	<p></p>.
	<expression>.</expression>
	</entry_body>
	</entry>

(4b)	<entry coordinate = “213_R_b”>
	<line_length>1</line_length>
	<entry_head></entry_head>
	<disambiguation>pāṟkukai</disambiguation>
	<Under>213_R_a</Under>
	<CATEGORY>INF1</CATEGORY>
	<entry_body>O olhar.</entry_body>
	</entry>

(4c)	<entry coordinate = “214_L_q”>
	<line_length>1</line_length>
	<entry_head></entry_head>
	<disambiguation>pātta</disambiguation>
	<Under>213_R_a</Under>
	<CATEGORY>PE</CATEGORY>
	<entry_body>C. que uio, ou uiſta.</entry_body>
	</entry>

In this rather verbose collection of tags,³⁹ only some items are a transcription of what is in the printed book, and everything else belongs to the interpretive layer. These three entries, which occupy 5 printed lines in the VTCSP represent approximately 0.0177% of the totality of the text which is encoded in XML, because the total number of printed lines occupied by the 16,215 entries is 28,227, with an average of 1.74 lines per entry.⁴⁰ Regarding the added interpretative layer, we can say the following:

• The coordinate attribute, which appears inside each entry_head element, has been explained in the previous section.

• The disambiguation element is here to provide my interpretation of the ambiguous printed Tamil text.⁴¹

• The Under element, which indicates that one entry is secondary with respect to another entry, named by means of its coordinate,⁴² allows us to make a virtual family out of several entries, if they are secondary with respect to the same primary entry, considered as the head of the family. This is the case here, and therefore {pāṟkiṟatu, pāṟkukai, pātta} are made into a family of entries,⁴³ in which the head is pāṟkiṟatu.

• The CATEGORY element, which in those three entries contains respectively the values ‘V’, ‘INF1’ and ‘PE’, is used here for indicating which element in the paradigm an entry represents. We shall come back to this when explaining in which manner the VTCSP continues HH’s Arte, and in which manner it goes beyond it.

• The p element, which stands for ‘paradigm’, and which appears inside the entry_body element, was intended for giving morphological information. However, it must have been difficult to use because the information given here, about the past and future forms of first person, is truncated. We simply read ‘ttēṉ’ and ‘ppēṉ’ —after disambiguation of the ambiguous sequences « » and « »— and have to guess whether it means that the past tense 1^st person singular is pāttēṉ and that the future tense 1^st person singular is pāppēṉ, which are the A-variety forms, or whether it means that those two forms are pārttēṉ and pārppēṉ, which are the B-variety forms.

• The expression element singles out an idiomatic expression which must have been considered as important but which is difficult to interprete now because it is not translated into Portuguese and no similar expression is found in modern dictionaries of Tamil. Since « » disambiguates as viyātiyaip pāṟkiṟatu, where viyātiyai is the accusative of viyāti ‘doença’ (436_L_j), i.e. ‘disease’, the litteral translation is ‘to see the disease’, which might be a medical expression referring to the performing of a medical diagnosis. Some external confirmation would be necessary.

In the remaining part of this section, I shall make a few remarks, intended as supplements both to what has been said concerning the CATEGORY element and to what has been said concerning the p element. These remarks will try to establish a link between the description of the Tamil verbal conjugation as seen in HH’s Arte, and the verbal conjugation as seen in the VTCSP. We shall revisit some of the apparent contradictions, connected with the tension which must always have been present in the description of diglossia, not to talk about the added layer of complexity which comes from the triglossia.

The first remark concerns my choice of pāṟkiṟatu as the head of what I have called the {pāṟkiṟatu, pāṟkukai, pātta} family of entries. What is the criterion for deciding that one item is the natural choice for such a role? As we have seen in the previous section, and notably inside Table 2, in the system used by HH in his Arte, the head of the ‘family’ in the description of a verbal paradigm was the 1st person present tense form, which I referred to as the primary α entry. In the case of the lexeme which we are currently examining, the head was, in HH’s transcription system, the form patħquiren ‘eu olho’ [‘I look at’], which in a disambiguation XML element would appear as pāṟkiṟēṉ. In that system, pāṟkiṟatu is ‘sub-tense’ α-J, if we make use of H&R’s terminology, slightly adapted. As for the form pātta, it is also a sub-tense, although it is part of another group because it is ‘preterite tense, subtense J’,⁴⁴ or, more compactly, the form β-J, enumerated as part of a succession of sub-tenses, morphologically derived from the tense β, i.e. the ‘preterite’ (HH’s preterito), which is paten ‘eu olhei’ [‘I looked at’], which would be transcribed inside the disambiguation element adopted here as pāttēṉ, or possibly as pārttēṉ, as discussed earlier in this section. To sum up, a system in which verbal forms are enumerated as { patħquiren, …, patħquiratu, …, paten, …, pata, …, patħpen, …} has been replaced by a system where the enumeration is {pāṟkiṟatu, …, pāttēṉ (or pārttēṉ), pātta, …, pārppēṉ (or pāṟpēṉ)}. To this, which might appear to the reader as speculative, can be added the massive evidence which results from statistical evidence made on a data base consisting of 16,215 entries, inside which every entry element whose entry_head is a verbal form ending in -kkiṟatu or -kiṟatu or -ṟatu has been tagged by means of a CATEGORY element containing the string ‘V’⁴⁵ (for ‘verb’) and in which other entries semantically related to them, in a manner to be explained, have a CATEGORY element with a string value which is ‘PE’ (for peyar eccam),⁴⁶ or ‘INF1’, or still many other possibilities. The statistic concerning those three types, plus another one to be discussed shortly, is as follows.

