It is often said that Porta clausa est can have two readings depending on the categorial nature of the participle: verbal (cf. clauditur/clausa est) or adjectival (cf. clausa est/clausa fuit), which in turn corresponds to a semantic difference between process and result/state, respectively. For example, in Spanish this distinction is also made clear by using different auxiliaries, ser vs. estar: cf. Sp. La puerta fue cerrada (sc. by someone) /La puerta estuvo cerrada (*/??by someone). A typical (although recently argued to be a bit problematic) test to distinguish verbal passives from adjectival ones is that the former can be used with an agentive by-phrase, while the latter can’t. As far as I know, this test has also been assumed to hold for Latin: cf. verbal passive (porta {clauditur/clausa est} a Gaio) with adjectival passive (cf. porta clausa {est/fuit} */??a Gaio).

Given the previous (typical?) approach, I think that the learner of Latin cannot often know if there is still a third reading available for porta clausa est or porta fracta est, the one corresponding to the so-called “anticausative” variant, e.g., 'The door broke' (e.g., cf. the so-called "causative alternation" in English: {John/The strong winds} broke the door (causative variant) vs. The door broke (anticausative variant)).

If so, porta fracta est can be said to have the following three (rather than only the typical two) readings:

1: verbal passive: the door was broken/has been broken (by someone).

2: adjectival passive: the door is broken (*by someone; *on this adjectival reading!).

3: anticausative: the door broke (*by someone). (NB: the use of * means ungrammatical).

Consider the following example, which can be taken as (good? What do you think?) evidence of the anticausative reading (which is, as noted above, different from the processive verbal passive reading (to be broken by someone) and from the resultative/stative adjectival one (to be broken *by someone)).

Postquam omnis res mea Ianum ad medium fracta est, aliena negotia curo, excussus propriis. (Hor. Sat. II 3.18-20)

‘After all my business collapsed at the central arcade of Janus…’

Now here is my question: Which is the nature of the participle in the anticausative reading of fracta est? Cf. the verbal-adjectival distinction drawn above. I would say that in this case the participle has a verbal nature (like in the first verbal passive reading above) but perhaps I’m wrong. More examples of (clearer) anticausative readings of periphrastic forms would also be greatly appreciated.
NB: here I’m mainly interested in the ambiguity of fracta/clausa est rather than in the one also involved in non-periphrastic forms like frangitur/clauditur, which can also have different interpretations. When dealing with non-periphrastic forms, further distinctions become relevant (e.g., cf. frangitur/se frangit, the pronominal form becoming even much more frequent in Late Latin).

What follows contains a relevant modification I should make thanks to the very useful feedback I’ve received from the commentators. Interestingly, I've been advised by the majority of them on the convenience of distinguishing what one could call the "uniformity" of the structure (e.g., the one corresponding to Porta clausa est) from its possible associated meanings (and "nuances", in Joonas's terms). Here is my personal view, which, against perhaps the opinion of some commentators, does crucially separate the three basic "meanings" (into types) from "nuances" (into subtypes). I consider that fuzziness is not involved at the type level (perhaps things are different at the subtype level...). For the sake of exposition, I also consider that it can be useful to use typical semantic predicates like CAUSE, BECOME, and STATE (which have been used in the specialized literature on lexical decomposition of verbs; e.g., Brutus killed Caesar can be decomposed into 'Brutus caused Caesar to become dead'). Next Porta clausa est can be analyzed into the following types and subtypes (NB: of course, some commentators and I disagree on the more convenient "labels" and on the most appropriate "taxonomy" but after all what is really important/relevant is what is behind them).

  1. Verbal passive: 'the door was caused to become closed' (by someone)

  2. Anticausative/"middle": 'the door became closed'.

    2.1. ‘the door became closed because there has been another previous/implicit event that caused the door to become closed’

    2.2. ‘the door became closed by itself'.

  3. Adjectival passive: ‘the door is closed’.

    3.1. Resultative adjectival passive: 'the door is closed as a result of a previous event of closing'.

    3.1.1 ‘the door is closed because there has been a previous event of causing the door to become closed’)

    3.1.2. ‘the door is closed because there has been a previous event of just becoming closed’, i.e., with no agent involved.).

