Prepositions and Perpetual Virginity

Since Christmas is approaching, let’s talk about what happened after Jesus was born.

Virtually all ancient interpreters considered the Perpetual Virginity of Mary to be a dogma or a pious–-even preferred––opinion. The biblical case usually involves stressing how the argument of the ἕως-phrase in Matthew 1:25 cannot receive an inchoative interpretation: 

  1. καὶ οὐκ ἐγίνωσκεν αὐτὴν ἕως οὗ ἔτεκεν υἱόν (THGNT)
    [He] did not have sexual relations with her until she gave birth to a son (CSB)

A good example comes from John Chyrsostom (Homilies on the Gospel of Matthew):

[Matthew] has here used the word until, not that you should suspect that afterwards [Joseph] did know [Mary], but to inform you that before the birth the Virgin was wholly untouched by man.

Chrysostom is an astute interpreter and appears to get the semantics right. For our purposes, let’s represent the logical form so we can distinguish the semantics (code) from the pragmatics (inference). The semantics are straightforward:

  1. ∃e ∃t [before Jesus was born(t)]  t < n  knew Mary (Joseph, e,t)  ¬ ∃e’ ∃t’ [t∈C  t'<t  knew Mary(Joseph, e’, t’)]
    There exists an event e and a time t, and e happened before t, and there does not exist an event e‘ and a time t‘ where the reverse happened.

The first thing to note about the until-phrase is that it requires the subinterval property: wherever an event e holds true at time t, e must also hold true at every interval of t: t2, t3, t4, etc. This means everyone must agree the ἕως-phrase in Matthew 1:25 describes e as not holding true for all sub-intervals of t: Before Jesus was born, Joseph did not know Mary––neither at t2, t3, t4, etc. But what happened after? Here is where it gets tricky. Chrysostom believes e did not hold true after t either. He gives Ps. 89:2 as evidence:

  1. ἀπὸ τοῦ αἰῶνος ἕως τοῦ αἰῶνος σὺ εἶ (Ps. 89:2 LXX)
    From age to age, you are

Chrysostom is an astute reader, and gets the semantics right again:

  1. ∃e ∃t [age to age(t)]  t < n  are (you, e,t)
    There exists an event e and a time t, and e holds true at t

Clearly, e (you are) is true not only at t (age to age) but also after: Where there is an age, God is. How can it be that God would cease to be? Therefore, Chyrsostom argues, the same reading should apply to Matthew 1:25. The reductio is effective. We are also tempted to trust him because he was a native Greek speaker. After all, we do not speak Ancient Greek and do not share his intuitions about the language. Who are we to argue with a Greek speaker? Each of these things seem to make a good case to assign the same interpretation to the ἕως-phrase in Matthew 1:25. But there are problems.

Let’s begin with interpretation and native speaker intuition. Moises Silva (2005: 27) puts it well:

Educated speakers are notoriously unreliable in analyzing their own language. If Chrysostom weighs two competing interpretations, his conclusion should be valued as an important opinion and no more. If, on the other hand, he fails to address a linguistic problem because he does not appear to perceive a possible ambiguity, his silence is of the greatest value in helping us determine how Paul’s first readers were likely to have interpreted the text.

Search your feelings. You know what Silva says is true. 

As this blog’s esteemed linguist overlord has noted, appeals to native speaker intuition actually prove the opposite: Chrysostom felt need to go into detail defending this particular meaning precisely because it was counter-intuitive to his audience. In fact, his entire defense is meant to persuade his audience that their intuitions are wrong! Anyone who has done field work knows that educated native speakers can be highly unreliable. But this isn’t the main problem. 

Language is fascinating and complex, and meaning cannot be reduced to words. Perhaps the vast part of the meaning that we communicate is inferred. Consider this example:

  1. Do you want to come to tea tomorrow at 5?
    Oh, my sister is here.

The two sentences have no meaningful relation beyond their context. And yet we all understand that the intended meaning of the sentence, “Oh, my sister is here” was “I can’t.” How? The answer is a conversational implicature––an inferred meaning that was intended by the speaker but not directly coded in the utterance. It has to be recovered from speaker intention. Inference is a rich design feature of natural language. It solves an important problem that Levinson (2000: 6, 27-30) calls the “articulatory bottleneck”: we compute information inestimably faster than we can articulate it in speech production. To solve this problem and speed things up, we use inference: “[I]nference is cheap, articulation expensive, and thus the design requirements are for a system that maximizes inference.” Natural language is minimally-specified and relies on interpretations that assign maximal informativeness to each utterance. Basically, we squeeze as much meaning from an utterance as can be licensed by the grammar and context. 

Returning to Matthew 1:25, ancient interpreters were right to claim the point of the text was to delimit the state of affairs: before time t where t is the time at which Jesus was born, Joseph and Mary had not consummated the marriage. Event e was not true for all subintervals prior to and leading up to t. But they miss the fact that this is only the semantics of the utterance. When attached to a durative event under negation, the ἕως-phrase triggers a scalar implicature meaning after but not before (Iatridou and Zeijlstra 2021). If I say in English, “I wasn’t sure where I was until I used Google maps”, I don’t mean that I continued to be unsure after I used it. I mean e was not true for all subintervals of t, although it was true after. This is called a Scalar Implicature. How do we know it’s there? Notice the infelicity of (7):

(7) #I wasn’t sure where I was until I opened Google maps. And I was unsure after too.

The contradiction exists because under negation the until-phrase embeds a separate proposition (8b):

(8a) I was not sure where I was.

(8b) I became sure of where I was.

The same holds true for Matthew 1:25. There are two propositions: Joseph did not know Mary before she gave birth, and Joseph knew Mary after she give birth, resulting in the siblings described in Matthew 13:55. Think about it in English. (9a) is a contradiction:

(9a) #Joseph did not know Mary until Jesus was born. And he didn’t know her after either.

Why? Because not all meaning is semantic meaning. Much of what we mean is inferred rather than coded. Yet you might say: Okay, if you’re right, why didn’t he just outright state that Joseph knew Mary after? Why does he leave it as an inference? That’s the beauty of the until-phrase. An exhaustivity operator like until negates alternatives, triggering a scalar implicature that locates an assertion along an informativeness scale. Matthew has made the strongest statement consistent with his personal knowledge. The truth value of the sentence is the same whether he explicitly stated the second proposition or whether he embedded it as an implicature.

A doctrine of perpetual virginity may be supported on other grounds, but it cannot be Matthew 1:25.