At the end of this already long section, and before examining other aspects in the coming sections, I shall briefly comment on the items contained in the last two lines of that chart. Concerning the label INF1, which I have provided for the form pāṟkukai ‘O olhar’ [‘the looking at’],⁴⁷ I should first observe that this form does not seem to be mentioned inside HH’s Arte. It might be more typical of the B-variety inside the triglossia. Nevertheless, HH himself has made use of it in a book which he composed, Flos Sanctorum , which is a collection of the lives of Saints. It is also found in the C-variety. As can be seen from the statistics in Table 3, items having this CATEGORY label form appear 1,121 times inside the VTCSP, as per my most recent count. This is certainly an argument for saying that the VTCSP is a fully diglossic dictionary.

Table 3

Statistics concerning four of the possible values of the CATEGORY element in the VTCSP.

CATEGORY	Examples	Total number of such items in the VTCSP
V	pāṟkiṟatu (213_R_a), iṭukiṟatu (94_R_j)	3,168 items	A or B
PE	pātta (213_R_b), iṭṭa (94_L_n)	719 items	A, B or C
INF1	pāṟkukai (214_L_q), iṭukai (94_R_l)	1,121 items	A, B or C
INF2	iṭutal (94_R_k)	168 items	C only

As for the last row in Table 3, it contains the statistics for a much less frequent form, which is labeled as ‘INF2’ and is found 168 times inside the VTCSP. That form is characteristic of ancient poetical texts, belonging to the C-variety, i.e. ‘Classical Tamil’. The example chosen is taken from the family {iṭukiṟatu, iṭṭa, iṭukai, iṭutal}, which we have already encountered on the row C7 of Table 2, representative of a time when the head of the family was jrhugiRen. The fact that such a form as iṭutal has its own autonomous entry inside the VTCSP shows that Proença had not completely separated himself from the explorations made by one of his predecessors, Ignacio Bruno, to whom he refers in in his introduction.⁴⁸

6. The one-sided debate between Ziegenbalg and Beschi concerning the two types of infinitivus

In this section, we shall move forward in time to a brief examination of the one-sided debate⁴⁹ between Ziegenbalg and Beschi, which has already been alluded to inside the section 2 of this article, as an anticipated comment on the question numbered (3a) in that section, which refers to the modern unanimity among English-speaking descriptors of Tamil, in their use of the English technical word ‘infinitive’, which was in the beginning simply a translation of an earlier use of the Latin technical word ‘infinitivus’.

Having now examined the prehistory of the {L\t} type grammars of Tamil, thanks to our brief explorations of HH’s Arte and AP’s VTCSP, which both fall under the {P\t} type, we know that the use of infinitivus (with respect to Tamil) was itself preceded by the use of infinitivo, a term which occurs rather frequently in HH’s Arte, because it is considered as one of the main tenses (see the 8^th column inside Table 1), from which 18 sub-tenses are morphologically derived in the scheme presented by HH’s Arte.

If we now consider the VTCSP, the use of the infinitivo label is more difficult to detect, but having now at my disposal the integrality of its text, I have located one occurrence of the abbreviation ‘inf.’ (as a label for the form uṇṇa inside entry 391_R_l, uṇkiṟatu) and two occurrences of the abbreviation ‘infin.’ (as a label for the form tiṉṉa inside entries 361_L_d tiṉkiṟatu and as a label for the form vayya inside entry 405_L_o vaykiṟatu). See for instance the entry 361_L_d, as reproduced in Figure 10, where we can see the use of four of the main tenses mentioned in Table 1, namely ‘Preterito’ (pr°) tiṉṟēṉ, ‘imperatiuo’ (imp.) tiṉ or tiṉṉu, ‘negatiuo’ tiṉṉēṉ and ‘infinitiuo’ (infin.) tiṉṉēṉ. On the basis of those three examples, we can conclude that the usage of the technical term infinitivo is identical between HH’s Arte and the VTCSP, and has not changed during the more than 100 years which have elapsed between the two works.

Figure 10

VTCSP Entry 361_L_d (tiṉkiṟatu ‘comer’).

However, contrary to what we have seen for other elements of the morphology, such as PE, or INF1, the number of entries which could be labeled as VE2, which is my {T\t} label for the standard infinitiuo is very small and seems to be restricted to a number of adverbs (currently 18 items) which are in fact grammaticalized uses of infinitives.⁵⁰ What we find on the other hand is the use of infinitives as components in entries which I have labeled as VE2-AUX2 constructions.

