Six Centuries of βάρβαρος: A Diachronic Embedding Analysis

The four previous notes in this series used per-author fastText models to compare how βάρβαρος is distributed across the vocabularies of Herodotus, Euripides, Plato, and Xenophon. The CADE (Compass-Aligned Distributional Embeddings) framework extends that analysis longitudinally: a compass model trained on the full ancient Greek corpus serves as an alignment anchor, and separate models are trained on century-sliced sub-corpora — 6BC through 1BC in this dataset. The century slices allow a question the per-author models cannot answer: whether the differences between Herodotus and Plato are specific to those authors or representative of their periods.

The corpus sizes per century are uneven in ways that reflect the survival record: the 6th century slice is small (18,217 tokens — the archaic period is less attested), the 5th and 4th centuries are the largest (42k and 41k, the classical peak), the 3rd century is moderate (30k), and the 2nd century smaller again (20k), with the 1st century large (39k, reflecting the volume of preserved Hellenistic prose). All key terms are in every slice’s vocabulary, so the comparisons are readable across the full range, though the 6BC and 2BC figures should be treated with more caution than the densely attested classical centuries.

The most important result can be stated immediately. φύσις/βάρβαρος across all six centuries:

Near zero in every century, ranging from −0.005 to 0.068 across six hundred years of text. The naturalization of the barbarian label through φύσις — the claim that Greek discourse progressively grounded ethnic difference in inherent natural constitution — does not appear as a distributional signal at any point in this record. This is not a feature of any particular author’s register; it is consistent across the archaic, classical, and Hellenistic periods, through the full arc from Ionian prose to Stoic philosophy.

The full diachronic similarity matrix for βάρβαρος against the key conceptual terms:

ψυχή and λόγος are consistently near zero or negative across all six centuries. The vocabulary of soul and reason does not organize itself around the ethnic antithesis at any period — not in the 5th century when the Sophists were making arguments about φύσις and νόμος, not in the 4th when Plato and Aristotle were theorizing rational nature, not in the 3rd when Stoic cosmopolitanism nominally dissolved the Greek/Barbarian distinction in favor of the rational/irrational one. Whatever is happening in ancient Greek ethnic discourse, it is operating in a different conceptual register from the philosophy of mind and reason.

τύραννος peaks in the 4th century (0.387) — the height of the panhellenic propaganda effort associated with Isocrates, Demosthenes, and eventually Philip II of Macedon. The claim that Barbarians are by disposition subjects of tyrannical rule (while Greeks require and are capable of self-governance) reaches its distributional maximum in the period when that claim was doing the most political work. It does not disappear afterward: τύραννος holds at 0.309 in the 3rd century and 0.298 in the 1st, persisting through the Hellenistic period well after the specific political context of the Persian Wars had receded.

δοῦλος drops sharply in the 3rd century (from 0.297 in the 4th to 0.107), then partly recovers. The slave/barbarian equivalence — a central feature of 5th and 4th century political rhetoric — weakens in the Hellenistic period. The βάρβαρος nearest-neighbor clusters in the 3rd century suggest why: the word has acquired a new set of neighbors. Where the 5th century clusters included Persian and Median ethnic terms, the 3rd century clusters include Γαλάτης (Galatian), Κελτός (Celt), Ἰλλυριοί, and μισθοφόρος (mercenary). The geographic reference of βάρβαρος has shifted from the Persian Wars east to the Hellenistic frontier, and the relationship has shifted accordingly: Galatians invading Asia Minor are military adversaries and hired soldiers, not the enslaved populations whose condition grounded the slave/free antithesis.

ἐλεύθερος peaks unexpectedly late, in the 2nd century (0.336). The freedom discourse around the barbarian label is strongest in the period of Roman conquest, when Greek freedom became an explicit political rallying cry — the slogan under which Rome itself claimed to be liberating Greeks from Macedonian domination in 197 BCE. Freedom as a concept defined in opposition to the barbarian condition is most active precisely when Greek political freedom itself was most at stake.

