Voice, lexical meaning, and deponency

What is voice?

In her introduction to voice in De Gruyter’s 3-volume handbook on syntax, Edit Doron begins her discussion by noting that voice has been regularly associated with alternations in the arguments of a verb, and her initial example comes from Attic Greek (2015:749–750): λούσω ‘I will wash (somebody)’, λούσομαι ‘I will wash (myself)’, and λουθήσομαι ‘I will be washed’. After noting nuanced differences on how to define voice, she then gives her own definition: “voice denotes changes in the grammatical function of the so-called external argument (typically the subject of the active verb), including the reduction of this argument” (italics original; 2015:750). Let us unpack this definition with the example she has given.

First, Doron says that something happens to the external argument of the verb. The notion of an external argument goes back to Williams (1981) who divided arguments of a verb into a binary: internal and external. Internal arguments are normally affected by the action of the verb whereas external arguments typically perform the action of the verb. They correspond to arguments that are within the verb phrase (internal) and those outside of the verb phrase (external). In the verb λούσω, the argument that performs the action of the verb (the external argument) would be the person who will wash (the “washer”), while the argument that is affected by the washing event (the internal argument) is the person who is being washed (the “washee”).1

Second, Doron says that the external argument undergoes changes in its grammatical function. In other words, we should see differences in how the washer functions grammatically in middle and passive voice. In middle voice, the washer remains the subject but is also interpreted as having the same semantic role as the grammatical object in active voice, namely the washee. Thus, the external argument has two semantic roles in middle voice (at least with the verb λούω ‘wash’), as both washer and washee, in contrast with the singular semantic role of washer when the external argument is the subject in active voice. In passive voice, the washee is the grammatical subject of the verb, and the washer, the external argument, is suppressed. Doron refers to this as the “reduction of this argument.” It should be noted that we would not call the subject of the passive voice an external argument. It remains the internal argument of the verb and has the same semantic role (as “washee”), but its grammatical function in the sentence changes from object to subject when moving from active to passive voice.

The system introduced by Doron is a three-way contrast between active, middle, and passive. If we look at voice typologically considering more of the world’s language (as Doron does in her article), we would find that some languages have more fine-grained voice distinctions than Greek and some have less. Even within NT Greek, we find that voice is not consistent across the different verb forms:

> More precise syntactic definitions of internal and external arguments are given by Williams (1981:84): “the external argument of a lexical item is located outside of the maximal projection of that item.” Since we have not introduced the idea of maximal projection (or the many syntactic building blocks that are presupposed in that notion), I rely upon the semantic definition here. Internal arguments, in contrast to external arguments, would be inside the maximal projection of the lexical item, namely the verb.

From a morphological standpoint, the medio-passive paradigms look more like the middle forms of Luso and the Aorist rather than the passive (and they are sometimes called middles because of this). However, we need to be careful not to equate form with function. In other words, the -ομαι paradigm might not be the same in Luo, where it is only contrasted with -ω (act), as compared to Luso, where -ομαι is contrasted with both -ω (act) and -θη (psv). Indeed, there is a difference in the interpretation of “middle” morphology when there is a 2-way contrast rather than a 3-way contrast. In the 2-way contrast, a passive interpretation is readily available with the middle forms, as evidenced by the possibility of adding in a by -phrase with ὑπό (the hallmark of a true passive form):

(284) Passive interpretation with medio-passive form

ὁ κύριος διήνοιξεν τὴν καρδίαν προσέχειν τοῖς λαλουμένοις ὑπὸ τοῦ Παύλου (Acts 16:14)

The Lord opened her heart to pay attention to the things spoken of by Paul

While this is a regular use of the Luo and Leluka medio-passive forms, the middle Aorist and Luso forms are not used with ὑπό introducing the agent of the event. Thus, we have labelled the -ομαι paradigm a medio-passive (since it can have a passive interpretation) when there is a 2-way contrast, and we have labelled it a middle (and not a passive) when there is a 3-way contrast. The precise interpretation of the voice forms, whether active, middle, or passive, is highly dependent upon the contrasting voices available as well as the meaning of the verb that voice combines with (on which, see below).

