The present contribution focuses on the behavior of the prefix wy- in Polish, as illustrated in (1a)-(1b). This prefix is associated with an unlicensed postverbal NP, a fixed syntactic frame and a specified meaning, all of which call for a constructional analysis. In terms of its syntax and semantics, the Polish prefix is similar to the prefix er- in German (example 1c) and a number of verb particles in other languages. These and other such examples will be viewed as instantiations of constructions related to the numerous family of resultatives.

The above examples are taken to share the meaning “Subject obtains/produces Object by V-ing”. These sentences exemplify a special group of constructions, which will be referred to as the “Manner of Obtainment Constructions” (MOCs). MOC sentences contain morphologically complex verbs composed of a root and a prefix, which are otherwise common forms in both languages. Although most examples in the paper come from Polish and some from German, giving the impression that a single construction is being discussed, similar constructions exist in other languages, and these are governed by slightly different constraints. Therefore, the constructions under consideration will mostly be referred to in the plural.

To appreciate how verbs contribute to the interpretation and form of MOC sentences, it is helpful to briefly review some relevant facts about verbal morphology in the two languages. In German, verbal prefixes come in two types: separable and non-separable particles. The difference between the two types is that, in finite verbs, the former detach from the verb and move to the position following the object NP. For example, in main clauses, the verb aufmachen (‘open’) will split into the root machen (‘make’) and prefix auf, as in Sie machen die Augen auf (‘They open their eyes’). Prefixes of the other type remain attached to the verb, as in Sie zer brechen den Spiegel (‘They break the mirror’), where the prefix zer- does not move to the position after den Spiegel. But what both types have in common is that the meaning of a particle verb is different from that of the root alone, sometimes only slightly so, and sometimes quite considerably. The same is true of prefixed verbs in Polish, where a single verb root can give rise to a few dozen forms. One important difference between the two languages is that Polish lacks separable particles, so prefixes never migrate to the position after the object NP, and such non-separable particle verbs are found in Polish MOC sentences. Interestingly, German does not make use of its ample inventory of separable particles in MOC sentences and, like Polish, only employs non-separable particles.

The interpretation of obtainment is an integral semantic part of the meaning in the above examples. In Polish and German, MOC sentences do not convey meanings along the lines of “try to obtain”, but “succeed in obtaining”. For example, sentences (2a) and (2b) do not mean “fought in order to score a goal”, but “managed to score a goal by fighting.”

It is quite clear that the scoring of the goal is not a mere implicature but a true entailment, because it passes the test of non-cancelability. The obtainment reading cannot be canceled by saying (2c)/(2d).

For comparison, in non-MOC sentences, where obtainment is not entailed, but merely implicated, it is possible to cancel it; sentences (2e)-(2f) are by no means anomalous.

The prefixes wy- and er- are similar to English aspectual particles like up (Jackendoff 2002) or on (Toivonen 2006; McIntyre 2004) in that they combine freely with any verb meeting their selectional criteria and produce predictable semantic effects. The particle up is used to impose a telic reading on sentences with telicly ambiguous verbs which can take either telic or atelic temporal phrases:

Additionally, the presence of up in such sentences signals the completeness of the action expressed by the verb:

Similarly, the prefix wy-, when attached to a verb with an object NP, consistently signals the element of obtainment (parallel to the completeness of up) in the interpretation of the sentence and gives it a telic reading, as in (5).

In Polish, this Aktionsart feature determines the grammatical aspect of wy- verbs as perfective, with the action directed toward obtainment being interpreted as having ended.

The construction under consideration is also similar to the Time-away Construction (Jackendoff 1997, 2002). In both cases, the semantic head of the phrase is other than the verb (the verb merely names a manner modifier), and a non-subcategorized object follows the verb.

What these verbs have in common is that they are either intransitive or optionally intransitive, and when they do take complements, the category of possible objects is limited. The verb drink can take a limited number of objects, and so can fish; even if it is possible in theory to fish for any kind of thing, there is one traditional object that the verb selects. The same goes for dance and kiss. Consequently, when their complement is an understood object, these verbs often appear without it.