γένος/βάρβαρος is highest in the 6th century (0.206) and then falls to near zero in the 4th (0.011). This is the inverse of what a simple naturalization narrative would predict. If γένος was progressively naturalized from political dynasty toward biological descent and then ontological natural kind across the classical period, one might expect its association with ethnic labels to rise. Instead it falls as the philosophical elaboration becomes most intense. The reason is probably that γένος in the 4th century acquired a more technically precise vocabulary: Platonic and Aristotelian uses of γένος as a natural-kind concept pulled the term toward εἶδος, ζῷον, γένεσις, and ἰδέα. Ethnic uses of γένος — lineage, descent, ancestral kind — still existed, but in the 4th century corpus they were diluted by the philosophical-technical register, weakening the term’s co-occurrence with βάρβαρος.

The βάρβαρος/Ἕλλην dyad shows a different trajectory:

The antithesis tightens from 6BC through 3BC, peaks in the Hellenistic period, then partially relaxes. The pairing is not most consolidated in the classical period — it is most consolidated in the 3rd century BCE, after the Persian Wars antithesis has been generalized into a standing cultural and political formula. The 3rd century is precisely when ἑλληνίζω — to Hellenize, to become Greek — appears as a significant term, and the model captures the mechanism: ἑλληνίζω’s nearest neighbor in the 3rd century slice is μηδίζω (0.658), to go over to the barbarian side, to collaborate with the enemy. Becoming Greek is defined against its structural opposite, going native, in a world where Greek dynasties governed mixed populations across Asia and Egypt. The antithesis is tightest not in its original political context but in the cultural-assimilation discourse of the successor kingdoms.

Perhaps the most unexpected finding is what the diachronic slices reveal about φύσις itself. The concept undergoes major transformations across the period, but these transformations are visible as shifts in its discourse field rather than as changes in its relation to ethnic labels. φύσις per century:

• 6BC / 5BC: ζῷον, εἶδος, ἰδέα, δύναμις, γένεσις — Platonic ontological vocabulary emerging in and around the 5th century
• 4BC: φυτόν (plant), μόριος (part, member), κίνησις (motion), φυσικός — Aristotelian biological and physical vocabulary
• 3BC: ψυχή (0.534), οὐσία (0.517), λογικός (0.498), ποιότης (0.486), ἀρετή (0.479) — Stoic vocabulary, where nature is defined through rationality and substance
• 1BC: ἰδέα, λογικός, λογισμός, ἦθος — mixed Platonic and Stoic inheritance

Each century’s dominant philosophical school appears in what φύσις clusters with. The concept is intellectually active throughout the period; its content changes substantially, from Platonic form-theory to Aristotelian biology to Stoic cosmology. But across all these transformations, φύσις never intersects with βάρβαρος. Whatever ancient Greek philosophers were doing with the concept of nature, and they were doing a great deal, the category of the barbarian was not where they were doing it.

The γένος trajectory across centuries confirms and extends the per-author findings:

The γένος/εἶδος association peaks sharply in the 5th and 4th centuries (0.453–0.454) — the Platonic and Aristotelian moment — then drops. This is γένος functioning as a logical-ontological natural-kind concept, the sense in which it appears in Plato’s division dialogues and Aristotle’s biological treatises. εὐγενής (noble birth) is high in the 6th century and again in the 1st, bracketing the classical philosophical elaboration with an archaic aristocratic sense and a late Hellenistic revival of the same. The biological reproductive vocabulary (φύω, σπέρμα) persists through the Hellenistic period. γένος does not travel along a simple arc from political to biological to philosophical; it moves through all three registers simultaneously, with different ones dominant in different centuries.

The picture that emerges from the diachronic data is one of stable dissociation between the ethnic label and the philosophical vocabulary for nature and mind, alongside substantial internal variation in what the ethnic label is doing and who it refers to. βάρβαρος across six centuries is consistently a term of political and military discourse — tyranny, slavery, freedom, military adversaries — and consistently not a term of philosophical naturalization. The naturalization that Hall, Saïd, and others identify in ancient Greek discourse is real, but it operates at the level of γένος as a conceptual framework for thinking about what makes a group a group, not at the level of βάρβαρος as an ethnic label. What shifts across the period is the ontological scaffolding available for group-talk; whether and how ancient authors apply that scaffolding to ethnic categories is a question the embedding models flag but cannot answer — which is the point at which the computational analysis hands off to annotation and close reading.