Remarks on Spanish conmigo, contigo, consigo

Contents

• 1. Introduction
• 2. A synchronic account
• 2.1. Analysis I
• 2.2. Analysis II
• 2.3. Analysis III
• 3. A diachronic account
• 3.1. A look at the more recent past
• 3.2. A look at a rather distant past

1. Introduction

As is well known, the grammar connected with the Spanish preposition con ‘with’ differs from that of the remaining prepositions. What is usual for prepositions is that they are construed with various kinds of NPs (noun phrases), and if a personal pronoun is intended, this is put into its object form; for Spanish yo ‘I’, tú ‘you (sg.)’ and in the case of reflexivity this means that mí, ti, sí respectively are used, e.g. a mí ‘to me’, para ti ‘for you’. The preposition con is different in this respect, however, since in these three instances grammar requires the use of certain special forms; these are:

As can be seen, the 1st and the 2nd persons of the singular as well as the reflexive pronoun are affected.

The origin of this peculiarity is known, see Lathrop (1980: 117), Penny (1991: 120f.), Ranson / Lubbers Quesada (2018: 254f.), Lara (2019: 51) and various other sources. In Latin, cum ‘with’ was mainly a preposition, as any other preposition, but with personal pronouns of the 1st and the 2nd persons (of both numbers) as well as with the reflexive pronoun it had been used as a postposition and had then become a suffix, which yielded:

In varieties of Early Romance, to these forms the preposition was added. Seen from a historical point of view, from then on ‘with’ existed twice in these forms – a fact that has repeatedly been noticed by researchers (e.g. by Lathrop 1980: 117, Penny 1991: 121). Stolz / Stroh / Urdze (2006: 501, n. 327) put it like this: “Comitativity is marked twice in the word-forms of Modern Spanish, etymologically speaking.”

Usual events of language history, such as sound change and morphological change, occurred as well. These led to the intervocalic “weakening” of -c- > -g- and established the vowel i (rather than regular e) in the singular forms due to analogy with mí, ti, sí. Old Spanish still had five forms: comigo, contigo, consigo, con(n)usco, convusco. Modern Spanish has only retained the first three (with the first one being spelled conmigo now). Also some other Romance languages and dialects show such forms: Portuguese still exhibits all five, and so does Galician; Sardinian (Nuorese variety of the Central Sardinian dialect) has chin mecus ‘with me’, chin tecus ‘with you (sg.)’ (Mensching 1992: 62).

At first glance it looks as if hardly anything new could be said about these facts. It turns out, however, that there do exist possibilities to say something new and interesting about the three Spanish special forms – this applies especially when grammatical, typological and diachronic considerations are brought into play. Section 2 will have a look at the synchronic structure of Spanish and will seek to do justice to it, while section 3 will make new points about the diachronic background of the phenomenon.

A brief note on terminology is perhaps in order. This paper will use the terms preposition, postposition and adposition, as is general practice in linguistics today. However, one should be aware of a few facts discussed by Holst (2001: 201f.). Prepositions and postpositions are one part of speech, not two. This can clearly be seen by comparison with adjectives, for instance, which are always called adjectives irrespective of whether they usually precede the noun (or NP), as in English, or follow it, as (mostly) in Spanish. The term preposition was coined with typical European languages in mind where these words precede the nouns (or NPs) they refer to. It was only modern linguistics that created the uniting term adposition, which is meanwhile standard in typological and other works, e.g. Croft (1990), Velupillai (2012), Herce (2023). (It is true that the term adposition is usually hardly required in Romance studies since in the Romance languages adpositions are usually prepositions. A postposition can be seen, for instance, in Italian due anni fa ‘two years ago’, as opposed to Spanish hace dos años.)

2. A Synchronic Account

Traditional grammars often express the facts in such a way that one does not combine con with mí, but uses conmigo instead (and correspondingly with the other persons). Some authors state that con and mí “become” conmigo together, though actually in a synchronic system nothing “happens”, hence nothing “becomes” anything; rather, there are nothing but facts and structures existing next to each other, side by side, and relationships between these facts and structures. It is just that explanations using the concept of “becoming” help the human mind. (For how grammars from the 16th and 17th centuries deal with conmigo etc. see Satorre Grau 2002: 376–379.)