Moving now to the central topic in this section, we can say that the formal type which I have referred to until now, following Hein & Rajam, as the sub-tense α-J, starting at Table 2, did not have a convenient designation in HH’s Arte but rose to such a prominence, for reasons which could be further debated, that we find 3,168 entries in the VTCSP falling under that type, a proportion which is almost 20% of the total number of entries.⁵¹ It seems that Ziegenbalg proposed to innovate⁵² by referring to forms such as VICUVĀCIKKIṞATU by means of the Infinitivus Subſtantivus label, as can be seen in Figure 11, where, at the same time, he proposes to use another technical designation, namely Infinitivus Abſolutus, for the sake of referring to forms such as VICUVĀCIKKA, which is what HH’s Arte refers to by means of the infinitiuo label (see the ζ column inside Table 1).

Figure 11

Ziegenbalg (1716: 72) making the distinction between two types of infinitive.

Beschi’s disagreement with this suggestion can be clearly seen in his 1738 grammar, on page 95, inside parag. 107, which is found inside §. II, De Nomine, which is itself contained in Caput IV, De Syntaxi. Beschi writes the following (accompanied here by the 1848 translation by Mahon):

(5a)	Ex quolibet verbo fiunt prætereà plura verbalia, quæ ad modum nominum declinantur, et … (Beschi 1738: 95, parag. 107)
	From every verb are formed moreover many verbals, which are declined in the manner of nouns, and … (Mahon 1848: 79)
(5b)	… 1) ex participiis præſentis et præteriti fiunt verbalia , et &c. … (Beschi 1738: 95, parag. 107)
	… 1. From the participles of the present and præterite come the verbals ceykiṟatu, paṭikkiṟatu, and ceytatu, paṭittatu, &c. … (Mahon 1848: 79)
(5c)	… Hinc apparet, quàm improprie hæ voces v.g. &c. dicantur voces infinitivi: quod certè falſum eſt, … (Beschi 1738: 95, parag. 107)
	… Hence it appears, how improperly these words, e.g. ceykiṟatu, collukiṟatu, &c. are called words of the Infinitive: which certainly is not true, … (Mahon 1848: 79)
(5d)	… Cùm autem has dicunt eſſe, ut ſunt, voces participii præsentis, cum addito pronominis , ex quo fiunt nomina appellativa hominum: ſic et eſt vox ejusdem participii cum addito pronominis , ex quo fit nomen verbale, quod per ſuos caſus declinatur et non correſpondet infinitivo Latinorum … (Beschi 1738: 95, parag. 107)
	… But since they say, that these are, as they are, words of the participle present, with the addition of the pronoun avaṉ, avaḷ, from which are formed appellative nouns of men: so also, ceykiṟatu is a word of the same participle, with the addition of the pronoun atu, from which comes the noun verbal, which is declined throughout its cases, and does not correspond with the Latin Infinitive, … (Mahon 1848: 79)
(5e)	… niſi quando hoc ſumitur tanquam nomen, ut in hâc propoſitione, ſtudere bonum eſt, quam vertam, . At in hoc ſenſu infinitivo Latinorum correſpondent cætera quoque verbalia, de quibus infra, &c. unde illa propoſitio poteſt reddi, , &c. Numquid proptereà hæc omnia voces infinitivi dicenda ſunt? (Beschi 1738: 95, parag. 107)
	… except when this is taken as a noun, as in this proposition, to study is good; which I may translate, paṭikkiṟatu nallatu. But in this sense other verbals also, of which hereafter, correspond with the Infinitive of the Latins; paṭittal, paṭippu, &c., whence that proposition may be rendered, paṭittal nallatu, paṭippu nallatu, paṭippatu nallatu, &c. But pray, are all these therefore to be called words of the Infinitive? (Mahon 1848: 79)

Although it is difficult to comment on this, because there is a large degree of arbitrariness in terminological choices, it seems that Beschi was successful, in front of the audience which he addressed. We shall not however directly comment now on what he said. We shall rather, in the following section, revisit some of our former steps, both in connection with the 16^th cent. Arte and the 1679 VTCSP, in order to clarify Beschi’s references to participles.

7. Understanding relative constructions in Tamil, as seen through the VTCSP and through HH’s Arte

In this section, we shall first revisit the item (4c) which was provided in section 5 but on which we have not yet commented, except when relating its entry_head, namely pātta, with the example given for sub-tense β-J of conjugation 3a in the Arte and except when indicating that the CATEGORY label chosen for it, namely PE, was an abbrevation of a traditional Tamil technical expression which is peyar eccam, known in English as ‘Relative Participle’. I shall now add to this that inside traditional {T\t} grammars, the PE label is the symmetrical of the VE label evoked in (3a) and (3b), because if the VE is an incomplete form (Eccam) which needs a Viṉai (verb) in order to become syntactically complete (muṟṟu), similarly the PE is an incomplete form (Eccam), but it needs a Peyar (noun) in order to form a complete syntagm. To which I shall add that if we choose a noun N in order to make the PE pātta ‘complete’, that N will come after pātta, resulting in the Noun-Phrase ‘pātta N’, which we still have to translate, but we shall make use of the VTCSP entry 214_L_a, alias (4c), for that purpose, reproduced below in a simplified, tag-less form, (6a), which is as follows:

