The Proto-Indo-European language, or PIE in short, is the ancestor of most European languages, apart from Basque, an isolate, and Finno-Baltic and Hungarian, which are Uralic. English derives from Proto-Germanic, itself from PIE.
According to the traditional reconstruction PIE had three types of consonants: (1) voiceless ones like k, p, t, (2) voiced ones like g, (b), d, and (3) voiced aspirate ones like gh, bh, dh. The voiced labial *b was extremely rare. It has long been noted that this system of stops is impossible and does not fulfil the requirements of typology and theoretical phonology. For that matter phonologists have proposed to reinterpret the traditional features in a new framework called the Glottalic Theory: (1) voiceless stops remain the same: k, p, t, (2) voiced stops are relabeled glottalized: like g < kˀ, (b < pˀ), d < tˀ, (3) voiced aspirate stops are just voiced: g, b, d. Aspiration is irrelevant in the Glottalic framework.
Most Indo-Europeanists do not adhere to the Glottalic Theory, but in the past days we have examined several loanwords that were adjusted in Indo-European languages with aspirates instead of “voiced” phonemes when they should not if so-called “voiced” phonemes were really voiced. (1) *ib ‘elephant’ > Latin ebur, Sanskrit ibha, (2) Berber *bid ‘monkey’ > Greek pithekos, (3) Luwian *barz ‘iron’ > Latin ferrum, Englis brass, (4) Sumerian gude ‘lute’ > Greek kithara, (5) Hurro-Urartean dabiri ‘blacksmith’ > Latin faber.
These numerous examples show that the so-called “voiced” phonemes cannot have been voiced originally and that the so-called aspirates were indeed voiced.