    3.2. Stative adjectival passive: 'the door is closed (with no previous event of closing involved)'.

Notice that in this classification it is important to distinguish the three types from further subtypes/nuances. My impression is that the three basic types are clear, whereas some subtypes are perhaps not necessary. Some further observations/remarks/caveats/... could also be made but here I will limit myself to commenting a distinction that can be drawn within the domain of adjectival structures: cf. the resultative and the stative readings. For example, in Porta clausa est the resultative one is the most prominent one. However, the stative one is not excluded (e.g., imagine that the door was built closed). As for the adjectival reading of porta fracta est , in this case, due to our encyclopedic knowledge, the resultative one seems to be the only possible one.

  • Please, Sp: "La puerta fue cerrada (sc. by someone)/ La Puerta estuvo cerrada (*/??by someone)", what is "sc." & "*/??". When something is described as "ungrammatical" is it violating grammatical rules and is therefore incorrect; "the door broke" should be "the door was broken", presumably by an agent?
    – tony
    Commented Aug 27, 2019 at 9:59
  • "sc." is a Latin abbreviation for scilicet 'specifically' or' namely'. The symbols * and ?? mean ungrammaticality and marginality, respectively. So for example *The door broke by John is ungrammatical (cf. ok. The door broke by itself). An Spanish example of a verbal eventive passive like La puerta fue cerrada (por alguien) 'The door was closed (by someone)' can be compared to an adjectival resultative passive like La puerta estuvo cerrada (*por alguien) 'The door stood closed (*by someone)' (NB: Sp. estar is a copulative verb similar to be but with an aspectual meaning added).
    – Mitomino
    Commented Aug 27, 2019 at 17:01

2 Answers 2


You have already explained the difference between verbal and adjectival. Another useful distinction is between passive and middle.

The middle is a voice between passive and active where something does something upon itself. Latin has no separate middle forms, but the Latin passive can have middle meaning. For example, fractum est = "it broke itself" = "it broke". It seems that what you call anticausative is very close to the middle voice. In Latin middle can be seen as a semantically slightly different reading of passives.

As with passive, the participle in middle reading can be either verbal or adjectival. I would split your third point in two:

  1. Verbal middle: "the door broke"
  2. Adjectival middle: "the door is broken" (in itself, not by anyone)

The difference between 3 and 4 is exactly the same as that between 1 and 2. Now both are middle, not passive, so the door breaks in itself without any kind of agent. The distinction between verbal and adjectival reading of participles makes sense for all voices, not just passive.

Finally, I have to remind that the two distinctions (verbal–adjectival and passive–middle) are not always clear at all. While these distinctions are useful, do not try to forcibly classify every occurrence in a clear category. The middle voice as a separate entity is found in Greek, but not really in Latin. Doing something upon oneself is often best achieved with se/sibi.

As Draconis points out in a comment, Proto-Indo-European did not distinguish the middle from the passive voice, and Ancient Greek barely did. In Latin the two voices never separated in the first place.

And more importantly, the passive voice in Latin does not imply that the there is an agent. (In Finnish it does, but the Finnish passive is very different anyway.) Perhaps a "fully agentless passive" is another way to phrase what you call anticausative and I call middle here. As Tom Cotton reminds in a comment, it is all too easy to be drawn into a detailed analysis that has no bearing on actual usage. Different points of view are useful in analyzing various uses of the Latin passive, but only as rough guides, not as strict categories. I would actually say that porta clausa est is not really ambiguous, but it can be interpreted with different nuances (four in my counting) — but I recognize that some may disagree.

  • 3
    I think it a mistake to try describing Latin in terms similar to those reserved to Greek, in which we distinguish a middle (rather than medium or medial) voice whose sense in Latin (as in English) is achieved in other ways. It's sometimes all too easy to be drawn into this kind of minute analysis with its clumsy descriptive terms, forgetting as we do so that the desirable objective called 'clear communication' can easily be turned into its very antithesis in that way. But +1 for the answer.
    – Tom Cotton
    Commented Jan 3, 2019 at 12:05
  • @TomCotton I agree; the middle is a more useful concept for Greek than Latin. I added details on the Latin middle and the perils of overanalysis. Also, thanks for the reminder about the name. I tend to forget to call it 'middle' instead of 'medium' due to the terminology used in Finnish and my limited exposure to Greek.
    – Joonas Ilmavirta
    Commented Jan 3, 2019 at 13:02
  • 1
    It might also be worth adding that Proto-Indo-European didn't distinguish the middle from the passive voice, and Ancient Greek barely did. It's not that they merged in Latin so much as that they never separated in the first place.
    – Draconis
    Commented Jan 4, 2019 at 4:35
  • 1
    @Mitomino I added an explanation under 3 and 4.
    – Joonas Ilmavirta
    Commented Jan 4, 2019 at 11:01
  • 1
    Yes, as you'll have seen from most of my questions, as a theoretical linguist, I'm also interested in the grammatical approach to Latin. Indeed, the grammatical perspective should not be necessarily incompatible with a usage-based one. In any case I fully undertand that most of people here are interested in Latin grammar if and only if it has consequences for usage. But in my professional discipline some people (mostly those ones who adopt a formal/generative approach to grammar) think that "grammar is grammar, and usage is usage": e.g., cf. muse.jhu.edu/article/50267/summary
    – Mitomino
    Commented Jan 5, 2019 at 20:32