In the MOCs, too, only those verbs are allowed which are either intransitive like chodzic in Polish, wandeln in German (both ‘walk’), or optionally intransitive - those with “less than obligatory” implicit objects such as wedkowac (‘fish’) or tanczyc (‘dance’) (for a discussion of similar constraints in German, see also Kolehmainen 2006: 284). In (7), the postverbal NPs are not objects of the verbs lowic/wedkowac (‘fish’) or chodzic (‘walk’); these verbs name the manner in which the objects were effected.

The above properties of the MOCs are not captured by Generative Grammar principles. The telic interpretation of obtainment induced by means of intransitive verbs poses two problems. First, the meaning conveyed by MOC sentences cannot be derived entirely from the compositional contributions of the lexical items found in these sentences. Second, intransitive verbs taking object NPs raise obvious licensing questions. These two problems will be reviewed in turn.

Syntax has traditionally been assumed to be separate from semantics, and syntactic constructions have been taken to be devoid of meaning. This generativist assumption was exposed by Jackendoff (1997: 48) who called it “syntactically transparent semantic composition”. Under this assumption, one should be able to account for the meaning composition in sentences only in terms of individual semantic contributions of the lexical items found in those sentences. However, in MOCs, some part of the meaning of a sentence, indeed the most interesting part, seems to arise out of nowhere. The semantic element “obtain” is not featured in any of the lexical items composing MOCs; this meaning cannot be attributed to the object noun, the verb, the prefix alone, or even the prefix and the verb combined. If used in a sentence without an object, the prefixed verb in (8) does not carry the idea of obtainment, and this is true of both Polish and German.

It should be stressed that sentences with verbs like wyplakac or erweinen lacking objects would be anomalous both syntactically and semantically. They would be ungrammatical because wy- and er- are transitivizing particles and call for an object. But apart from being ungrammatical, such sentences would also be deprived of the obtainment interpretation. By comparison, when the verb put is used without an object, a sentence like The boy has put into the box will be ungrammatical, but its “putting” interpretation is preserved. For an er- or wy- verb alone to evoke an obtainment interpretation, a speaker would need to associate it with a similar er-/wy- verb followed by an effected object, and it is questionable that all native speakers will be able to make that association, especially with new or rare er- or wy- verbs like ermogeln (‘get by cheating’) or wyfaulowac (‘obtain by committing a foul’). When faced with an objectless wyplakal, Polish native speakers cannot provide even an approximate interpretation to the effect of “he obtained something by crying and begging, but I don’t know what it was”. Therefore, whenever wy- or er- verbs are glossed in this paper as ‘obtain by doing something’, these glosses should be taken as shorthand approximating the meaning of the whole verb phrase, not as meanings of these verbs alone.

To further illustrate the insufficiency of a verb and its particle alone to produce a meaning associated with a given construction, similar examples from English could be provided. For instance, in the Time-away Construction, the verb and the particle away (without an object NP) are not sufficient to convey the interpretation of “spend time (frivolously) by V-ing”. A sentence like Karen and George danced away will not be interpreted as Karen and George having spent the whole time dancing. This meaning does not appear until a postverbal time NP is added.

The second problem presented by the MOCs is that of licensing object NPs. Under the Projection Principle, intransitive verbs should not take direct objects. Intransitive verbs, such as the unergative biegac (‘run’) or skakac (‘jump’), do not have internal arguments in their subcategorization frames, so the Projection Principle should not build verb phrases containing obligatory direct objects. Although the verbs in question are preceded by transitivizing prefixes (which make them seem like legitimate object-taking verbs), I will show that they are not transitive prior to entering the constructions. Therefore, it is impossible to account for where their objects originate, if one considers only the subcategorization frame of the verb without assigning a subcategorization frame to the construction.

The extra-lexical meaning and licensing problems could, in theory, be dismissed if the wy- verb forms were recognized as separate lexemes, complete with “obtain” meanings and independent subcategorization frames, as illustrated in (9), allowing for the generation of VPs in full compliance with the Projection Principle.

