Skip to content
Jan 20 / Steve Runge

Meaningful distinction between ἀλλά and εἰ μή, pt. 2

Making a clear distinction in meaning between function words like ἀλλά and εἰ μή can be difficult. They are function words providing instructions about how to relate two textual elements. Our tendency is to link them to the closest counterpart in English, but this means we are understanding them through a perhaps inaccurate filter. This is why I have found the idea of cognitive constraints to be so useful; it removes the need for assigning a gloss and focuses on what the word signals. In the last post, I summarized:

Both exceptive and restrictive constructions have something in common. Both constructions signal that what follows the ἀλλά and εἰ μή (or English rather and except) is to correct or replace some comparable element in the preceding context. It qualifies the preceding statements by adding some other piece of information. This piece of information either corrects some overstatement or replaces one option for another, in the case of ἀλλά.

The last post focused on the function of ἀλλά to correct or replace. So what exactly is the difference between these two connectives if they both do this replacing thing? The key distinction between them is whether the correction/replacement was part of the preceding set or not. Here’s what I mean.

In the case of ἀλλά, the following statement introduces information that was not present in the original statement. The original statement had one or more elements that were incorrect or incomplete, and ἀλλά introduces a new element that was not under consideration. This new element can be added to the preceding set to make it complete or correct (i.e. correcting) or it may stand in the place of some incorrect element(s) (i.e. replacing). Here is what it looks like graphically.

Corrected diagram

 Let’s take another look at an example from the last post

1 Peter 3:13–16 (SBLGNT)

13 Καὶ τίς ὁ κακώσων ὑμᾶς ἐὰν τοῦ ἀγαθοῦ ζηλωταὶ γένησθε; 14 ἀλλʼ εἰ καὶ πάσχοιτε διὰ δικαιοσύνην, μακάριοι. τὸν δὲ φόβον αὐτῶν μὴ φοβηθῆτε μηδὲ ταραχθῆτε,

15 κύριον δὲ τὸν Χριστὸν ἁγιάσατε ἐν ταῖς καρδίαις ὑμῶν, ἕτοιμοι ἀεὶ πρὸς ἀπολογίαν παντὶ τῷ αἰτοῦντι ὑμᾶς λόγον περὶ τῆς ἐν ὑμῖν ἐλπίδος, 16 ἀλλὰ μετὰ πραΰτητος καὶ φόβου, συνείδησιν ἔχοντες ἀγαθήν, ἵνα ἐν ᾧ καταλαλεῖσθε καταισχυνθῶσιν οἱ ἐπηρεάζοντες ὑμῶν τὴν ἀγαθὴν ἐν Χριστῷ ἀναστροφήν.

Verse 13 asks who there is who’d harm you for being zealous, with the implied answer being “no one.” Verse 14 now introduces a new element, the idea that such a person really does exist. The ascensive καί casts this hypothetical person as though it were a “least likely possibility” comparable to our use of “even” in English. So in terms of the diagram above, the Xs stand for the people who’d harm you for being zealous. Are there any? No, well, maybe. The Y is adding a caveat that even if they do exist and this does happen, you shouldn’t let them affect your behavior.

The second one is a bit more tricky, since the ground rules for how you make your defense are not explicitly mentioned in v. 15. Nevertheless, v. 16 introduces a new element that (most likely) was not under consideration by the hearer. If Peter was merely adding another thing to do (“Oh, and be sure to do it with gentleness and reverence”), whatever negative thing that he was trying to prevent would still have been admitted. The use of ἀλλά constrains what follows to replace whatever negative behavior the preceding may have conjured up, defined as the opposite of gentle and reverent.

So although ἀλλά and εἰ μή both do the same sort of thing, and both can quite often be translated using a generic “but,” they nevertheless have a meaningful distinction which differentiates them from one another. The distinction is present in Greek regardless of how it might be translated. I feel like this point is dismissed by some using a sense-based explanation: “It’s the A sense of the word in this context, not the B sense.” Regardless of the translation used, the same distinction between them will be present.

This means that Paul’s use of εἰ μή in Gal 2:16 rather than ἀλλά  was meaningful and intended. The lack of textual variants for this reading also points toward Paul intending to communicate something with εἰ μή that would not have been conveyed with ἀλλά. I’ll move on to εἰ μή in the next post.

Understanding grammar is a double-edged sword. It can improve the precision of our exegesis and understanding of the text. But it also makes it much harder to hide from what appear to be unpleasant things. If Paul had intended for “works of the law” to have been replaced by “through faith in Jesus Christ,” he surely could have done so using ἀλλά. To be blunt, I think a lot of reformed folks would have preferred Paul has used ἀλλά, but he didn’t. An unpleasant thing this, but not the end of the world that some have made it out to be.

Instead Paul used εἰ μή in order to bring about some constraint that ἀλλά would not have achieved. Translating with the generic “but” in English–which can convey either constraint–only masks what is going on. I am not arguing here for the proper gloss, but for a recognition of what is going on in Greek.

Next post will spell out the constraint of εἰ μή and pave the way for heading back to Gal 2:15-16.


  1. Kris / Jan 29 2013

    Can you explain what you mean by:

    “The distinction is present in Greek regardless of how it might be translated. I feel like this point is dismissed by some using a sense-based explanation: ‘It’s the A sense of the word in this context, not the B sense.'”