The description of voice we have given defines it in terms of how the argument of the verb changes grammatical function. It is thus concerned with linking the grammatical function(s) of a verb’s argument(s) with the morphological reflection we see in the verb itself. Whereas the ending -ω provides us with one grammatical function of the external argument, the ending -ομαι provides us with a different grammatical function of the external argument. This kind of definition, however, does not adequately account for cases where the argument of a verb does not change, but there is still a difference in morphology. To illustrate, the verbs πορεύομαι ‘go’ and περιπατέω ‘walk’ are both intransitive motion verbs, and they differ with respect to voice. Since neither verb has a voice counterpart in Koine Greek (πορεύω and περιπατοῦμαι are not found), there is no change in grammatical function of the external argument with a change in morphology, there being no change in morphology. Between the two verbs though, there is indisputably a morphological distinction, one being medio-passive and the other active. I argue below that this is due to the nature of the verb selecting for either an internal argument or an external argument. If this suggestion is correct, voice does not only reflect intra-lexical changes in the function of the external argument (i.e. λούσω vs. λούσομαι), but it also reflects inter-lexical differences in verb meanings (i.e. πορεύομαι vs. περιπατέω). Voice morphology is not random – it is a window into the meaning of verbs.

Agency, voice, and verb-meaning: βαπτίζω

In this section, we discuss the relationship between verb-meaning and voice by focusing on the verb βαπτίζω. Beth Levin begins her seminal work English Verb Classes and Alternations with the following statement: “This work is guided by the assumption that the behavior of a verb, particularly with respect to the expression and interpretation of its arguments, is to a large extent determined by its meaning” (1993:1). For students of an ancient language, this is an incredibly important statement (already hinted at in Semantics A). If the meaning of a verb determines (at least to a large extent) its syntactic behavior, we should find a correlation between the two. The reason why this is so significant for students of ancient languages in particular is that we have syntactic data, but we do not have semantic data. True semantic data would consist of native speaker intuitions about a particular verb, especially negative data, i.e. statements like “you cannot say Paul baptized with the meaning Paul baptized himself .” Such data would, of course, be invaluable for textual analysis. Though we lack this kind of semantic data, we do have syntactic data, and if Levin is correct that the two correlate, we can use syntactic data as a window into meaning. In other words, it should be possible to move from a verb’s syntax to its meaning. To demonstrate this idea, we begin with English:

• (285) (a) The child/?Natural causes broke the vase.
• (b) The vase was broken (by the child/?from natural causes).
• (c) The vase broke (*by the child/from natural causes).

Both the (285a) and (285b) examples are fundamentally similar in that the external argument is interpreted as an agent, i.e. as a person who is intentionally acting. In (285a), the external argument/agent is given explicitly in the grammatical subject the child, while the agent is implicit in (285b) and can be added back in with a by -phrase. In contrast, (285c) does not allow for the addition of the by -phrase, suggesting that the external argument (the agent) is not present. This contrast shows that the event referred to by (285a–b) differs from (285c)—in the active and passive variants of break an agent is involved in the breaking event, whereas there is no agent in (285c). What we see allowed in (285c), and what is not as natural in (285a–b), is a causer adjunct introduced by a from -phrase. This causer adjunct is not readily allowed as the subject of the active voice verb, so it is not an external argument. In these examples, voice reflects different kinds of breaking events – in (285a–b), there is an intentional breaking event that occurs, while in (285c), there is a breaking event that occurs with an unspecified causer.

Our suggestion is that actives and passives represent fundamentally similar events in that both have agents involved in the event, even if the agent is implicit in the passive (at least, that is the interpretation of the voice variations with break). Whether or not an agent is present may have exegetical significance. In 1 Cor 1:13–17, Paul is in the middle of lamenting the Corinthians’ divisions, and he thanks God that he did not baptize any of the Corinthians himself (other than the exceptions he lists).4 The verb βαπτίζω is used 5 times in 1 Cor 1:13–17 and alternates between passive and active voice. In the Corinthian church, there seems to have been factions created over who was physically involved in baptizing particular individuals. Some were saying that they were Paul’s followers, others Apollos’s, others Cephas’s, and others Christ’s (vs. 12). Paul is glad that he was not the agent of baptism for many people (vs. 14) because he did not want anyone to be following him (vs. 15), suggesting that some within the predominantly gentile church may have misunderstood something about baptism – the person performing the rite was not the leader to whom the “baptizee” should then be faithful. Rather, the baptizee was called to be faithful to the one whose name he or she was baptized into, not the agent of the physical act of baptizing (though those who were “Christ’s” seemed to understand baptism correctly). Thus, the passive forms of βαπτίζω in 1 Cor 1:13, 15 imply an agent just as the active forms of βαπτίζω specify an agent in 1 Cor 1:14, 16–17. The event structure is the same for active and passive voice verbs.