This approach would require listing such meanings in the lexicon, a possibility advocated by Svenonius (2004: 227) who argues that when spatial or directional prefixes are used for metaphorical extensions, the meanings they yield are idiosyncratic, as is illustrated by the Russian example in (10).

Svenonius’ conclusion that such idiosyncratic meanings must be listed in the lexicon is contested in the present paper. It is argued that such meanings are not idiosyncratic, as they can be generated systematically by a rule associated with the MOCs. In fact, the Russian prefix vy- signals the exact same semantic function as is found in the Polish wy- and German er-, namely that of “obtain/produce x by V-ing”. The reason why Svenonius classified (10) as an idiosyncratic meaning is perhaps that the example was analyzed in isolation from other similar examples, and without reference to the rule responsible for generating such forms.

Thus, the “separate lexeme” approach comes at the price of postulating hundreds of potential wy- forms which are not even stored in the lexicon. The forms in question can be generated freely more or less the way English aspectual particle verbs can be added to verbs to produce forms like glue up, boogie on, or sing away. Although some of such forms, especially the more frequent ones, may be represented in the memory (Bybee 1985; Pinker 1999), it would be absurd to envisage all of them as separate items in the lexicon, including the myriads of potential combinations like the Polish wypogowac (‘obtain by pogoing’) or German erdribbeln (‘obtain by dribbling’). There is something implausible about the idea that lexemes, which are otherwise thought of as elements stored in the long-term memory, can be mass-coined on the spot, in working memory. MOC verbs are similar to regular past tense forms of low-frequency verbs like vitiated or malingered in that neither these nor MOC verbs need to be listed in the lexicon to be used or comprehended. Instead, they are assembled online more like inflectional forms or syntactic phrases rather than bona-fide lexical derivations.

The classification of the wy-/er- forms as derived independent lexemes is further complicated by the absence of a property normally observed in derivation, namely semantic opacity. In the process of derivation, a single morpheme may have a number of related but different word-formation effects with the resulting coinages standing in dissimilar relationships to their bases. For example, Plag (2003: 15) shows that adding the morpheme -ity to an adjective like curious yields two related but independent meanings (‘lust for knowledge’ and ‘object of interest’). A derivational morpheme does not have a single fixed derivational function identical in all instances of its use. On the other hand, it is hard to classify wy- verbs as derived lexical items, complete with their own subcategorization frames, because wy- is 100% regular wherever it combines with a verb. A brief survey of wy- verbs, those listed in dictionaries as well as novel verbs, shows a fixed semantic pattern. Established verbs such as wycwiczyc (‘obtain by practicing’), wymeczyc (‘obtain by tiring’), wypracowac (‘obtain by working’), wyprosic (‘obtain by asking’), or wysiedziec (‘obtain by sitting’), which are listed in major Polish dictionaries are all defined as expressing obtainment. Novel verbs, which are not normally included in dictionaries but can be found on the Internet, such as wymantrowac (‘obtain by chanting mantras’), wywioslowac (‘obtain by rowing’) or wyzonglowac (‘obtain by juggling’) also carry the obtainment interpretation which is clear from the context: whenever they are followed by an object NP, the entity expressed by that NP will be understood to be obtained in a manner specified by the verb.

More indirect evidence suggesting that the wy- verbs are not separate lexemes comes from the observation that languages impose constraints on the contents allowed within the verb meaning. Levin and Rappaport Hovav (1991) suggest that manner and result cannot be lexicalized at the same time in one verb. When a verb lexicalizes manner (to wipe, rub, etc.), it says nothing about the result (what is wiped does not necessarily become clean), and conversely result verbs do not specify the manner (the verb to clean says nothing about how the cleaning was done). If a wy- form were in fact a single separate lexical item, it would be in violation of the constraint, as every wy- verb clearly expresses both manner and result.