Statements of the kind mentioned are unsatisfactory since they are not as explicit as one may wish, as well as for other reasons. One would be interested in an explicit description of what Spanish – or more specifically Spanish grammar – is like. What precisely is its morphology, and what precisely is its syntax? There are several options how to account for conmigo, contigo, consigo. (It needs to be added that there is nothing remarkable about the fact that there are several options. While in diachronic linguistics usually simply one solution is the one that matches language history, even though we linguists may be unable to identify it, in synchronic linguistics often several accounts are possible.)

In the following, three options how to analyze the facts of Spanish will be presented, and their pros and cons will be discussed.

2.1. Analysis I

The first option would be to insert conmigo into the paradigm of yo in some way, as well as the other two forms into the paradigms of tú ‘you’ and él ‘he’ etc. respectively. By this statement the following is meant. Typical modern western European languages, both of the Romance branch and of the Germanic branch, have reduced the Indo-European case system considerably. However, they still have case distinctions with personal pronouns. When in English, for instance, I and me (and he and him, etc.) are distinguished, this is best described as a difference in case. Spanish, too, is of course not a language without case; it is just that case is rather restricted in this language’s structure: to pronouns – while other nominals do not have case: nouns, adjectives, numerals; even many types of pronouns lack it.

Hualde / Olarrea / Escobar (2001: 144) distinguish four cases. The table they provide shall be reproduced here for the singular forms:

The authors are correct in positing a case named “prepositional” by them (“preposicional” in their Spanish text), the fact that this case has no direct antecedent in Latin of course being irrelevant. However, what strikes the reader is that for the 1st and 2nd persons the fields of the prepositional are filled with two forms each. (Schpak-Dolt 2012: 45 proceeds in the same way with having one cell each.) Usually such grammatical tables should be designed in such a way that a field contains only one form (unless the goal is to register free variants, dialect differences, or the like). This raises the idea that perhaps the grammatical structure of the language has not yet been captured in an adequate way.

A possible solution is to claim that we require another case. This additional case could be called comitative as in Uralic studies and also in various other philological traditions. The job of a comitative case is to express what in other languages would be expressed with the adposition ‘with’ and an NP. In Hungarian, for instance, hajó ‘ship’ forms the comitative hajóval ‘with a ship’ (cf. Rounds 2009: 106); also Estonian possesses a comitative, while Finnish does not. Another option would be to call this case sociative, which is the term used in Basque studies, see e.g. Santazilia (2013: 269–271). Spanish conmigo corresponds in function to Basque nirekin ‘with me’ (ni ‘I’), and Spanish contigo to Basque zurekin ‘with you’ (zu ‘you’). (Note also the areal implication, since Spanish and Basque are neighbours.) Since comitative is the more frequent term (in relation to sociative), it will be used in the following. The comitative is clearly distinct from the prepositional; it is a fifth case.

For Hungarian, Basque and other languages researchers had no problem with positing a case (and along with this, inventing a case name) since these languages exhibit large case systems anyway. With Spanish, by contrast, scholars seem to have been guided less by the Spanish data themselves and more by the Latin model that they had in mind. A fact that should not hold researchers back from speaking of another case is that the formation of the comitative implies the use of a circumfix, con- -go: circumfixes have been reported from various languages (Harris 2002, Harris / Xu 2009: 287f., Marušič 2023). It would also be possible to speak of simultaneous prefixation and suffixation. With the circumfix mentioned, the comitative can be formed from the prepositional – not from the accusative or any other case; in this way it is ensured that the vowel i is arrived at correctly.

While the morphology has now been accounted for, what remains to be clarified is the syntax. If we formally define conmigo etc. as comitatives, there is a consequence that must not escape our attention. In order to make a statement containing the meaning ‘with’, for almost any NP the preposition con is used: con el hombre ‘with the man’, con él ‘with him’, etc. But for the 1st and 2nd persons singular as well as for the reflexive, no preposition con is used – the fact that the comitative form begins with con- must not distract from this. The comitative corresponds in function not to an NP, but to a PP (prepositional phrase). The syntax thus becomes slightly more complex here, and this is a disadvantage of this solution – which is otherwise rather straightforward in many respects.