The middle, however, shows a different event structure. There are only two examples of βαπτίζω in the middle voice in the NT: Mark 7:4 and Acts 22:16. In Mark 7:4, the reference seems to be to the Pharisaic practice of eating non-sacred food in a state of purity, which is a “tradition of the elders” and is not required by the Torah. When the Pharisees came back from the market, they would “baptize” before eating. Given what we know about ritual purity, we can assume that the example in Mark 7:4 refers to people “baptizing” themselves when returning from the market, so that they ate in a state of purity. Thus, the middle is interpreted as a reflexive, one of its most common functions, and it refers to the event of ritual cleansing rather than being baptized by someone else into their group. In Acts 22:16, Paul is recounting to fellow Jews how Ananias told him that he should be baptized and have his sins washed away after having his sight restored. I suggest that Paul’s description in Acts is best understood as ritual self-immersion rather than being baptized by someone else. Paul’s direct speech in Acts 22 differs from Luke’s narrator-description of the event in Acts 9:18 where we find the passive verb ἐβαπτίσθη (where we can assume that Ananias baptized Paul). Why would Paul have changed his language? It seems that Paul wanted to be careful about how the mob would think about him being “baptized.” Ritual immersion, which would have normally been a reflexive act, would have been mainstream Jewish practice, but being baptized by someone else into the early Christ-community might not have gone over well with the mob that wanted to kill him. The middle verb βάπτισαι in Acts 22:16 presents the event as a reflexive and presents Paul as a Jew concerned with purity and not one who is baptized by someone else into another group. There might be various reasons why Paul portrayed the event in this way (including focusing on the important points of commonality between Paul and his Jewish audience), but the grammar forces us to conclude that Paul is presenting his baptism as self-immersion here.

The above discussion has focused on one verb, βαπτίζω, and how voice interacts with verb meaning. The NT data suggest that the verb in the active and passive voice means to be baptized by someone else, and this involved a transfer of allegiance from one group to another. The middle voice βαπτίζω referred to self-immersion, which was used in Jewish ritual purification. This discussion, however, is meant to demonstrate a larger point: voice can help us to determine lexical meaning. In the next section, we focus on motion verbs to see how this works for a class of verbs.

Unaccusatives and unergatives: Motion verbs as a test-case

As mentioned above, the verbs πορεύομαι ‘go’ vs. περιπατέω ‘walk’ are close in meaning, but the former is medio-passive and the latter is active. Although both are intransitive, the two verbs do not necessarily take the same kind of argument. One could take an internal argument that is similar to the object of a transitive verb, e.g. Κρίσπον in Παῦλος ἐβάπτισε Κρίσπον ‘Paul baptized Crispus.’ The other could take an external argument, which would be more similar to Παῦλος in the same sentence. Remember, we defined internal and external argument by which argument was subject (external) and object (internal) in a transitive sentence, so if the verb is intransitive, we do not know in principle whether the single argument of the verb should be analyzed as an external argument or an internal argument. In fact, these intransitive verbs have been given different labels in linguistics: verbs with a single internal argument have been called unaccusatives, and verbs with a single external argument have been called unergatives. How, then, do we determine whether πορεύομαι and περιπατέω are unaccusatives or unergatives?

To answer this question, we can compare the properties of typical subjects in a transitive verb with the properties of typical objects. In Παῦλος ἐβάπτισε Κρίσπον, Paul physically engages in the event named by the verb, and he causes something to happen to Crispus. Thus, Paul is an intentional causer who affects change in someone else. Crispus, on the other hand, does not do anything in the event. He is immersed by someone else, and he changes state, going from not wet to wet. We can now compare these properties with the properties of the single arguments in πορεύομαι and περιπατέω. If the argument has properties like Παῦλος (physical involvement in the event, volitionality, causing a change of state, etc.), it is more likely an external argument, and if the argument of the verb functions more like Κρίσπον (passive participant in the event, no volitionality, undergoing a change of state), it is more likely an internal argument.