Another possibility to be considered is that the verb is polysemous. If a verb like think additionally meant “produce by thinking”, the form of the construction in question would arise naturally via the application of traditional argument linking rules. However, the polysemy approach seems misguided. It is convincingly ruled out by Goldberg and Jackendoff (2004: 533-534), who insist that in unusual uses, the verb retains its original meaning. To make a case for unchanging verb meanings, they provide an example of a rather unlikely meaning shift: the verb belch used in a motion phrase like He belched his way out of the restaurant (example (4a) in Goldberg & Jackendoff 2004). Quite simply, it is beyond belief that the verb belch is converted to a motion verb in the lexicon. Besides, the benefits offered by polysemy would be heavily outweighed by the burden to be shouldered by the lexicon. The number of meanings to be stored would not just double – meanings would have to be multiplied to accommodate these and probably many more constructions that will be encountered in future research.

A possible way to account for the unlicensed postverbal NP is an approach which Teun Hoekstra (1988) developed for resultative phrases. He insists that “the licensing of the complement structure in the resultative construction should be determined by general principles” (Hoekstra 1988: 101). To defend the explanatory completeness of GB-theory, Hoekstra had to contend with the question of licensing NP complements that are clearly not arguments of the verb, as in example (11).

He applied Stowell’s (1981) theory of small clauses (SC), syntactic units hypothesized to have clausal status, expressing a subject-predicate relation, but lacking a finite verb. For example, in the sentence They consider John intelligent, the phrase following the verb is analyzed as a small clause, where the subject John is selected by the AP head intelligent. The small clause itself is a complement compatible with the selectional requirements of the verb consider. Thus, the appeal of the SC analysis is that it helps keep alive the possibility that the complement structure of a verb phrase can be accounted for by the verb alone.

According to Hoekstra, the small clause analysis is equally effective when applied to resultatives. Unsubcategorized NP complements such as him in (11) cease to be problematic if they are viewed as subjects (and therefore rightful arguments) of small clauses. In other words, it is no longer necessary to attempt to justify the NP him as being selected by the verb; it appears on the strength of the invitation from the PP head under the table.

However, explaining NPs as complements of small clauses in resultative phrases merely pushes the question a step back, because an explanation is required for what licenses small clauses themselves. Although a small clause like John intelligent seems an uncontroversial complement of the verb consider (after all, this verb typically takes clausal complements), it is much harder to justify the small clause him under the table as a complement of the verb drink or intransitive verbs taking unlicensed resultative complements such as the verb sleep in sleep your wrinkles away. These are not the kind of complements specified in the selectional requirements of these verbs.

Having admitted that SC predicates are not licensed by the verb, Hoekstra concludes that

we do not need a specific rule to introduce the SC, thereby licensing its occurrence. Rather, we claim that each activity denoting verb may be combined with a complement that denotes the state resulting from that activity. We might formally implement this by providing each activity denoting verb with an optional result argument. Alternatively, we may postulate a lexical rule stipulating that a verb of the category ‘non-stative’ may have a result argument. (Hoekstra 1988: 131)

In other words, to salvage the unassisted functioning of the Projection Principle, the burden of licensing is transferred to the lexicon, a problem discussed earlier in section 2.1. Unfortunately, the economy of this solution is compromised by the need to similarly tag lexical entries for a host of other configurations with unlicensed complements, including the MOC constructions. Furthermore, even if lexical entries do in fact leave an empty slot for optional arguments as postulated by Hoekstra, that would still be insufficient to determine the form of the construction under consideration. The secondary predicates observed in the resultative and other related constructions are of detailed, often conflicting architectures, which are not sufficiently blueprinted by a mere “optional result argument” slot appended to verbs in the lexicon. The small clause observed in the MOCs differs from that proposed for the resultative, and that difference should be specified somewhere.