2.2. Analysis II

The second option how to handle the forms conmigo etc. can now be explained by referring to the first option. It would be possible to say that actually we are dealing with two words each: con + migo, con + tigo, con + sigo. Of course, the Spanish orthography demands spelling these forms in one word each – but this is just a matter of the spelling. As is well known in linguistics, a spelling convention of course need not do justice to the structure of a language, and it must not prevent a linguist from positing any analysis that might suit the facts in a smooth way. (The relationship between profesor and profesores is the same as that between cuestión and cuestiones, even though the rules of placing the accent mark demand no accent mark in the first pair, and an interplay between the presence and the absence of an accent mark in the second pair.)

As to the elements migo, tigo, sigo, it would then be these that would be called the comitative forms (not the entire forms, as in analysis I), and it would be these that would be listed in morphological tables. These forms would not be formed with circumfixes, but only with suffixes – which is simpler. The forms have initial consonants that are known from other forms of the paradigms in question: m-, t-, s-.

This solution has the great advantage that it entails simpler rules in syntax. The complications arising with the first analysis, laid out above in 2.1., would be lacking; utterances containing the idea of ‘with’ would always contain the preposition con. While all other prepositions then govern the “prepositional” (to cite the name chosen by Hualde / Olarrea / Escobar 2001, see 2.1.), con governs the comitative. The fact that prepositions do not all agree as to which case they govern is of course well known from many Indo-European languages, e.g. German, Russian, Lithuanian, Albanian, etc.; German, for instance, has mit mir ‘with me’ with a dative form, but ohne mich ‘without me’ with an accusative form. Only a subset of (modern) Indo-European languages always exhibits the same case, after any preposition, and such languages are best known from the Germanic and Romance branches, e.g. English, Swedish, and Italian. English has me, for instance, for the 1st pers. sg. after any preposition: to me, at me, for me, with me. Spanish is usually perceived as a language in this typological category as well – but, considering what has been discussed, this can be questioned. When operating as proposed with two cases “prepositional” and “comitative” for Spanish, this would move this language from one like English into the group of German, etc. It should be added, though, that only a single preposition would govern the “comitative”, all others the “prepositional”, and in this way the complexity of German is not reached. Incidentally, even within Romance having more than one case after prepositions is not entirely unheard of: in Sardinian, prepositions usually go along with the forms mene ‘me’, tene ‘you (sg.)’ (Mensching 1992: 62), but a ‘to’ deviates by requiring mime, tibe (historically datives) (Mensching 1992: 62). Moreover, there is also a comitative mecus, tecus, see section 1.

Another analysis will be presented in the next subsection. All in all, however, possibly this second solution is the one with fewest disadvantages. (Herce 2023: 437 mentions this analysis as well, apparently taking it for granted.)

2.3. Analysis III

The third option that suggests itself should not go unmentioned. As underlined right from the start when discussing the previous option, the spelling is something that should not prevent the linguist from proposing an analysis. It turns out that another solution would be to segment into three words con + mí + go, and correspondingly for the 2nd pers. sg. and the reflexive. Again, this basic idea has consequences that require exploring.

This analysis has one surprising advantage. The elements in the middle of the phrases that this analysis would yield are simply mí, ti, sí, hence forms already known from the paradigms. (We ignore here the issue where accents marks are found in Spanish orthography, and where not.) These are the case forms called “prepositional” by some authors, cf. 2.1. Positing an additional case “comitative” is superfluous now, and it is even impossible. This constitutes an advantage over the two previous analyses.

In syntax, however, the situation is not as simple – which shows that whatever you do, there are always unsatisfactory effects somewhere, without there being an escape. The structure now contains not only one adposition, as in analysis II, but two: one comes before the pronoun and one after it. One is a preposition, con, and the other is a postposition, go. (The g in this postposition would always be not the plosive / stop, but the corresponding fricative or approximant, the “gamma”, but this can be attributed to sandhi.) Historically con and go are identical; synchronically we can choose whether we want to interpret them as the “same” element or as two different ones.