In a verb like περιπατέω, the subject is necessarily an intentional agent, which makes it similar to the subject of ἐβάπτισε. Moreover, the subject is physically engaged in the event, and though some kind of movement is assumed, a change of location is not possible with this verb on its own. It does not combine with the prepositions εἰς and πρός to refer to the subject coming to a different place. Rather, it takes the preposition ἐν to refer to the location in which the action takes place, as in ἐν τῷ ἱερῷ περιπατοῦντος αὐτοῦ ‘as he was walking in the temple’ (Mark 11:27). Whenever it is used in descriptions of movement to another location, it is accompanied by another verb like ἔρχομαι (Matt 14:25, 29; Mark 6:48), suggesting that it cannot contribute the idea of change of location on its own. Each of these semantic features suggests that the single argument of περιπατέω is an external argument, similar to the subject of the transitive verb ἐβάπτισε. This verb is, therefore, an unergative.

The verb πορεύομαι, on the other hand, has a different set of properties. Although the subject of the verb is normally an agent, there are examples where an inanimate thing is the subject, such as βολίς ‘arrow’ (Wis 5:21), θεὸν μάταιον ‘worthless god/idol’ (Wis 15:8), and ὕδωρ ‘water’ (1 Enoch 26:3).10 These examples suggest that the verb does not require an agent, so if we see agentive properties in an example (such as intentional engagement in a physical event), this does not seem to be coming from a requirement of the verb but from the nature of the subject itself. In other words, πορεύομαι allows agentive subjects, but it does not require them like περιπατέω. Again unlike περιπατέω, πορεύομαι necessarily refers to a change in location and regularly takes the prepositions εἰς and πρός to refer to the final location of the subject, similar to how the object of ἐβάπτισε necessarily undergoes a change of state. Each of these semantic features suggests that the single argument of πορεύομαι is an internal argument, similar to the object of the transitive verb ἐβάπτισε. This verb is, therefore, an unaccusative.

This brief analysis divides motion verbs into two broad classes: manner of motion verbs which require an agent and do not entail a change of location, and direction of motion verbs which do not require an agent and do entail a change of location. We cannot here engage in a comprehensive survey of motion verbs in Greek, but a broader consideration of motion verbs reveals much the same. The active voice verb τρέχω ‘run’ requires an agentive subject and refers to how one moves (manner), whereas the medio-passive ἔρχομαι ‘come’ refers to the direction (generally towards the speaker) and again allows for non-agentive subjects (e.g. σκάνδαλον in Matt 18:7; Luke 17:1). Even near-synonyms like ἅλλομαι ‘spring up’ and ἀναπηδάω ‘jump’ can be distinguished on this basis, the former allowing non-agentive subjects (John 4:14) and there being no evidence for the latter allowing the same. This is not to say that all direction of motion verbs are medio-passive, but there is a much stronger tendency for intransitive verbs with internal arguments to surface with middle or passive voice rather than active. Likewise, active voice is normally used for verbs with external arguments which display properties associated with the subject of a transitive verb rather than an object. Again, the utility of this is that it helps us to determine verb-meaning with less data and to predict the meaning of verbs based on their morphological and syntactic characteristics.

For the sake of clarity, it is worth distinguishing this brief outline of voice from what might appear to be a similar system as outlined in Mathewson (2021). According to Mathewson, who closely follows ideas put forward by Stanley Porter, voice grammaticalizes different kinds of causation: active voice grammaticalizes “direct causality”, middle voice “internal causality,” and passive voice “external causality” (2021:52). These descriptions are, of course, only as accurate as the definitions ascribed to direct/internal/external causality, which are as follows (following O’Donnell 2005:371; italics original):

Direct causality is meaning of the active voice form: the cause of the process is attributed to the actor of the process (grammatical subject for finite forms or logical subject for nonfinite forms) of the verbal form.

External causality is the meaning of the passive voice form: the cause of the process is attributed to some external entity, which may or may not be grammaticalized in the immediate clause structure or the surrounding co-text. Internal causality is the meaning of the middle voice form: the cause of the process is attributed to elements within the process itself, in which the actor is involved.

One of the main issues with these definitions is that what it means to be a “cause” is never defined, and the philosophical and linguistics literature evidences that defining causation is no small matter. Besides this general point, not all verbs

10 There is one example where δένδρον is the subject of περιπατέω, but in this case, a now partially-blind man describes people as looking like walking trees. It is not as if real trees walked.