Note in passing that it is irrelevant whether the extra unlicensed material should be analyzed as a small clause or something else. It seems that under the assumption of syntactic transparency, no alternative treatments eliminate the problem of licensing postverbal NPs. For example, Levin and Rappaport Hovav (1995) rejected Hoekstra’s Small Clause approach on the grounds that its uniform representation disregards the type of verb used in the resultative construction. Under their analysis, licensing depends on the verb in the resultative construction, with some NPs being regular arguments projected by the verb (in transitive-based constructions), and other NPs (in unergative-based constructions) being unlicensed complements. Apart from a postverbal NP, the construction involves the addition of a resultative XP which almost certainly does not originate as a projection from the verb. Whatever its adequate analysis, the formation of the resultative has to be specified somehow either by postulating special lexical rules or by dropping constraints of syntactically transparent composition. Either way, as Goldberg and Jackendoff (2004: 534) observe for constructions in general, it is necessary “to abandon the rigid view that the verb alone determines the complement structure of its VP.”

The obtainment reading could arise as a result of an unpronounced affix attached to the verb. A similar idea for resultative constructions was proposed by Kratzer (2005), who postulated a covert marker to account for the causation interpretation. She notes that “[this interpretation] does not seem to be contributed by either the adjective or the verb alone. Do we have to conclude then that constructions all by themselves can introduce meaning components as specific as causal relations?” (ibid: 2) Kratzer answers in the negative; she claims that “[r]esultatives do not force us to assume that syntactic constructions or semantic composition rules can introduce non-logical meaning components like causal relations” (ibid: 2).

Reluctant to ascribe the causative meaning to the operation of the structure, she proposes an analysis where the NP originates in the resultative phrase via raising. To account for the event structure found in resultatives, Kratzer postulates an “unpronounced morpheme consisting of an interpretable feature [cause],” (ibid: 32) which serves to introduce an event argument. Thus, under Kratzer’s analysis, the untypical structure of resultative sentences follows from straightforward syntactic principles, without extra semantic or structural specifications carried by resultative constructions.

If this stance could be sustained, then a host of other unpronounced affixes could be proposed, including one with an interpretable feature [obtain] present in the MOCs or an affix with a feature [spend] responsible for the Time-away Construction. A point of contention would then be whether the unpronounced affixes should be regarded as a thing of the lexicon or syntax. If they can be shown to come from the lexicon, then Kratzer’s solution should be sufficient not only for resultative constructions, but also for any construction with a meaning having an unknown source. Simply, that meaning could be imputed to a hidden morpheme which projects its structure and interpretation onto the sentence.

To determine the provenance of unpronounced affixes, it is useful to consider some of their characteristics. Kratzer proposes that they attach to adjectives (in adjectival resultative phrases), so they are bound morphemes rather than independent lexical items. But are they more akin to overt derivational or inflectional affixes? Kratzer gives reasons to classify them as derivational affixes with specific constraints on the kinds of adjectives they can be attached to. She also stresses that not all types of causation are expressed, and the resultative adjectives express only those states which are brought about by direct causation.

This last fact, however, makes it hard to consider unpronounced [cause] affixes as derivational morphemes, as they do not exhibit the property of semantic opacity mentioned in section 2.1. The relationship which holds between the source adjectives and resultative adjectives is completely regular and predictable (it is always direct causation). Such word-formational regularity is typical of inflection, not derivation. In other words, Kratzer’s unpronounced affixes are unlikely to be lexical material; instead they seem to originate from syntax, and so would any other unpronounced affixes with interpretable features like [obtain] or [spend]. While this does not seem to be an impossible burden for syntax, this approach does not, in principle, differ much from a constructional stipulation which Kratzer claims to reject by arguing “in favor of a very restricted set of general composition principles which apply freely” (Kratzer 2005: 25).

Yet another way to account for the obtainment meaning without recourse to constructional explanations is to suppose that it is a semantic effect similar to the directness effect or wholeness interpretation associated with direct objects (Pinker 1989: 66). Briefly, a relationship holds between a verb and its direct object to the effect that the argument expressed by the object is understood to be affected directly or completely. For example, the difference between I loaded the wagon with hay and I loaded hay onto the wagon is that in the first sentence, the whole wagon is loaded with hay, and in the second, nothing is said about how much space is left. The directness effect is clear in sentences I had my hair cut (indirect causation) and I cut my hair (direct causation – the object immediately follows the verb). But unlike the directness and wholeness effects, the obtainment interpretation is a more local semantic effect, and there are no general rules of syntax saying that combining a prefixed verb with an object yields this meaning.

