Advanced search

First, two things worth knowing

Greek changes the ends of words. English mostly does not: "virtue" looks the same wherever it stands. Greek changes the ending to show the word's job — ἀρετή, ἀρετῆς, ἀρετήν are one word doing three jobs. So a search for one spelling misses most of the places that word occurs. Several of these tools exist to fix that.

You can type Greek in plain letters. With no Greek keyboard, type the sounds instead — scholars call this Beta Code. texnh gives τέχνη. The search page has a key to it, and this page shows examples both ways.

Numbers like 34b are Stephanus pages, the standard way of citing Plato. Most editions print them in the margin, so the same number finds the same passage in any edition or translation. Results here use them throughout, and nothing finer: Plato is cited by page and section, never by line.

Wildcards

Use this when you know how a word begins but not how it ends.

End a word with * and it stands for whatever follows. fron* reaches 68 different words across the dialogues — φρόνησις, φρόνιμος, φρονεῖν and the rest of that family at once. Since Greek marks a word's job at the end, one wildcard gathers the whole paradigm.

A wildcard reaches further the more text you give it. In the Republic alone, l* reaches 302 different words; across all 36 dialogues it reaches 891, out of 46,389 distinct spellings in the corpus. The search reports occurrences rather than distinct words, so it answers l* with passages, not with a vocabulary count.

The wildcard works at the end of a word only. dikaio* finds the 30 spellings that begin that way, including δικαιοσύνη and its endings. A star in the middle of a word finds nothing at all. l*s looks like a narrower search, but the index is built on word beginnings and cannot answer a pattern about how a word ends; rather than quietly hand you every word starting with λ, which is not what you asked for, the search returns nothing. A star at the start is read differently again: in Beta Code it marks a capital letter, so *swkraths is simply Σωκράτης, not a wildcard at all. If you want every form of one dictionary word, do not reach for a wildcard at all — set the Greek box to Any form of this word and type the form you have.

Word endings

Use this when you know a phrase but want every form of it, not just the one you typed.

Since Greek changes word endings, an exact phrase search finds only the version you typed. Search a phrase with Exact phrase set and a button appears: Find this phrase in any inflection. It works out which dictionary word each of your words belongs to, then looks for every combination of those words in the order you gave.

The phrase index shows the same gathering, and you can see the size of it there. τὸ καλόν ("the beautiful") stands 93 times as those exact letters, in 14 dialogues. Read as dictionary words — ὁ καλός, which gathers τοῦ καλοῦ, τῷ καλῷ, τῶν καλῶν and the rest — it stands 385 times, in 32. To find those yourself you would have to know that τό, τοῦ and τῷ are one word. The tool knows it for you.

It invents nothing. πάνυ μὲν οὖν ("certainly") stands 333 times as written and 333 times read as dictionary words: those three words take one form only, so there is nothing to gather. A tool that always found more would tell you nothing.

Search by grammar

Use this to find a kind of word rather than a particular word.

Every Greek word here carries a grammatical description: singular or plural, past or present, and so on. You can search those descriptions instead of the spelling. Pick as many as you like; a word must match them all.

Grammar is one term inside a proximity search, never a search of its own. On its own it would not be a question: there are 19,702 genitive plural feminines in the dialogues, and a list of them tells you something about Greek rather than about Plato. Beside another term it earns its keep — this word near a verb in the optative is a real question, and it is the one the tool answers. Set a term's Kind to Grammatical description under Words near each other.

You need not know how any of these forms are spelled. You pick the description; the search finds the words.

Why some results say "one of several readings"

The same Greek letters can be two different forms. When that happens the tool marks the result instead of choosing one for you.

Plato's dialogues hold 7,212 optatives. Of those, 4,054 have one possible reading; a hit on any of the rest comes back marked, because you could read those letters another way. The mark shows no flaw in the search. It reports a fact about the Greek.

A clearer case: of the 473 genitive duals (the dual describes pairs of things), not one is certain. In the dual, genitive and dative share one ending. The search cannot tell them apart because Greek does not, and it says so instead of deciding for you.

You will find no option for "noun", "verb" or "adjective". We left it out on purpose. The descriptions this site uses do not record parts of speech, and we cannot work them out — a participle behaves as both verb and noun, and nothing separates a noun from an adjective. Offering the choice would mean making it up.

Words near each other

Use this to find where two ideas are discussed together.

Ask for two or more things close together in the text. Each term takes a word as written, a word in all its forms, an exact phrase, or a grammatical description, and you can mix them. "This word near a verb in the optative" is a question you can put here. Search either word on its own and you get every passage holding it; ask for both together and you get the passages where Plato is doing something with the pair.

