1. Origin of numbers in Proto-Indo-European.
Nowadays practically no one doubts that the number was one of the most ancient nominal categories and its appearance in the language took place long before genders and cases appeared. The need to count things was quite essential for ancient people already at the earliest stages of the language.
In every language forms connected with counting include several grammar types: the declension of nouns and other nominal parts of speech in number; conjugation of verbs which also have the category of number; numerals themselves with their peculiarities, and different adverbial derivatives from numerals.
The first part of speech which acquired changes in the number was the noun. Linguists tend to believe that the basic change that took place already on the ergativic stage of the Proto-Indo-European language, was the change of the ending -s in singular to -es in plural. Practically the ending denoted the same in both numbers - the nominative case of animate nouns (people, gods, sacred animals). But the difference laid in the grade of vowel Ablaut in stem - the shorter grade in -s, the longer in -es (here -e- must have been a long sound).
An interesting problem is concerned the relations between genders and numbers of nouns in Proto-Indo-European. It is known, that the earliest stage used only two genders - animate and inanimate, the second for things which do not speak, like trees and mountains. The animate gender (the ending was -s for it) evidently had the number and was subject to declension. The very term of plurality could be used only by animate nouns, while the inanimate gender (it used the endings -m, -d, -t, or no ending at all) had no cases, no declension and even no plurality.
So how can things be counted then? Most of scientists agree that the inanimate gender did not have the category of plurality, but that of collectivity. They did not act in the sentence, could be only objects action, so the plural form of them denoted not the several nouns acting together (as in two women) but the aggregate of objects (like in skies, forests). That is why the ending could not be the same in plural of inanimate gender nouns as in that of animate nouns. The inanimate inflection, which later became the ending of neuter nouns, turned to be -a, or -aH, as the supporters of the "laryngeal theory" state.
Note that this exactly ending can be observed in all Indo-European groups
of languages. While all masculine and feminine nouns usually have some
kind of -s or -es in plural nominative, neuter
nouns have -a everywhere:
Greek teknon (a baby, neut.), tekna (plural)
Slavic *nebo (sky, neut.), *nebesa (plural)
Latin caelum (sky, neut.), *caela (plural)
Hittite gienu (a knee, neut.), gina (plural)
The list can be made longer, but the fact is evident. By the way this -a ending is always repeated in plural accusative as well, which proves that neuter used to have no declension and the same ending for all cases in plural.
Ancient Indo-European languages preserved several examples which show that sometimes even plural neuter nouns required the verb in singular - another proof that inanimate nouns ignored numbers. Such examples are known from Greek, Avestan, Hittite, Vedic - all the most archaic languages of the family. So even though neuter acquired the declension, its plural was still weak in ancient tongues.
At first there were two numbers in Proto-Indo-European nouns. They were, naturally, singular and plural. At that time the language needed to make a distinction between one and several things. But later, as the time went, another new category appeared, called the dual number.
This number is interesting nowadays particularly because not many language have preserved it. Dual nouns and adjectives are sometimes used in archaic dialects of Lithuanian, some relics are known in Russian, Ukrainian and some other Slavic languages. In fact dual was always quite unstable and already in ancient times had a trend to coincide with the singular or with plural. In today's world only some 40 languages use the dual, of them more than a half have very limited use of it (including Philippine languages, some languages of Southern Asia, Baltic languages). Several tongues can have even special number forms for three and four people or things.
The origin of the dual number in Indo-European is simple - some things exist only in a pair: like ears, eyes, knees, legs and sides - they all can be only two. Many sacral objects could also exist only in a couple - for example, Asvines, divine twins in the Vedic religion. The wide use of the dual in religious cults resulted to an opinion that its origin is also sacral. But more probable is that pair things like hands and eyes required a new number and did not use plural.
The dual number never had a full paradigm of declension, and usually
nouns in dual had only two or three forms, compare:
Common Slavic: ro.k (two hands, nominative-accusative dual), ro.ku (genitive-locative), ro.kama (dative-instrumental)
Greek adelf (two twin brothers, nominative-accusative-vocative), adelfoin (genitive-dative)
Old Irish di thuaith (two nations, nominative-accusative-vocative), dib tuathaib (dative), da thuath (genitive).
The dual number appeared only in late Proto-Indo-European: Anatolian languages which separated from the Proto-language earlier than other groups do not show any signs of the dual, so this category was rather new. But already in ancient Indo-European languages, in about 1500 BC, the dual begins to disappear practically everywhere. Latin and Italic no longer use it; Celtic languages since the Old Irish epoch made it analytic; Old English preserves it only for several personal pronouns (wit - we two, etc.). This shows that the dual number was rather unnatural in the Indo-European language structure, so it is wrong to state that three numbers is the exactly Indo-European model. The opposition singular - plural is much stronger and survived even in such analytic languages as English: the inflection for plural nouns is still preserved here (tables, tables').
This two-number system of the language, marked by endings of nouns,
adjectives, verbs and pronouns, is common and stable in all Indo-European
languages and can be called an eternal Indo-European feature.
2. Indo-European numerals.
First we would like to present a table which points the first ten Proto-Indo-European
cardinal numerals in their reconstructed forms:
|*duwo / *dwo||two|
|*sweks / *seks||six|
These forms have varieties, but in principle all linguists agree that cardinal numerals must have sounded quite like this. Now let us look deeper at each of them and describe the system of the Indo-European numerals.