It has been observed in the literature (Talmy 1985, 2000; Slobin 1997) that when it comes to expressing motion and manner, languages encode motion paths in one of two ways. Languages like Spanish frame paths by means of verbs (25a); others, like English, frame motion paths by means of particles (such as in in into), also referred to as “satellites” (25b). This English option is not possible in Spanish (25c).

Thus with respect to this property, languages form two groups referred to as verb-framed languages (V-languages) and satellite-framed languages (S-languages) (Talmy 1985: 102; Slobin 1997: 439). An important consequence of the difference is that particles present in S-languages allow for the expression of path along with manner of motion in the same VP, an option unavailable for speakers of V-languages, where manner can only be expressed as an adverbial modifier in a non-minimal VP:

This difference was further specified by Slobin (1997: 441) who observed that what V-languages do not allow is conflating manner and path involving the element of “boundary crossing”. Motion along an unbroken path can be expressed next to the verb as in (27) expressing both path and manner, but the PP is an adverbial modifier, not a true complement in the event conflation:

When a boundary is crossed, V-languages stop the path particle being realized in a minimal VP. Slobin (1997: 441) observes that “[i]t appears to be a universal characteristic of V-languages that crossing a spatial boundary is conceived of as a change of state, and that state changes require an independent predicate.”

If there is a relation between resultative change of state, crossing a boundary and obtainment involving directional phrases, it should be reflected by cross-linguistic correlations to this effect. That is, one would expect V-languages to disallow conflating obtainment with manner (since they disallow conflating resultative meanings with manner) and S-languages to offer the option of conflating manner and obtainment by means of the same mechanisms as in literal motion event conflations described above. These expectations are confirmed by data coming from languages of the two types, as will be demonstrated in the subsequent discussion based on Slobin’s (2003: 162) division into the two groups.

It turns out that languages of the S-frame type allow for the simultaneous expression of manner and obtainment. Apart from the Polish examples listed before, similar forms are found in a number of typologically unrelated languages.

Similar examples for Finnish are provided in Kolehmainen and Larjavaara (2004: 17).

Of course, just how close the above forms are in terms of their syntactic distribution is an open question. It is quite obvious that the languages quoted impose different restrictions on the construction, and in fact the differences may be so great that it will be necessary to describe them as instantiations of different constructions. However, although they may turn out to be different, the constructions in these languages are not unrelated. The directional markers suggest that they have more in common than is possible by mere coincidence – in most (if not all) of them the causative interpretation of obtainment is conveyed by the metaphor of directional movement and boundary crossing. For example, the suffix pri- in Slovene expresses “bringing in” or movement “to oneself”, which is also the underlying function of the Chinese verb particle lai (28c). The Polish wy- conveying the meaning of “retrieving” or “pulling out” is similar to the Estonian välja (28d). Although at first look German does not seem to be a language with a directional phrase expressing obtainment, it should not be stricken from the sample prematurely. Elisabeth Rieken (pc) alerted me to the fact that er- is historically related to the Old-German uz-, us-, and ur- prefixes, all of which convey the meanings of “aus, hinaus, hinauf” (Krahe 1967: 39). Thus, although native speakers are not aware of the directional frame, it serves as a source domain for the obtainment reading. Whether the obtainment reading in Dutch (28e) is also metaphorically motivated is hard to establish with any degree of certainty, because no overt directional morpheme is used and it is a question for further research how productive such forms are in Dutch. English may be the odd one out among the S-languages, as it does not use a productive verb particle construction for the expression of obtainment meanings. The closest equivalent seems to be the X’s Way Construction, but this does not bear striking resemblances to the single-morpheme satellite particles found in the other languages. However, here too, obtainment is conveyed metaphorically by a directional phrase, as in Beckham dribbled his way to fame. Although more research is necessary to settle this question, as a preliminary approximation, the obtainment meaning can be assumed to manifest itself as effected objects “coming one’s way”.