It isn't clear to me that it makes sense to speak of an "anticausative reading" of the Latin passive. The English anticausative construction, as I understand it (I'm more used to seeing this called "unaccusative", but maybe you're drawing a distinction there?), rules out specifying an agent in the syntax, but it does not rule out that an agent may have been involved in the event. Consider:

  1. The boy punched the window and it shattered.

Here, shattered is anticausative, but nevertheless the boy is the agent of the shattering. The fact that *"The window shattered by the boy" is ungrammatical is due to the syntax of the English construction, not its semantics. But the Latin passive has no such constraint: if there's an agent, you can add an agent PP with a(b).* Of course, when such a PP is absent, there can be two readings according to whether you think that an agent was involved but is unspecified, or that no agent was involved at all. But it seems imprecise to call the latter an "anticausative reading", given that the English anticausative doesn't imply the semantic absence of an agent.

Your question about fracta est in the Horace quote seems straightforward, though: this is clearly a description of an event, not a state (as postquam strongly implies), so it's a verbal rather than an adjectival passive.

In a comment on Joonas's answer you ask "if Classical Latin ... can express the difference between The door was broken by the wind and The door broke with the wind". To me the latter is ungrammatical, but it can be repaired with a different choice of preposition, e.g. The door broke because of the wind. But this seems truth-conditionally equivalent to The door was broken by the wind -- as far as I can see there isn't a difference there to be expressed.

*As a possibly interesting side note, Greek is even less constrained in that it allows an agent PP with unaccusative verbs that are not formally passive: e.g. ἀπέθανεν ὑπ᾽ Ἀθηναίων "he was killed (lit. died) by the Athenians". This is rather as if English allowed *"The window shattered by the boy".

  • I agree that in your example The boy punched the window and it shattered, the verb shattered is anticausative (and yes, probably, as you say, it is unaccusative rather than unergative). However, I disagree with your claim that “nevertheless the boy is the agent of the shattering”. Truth-conditionally speaking, your claim is not wrong but, crucially, in your example the boy has not been linguistically construed as the agent of shattered.
    – Mitomino
    Commented Jan 4, 2019 at 4:36
  • That’s why you cannot continue your example with a final clause or with an instrument: The boy punched the window and *it shattered with a stone/in order to…. If there was an agent, the final clause and the instrument modifier would be expected to be licensed, contrary to fact. Cf. the previous anticausative construction with the following verbal passive: ok The boy punched the window and it was shattered with a stone/in order to…
    – Mitomino
    Commented Jan 4, 2019 at 4:36
  • As for your note on Greek, I consider it not only, as you say, “possibly interesting” but indeed extremely interesting! This point is for example addressed by Alexiadou & Anagnostopoulou (2008). lingref.com/cpp/wccfl/26/paper1653.pdf In particular, see Section 5 (“Differences between Greek and English/German”). However, I must say that I am not convinced by their analysis. But it does contain food for thought.
    – Mitomino
    Commented Jan 4, 2019 at 4:39
  • @Mitomino I'm not sure we're in disagreement about anything. The boy cannot be introduced syntactically into the clause as an agent, but semantically he is clearly the cause of the event. In any case, where English has two constructions (passive and anticausative) Latin has just one (passive), which does not necessarily imply an agent but can express one with a PP.
    – TKR
    Commented Jan 4, 2019 at 4:46
  • 1
    Yes, TKR, I'm very grateful to your comments since you've addressed the point I raised. That's why I gave an upvote to your answer. We agree that Porta clausa est is triply ambiguous but we disagree on to what extent the syntactico-semantic structure of my third reading can be collapsed with the first reading. Importantly, we agree on the verbal nature of the participle of first and third readings (vs. the second adjectival resultative/stative one). And finally I insist that your point on Greek is VERY interesting!
    – Mitomino
    Commented Jan 4, 2019 at 5:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.