The numeral "one" as you can see had in fact two different forms. But it looks as if it was an adjective in Early Proto-Indo-European, not exactly a numeral. This adjective practically was not used for counting - logically, what to count with one thing? That is why the meaning of this word in Proto-Indo-European, as well as in many ancient IE languages, was not "one". The word *sems meant "joint", "united" and was preserved in Latin semel (once) and in some other languages; the same stem is English same. The stem *oi- evidently meant "single", "the only", but could rarely exist just as it was, it usually added a suffix: thus, *oi-k-os existed in Indo-Iranian (Sanskrit eka, Kurdish yak), *oi-n-os was developed in most European languages (like Greek en, Latin unus), and some languages had the form derived from *oi-w-os (for instance, Avestan aeva). This was the matter of a dialect which existed within the Proto-Indo-European language community, but the stem was the same.
There is a theory that the count started with "two" in Proto-Indo-European, and "one" was not considered a cardinal numeral. This is proven by the fact that the word for "one" is always declined as a simple adjective, though having only the singular forms. The origin of the stem *duwo which mutated into *dwo sometimes is unknown. But it is known that it was declined in dual number only (which is natural), and its feminine and neuter form was *duwoi / *dwoi. It is easy to trace the original sounding of this numeral, for even many nowadays languages did not go far from it: Russian dva, English two, etc.
The next numeral is *treyes, feminine *trisres, neuter *tri. The last became the most widespread within the Indo-European family, though the original form with -s exists in Lithuanian trys, Romance languages (tres, trois) and existed in Classical Greek. For some reason I do not know this particularly numeral became the most stable among all the ten ones, and the form sounding somehow like tri exists practically in every Indo-European tongue.
This cannot be said about *kwetwores, which was too complicated in comparison with *tri. But the structure of it can be recognized everywhere according to what the Indo-European *kw turns into in each language (Latin quattuor, but Oscan petir). Gamkrelidze and Ivanov in their book "Indo-European and Indo-Europeans" state that the count in the Proto-IE language was quaternary (in fours), not decimal (in tens), so the cycle ended in 4. The next one represented the next cycle:
It was *penkwe which probably meant "5 fingers" on the hand. Compare the Slavic word pyad' meaning the hand, and pyat' meaning five. The word *penkwe and all numerals after it were not declined.
The origin of *sweks (the dialectal form *seks) is again unknown, and the next one - *sept@m - is believed to have been borrowed from Semitic on a very early stage of the PIE language. It was present in all dialects, so at the moment it was borrowed the Proto-language was nor divided into dialects yet. Compare the Proto-Semitic *-b-tu with the PIE form - here is another proof for the Asiatic homeland of Indo-Europeans.
The word *okt is strange enough to be discussed for ages already by linguists. Scientists noticed that in fact this word is dual in number, so meaning two things. A version says it denoted two furrows made by the plough, another believes that is meant "two fours" so denoting the end of the second quaternary cycle. Maybe it was also borrowed from somewhere, but it was too far away to be sure.
The second version is witnessed by the numeral *new@n, probably a cognate to the stem *newo- (see the analysis) meaning new - the new cycle after "eight" is a possible meaning. Or it is just a coincidence.
And finally the word *dek@mt, with the possible meaning "two hands", if it is composed of *dwe + *k@mt (the latter meaning "a hand", cognate to Gothic handus).
At first the majority of numerals were just nouns and behaved like nouns
in the sentence. But later as many of them lost their declension, the usage
also changed and a special part of speech appeared in the language. But
still in all Indo-European tongues they remain nominal words, and the first
four of them are declined, not with the full paradigm however.
Now a little bit about the cardinal numerals in Proto-Indo-European and their successors in late IE languages. We should say that in fact ordinal numerals are fully adjectives, and both their use and their declension are adjectival. In most cases these words were formed from the cardinal numerals with certain suffixes added. The most widespread of such suffixes were -o- and -to-. The second is more habitual in European languages (Old Irish coiced, Greek pemptos, Latin quintus, Common Slavic *pe.ty), though sometimes other ones are used.
The exception was made for the words "first" and "second". Note that
they do not derive from "one" and "two" respectively in any language: Lithuanian
vienas and pirmas, Russian odin and
pervy, Latin unus and primus,
etc. The reason for this strange but characteristic Indo-European feature
is that they were not numerals, just like the word "one", in Proto-Indo-Europeans.
Adjectives which were used as "1st" and "2nd" in fact meant "forward" and
"other". Remember that the Old English language shows the same interpretation:
the word fyrrest means both "first" and "farthest", and ther
means "two" as well as "other, another".
The problem with the interpretation and reconstruction of the Proto-stage of the language is that we nowadays do not fully realize how ancient people thought, the sociolinguistic aspect of the language and the mentality itself. It is very simple for us to say "one", "two", "three", but it may have meant quite another thing for those who spoke Proto-Indo-European. Many features of the language were connected with the religion or mythology, with the understanding of the world which existed at the time. The very basis of life there could be totally different from ours, and not much was preserved even in those languages we can read - Ancient Greek, Hittite or Vedic. In fact, through the centuries the language has been simplified much - we understand words just like a morphological means of expressing our thoughts, we care only about the meaning of the word, that's all. But the ancient person took words as symbols, and all the language was full of symbolic signs. We have forgotten much of the etymology of this or that word, and numerals mentioned here sound jus like simple numerals for us today. But ancient Indo-European words and forms used for counting could in fact mean quite different things.
One of the tasks which comparative linguists face is the task to understand the mentality of people who used to speak the very Proto-language. And maybe this problem is the most important for the right reconstruction and comparative analysis.