The V-languages on the other hand do not express obtainment along the same lines. Manner verbs do not appear in causative frames and are not interpreted as taking constructional arguments; when faced with a hypothetical sentence like (29a), native speakers of Spanish consider it not only ill-formed but quite incomprehensible.

These correlations can additionally be argued to bear on the question of whether or not directional phrases are resultative in nature. Kratzer (2005) argues that resultatives should be separated from directional phrases, and shows that despite superficial similarities with the resultatives, directional phrases exhibit properties defying generalizations proposed for resultative phrases (like the Direct Object Restriction). This objection is probably valid for “pure directionals” which convey mere “toward” senses, but boundary-crossing directionals of the “into/out of” type are resultative in nature. Note that the division of S-languages and V-languages is at the same time a division into languages which possess and lack resultative phrases respectively.

I will now present a constructional analysis of the MOCs following Goldberg and Jackendoff’s (2004) approach to the resultative constructions. They argue for an analysis allowing the syntax of the resultative construction to be predicted on the basis of its semantics by breaking up the interpretation of idiomatic structures into two subevents, constructional and verbal. The interpretation of (30) is analyzed as consisting of two semantic subcomponents illustrated below.

According to Goldberg and Jackendoff, the form of the construction is derived from the semantics of the constructional subevent; the construction “borrows” the argument distribution properties from the deep verb MAKE found in the constructional subevent. This claim is formalized in point (31) (15 in Goldberg & Jackendoff 2004).

This approach can also be adopted for event conflation uses of motion verbs. The structure is generated on analogy with directional verbs like put or move. If it is assumed that (32) accurately captures the constructional and verbal subevents of sentences like Sven threw the ball into the basket, then the argument realization involving a PP complement is a straightforward consequence of (31).

A similar analysis is offered in Svenonius (2004), who points out similarities between Slavic verbal prefixes and Germanic directional particles. On his view, the Ground PP is a predicate triggered by the directional verbal prefix. As a parallel with Goldberg and Jackendoff, this predicative prefix is taken to originate from the semantics of the constructional subevent associated with the deep verb PUT, which also selects the PP complement. This approach also works well for Polish, which has the same patterning of the lexical prefix signaling a directional use of the verb.

In (33), the prefix w- functions as a particle introducing the PP do kosza, which is part of the constructional subevent “Sven put the ball₍ FIGURE ₎ into the basket₍ GROUND ₎”.

When applied to MOC sentences, the subevent approach captures their form, including the effected object NP. The argument structures of MOC sentences are analogous to the single-complement structure projected by the verb obtain, which provides the semantics of the constructional subevent. Consider the following sentences in German and Polish.

A constructional subevent based on the “obtain” meaning serves as a subcategorization frame specifying the arguments in the verb phrase. If it is assumed that the MOCs carry a deep verb OBTAIN, it should follow that those constructions come with a subcategorization frame dictating the form of a MOC sentence.

The above analysis leaves one question unanswered. It was argued earlier that the obtainment reading is a metaphorical extension of a path across a boundary involving agentive movement interpreted as AMOVE (“take out”). Why then is a directional phrase “out of a SOURCE” nowhere to be found in the constructional subevent? In fact, it cannot even be added optionally:

Although this question may require further research, at this moment it is tempting to explain the blocking of a GROUND PP complement in terms of the Unique Path Constraint: specifying a source of obtainment could represent a distinct path conflicting with the resultative element present in the interpretation of obtainment. If obtainment is resultative in nature, then it already consists of a path. That no further path phrases can be added should then be a natural consequence of the UPC.

The absence of the PP complement in MOC sentences is significant for the analysis of the construction. It turns out that the constructional subevent contains a predicate OBTAIN. Although this predicate should, at some level of representation, be decomposed into AMOVE and a directional, it is OBTAIN, not AMOVE, that dictates the semantics and the subcategorization frame of the constructional subevent.