See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/239563243

Benefactives in English: evidence against argumenthood

Article · January 2005

CITATIONS READS

5 546

1 author:

Tim Nisbet
Coventry University
2 PUBLICATIONS   7 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Tim Nisbet on 05 September 2017.

The user has requested enhancement of the downloaded file.

 Benefactives in English: evidence against
argumenthood

Tim Nisbet

School of Linguistics & Applied Language Studies, The University of Reading

Abstract. This paper addresses the problem of how benefactive clauses in

English involving an indirect object (Mary baked John a cake) are generated.
Two major hypotheses are considered in turn: (1) that Beneficiaries are simply
listed as ‘optional’ indirect objects in a verb’s subcategorisation frame (a view
defended by Emonds 1993); (2) that a variety of ‘lexical rule’ adds a
Beneficiary argument to a specified class of verb as proposed by Larson
(1990) and suggested by the general approach to lexical semantics of B. Levin
and M. Rappaport Hovav. I demonstrate that the evidence strongly
disconfirms the first hypothesis and that lexical rules either fail to be
adequately predictive or necessarily overgenerate ‘redundant entries’. I
conclude that another solution must be sought that does not depend on
Beneficiary NPs being either stipulated or rule-generated arguments.

1. Introduction

The question of how to generate the English indirect object benefactive has
always been a tricky one for generative theory (see for example Allerton
1978). In this paper I examine the evidence for two possible approaches to
a solution: (1) that benefactives involve ‘optional’ indirect objects of a
subset of verbs, as for Emonds (1993); (2) that benefactives are generated
by a lexical rule.
The paper is structured as follows. In section 2 I introduce the
syntax of benefactives in English, illustrating the superficial similarity of
the benefactive to the dative construction, facts which have undoubtedly
motivated those approaches in which Beneficiary NPs are (optional)
indirect objects of the verbs concerned. I then marshal the evidence against
this position and conclude that Beneficiaries cannot be listed in lexical
entries (section 3). In section 4, I examine two alternative ‘lexical’
hypotheses, one proposed by Larson (1990), the other modelled on the
lexical semantics framework of B. Levin and M. Rappaport Hovav, both of
which involve a rule adding a Beneficiary argument to a transitive lexical
entry. I conclude, however, that lexical rules are either too constrained and
stipulative or far too unconstrained to adequately account for the
phenomena illustrated.

Reading Working Papers in Linguistics 8. 51-67.

© 2005 T. Nisbet
52 T. NISBET

2. The syntactic form of benefactives

Benefactive clauses in English may take two syntactic forms, with the NP
interpreted as the ‘Beneficiary’ of the verbal event being realised either as
an immediately postverbal NP, as in (1a), or within a PP headed by for
(1b). Their syntax parallels the so-called ‘dative alternation’ displayed by a
large number of verbs with ‘Recipient’ indirect objects as in (2), the only
overt difference being in the selection of preposition to for the Recipient
NP (2b).

(1a) John found the man a map

(1b) John found a map for the man

(2a) John gave the man a map

(2b) John gave a map to the man

Purely for convenience, I will henceforth refer to the bolded NP in the (a)
clauses as the ‘nominal indirect object’ (NIO), and the corresponding NP in
the (b) clauses as the ‘prepositional indirect object’ (PIO). I call the clause
type of (1) a ‘benefactive’, that of (2) a ‘dative’, and where necessary I
refer to the roles of the bolded NPs as ‘Beneficiary’ and ‘Recipient’
respectively.
The syntactic parallelism of the benefactive and dative extends to
various movement phenomena, for example, a wh-NIO is resistant to
fronting (3), whereas an embedded wh-PIO freely accepts fronting (4).

(3a) ??Who did John find/buy a map?

(3b) ??Who did John give a map?

(4a) Who did John find a map for?

(4b) Who did John give a map to?