This raises the question whether it would be unnatural to have such an element twice. What can be clarified relatively quickly is that in general it is very well possible for grammatical elements to occur twice. This can be seen from an Icelandic example, for instance (data and analysis from Holst 2023: 106). The North Germanic languages mark definiteness in the noun, so that Danish hestens ‘of the horse’ consists of the stem hest-, the definiteness suffix -en- and the genitive suffix -s. In related Icelandic, the corresponding form is hestsins ‘of the horse’, consisting of the stem hest-, the genitive suffix -s-, the definiteness suffix -in- and again the genitive suffix -s; hence the genitive suffix occurs twice. Icelandic represents an older state; the form originated from a noun in the genitive followed by a definite article also in the genitive which were then univerbated. Since each of the two former constituents was marked for genitive, the suffix for this case -s eventually appeared twice in the resulting form. Danish removed the first genitive marker. The example shows a suffix that appears twice, but it is also possible to have the same morpheme both as a prefix and as a suffix: in Berber, for instance, frequently the simultaneous occurrence of prefixed t- and suffixed -t marks feminines.

However, these examples are morphological, and seeking to back up the analysis con + mí + go typologically, the question of a double occurrence should rather refer to syntax. To be more specific, it narrows down to: is it possible that an adposition appears twice in a PP? This is indeed possible, as shown by Hungarian. In this language demonstrative pronouns added to an NP require the postposition to be doubled. An example, cited from Rounds (2009: 151), is ez alatt a szék alatt ‘under this chair’, literally “this under the chair under”. Step by step, the structure of this PP can be understood as follows. The NP a szék is ‘the chair’, consisting of the definite article a ‘the’ and the noun szék ‘chair’. Demonstrative pronouns precede their heads and they require the definite article to be present (unlike in English and in fact most European languages with articles, where the demonstrative pronoun would replace the definite article), hence with ez ‘this’ one obtains ez a szék ‘this chair’. Moreover, Hungarian is a language with postpositions, so that alatt ‘under’ must follow the NP: a szék alatt is ‘under the chair’. If now an NP with both a demonstrative pronoun and a postposition is formed, hence semantically speaking ez a szék and a szék alatt are fused, it will not do to simply merge them but after the demonstrative pronoun the postposition must occur as well: ez alatt a szék alatt ‘under this chair’. It would be ungrammatical to omit the first alatt.

As can be seen, it is possible that an adposition occurs twice in a PP. However, the syntax of the analysis of the Spanish remains slightly awkward in so far as both a preposition and a postposition are used under this analysis (one might speak of a “circumposition”). More importantly, however, a disadvantage appears again which already showed up in a similar way with analysis I. The 1st and 2nd persons singular as well as the reflexive do not behave in the same way as all other NPs: whereas the former require a preposition and a postposition, the latter require only a preposition.

3. A Diachronic Account

In the previous section three synchronic analyses were presented. It turned out that achieving an advantage in one issue goes along with a disadvantage in another. Therefore there is perhaps no optimal solution how to analyze Spanish conmigo etc. However, no decision needs to be made: it is a general fact of synchronic linguistics that often several analyses are possible. An important issue is that awareness has been raised about the facts of the Spanish language regarding the forms in question. As has been seen, it is possible to go beyond previous treatments of Spanish and be more explicit about how this language is structured.

But we will now turn to diachrony. This area is different in nature from synchrony since in diachrony there usually is one historical truth and the linguist’s task is to point this out or reconstruct it, or at least make reasonable proposals on how structures may have developed. In the following, first a look will be taken at the more recent past in 3.1., which means the connection to Latin. Subsequently a rather distant past will be explored in 3.2.

What will not be done is to address the possible future, this being hardly predictable. It is interesting to note, though, that occasionally forms such as sinmigo for ‘without me’ have appeared, for regular sin mí. Bruyne (2002) cites an attestation from almost half a millennium ago and calls it humoristic. Ranson / Lubbers Quesada (2018: 255) report that “one sometimes hears” sinmigo and sintigo and cite as an example a song Sinmigo by Mexican artist José Madero Vizcaíno. It is obvious that the non-standard forms sinmigo and sintigo arise by analogy with conmigo etc., a fact that highlights the semantic connection between the concepts ‘with’ and ‘without’. In Finnish grammar the case indicating absence (‘without’) is called abessive; raha ‘money’ forms the abessive rahatta ‘without money’. Theoretically, if one day sinmigo etc. should become standard usage, in analysis I a case abessive would have to be added, while in analysis II a form such as migo could not simply be called comitative but a compound comitative-abessive would be the appropriate designation (and further modifications of the three analyses would be required).

