The English word 'two' is one of the most ancient and stable words in human language, traceable with remarkable certainty to the Proto-Indo-European root *dwóh₁. This numeral has survived largely intact across six millennia and dozens of descendant languages, making it a cornerstone of comparative linguistics — indeed, the regularity of the 'two' cognate set across Indo-European was among the first evidence used to establish the existence of the proto-language itself.
The Old English forms of 'two' were grammatically complex in a way that has been entirely lost. Old English had three gendered forms: 'twēgen' (masculine), 'twā' (feminine and neuter), and a genitive 'twēgra' and dative 'twǣm.' This gender distinction in numerals was inherited from Proto-Germanic and ultimately from PIE, which also inflected its low numbers for gender and case. By Middle English, these forms had collapsed into the single word 'two' (or dialectal 'twa,' still preserved in Scots).
The Proto-Germanic form *twō descended regularly from PIE *dwóh₁ through a series of well-understood sound changes. The initial PIE *dw- cluster became *tw- in Proto-Germanic, a change paralleled in other words (PIE *dworh₂ 'door' became Proto-Germanic *durō). This *dw- to *tw- shift is one of the defining features that separates Germanic from other IE branches.
The PIE root produced an extraordinarily wide cognate set. Latin 'duo' (the source of English 'dual,' 'duel,' 'duet,' and 'doubt' — the last from Latin 'dubitāre,' literally 'to be of two minds'), Greek 'dúo' (source of 'dyad'), Sanskrit 'dvā,' Old Church Slavonic 'dŭva,' Lithuanian 'du,' Old Irish 'dá,' and Armenian 'erku' (with regular sound changes) all descend from the same PIE numeral. The word is so stable that even Tocharian, the most geographically distant known IE language (spoken in western China), preserved it as 'wu' (Tocharian A) and 'wi' (Tocharian B).
English has inherited or borrowed a remarkable number of words from the 'two' root. Native Germanic derivatives include 'twin' (from Old English 'twinn,' originally meaning 'double' or 'twofold'), 'twice' (from Old English 'twiges,' with an adverbial genitive), 'twelve' (from Old English 'twelf,' literally 'two left over' after counting to ten), 'twenty' (from Old English 'twēntig,' literally 'two tens'), 'between' (from Old English 'betwēonum,' 'by the two of'), 'twilight' (the 'twi-' prefix meaning 'two' or 'half,' referring to the half-light), 'twine' (something twisted from two strands), and 'twist' (related to the concept of two things wound together).
Through Latin, the 'two' root entered English in learned borrowings: 'dual,' 'duality,' 'duel' (originally a fight between two), 'duet,' 'duplex,' 'duplicate,' 'double' (from Latin 'duplus,' twofold), 'doubt' (from 'dubitāre,' to waver between two options), 'dubious,' and 'diploma' (from Greek 'díplōma,' a folded-in-two document). Through Greek came 'dyad,' 'dilemma' (a choice between two propositions), and the prefix 'di-' as in 'dichotomy' (a cutting in two).
The modern pronunciation /tuː/ conceals the former presence of a /w/ after the /t/. In Old English, 'twā' was pronounced with a clear /tw/ cluster, as 'twin' and 'twelve' still are today. The /w/ was lost before the back rounded vowel /uː/ during the Middle English period, but the spelling 'two' preserved the ghost of the lost consonant. This same loss of /w/ occurred in 'sword' (Old English 'sweord') and 'answer' (Old English 'andswaru'). Dialects vary: some Scottish and Northern English dialects preserve 'twa,' closer to the Old English feminine/neuter form.
The special status of 'two' in human cognition is reflected in grammar. Many languages, including Old English, Arabic, Sanskrit, and classical Greek, had a 'dual number' — a grammatical category specifically for pairs, distinct from both singular and plural. The loss of the dual in most modern European languages (Slovenian and Sorbian are notable exceptions) represents a simplification that obscures how central the concept of 'twoness' once was to the structure of language itself.