The two features on which I must comment now are the presence of the place-holder ‘C.’ (for Portuguese cousa, i.e. ‘thing, entity’) and the presence of a double translation. I shall react to the double translation by duplicating the entry as (6a1) and (6a2). Additionally, I shall make the place-holder role of ‘C.’ explicit by inserting a phantom (N) in the Tamil, Portuguese and English sequences which translate each other. The result is:

What this linguistic situation may have suggested to the compiler of the VTCSP is the fact that a form like pātta was the close equivalent of both the Latin Past Passive Participle visum and of the Latin Present Active Participle videns (if we overlook the tense difference) because in (6a2) the head noun N stands for the object of the action whereas in (6a1) the head-noun N stands for the agent of the action. The ambiguity caused by this double equivalence must have appeared to the successive compilers of dictionaries as very striking. This must be the reason why, as we saw in Table 3, the VTCSP contains 719 entries tagged as PE, to which we can add 67 entries tagged as PE_pr (Present relative participle) and 77 entries tagged as PE_neg (Negative relative participle). I could of course have CATEGORY-tagged the PE as PE_past (Past relative participle), but since it is so much more frequent than the other two, that would have been a misleading symmetry.

How, in a given situation, we must translate the noun-phrase pātta N depends of course on a number of parameters, such as the nature of N (is it agentive? is it inanimate? etc.) and such as the context. I must however hasten to add that giving the choice between these two translations is, as a matter of fact, insufficient: there are other possibilities, in addition to these two, as was in fact clear already to HH, when he composed the Arte. We can see that in the following example, extracted from folio 43r (see Figure 12), where HH explains the behaviour of the form which I have tagged as PE_pr and to which H&R refer to as subtense α-D, by means of two examples from which I shall now reproduce the second:

Figure 12

Extract from folio 43r in Codex 3141 (BNP, Lisbon) [containing text of example (6b), preceded & followed by other words].

This example can be written in modernized transcription as follows:

Inside this example, the present relative participle (alias PE_pr) form pōṟa is followed by a head, which is the noun-phrase inta ceyti (‘this fact’, or ‘this action’, or ‘this information’). As for the relationship between the relative participle and its head, the easiest way to explain it seems to use the language of traditional {T\t} grammars, such as the Tolkāppiyam (see Figure 5). Those grammars explain that the noun which is the head-noun for a PE (peyar-eccam) ‘relative participle’ can have as a referent:

• the agent of the action expressed by the PE, as is the case in (6a1)

• the object of the action expressed by the PE, as is the case in (6a2)

• the time of the action, as would be the case in […] pātta nēram (‘the time when […] saw’)

• the location of the action, as would be the case in […] pātta ūr (‘the village where […] saw’)

• the instrument for the action, as would be the case in […] pātta kaṇṇāṭi (‘the pair of glasses with which […] saw’)

• the action itself, as would be the case in […] pātta ceyti (‘the fact that […] saw’) and as is the case in example (6b2)

More details can of course be obtained by studying the commentary to these grammars,⁵³ which were written for describing the C-variety of Tamil, but which are nevertheless applicable in many respects (but not all) to the B-variety and the A-varieties of Tamil, even now.

These points being established, we are now in a better situation for understanding some of the elements evoked by Beschi, which were reproduced in Section 6, from (5a) to (5e). When he refers in (5d) to ‘words of the participle present, with the addition of the pronoun avaṉ, avaḷ, …’, he is in fact proposing an analysis of words such as ceykiṟavaṉ, ceykiṟavaḷ, … as a combination of the PE_pr form ceykiṟa with Human pronouns such as avaṉ ‘he’, avaḷ ‘she’, …, considered as the head-nouns for ceykiṟa. Since he does not translate, it is not completely clear which role he has in mind for the head pronoun. Are those heads meant to refer to the agent of the action, in which case we must translate them by ‘He who does’, ‘She who does’, … ? Are those meant to refer to the object of the action? This is much more rare, if the head noun refers to a human being but not strictly impossible. When however Beschi continues his analogy and declares in (5d) that ‘ceykiṟatu is a word of the same participle, with the addition of the pronoun atu,’, we are now faced with a greater number of possibility. The atu ‘that-NEUTER’ which is the head pronoun incorporated inside the fusion form ceykiṟatu (= ‘ceykiṟa + atu’) can have any of the roles enumerated above.⁵⁴ It can refer to the agent and mean ‘that which does’. It can refer to the object and mean that ‘that which is done’ … It can refer to the action itself and mean ‘that which is the fact of doing’.⁵⁵ This is the reason why Portuguese lexicographers thought it was a good equivalent for the Portuguese infinitive and chose it as the citation form in their dictionaries, a tradition which continued for several centuries.⁵⁶

8. How should one write pāṟkiṟatu? (between phonetics & orthography)

In this section, we shall return to the topic of the Tamil triglossia, through an observation concerning changing orthographies. The occasion for this remark, is the observation that there is a difference of spelling between Ziegenbalg and Beschi, concerning the word pāṟkiṟatu, which has already been one of our targets of observation, inside section 5, where it can simultaneously be seen in Figure 8, which contains an image extracted from VTCSP page 213, and in (4a), which is a transcription of the content of that image. Additionally, the word also appeared on row C3a of Table 2, where it was spelled patħquiradu, a spelling which tries to be faithful to what is observed in the Lisbon BNP MS of HH’s Arte, here visible on this page as Figure 13, where it is accompanied by another form, patħquiraducu, which would be written in normalized transcription as pāṟkiṟatukku, and is almost identical, except for the final additional suffix.