3.1. A Look at the More Recent Past

As laid out in section 1, the three modern Spanish forms under study go back to a slightly larger system in Latin, comprising the five forms mēcum, tēcum, sēcum, nōbīscum, vōbīscum. These forms are unusual within the system of Latin grammar since cum ‘with’ is otherwise a preposition. It must have been used as a postposition, however, after personal pronouns; this is the only explanation that can account for the five special forms – via univerbation.

Upon a closer look one discovers a link to the Silverstein hierarchy, or animacy hierarchy, as it is called by Croft (1990: 111–117); the former term has a slight disadvantage in that it was not only Silverstein who did research on it (with his seminal paper Silverstein 1976), while the latter term is probably still more misleading since it suggests a simple distinction between animate and inanimate – which is not what is intended. The Silverstein hierarchy is a hierarchy for nominals (or whole NPs) which helps describe and understand various phenomena in the languages of the world. From top to bottom, this hierarchy approximately consists of (it can vary from language to language): first and second person pronouns, third person pronouns, humans, other animates, inanimates, masses. If a language has two different behaviours of nominals in grammar, there usually is a point somewhere in this hierarchy, and all nominals above behave in one way, whereas all nominals below behave in a different way. See recently Holst (2019: 74f.) with further references. In language history this point can move downwards or upwards.

Apparently the positioning of the adposition Latin cum depended on the Silverstein hierarchy: the very upper end (personal pronouns) required postposing, while the remainder required preposing. On the way to Modern Spanish the point on the hierarchy shifted upwards, limiting the use of special forms to singular forms and the reflexive form, which is indifferent with regard to number; this presupposes that the singular is higher on the hierarchy than the plural. Recently a shift upwards on the Silverstein hierarchy has been shown to have occurred in the history of neighbouring Basque concerning the suffixes -e and -en for the genitive (Holst 2019: 193–195). Also the morphological and syntactic rules that the three synchronic analyses in section 2 imply can be formulated by referring to the Silverstein hierarchy.

3.2. A Look at a Rather Distant Past

Turning back to Latin, the question can be raised how the conspicuous syntactic system with prepositional and postpositional use of cum ‘with’ came into being, and whether the Latin facts perhaps tell something about a distant past. Note that learning about the past of Latin implies also learning something, indirectly, about the past of Spanish. In the following it will be argued that indeed something rather interesting can be gathered from the Latin facts concerning cum.

It will first be noted that Latin cum has cognates in other branches of Indo-European, e.g. Germanic and Slavic (Pokorny 1959: 612f.); these are simply prepositions, or the corresponding prefixes. As to the Italic branch of Indo-European (that Latin belongs to), we learn about Umbrian kum that it is attested both as a preposition and as a postposition (Untermann 2000: 409). According to Weiss (2009: 460, n. 45) Umbrian kum is postposed mostly with the meaning ‘at’, and Latin has some other, minor, instances where adpositions otherwise known as prepositions are postposed. However, while all of these facts are worth noting, they do not yet shed light on why Latin cum exhibited the syntactic behaviour it did.

In order to arrive at a really interesting insight, it is necessary to turn to Tocharian and Ossetic (from the Indo-European language family). The reason for doing this will be explained subsequently.

1. The two extinct Tocharian languages, Tocharian A and B, were once spoken in what is now the north-west of China. They clearly are languages with postpositions (Winter 1998: 156). However, two adpositions stick out in that they are prepositions: the semantically opposed A śla B śale, śle ‘with’ and A sne B snai ‘without’ (Krause 1971: 20). Tocharianists are aware of why this is so. The Indo-European languages originally had prepositions, as is still the case in Latin, Greek, Old Indian and many other key languages. Some branches or languages, however, switched to postpositions as the norm. To these belongs Tocharian, and ‘with’ and ‘without’ are remnants; the fact that they are used as prepositions is a syntactic relic. (There is also evidence that ‘in’ once was a preposition since it occurs as the first element in the Tocharian word for ‘today’, Krause 1971: 20.)

