I'm milk

« previous post | next post »

This has been making the rounds:

1. Go to Google Translate.
2. Set the input language to Spanish.
3. Paste in "soy milk"
4. Set the output language to English or X language.
5. Hilarity ensues.

The obligatory screen shot:


This works for Japanese, Chinese, German, Russian, Nepali, and every other language I tried it on — except for Spanish, where it properly yields "leche de soja".

Hint:  the Spanish for "I am" is "soy" (or "estoy").

Selected reading

[Thanks to Nathan Hopson]


  1. Ferdinand Cesarano said,

    February 10, 2021 @ 11:46 am

    For at least 40 years I have been referring to soy sauce as "yo soy sauce", thereby transforming a simple condiment into a kind of existential elixir.

    Perhaps one day I will encounter someone who finds that as amusing as I do (though probably not).

  2. Victor Mair said,

    February 10, 2021 @ 11:57 am

    @Ferdinand Cesarano

    I find it extremely amusing and deeply profound. Thank you so much for sharing your existentially transformative elixir with us.

  3. Joshua K. said,

    February 10, 2021 @ 12:01 pm

    This is a version of a common joke: "What if soy milk is just regular milk introducing itself in Spanish?"

  4. Cervantes said,

    February 10, 2021 @ 12:34 pm

    Technical point, in this case it would be soy rather than estoy because it's presumably expressing the speaker's basic nature rather than a transient state. This does not create any problem in Spanish because the Spanish word for soy is soja, so it's the user's error, not Google's.

  5. Daniel Hirst said,

    February 10, 2021 @ 1:14 pm

    If you replace soy milk by soy beans you get: I am beans
    But if you put soybeans then you get: soybeans…

  6. Sean Richardson said,

    February 10, 2021 @ 1:19 pm

    @Cervantes my first thought was, of course no language would have homonyms for a core grammar word. Second thought, but, Popeye: I yam what I yam. Third thought: right, too much potential for unintentional hilarity to allow.

  7. Michael Watts said,

    February 10, 2021 @ 1:37 pm

    What's the hilarity? You put in "I'm milk" in Spanish and you get out "I'm milk" in English. We're laughing at Google Translate for doing its job correctly?

  8. Michael Watts said,

    February 10, 2021 @ 1:59 pm

    my first thought was, of course no language would have homonyms for a core grammar word.

    是 [to be]: homophonous with several things, but most particularly 事 [situation; affair]

    的 [many grammatical functions]: perfectly homophonous with two other particles 地 and 得 which are distinguished in formal orthography; native speakers routinely confuse them. Probably not really homophonous with other "de" non-grammatical words due to clitic tone.

    把 [preposition for marking the direct object of a verb, used for rearranging the layout of the various parts of a sentence]: perfectly homophonous — and homographic — with 把 [measure word for various things like umbrellas, swords, and chairs].

    被 [passive voice marker]: homophonous with 背 [the back of something; also a verb meaning "memorize; recite from memory"]

    are [to be]: homophonous with R [the name of the letter], also homophonous with "our".

    will [auxiliary verb]: homophonous with the historically related "will" [document specifying how the property of a deceased person should be distributed].

    can [auxiliary verb]: homophonous with "can" [metal container].

    have [auxiliary verb]: homophonous with "have" [possess].

    be [passive voice marker]: homophonous with be [to be]. They're both highly grammaticalized meanings! Also homophonous with bee [insect known for producing honey].

    The particles that attach to verbs ("pick up" / "throw away" / "ask out") are quite transparently related to the prepositions with which they share spelling, pronunciation, and often semantics, but they're pretty easy to distinguish grammatically.

    It seems like a safe bet that every language has plenty of homonyms for core grammar words.

  9. Michael Watts said,

    February 10, 2021 @ 2:25 pm

    Expanding a little on the above, "be" and "bee" are perfect homophones, but there is essentially no risk of confusion because "be" is a core grammar word. (Actually, several core grammar words.) Substituting "be" for "bee", or the other way around, will almost always result in an ungrammatical sentence.

  10. Philip Taylor said,

    February 10, 2021 @ 2:49 pm

    « are [to be]: […] , homophonous with "our" ».

    Perhaps in your topolect, Michael, but in British English as different as soya milk and dairy cheese (/ɑː ǁ ɑːr/ v. /ˈaʊ‿ə ǁ ˈaʊ‿ər/).

  11. KevinM said,

    February 10, 2021 @ 3:03 pm

    I am a (coconut) milk sincero/De donde crece la palma

  12. EmilyPigeon said,

    February 10, 2021 @ 3:08 pm

    Soy beans– supposedly Pythagoras believed that!