These striking syntactic parallels, together with the (virtual)

synonymy of the NIO-PIO pair of each type, strongly suggested parallel
syntactic derivation in the early days of generative grammar. In Fillmore
(1965), benefactives such as in (1a) are claimed to be generated by a
transformational rule operating on underlying strings which is essentially
 BENEFACTIVES IN ENGLISH 53

similar to one relating the datives in (2a) and (2b). It is taken for granted
that both the Beneficiary and Recipient NPs are selected items, Fillmore
(1965:11) noting merely that ‘the choice of the preposition seems to
depend on the particular transitive verb’.
Since the demise of transformational rules, a fairly general
consensus has been that the relationship between (2a) and (2b) is to be
captured not by a syntactic rule but ‘in the lexicon’ (following Oerhle
1976). That notwithstanding, there have been more recent attempts to
revive a derivational account of the dative alternation subsuming the
benefactive (Larson 1988; Emonds 1993): both these accounts assume, as
did Fillmore, that both the NIO and PIO of datives and benefactives alike
are selected as arguments by the verbs involved.1 In Emonds’s account, for
example, the Beneficiary appears in a subcategorised verb’s lexical entry
as an ‘optional’ indirect object, and the two structures are then derived by
the application of a ‘structure-preserving’ rule (see Emonds 1993 for the
technical details). However, several researchers, from different theoretical
perspectives, have held this assumption to be incorrect (e.g. Goldberg
1995; Jackendoff 1990a, 1990b; Wechsler 1995), and that contrary to
surface appearance, benefactives are unlike datives in many respects.2 In
the following section, arguing primarily against Emonds (1993), I review
the evidence for why Beneficiary NPs, in contrast to Recipient NPs of
regular dative verbs, cannot be treated as verbal arguments.

3. Evidence that benefactives and datives are nonequivalent

3.1 No evidence for ‘implicit arguments’ with benefactive verbs

Jackendoff (1990a:448) remarks that, although one can ‘sing someone an

aria’, ‘peel someone a grape’, etc., there is nothing in the monotransitive
uses of these and similar verbs that implies an intended Beneficiary:
‘…one could just be doing these things for the hell of it.’ We can test this
intuition by contrasting the implicatures of dative verbs with omitted

1
Larson (1990) recants on selection for benefactives, in the face of Jackendoff’s
(1990a) critique, suggesting a lexical rule which ‘augments’ the argument structure of
the relevant transitive verb. I discuss this proposal in section 4.1.
2
Other researchers, despite arguing that Beneficiary NPs are not selected arguments,
nevertheless end up stipulating that they are (mysteriously) ‘realised in the lexicon’
(Hoffman 1995), or that the verb involved somehow manages to ‘assign’ a Beneficiary
role (Marantz 1984). Such proposals are too question-begging to merit further
consideration here.
54 T. NISBET

indirect objects (5) and verbs that can take Beneficiaries but appear in their
monotransitive form (6) (I use the symbol # throughout to mean
‘semantically anomalous’, examples mine).

(5a) #Although John lent some books, he kept them all

(5b) #Although John passed some beers, he kept them all

(6a) Although John bought some books, he kept them all

(6b) Although John opened some beers, he kept them all

We see that the dative verbs lend and pass in (5) entail the existence of
some kind of Recipient even if it is not syntactically present in the clause:
it is part of the meaning of these verbs that the Theme moves away from
the Agent to or towards a Recipient. However, the absence of a Beneficiary
from the buy and open clauses in (6) results in no anomaly, since ‘opening’
and ‘buying’ minimally involve only an Agent and a Theme/Patient. In
fact, even when the Beneficiary NP is present, the verb does not entail
actual transfer of the direct object:

(7a) John bought Mary a book, but then decided to keep it

(7b) #John lent Mary a book, but then decided to keep it

(7a) means that John did buy a book (with the intention of giving it to
Mary), but Mary ended up not getting it. In contrast, the dative clause (7b)
cannot mean that the book was lent but that Mary never received it.
The conclusion is twofold: (1) that the Beneficiary is not a
participant in the event denoted by the meaning of the verb; (2) that the
notion of ‘transfer’ is part of the meaning of lend and pass but not of buy
or open.

3.2 Asymmetry of the benefactive alternation

A further problem for the assumption that Beneficiaries feature in lexical

entries is a syntactic asymmetry, which in fact reflects not properties of
particular verbs, but semantic properties of the Beneficiary itself, as I will
show.