Figure 13

Patħquiradu on folio 83r of Codex 3141 (BNP Lisbon).

Coming now to Ziegenbalg, we have an attestation of the same word on page 101 of his grammar, where we can see, if we examine Figure 14, a rule, formulated by Ziegenbalg, which states.

Figure 14

Ziegenbalg 1716: 101, Top.

(7a)	Regula ſeptima.
	Omnia verb, quæ ante habent accipiunt in Præterito ut:
	Videre
	Vidi (Ziegenbalg 1706: 101)

Interestingly, a 2010 book, Tamil Language for Europeans: Ziegenbalg’s Grammatica Damulica (1716), by Daniel Jeyaraj which presents itself as a translation of Ziegenbalg’s grammar, contains the following translation, which can be usefully compared with my translation.

(7c)	Seventh Rule [Page 101]
	All verbs which have -[-tal] before [r] take - [-tēṉ] in the past, like this
	[pārttal]171 to see, look at
	[pārttēṉ] I saw
	[Footnote 171: ‘’[‘pāṟkiṟatu’] ] (Jeyaraj 2010: 133)
	[[NB: the transliterated items between square brackets have been added by me (jlc)]]

The reader of this article may try to imagine the nature of a collective linguistic pressure for correctness which is so strong that it induces the 21^st century translator of an 18^th century important historical text to modify that text, while translating it, in such a manner that the translation becomes misleading for anyone who does not have access to the original. We can indeed verify that inside (7c):

• the rule has been changed, but we are not informed of the change

• the original target of the rule has been replaced by another target, and put in a footnote.

• the result of the rule has been changed, but we are not informed of the substitution.

I can only conclude that the translator must have thought that the prestigious original was written in such a sub-standard form of language, that one should not run the risk of having the prestigious original setting a possibly contagious bad example. Without making additional comments on (7c), I shall now turn to the text of Beschi’s 1738 Grammatica Latino-Tamulica, as it was printed in Tranquebar. We find inside this text, on page 50, in paragraph 55, a rule which is reproduced here in Figure 15. That Rule explains how to form the preterite of the verbs whose stem ends in y, r, i, u or ai. Inside the image, on the 7^th line, we can see one of the targets of the rule, which is [to be normalized as pārkkiṟatu]. The result of the application of the First preterite rule appears on the 7^th and 8^th lines and is [to be normalized as pārttēṉ]. As we can see, the target is identical with the value which Daniel Jeyaraj has substituted to Ziegenbalg target. which was [to be normalized as pāṟtēṉ]. I should add that there are several levels in ungrammaticality from the point of view of Traditional Tamil grammarians. For them:

Figure 15

Beschi 1738: 50, (Cap. III, §. II, parag. 55, Præteritum, Regula 1).

• both pārkkiṟatu ‘to look at’ and pārttēṉ ‘I looked at’ are acceptable phonetically and are the correct forms in the B-variety

• pāṟkiṟatu is a phonetically acceptable combination of sounds, which is distinct phonetically from pārkkiṟatu, but which is NOT the correct form, inside the B-variety, if the intention is to convey the meaning ‘to look at’, and can only appear as a mispronunciation.

• pāṟtēṉ is a succession of written symbols, which is the transcription of a forbidden combination of sounds, because (alveolar) ṟ is NEVER followed by (dental) t, according to the rules of the C-variety, as has been stated in the Tolkāppiyam (see Figure 5), and because the phonetic rules of the C-variety are supposed to govern the B-variety. The educated guess of an educated Tamilian is that whoever wrote pāṟtēṉ must have made a mistake and wanted to write pārttēṉ.

Of course, a descriptive linguist may have a different point of view and remark that since the A-variety component of the Tamil triglossia is in fact a collection of dialects and that each of them is likely to have its own phonological system, it should be possible to make sense of at least some of the aberrant combinations of letters found in ancient written documents but that is of course a tall order, which will not be attempted here, although that should be at least one of the possible outcomes of my forthcoming edition of the 1679 VTCSP.