  13. Michael Watts said,

    February 10, 2021 @ 3:22 pm

    Perhaps in your topolect, Michael, but in British English as different as soya milk and dairy cheese

    Sure; I've heard that in Australia they distinguish "can" from "can". But that's not relevant to the claim that "no language would have homophones for a core grammar word"; that claim does not require several related languages to agree that two words are homophonous in order to be falsified. The quorum for a counterexample to "no language" is one language.

  14. Michael Watts said,

    February 10, 2021 @ 3:24 pm

    (As to the specifics of the pronunciation of "our" vs "are", I would say that "our" permits a certain amount of lip rounding, and "are" doesn't really, so in many but not all cases, they will come out phonetically distinct. However, lip rounding is never phonemic.)

  15. Peter Taylor said,

    February 10, 2021 @ 3:32 pm

    Sean Richardson, have?

  16. Dara Connolly said,

    February 10, 2021 @ 5:23 pm

    "Soy sauce" means "I am willow".

  17. Thomas Rees said,

    February 10, 2021 @ 5:29 pm

    Chava Flores song «Tomando té»:

    No quiero tomar café
    porque el café quita el sueño,
    lo que quiero es tomar té,
    pues tomando té me duermo.

    La primera vez que te tome,
    ¡ay! que cosas disfrute,
    que desde entonces quiero estar
    tomando te, tomando te.

    El doctor que a mi me ve,
    de este mal que aqui me inflama
    recetóme tomar te
    y ahora te tomo en la cama.

    Una vez que te tome,
    tan cansado me quede,
    que desde entonces quiero estar,
    tomando te, tomando te

    Written by: Salvador Flores Rivera

    This is called the tilde diacrítica. “Tomar té” means “drink tea”; “tomar te” is “take you”. The Google translation isn’t bad.

  18. Jerry Friedman said,

    February 10, 2021 @ 7:37 pm

    Dara Connolly: I was going to say that!

    If you use GT to translate "I am willow" into Spanish, you get "yo soy sauce".

    Michael Watts: I'm told the verb "can" and the noun "can" are also not homophones in New York City English. (This is called the can-can split.) It may be true in other dialects from Albany to Baltimore. I can't quite tell from the Wikipedia article. Just a language fact, not an argument against your example.

    Along with "have", you might mention "halve" (a verb many of my students replace with "half"), at least in most dialects of American English. However, I don't always pronounce the two "have" meanings the same: "I have to go" with /f/ versus "I have to-go orders to take care of" with /v/.

    How many words are pronounced /e/ in French, most if not all being core grammar words? Although some of them may be different before a vowel.

  19. Counterbander said,

    February 10, 2021 @ 8:39 pm

    Eh bien, La Haye (The Hague) et la haie (hedge)

  20. Michael Watts said,

    February 10, 2021 @ 11:10 pm

    I don't always pronounce the two "have" meanings the same: "I have to go" with /f/ versus "I have to-go orders to take care of" with /v/.

    There are more than two haves.

    There is the non-grammatical have, meaning possess, which in American English is not auxiliary in any way; the opposite of "I have change" is "I don't have change", not "I haven't [any] change". Similarly, the interrogative form of "I have a cat" is "Do you have a cat?" and not "Have you a cat?".

    There is the perfect auxiliary have.

    And there is the have of obligation, which is always followed by the particle to. This is the only one to be devoiced, and it is devoiced in all forms; /f/ in "have to" and /s/ in "has to". I assume the devoicing arises from the fact that it is followed by devoiced /t/ 100% of the time.

  21. Michael Watts said,

    February 11, 2021 @ 12:01 am

    (I overspoke; obligational have is not devoiced in the extant form having to. That's not a problem for the theory that it long ago assimilated the (lack of) voicing of the following /t/, since another syllable is interposed between the "v" and the /t/ in that form. The past form had to is trickier; I somehow feel that it preserves a /d/ there.)

  22. Philip Taylor said,

    February 11, 2021 @ 2:51 am

    « [the non-grammatical 'have', meaning possess] in American English is not auxiliary in any way; the opposite of "I have change" is "I don't have change", not "I haven't [any] change" ». Can you not also say (in American English), "[Sorry,] I have no change" ? This would be my normal response when asked for change if I have none.

  23. David Morris said,

    February 11, 2021 @ 5:52 am

    (AusEng speaker) I *can* pronounce grammatical 'can' the same as lexical 'can', but I c'n also reduce it (and usually do). I c'n can tomatoes. Yes, I can. I *can* can them.