12 Although the brief analysis of voice given here has some points of contact with the system based on causation outlined in Mathewson (2021), it is not the same. Mathewson’s description does not capture the data in that the definitions of causation given (the supposed meaning of the voices) does not always map onto the morpheme.

are causative, but all have voice. The verb ἔχω ‘have’ does not have a subject that is the cause of a process because it refers to a state and not a causative process. Thus, the subject in δαιμόνιον ἔχει ‘he has a demon’ (Matt 11:18) is not the cause of having a demon (there is no cause in states like this, but if one were to argue for one, it would be the demon and not the person possessed by the demon who caused the possession). The definition of external causality faces the same problems, as well as others – the participle in ἐγερθεὶς ἀπῆλθεν εἰς τὸν οἶκον αὐτοῦ ‘he got up and went to his house’ (Matt 9:7) does not have an external causer, but the causer is the subject (i.e. ἐγερθείς should be semantically classified as direct causality). The notion of internal causality also does not fit the data. A verb like πορεύομαι involves the subject moving from one location to another, but there is no element within the process of moving that causes movement outside of the agent. Even with non-agentive subjects (on which, see below), it is not as if something in the process of moving causes movement – something external to the process must cause the process. Although voice and causation are linked in several important ways, the description given in Mathews (2021) is too simplistic and does not capture the data.

Deponency as the morphological expression of a single internal argument

We have already mentioned deponency and how some Greek scholars have rejected the term (8.2.3), and Campbell (2015:97) notes a general consensus to abandon the term at the 2010 SBL annual meeting. Scholars seem to take issue with the Latin term “deponent” meaning to “set aside” stating that the middle voice has not truly “set aside” the active voice, there having been no active voice to begin with (see especially Campbell’s 2015:93 description of Bernard A. Taylor). There are also arguments that deponents are “true middles” and are not active semantically, as defined by Wallace (1996:428): “A deponent middle is a middle voice verb that has no active form but is active in meaning ” (italics original). In my reading of the debate, this seems to be the main issue – deponent is used by some scholars to refer to middle voice with active meaning, but other scholars contend that these purported “deponent” verbs have a middle voice meaning rather than active voice meaning. Because deponency has been associated with “middle in form, active in meaning,” the term has been rejected.

However, linguists have defined deponency differently. Artemis Alexiadou, possibly the foremost Greek syntactician in the world, defines deponents as follows: “intransitive verbs which lack active forms” (2019:107).14 Throughout her article, she cites numerous other linguists who use the term deponent with no reservation and no misunderstanding of its meaning. Deponents are not middle in form but active in meaning; they are middle (or passive) in form with no active counterpart form. Defined as the linguistics world defines it, the category itself is simply a fact about forms. Rejecting the term cuts off NT Greek grammatical studies from the wider world of linguistics which does use the term in a meaningful, coherent way that helps us to talk about a category of middle or passive verbs that share a certain feature, namely not having an active counterpart. It is an “advancement” in the field to recognize that deponents are middle in meaning, but it is not an advancement to get rid of a term used in the field of linguistics because some NT Greek grammarians have defined the term wrongly. If anything, we should go back to the linguistics literature to define our terms more carefully.

Our preceding discussion of the difference between internal and external arguments and how those map onto transitive and intransitive verbs hints at how deponents should properly be understood. Middle deponents are “true middles” in that they behave semantically like other middle verbs. Although we have not given a comprehensive definition of “middle meaning,” we have shown that middle verbs refer to an event that is distinct from active verbs, and the arguments associated with middle verbs have distinct properties. These properties align with internal arguments of transitive verbs, and

> Proto-agent characteristics:

> Proto-patient characteristics:

> a. Undergoes change of state this is why the verb surfaces with middle or passive morphology. The idea is that a verb like ἔρχομαι has a root √ερχ that refers to an event of change of location without a causer. Since this kind of event involves a change in the argument of the verb and this property is normally associated with the object of transitive verbs, the verb has a single internal argument. In transitive verbs like βαπτίζω, internal arguments like the baptizee would only appear as the grammatical subject in middle or passive voice (e.g. Crispus was baptized has Crispus as the subject). Since ἔρχομαι only has an internal argument, its subject surfaces with medio-passive voice rather than active, just like the internal argument of a transitive verb. Similar logic holds for why περιπατέω would be in active voice and not medio-passive – the root refers to an event that requires an agent (like the subject of the active voice Paul baptized Crispus) rather than something that necessarily changes location, so this verb uses active voice to introduce its subject. Both ἔρχομαι and περιπατέω are a true medio-passive and active verb respectively. The meaning of each root determines which voice the verb surfaces with. Some roots have a meaning that allows for only a single internal argument and no external argument. At least some of these verbs surface with only non-active morphology, since that is how internal arguments are introduced as subjects. These are deponent verbs.