A final question must now be dealt with in this section concerning pāṟkiṟatu: is it possible that this usage is limited to Christian documents and is due to the (bad) influence of Christian Missionaries? In order to answer this question, I have been lucky enough to receive the help of a doctoral student of Hamburg University, Neela Bhaskar, who is currently preparing a Ph.D. on a śaiva collections of stories known as the Tiruviḷaiyāṭaṟpurāṇam, which narrates the Sixty-four ‘Sacred Sports’ (tiruviḷaiyāṭal) of Śiva in Madurai. She has easily located for me several occurrences of the spelling pāṟkiṟatu inside a group of śaiva MSS on which she works.⁵⁷ I shall now display one of those examples in Figure 16. Although the MS is not very easy to read because it is incised but not inked, one can clearly see that the first of the two lines contains the written form [pāṟkiṟatu] preceded by the sequence [ttaip] and followed by the glyph [cu].⁵⁸

Figure 16

Pāṟkiṟatu in extract from Folio 55 r in MS Indien 291 (BnF, Paris).

9. Going beyond families: the case of the causatives

In this section and in the following one, I shall return to what could be called the global numerical perspective, where we attempt to measure the progress made by Portuguese-speakers between the 16^th and the 17^th century in the twin tasks of learning and describing Tamil. That progress is seen in the comparison between HH’s Arte and Proença’s VTCSP. A preliminary evaluation of the content of the VTCSP, as a whole, has already been provided here in Table 3, inside Section 5. That table contains basic statistics on the number of occurences of four CATEGORY labels, starting with the omnipresent V label, which is used 3,168 times in my tagging of the VTCSP in order to characterize those forms which BZ referred to as Infinitivus Subſtantivus (see section 6), and continuing with three other labels, PE (719 occ.), INF1 (1,121 occ.) and INF2 (168 occ.), which were also commented upon in the previous sections. When we compare this with the fact that Vermeer’s 1982 critical edition of HH’s Arte includes as ultimate section, on pp. 155–166, an ‘Index of the Tamil words occurring in the grammar’ which contains 309 entries, among which ca. 120 are identifiable as verbs, we can see indeed how much more information had been made available.

The huge progress, obtained within one century of descriptive efforts, when going from a list of 120 verbs in HH’s Arte to a list of 3,168 V-tagged VTCSP entries, which we can also call ‘preliminary verb candidates’, can be evaluated more precisely by providing more information on the grouping of items in ‘verb families’ which was first introduced inside section 5, and which was performed by means of the Under element, seen for instance in (4b) and (4c). More specifically, my current XML file of the VTCSP, contains 3475 occurrences of the Under element, pointing towards 1,421 distinct family head. This means that a family created in this manner contains one head accompanied by an average number of 2.45 items. To this must be added the fact that there are also currently 1478 entries which are tagged as V, but which are not related to other verbal forms, either as target or as source of an Under element. Each of those 1478 entries is the only member in a singleton family. This is summarized here in the form of a chart, which contains my current counts, inside the VTCSP edition work in Progress.

Although this table may appear complex, the position of the 6,374 entries which it describes is dermined simply by answering three questions, the first one being purely morphological, and the other two requiring a greater command of grammar.

• does the entry end in -kkiṟatu or -kiṟatu or in -ṟatu? (if yes, it is labelled V)

• does the entry belong to the paradigm of another entry, considered as main citation form?⁵⁹

• is the entry the target of an Under element link contained in another entry?

As soon as we introduce other parameters however, things can become more complex. My purpose in this section is to illustrate that it would be premature on the basis of Table 4 to conclude that the VTCSP contains 2,899 distinct verbs. I shall for that purpose provide basic information concerning the elements of a large subgroup which is contained inside the group of 2,899 heads of families VTCSP entries. This subgroup, which possesses 554 elements, contains all the VTCSP entries which I have labelled CAUSAT (for causative).⁶⁰ It will be briefly illustrated here, in a convenient manner, by the entry 213_L_k, alias pāṟpikkiṟatu ‘Mandar ver’ [‘to tell to see’] (See Figure 17). That entry is related, in a natural manner, with the entry 213_R_a (pāṟkiṟatu ‘Olhar’ [‘to look at’]), which has been the focus of our observations in some of the preceding sections, starting with section 5.

Figure 17

Entries 213_L_k to 213_L_m in the 1679 VTCSP.

Because a satisfactory treatment of the group of 554 causatives in the VTCSP would require a full-length article, I shall simply enumerate, in the rest of this section, a few basic facts, which have to do with the morphology of causatives and with the strategies which we can observe, inside the Portuguese glosses which explain the meaning of those entries. From the morphological point of view, it can be said that all the elements tagged by me as CAUSAT inside a CATEGORY element, end either in -vikkiṟatu or in -pikkiṟatu. The group is already identified, morphologically and semantically, inside HH’s Arte and we can conveniently refer to those forms as being the subtense γ-G (see Table 1), following the conventions introduced in Hein & Rajam (2013: 99). As far as the meaning is concerned, I shall simply reproduce, as a preliminary explanation, a small extract from from the two paragraphs devoted to the topic in HH’s arte, where they take up the second half of folio 54v and the whole of folio 55r.

When however we compare what looks like a simple and straightforward explanation, occupying two paragraphs in a grammar, with the more complex linguistic reality captured inside a set of 554 entries in the 1679 VTCSP, we find that several strategies have been used by the Portuguese translators for approximating the relationship between the base form and its causative, although there is very clearly one which is used in a majority of cases. More precisely, on the total of 554 pairs of VTCSP entries consisting of a base verb and its associated causative.