  24. Rodger C said,

    February 11, 2021 @ 8:01 am

    are [to be]: homophonous with R [the name of the letter], also homophonous with "our".

    In the public schools of West Virginia I was taught not to do that. It was a stricter time.

  25. Rodger C said,

    February 11, 2021 @ 8:06 am

    And in WV, "can" (noun) and "can" (verb) are [kajən] and [kiən].

  26. Robert Coren said,

    February 11, 2021 @ 10:48 am

    Years ago, we had put some alternative flours in old wheat-germ jars, and so there these two jars side by side on a pantry shelf, on which had been written in magic marker "SOY" and "CORN". A clever friend who was visiting spotted them, grabbed the marker, and inserted an E between the last two letters of the latter, and held the two jars out to suggest that I use them to proclaim my identity.

  27. Michael Watts said,

    February 11, 2021 @ 11:33 am

    Can you not also say (in American English), "[Sorry,] I have no change" ?

    Yes, you can say that.

  28. Michael Watts said,

    February 11, 2021 @ 11:43 am

    David Morris:

    I think reduction of modal can is fairly normal in the US too, but it's not what I was trying to refer to. https://en.wikipedia.org/wiki/Pronunciation_of_English_%E2%9F%A8a%E2%9F%A9#Bad%E2%80%93lad_split :

    The bad–lad split has been described as a phonemic split of the Early Modern English short vowel phoneme /æ/ into a short /æ/ and a long /æː/. This split is found in Australian English and some varieties of English English in which bad (with long [æː]) and lad (with short [æ]) do not rhyme.


    Outside of England, can meaning 'able to' remains /kæn/, whereas the noun can 'container' or the verb can 'to put into a container' is /kæːn/; this is similar to the situation found in [/æ/ raising æ-tensing] in some varieties of American English. A common minimal pair for modern RP speakers is band /bæːnd/ and banned /bænd/. Australian speakers who use ‘span’ as the past tense of ‘spin’ also have a minimal pair between longer /spæːn/ (meaning width or the transitive verb with a river or divide) and /spæn/, the past tense of ‘spin’ (/spæn/).

    That said, this is all hearsay for me.

  29. Jonathan said,

    February 11, 2021 @ 1:53 pm

    I always thought 'yo soy marinero' in La Bamba meant that he was a marinated soybean.

  30. Ben Zimmer said,

    February 11, 2021 @ 5:30 pm

    @Jonathan: And in Beck's "Loser," the line "Soy un perdedor" has been misheard as "Soy on my candy corn" (among many other things).

  31. Chester Draws said,

    February 12, 2021 @ 1:43 am

    "Eh bien, La Haye (The Hague) et la haie (hedge)"

    Despite the archaic spelling, they are exactly the same word.

    The Hague is literally "the hedge", originally the Count's hedge. Den Haag, likewise is "the hedge".

  32. Batchman said,

    February 12, 2021 @ 3:57 pm

    The can-can split in NYC and similar dialects also serves to distinguish verbal "can" from "can't", since the final T of the latter is often elided, especially before a subsequent alveolar/dental. So the only difference between "I can do this" and "I can't do this" may be the quality of the A.

    There may be stress distinctions as well, but there are plausible utterances that may stress or fail to stress either "can" or "can't."

  33. Alyssa said,

    February 12, 2021 @ 3:59 pm

    Can you not also say (in American English), "[Sorry,] I have no change" ?

    You can say it, but it's not the most natural wording in this context. I think "have no" is mostly used for emphasis, as in phrases like "I have no idea". It's a stronger negative than "don't have".

  34. Philip Taylor said,

    February 13, 2021 @ 6:40 am

    That is interesting, Alyssa, because for me the stress patterns are very different in the two utterances — "Sorry, I HAVE no change", but "I have no IDEA".

  35. Michael Watts said,

    February 13, 2021 @ 3:11 pm

    That's odd, I more or less agree with the sentence stress in "Sorry, I HAVE no change", but I would say "I have NO idea".

  36. Philip Taylor said,

    February 14, 2021 @ 6:19 am

    Yes, I can certainly imagine stressing it as "I have NO idea", and it seems to me that the implied difference between that utterance and "I have no IDEA" is that in the former one is tacitly eliding "absolutely" as in "I have absolutely no idea" while in the latter one is implying "not only do I have no factual answers to the question asked, I do not even have any suggestions". But clearly they are on a continuum, and not very far apart at that.

RSS feed for comments on this post