The controls

• Window — how close is close. Five words by default. You can widen it to 50.
• Measured in — count that distance in words, or ask instead for the same line of Greek or the same speech. A speech here is one turn: everything one speaker says before the next answers. That is the unit a dialogue is jointed at, and below the book it is the only division Plato's text has.
• Relative to term 1 — you can require a term to come before or after the first one, rather than simply near it.
• In this order — everything must appear in the order you typed it.

A match never crosses from one book into the next, however wide the window: the last words of one book and the first of the next are not near each other in any useful sense. Matches do cross from one speaker's turn into the next by default — a question and its answer belong together — but you can turn that off.

The phrase index

Use this when you do not yet know what to look for.

Search answers "where does this occur?" The phrase index answers a different question: what does Plato repeat? It lists every phrase of two to five words that appears at least twice anywhere in the dialogues — 84,894 of them as written, 219,287 when forms of the same word count together, and 187,774 in the English translations. (If you have met the term, these are what linguists call n-grams.)

This matters because a dialogue is built as much out of set phrases as out of single words, and no word search will show them to you: ἦν δ’ ἐγώ ("said I", Socrates narrating) occurs 860 times across 11 dialogues; πάνυ μὲν οὖν ("certainly") 333 times; ὡς ἔπος εἰπεῖν ("so to speak") 74 times in 20. Not one of those is a word. Each is a habit of thought, and browsing is the only way to meet them.

Reading a row

Every row is one phrase and four numbers. This is the row for ἦν δ’ ἐγώ:

The same phrase, counted two ways

The first two buttons at the top of the page are the same corpus read differently, and it is worth seeing what changes. Word as written counts the letters on the page: change one ending and it is a different row. Word in any of its forms counts the words behind the letters, so every inflection of the phrase gathers into a single row. Either takes the phrase as it stands on the page — type ἦν δ’ ἐγώ and the second finds it under ἠμί δέ ἐγώ, which is where it is stored, and tells you it read the words that way.

The dictionary row is never the smaller of the two: it gathers the formula's other endings in with it. It also asks more of you: the same dictionary words can be assembled into phrases that are not the one you wanted, and the tool cannot tell the difference. Use it to discover what you would otherwise miss, then read as written to see what you have.

The score changes between the two rows for what is, in the main, the same expression. Nothing has weakened. Distinctiveness measures surprise against the words a phrase is made of, and the two readings are made of different words: ἠμί and δέ are commoner than the exact shapes ἦν and δ’, so a phrase built from them is less surprising.

Reading the list

• Distinctiveness orders the list by default. It asks whether a phrase occurs more often than the commonness of its words would lead you to expect. Sort by plain frequency instead and you see the grammar of Greek rather than the thought of Plato.
• Filters narrow the list by phrase length, by how often a phrase occurs, and by which dialogue it appears in. Sorting never removes anything — it only changes the order.
• The English list indexes the translation each dialogue carries, which here is the Loeb throughout — 4 translators working to one editorial standard rather than a different hand for every dialogue. So an English phrase that recurs across several dialogues is more likely to be Plato recurring than one translator's habit, and the Works column is the thing to read. It is still English: a phrase in this list is a place to look, not evidence about the Greek.

Click any phrase to see every place it occurs, grouped by dialogue, each one linked to that passage in the reader.

A way to use it

Pick the dialogue you are reading and a phrase length of four or five, then sort by distinctiveness. You get that dialogue's set expressions — often the working vocabulary of its argument, including phrases no glossary would have sent you to look for. Click one to see where else in Plato it stands.

What these tools cannot tell you

All of this depends on a word-by-word grammatical description of the Greek, and that description is often uncertain. Where it is, these tools say so rather than choose an answer and present it as fact. The limits worth knowing:

• A result marked as having several possible readings comes from a form the description cannot settle. Of the 7,212 optatives, 4,054 are settled and the rest are not.
• Searches ignore accents, so λόγος and λογός count as one word. The search page has a setting to match accents exactly. Of the 46,138 spellings the corpus knows, 6,730 belong to more than one dictionary word, and no accent-blind search can separate them.
• The phrase index shows Greek without accents, because it does not record them. We do not guess them.
• Where a speech begins is worked out from the text, and in a few places only to the nearest line of Greek rather than to the exact word. Any search that turns on speaker turns says so when those dialogues are involved.
• An exact phrase search cannot match across the boundary between two Stephanus sections; the phrase index and the proximity search can. So the index will sometimes count a phrase once more than an exact search finds it.
• None of these tools reads Greek. They match forms and count them. What a passage means remains your work.

The Greek text is the TLG. The grammatical descriptions come from Morpheus, by way of Diogenes; the dictionary is Liddell–Scott–Jones.