• in 494 cases, the causative is translated by a periphrastic form containing fazer combined with a Portuguese infinitive which has been used as one of the possible translations for the base verb. For instance 486_R_q (cirippikkiṟatu) is translated by fazer rir [‘to cause to laugh’] whereas 486_R_s (cirikkiṟatu) is rendered by rir [‘to laugh’]. This corresponds to the strategy described in (8a–b) for ariquiren ‘espamco’ and aripiquiren ‘faço espamquar’, although a more faithful transliteration of the corresponding entries in the VTCSP is aṭippikkiṟatu (36_R_l ‘fazer eſpancar’ [‘to cause to beat’] and aṭikkiṟatu (36_R_p ‘espancar, […]’ [‘to beat’])

• in 22 cases, we have a pair where the causative is translated by a periphrastic Portuguese form containing fazer while the base verb is a reflexive/pronominal form. For example, kūṉivikkiṟatu (321_L_g) is explained by ‘Fazer corcouar’ [‘to cause to bend’], whereas kūṉukiṟatu (321_L_j) is explained by ‘Corcouarse’ [‘to bend’].

• in 8 cases, instead of having a periphrastic form with fazer, we have a periphrastic form with mandar ‘to order, to tell’. This is for instance the case with the example 213_L_k, alias pāṟpikkiṟatu ‘Mandar ver’ [‘to tell to see’] already provided in Figure 17.

• in a few cases, unrelated verbs are used in the Portuguese translations for the causative and for the base verb. This is the case for instance for paṭippikkiṟatu (204_R_c, ‘enſinar’ [‘to teach’] and paṭikkiṟatu (204_R_i, ‘aprender, eſtudar’ [‘to learn, to study’]. The translator may have thought that it would be too clumsy to use a periphrase in the translation.

• occasionally, additional grammatical technical terms (Active, Neutro, …) are seen in some entries.

However, as a last observation before going to the next section, I shall remark that, if we leave the 17^th century and move to an examination of the Madras Tamil Lexicon (MTL), we seem to find in it very few entries for causatives. A rapid search seems to locate only 19 causatives having a stem ending in -pi (such as kaṟpittal ‘to teach’)⁶¹ and 24 causatives having a stem ending in -vi (such as terivittal ‘to cause to know’).⁶² Given the fact that the VTCSP has 16,215 entries whereas the MTL has more than 100,000 entries, the densities of causatives are respectively 3.4% and 0.043%. Before moving to the next section, I shall make two additional remarks in connection with this observation:

• the first remark is that we should probably not continue to count the causatives in the count of 2,899 verb candidates, which was given in Table 4. If we substract from that total the 554 causatives, reclassified as another type of secondary entries,⁶³ we would now have a reduced total of 2,345 potential heads of families, whereas the total count of secondary entries would now increase to 4,029.

• the second remark is that it is possible that the high visibility of causative forms in the VTCSP is due to the fact that part of the corpus analysed by Portuguese missionaries was heavily sanskritized prose. Such a hypothesis can only be verified by a careful examination of early prose. I shall simply remark that HH declares, at the end of the second paragraph devoted to causatives E vsaõ muito deste modo de cõpoer os que bẽ sabẽ falar (Vermeer, 1982: 47).⁶⁴ Such a remark may mean that the missionaries were aware of the fact that causative forms were not used very spontaneously in ordinary language. This is a topic for further investigation.

We shall now move to a brief examination of another feature, which is probably more central in a typological characterization of the signature features of Tamil (and of other related languages).

10. How the VTCSP dealt with paired verbs

In this last section before the conclusion, I shall examine very briefly the elements belonging to two twin subsets of the set of primary V-tagged entries in the VTCSP, referred to hereafter as AV (i.e. set of ‘affective verbs’) and as EV (i.e. set of ‘effective verbs’). These sets both contain 167 entries and a one-to-one natural correspondence exists between them. When using the word natural, I am hereby postulating⁶⁵ that the correspondence is intuitively recognized by native speakers of Tamil but that it took time for the Portuguese-speaking missionaries to master it. In the modern period, the elements of these two sets have been described as ‘paired verbs’, each pair containing an affective verb and an effective verb. This topic has been studied for the first time in depth in Paramasivam (1979) and is also the topic of study chosen by Cristina Muru, in this issue of the JPL (See Muru 2021). In this section however, I shall limit myself to giving a brief overview of the visibility of those 167 pairs in the VTCSP, and shall also briefly give an account of HH’s strategy for coming to terms with their existence.