    I’m not sure I understand what you mean by a “sense-based explanation”…

    Nonetheless, really great post. Isn’t it crazy (and scary) how much translations can “mask” what’s really going on! In reading the original languages—but having English translations in the back of my mind—I’m always surprised at how hard it is to let the text speak for itself.

    I knew one man who became a Christian later on in life, and actually learned the Biblical Languages, and was always amazed out how critical/objective/outside-the-box he could be when reading the originals since he had no real cemented default-translation in the back of his head. It takes a lot of discipline and self-awareness to approach the text like this if one’s “grown up in the church” though. Thanks for encouraging this process with these posts.

  2. Eeli Kaikkonen / Oct 17 2013

    We are still waiting for the part 3… Especially because there’s a Galatians reading group going on in B-Greek. I just linked to your posts from there.

  3. Eeli Kaikkonen / Apr 22 2015

    OK, you did a google hangout out of this. Unfortunately I didn’t find a way to write comments there.

    There are examples in LXX where ει μη quite clearly doesn’t mean “except” but “only” or “but only”. 1Reigns 21:23 (=1Kings 20:23), 4Reigns 7:10 (=2Kings 7:10), 4Reigns 23:9 (=2Kings 23:9). Therefore I believe your interpretation of Gal 2:16 is also wrong and the translations are correct.

    My theory: if the thing after ει μη doesn’t belong to the set expressed before it, it creates a larger set in which the original set belongs to. It takes another small subset out of the larger set and gives it as an alternative to the original set. So in Gal 2:16 “salvation by works” is the original set, “all ways of salvation” is the larger set, and “by faith” is the smaller alternative set. It’s not an exception to “no salvation by works”, it’s an exception to “no salvation by any means”.

  4. Steve Runge / Apr 23 2015

    Hi Eeli,

    Thanks for the comments. To clarify about applying your theory to Gal 2:16, are you saying that faith is not included in the means by which someone might be justified (by works of the law, from my reading)? Your claim is unprovable and unfalsifiable, in that you say if it isn’t included, it will become included in a superset. I’d want to see it worked out in more data.

    My primary point is not to claim that “but” is wrong, but that it is unnecessary. Reading the clause it its larger context makes this clear, in my view. You are free to disagree (as you do), but I would have hoped you’d address my argument. I am not claiming another reading is wrong, simply that the reaction against what looks like a contradiction is misguided.

    Regarding your LXX examples, 3 Reigns 21:23 εἰ μὴ κραταιώσομεν is the matrix clause, not a dependent exception, based on ἐὰν πολεμήσωμεν αὐτοὺς κατʼ εὐθύ being a dependent conditional clause, so not really a useful basis of comparison. Most translators understand the εἰ μὴ as an intensive, “surely.” The other two read ὅτι εἰ μὴ. You’d need to account for the role that ὅτι might play in changing the constraint of the collocation. You’d also need to determine what, if any, relationship this collocation has with Paul’s usage; not something easily done.

    In your 4 Reigns 23:9 example the priests of the high places actually did go up to the altar in Jerusalem, viz when they ate unleavened bread. This seems like a prototypical exception to me, not sure how it makes your point. But the 4 Reigns 7:10 example is much more interesting, as it illustrates the point at stake here. There are two possibilities to resolve it. One would be to take a strict and narrow reading, i.e. that the original set only included men and their noise and nothing more, no donkeys or anything else. This reading would require us to either expand the prototypical range of meaning for the collocation, or treat it as an exception as you seem to do. Another possibility is to understand the initial statement as a “no one / nothing” statement that pragmatically entails the exception that follows about livestock. In other words, it’s not just that there were no men, there were only two animals left. I would opt for the latter, as it respects the writer’s decision to use X as a connective instead a Y we might have preferred. Both are possibilities; the resolutions differ more in methodology than truth-conditional status. Neither can be definitively proven. Thanks again for the input.

  5. Eeli Kaikkonen / Apr 23 2015

    My impression was that you think that exegetes say that Gal 2:16 is somehow exceptional so that ει μη there means something it normally doesn’t mean. I tried to prove that “only” or “but only” is a normal and legitimate, if somewhat rare, meaning (or translation) for ει μη and your, in my opinion quite complicated and difficult, interpretation of the passage is unnecessary. Here I must say that I know how an explanation may at first feel difficult and later, when one grows in knowledge, it may feel natural. It can happen to me here, too, but this is how I feel about it now.

    As for 4 Reigns 23:9, NETS LXX translation translates it “Yet the priests of the high places did not go up to the altar of the Lord in Ierousalem but only ate unleavened bread in the midst of their brothers.” Likewise 7:10: “there is no man nor human voice there but only a tied horse and a donkey and their tents as they are.”

    About my “theory” (or a quick construction of logic with idea of sets), I would give examples like this:

    “Salvation doesn’t come from works. Actually it does come in no way – except by faith.” “Salvation by some way” is the larger set, “from works” is a subset, “by faith” is a subset.

    Likewise in Gospel of John: “Jesus does nothing by himself. Actually he does nothing – except what he sees his Father doing.” The larger set is “Jesus does something”, the subsets are “by himself” and “what he sees his Father doing”.

Comments are closed.