2. A similar situation is found in Ossetic, an Eastern Iranian language spoken in the Caucasus. This language, too, has postpositions as the norm. According to Abaev (1964), there are only two prepositions: æd ‘with’ and ænæ ‘without’ (data from the Iron dialect). Thus also in Ossetic these two prepositions withstood the syntactic change and remained in their position. Subsequent research made Thordarson (2009: 174) point out that there are some other prepositions, but nevertheless the role of ‘with’ and ‘without’ is striking and parallels that in Tocharian, as pointed out by Holst (2013: 26, n. 13).

To sum up, there are at least three languages (Tocharian A, Tocharian B, Ossetic) that indicate that there may be a special role of ‘with’ and ‘without’ among adpositions. In the cases just discussed, seen historically ‘with’ and ‘without’ indicate the former location of the adpositions; they are valuable relics permitting a view at earlier syntax.

It is now possible to try to apply this insight to Latin – and more broadly to Italic, since Umbrian shares properties of Latin. Here it becomes apparent why having discussed Tocharian and Ossetic is useful: if in these languages ‘with’ and ‘without’ can tell something about the former syntax with regard to the typological opposition of prepositions and postpositions, perhaps the Latin facts concerning ‘with’ can do so also. This holds even if in Latin ‘with’ is not paralleled by ‘without’ (which is the more marked preposition of the two and hence of lesser importance; in Latin this is sine), and even if postposed cum was restricted to the upper end of the Silverstein hierarchy. It makes sense to interpret the position of ‘with’ as a relic again. In Tocharian and Ossetic, basic constituent order requires postpositions, the prepositions ‘with’ and ‘without’ are the exception and point to former prepositions. In Latin (and Italic) the situation is converse: prepositions are the norm, the role of ‘with’ as a postposition is the exception sticking out, and consequently one should infer a former state with postpositions. (There is an entirely legitimate method of arguing in syntactic reconstruction involved here.)

This may come as a surprise at first. The Indo-European languages typically use prepositions, as laid out, and this must have applied to Proto-Indo-European as well. This does not exclude the possibility, however, that relics of postpositions were part of the linguistic system, and these relics may have been handed down to daughter languages, even if in attenuated form. The Latin and Italic evidence concerning ‘with’ points to precisely that. Germanic and Slavic would have innovated in making ‘with’ exclusively a preposition (Czech s ‘with’, for instance, with s in this satem language corresponding to k in Umbrian kum, etc.). If the hypothesis laid out is correct, it would mean that Latin -cum in mēcum etc. is an archaism of great salience.

It is now time to draw attention to why this discussion is so important. A challenge which has been discussed for a very long time now (from the 19th century onwards) is the question whether the Indo-European languages have further relatives, or to put it in a different way, whether they are part of a larger “macrofamily” located in Eurasia. Evidence from syntax should not be forgotten in such discussions. It turns out that families such as Uralic, Turkic, Mongolic and others in Northern Asia fit the earliest Indo-European rather well with regard to constituent order. To be more specific:

There had been a problem so far, however, with adpositions. Language families such as Uralic, Turkic, Mongolic and others in Northern Asia have postpositions, while Indo-European had prepositions. If now, based on the considerations in this paper, evidence has appeared from within Indo-European that postpositions occurred, this moves Indo-European towards the other language families in this respect.

It is interesting to become aware of what all this implies. It means, in fact, that Spanish still retains traces of an archaism which is many thousands of years old. In the Indo-European language family one would rather expect such an archaism in Lithuanian, famous for its often conservative traits, and one would not be surprised to find it in Modern Greek, given that this language has preserved many structures from Ancient Greek. However, one would not particularly expect it in Spanish, as a Romance language, and as a language rather remote from Proto-Indo-European, let alone from stages before Proto-Indo-European. This shows how surprising historical linguistics sometimes can be.