One of the difficulties in the initial evaluation of the situation is that all the pairs do not follow the same morphological pattern. In the largest subgroup which is observable inside the set of 167 pairs, both verbs in the pair have the same stem but they differ by the ending, whereas in other subgroups, the two verbs in the pair have different stems, which combine with the same endings.⁶⁶

• the first situation is illustrated by the elements of the pair consisting of kuṟai-kkiṟatu [315_R_o] and kuṟai-kiṟatu [315_R_j], for which the Portuguese translations given in the VTCSP are ‘diminuir’ (effective verb) and ‘diminuirse’ (affective verb)

• the second situation is illustrated by the elements of the pair consisting of kalakk-iṟatu [255_L_g] and kalaṅk-iṟatu [254_R_n], translated respectively as ‘miſturar’ (effective verb) and ‘miſturarse’ (affective verb)

When HH first tried to describe the phenomenon in his Arte, in a section called ‘Dos verbos passiuos’ (Vermeer 1982: 107), he started his presentation by the following declaration:

I shall not however elaborate much further on the topic because this same issue of the JPL already contains an article by C. Muru dedicated to paired verbs. I shall simply observe that the current conclusion of several centuries of observations, as far as paired verbs are concerned, is that, unlike what is done in the treatment of causatives, it is currently considered preferable to treat the two elements of a pair as two distinct lexemes, rather than to consider the affective verb as a kind of passive form with respect to its corresponding paired effective verb. Finally, from a statistical point of view given the fact that my temporary conclusion (at the end of section 9) is that the number of (potential) heads of (verbal) families does not exceed 2,345, and given the fact that we have 167 pairs, associating one effective verb and one affective verb, we can conclude that the proportion of paired verbs recognized by the VTCSP is at least 14%. This figure is certainly lower that the estimation found in Paramasivam (1979: 5, fn.8). However, since the figure of 2,345 is an upper bound, which will decrease as the classification of entries is refined, the real percentage of paired verbs is certainly higher. As a final hypothesis, we can also consider the possibility that paired verbs are under-represented in the VTCSP because of the factor which I have already evoked at the end of section 9, when trying to explain why causatives are over-represented in the VTCSP, namely the fact that it it partly based on the analysis of heavily sanskritized prose.⁶⁷ Additional research is necessary.

11. Conclusion & future challenges: from Figure 5 to Figure 18

Figure 18

Knowledge is always mediated by at least one language, or by more … [There is no language-independant knowledge about languages] (A Battle Plan for future explorations).

I believe I have now reached the limit of what can be exposed on such a complex topic in a single article. In this concluding session, I would like to try to sketch the future challenges, which are implicit in the shuttling between several points of view which has taken up a great part of the substance of this article in which I am trying, using English as a scientific Lingua Franca, to explain to readers, who could have any language X as a mother tongue, why it is relevant to examine the content of several texts —and especially one, the VTCSP, in which I have invested a lot of efforts— which were created at a time when some human beings, who were using Portuguese as a Lingua Franca, were trying to learn, to teach and to use another language. This represented a huge effort, which should not be underestimated because, even after 200 years of successive attempts, they were producing books in which native Tamil teachers —being the exponents of the B-variety and the admirers of the C-variety of Tamil— would have found many faults, as should be clear from what is illustrated inside section 8, and notably by the group (7a)–(7c).

When using the word ‘relevant’ in the previous paragraph, I was having in mind several types of possible readers of this article, who are indirectly evoked by the elements of the small graph (or network) which is sketched inside Figure 18, above. I see primarily three possible main scenarios for the readers of this article, with an additional fourth one, depending on the language X which is a primary language for them, in addition to their compulsory command of E:

• If X refers to a language actively used by the reader of this article, and possibly a mother-tongue …

• Scenario 1 (X = T). For those readers, who are already in control of the B-variety of Tamil and of at least one (A-variety) dialect of Tamil, the primary challenge consists in discovering other (A-variety) dialects, while learning enough archaic Portuguese for that purpose. Another possible challenge would be to make 17^th century Portuguese into the primary target language.

• Scenario 2 (X = P). For those readers, who are already in command of Portuguese, the primary target will be Tamil and the main difficulty for them will be to understand what exactly the Tamil triglossia is. They can also decide to invert the object of study and verify whether for instance the implicit comparison between the structures of Tamil and of Portuguese which is partly visible in many of the Portuguese translations given for Tamil examples, can decenter their perspective on their own language. They might also try to go into a deeper exploration of the phonetic side.

• Scenario 3 (X is neither T nor P). For those readers, everything is difficult, because there are two languages to study at the same time, namely T and P, and the object of study is a forgotten path between those two languages, as shown on Figure 18. However, I believe that even for these readers (who are like me), although the challenge is, in the absolute, greater than what it is for those belonging to the first two groups, the task is as beautiful.

• Scenario 3b (X = L). Finally, inside the group described in Scenario 3, there is a subgroup of researchers who are fluent users of Latin. Those users also have a unique perspective, as can be seen on Figure 18, but I shall not elaborate further.

I must of course not choose between all those potential readers and must now return to the task of preparing several entry points (and human interfaces) for the future consultation of the forthcoming database —i.e. the electronic edition of VTCSP— keeping in mind the various possible users which it may have, and remembering that, if one has only one chance of making a good first impression, the clarity of the interface navigation can hopefully go a long way in this new age of digital humanities, where content accessibility may have become the main challenge for the transmission of knowledge, if the impressive efforts made by AP’s colleagues in 1679 are to be retrieved from the quasi-oblivion where they are now.

Competing Interests

The author has no competing interests to declare.