(8a) I’ll fix a sandwich for you

 BENEFACTIVES IN ENGLISH 55

(8b) I’ll fix you a sandwich

(9a) I’ll fix the radiator for you

(9b) *I’ll fix you the radiator

Emonds’s (1993:239) explanation for these facts is that fix (‘make’) is

subcategorised for an optional indirect object, i.e. it can take a Beneficiary
object. However, homophonous fix (‘mend’) is simply transitive, and for
you in (9a) is a syntactic adjunct (not a sister to V) and cannot exchange
with the direct object the radiator (which is a sister to V) to become an
indirect object as in (9b).
First let us note that there is little if any unambiguous structural
evidence for a distinction between the PPs of (8a) and (9a). In a footnote,
Emonds (1993:255) claims that the structural difference comes out in their
respective tolerability under (1) preposing with a comma; (2) long distance
extractability. He offers no corroboration of this claim, but in the following
examples I am unable to detect any difference in acceptability:

(10a) For Mary, John bought a watch (= ‘object’)

(10b) For Mary, John fixed the radiator (= ‘adjunct’)

(11a) ?Who did John wonder whether to fix a sandwich for?

(= ‘object’)

(11b) ?Who did John wonder whether to fix the radiator for?
(= ‘adjunct’)

Emonds is, however, correct in stating that the two PPs in (8a) and
(9a) above differ in their interpretation: (8a) implies that the sandwich is to
be made and then given, whereas the benefactive of (9a) implies a service
or favour (for example, performed in someone’s stead). The semantic
distinction is reflected syntactically in that only (8a), with the ‘intended
recipient’ reading, is synonymous with the NIO in (8b): no other reading
of the Beneficiary licenses an NIO. But this distinction is insufficient
evidence on its own to justify treating (9a), but not (8a), as a structural
adjunct. Moreover, the distinction English makes is not universal: in
German, both interpretations of the Beneficiary permit either a PP or a
dative-marked NP:
56 T. NISBET

(12a) Ich mach’ dir ein Käsebrot

I make you-dat a cheese-sandwich

(12b) Ich mach’ ein Käsebrot für dich

I make a cheese-sandwich for you

(13a) Ich repariere dir den Radiator

I repair you-dat the radiator

(13b) Ich repariere den Radiator für dich

I repair the radiator for you

The evidence so far leaves us with the descriptive fact that in

English, the NIO benefactive is licit with only one subtype of Beneficiary
role, that of ‘intended recipient’. Further data show, however, that the
acceptability of this NIO depends not on whether a particular verb selects
it, but on the reading of the verb plus its particular direct object in the
clause. This means that the same verb can lead to different acceptability
judgements (example (14) is from Langacker 1987):

(14a) Joe cleared Sam a place on the floor

(14b) Joe cleared (*Sam) the floor

(15a) John cut Mary some string

(15b) John cut (*Mary) the string

(16a) John opened Mary a beer

(16b) John opened (*Mary) the door

Clearly, verbal subcategorisation cannot refer to such variables as the

reference of the direct object to the whole versus a part of something (14),
the choice of determiner (15), or the kind of thing referred to by the noun
(16). In any case, these examples cannot reflect selectional restrictions
imposed by the verb on the direct object, since the omission of the NIOs in
the (b) sentences renders them grammatical, as indicated. Yet, in contrast
to the behaviour of benefactive NIOs, the indirect objects of dative verbs
are acceptable regardless of the above factors:
 BENEFACTIVES IN ENGLISH 57

(17) Joe promised Sam {a place on the floor/the floor}

John gave Mary {some string/the string}
John showed Mary {a beer/the door}

If dative and benefactive verbs differ only trivially (i.e. in the

‘optionality’ of an indirect object in the subcategorisation of the latter),
then the above contrasts are unexpected, and unexplained on Emonds’s
account.

3.3 Only dative verbs idiosyncratically disallow ‘dative shift’

One very well-studied phenomenon concerning dative verbs is the inability

of many to undergo ‘dative shift’, i.e. to allow an NIO in addition to a PIO.

(18a) John {gave/lent/donated/returned} the book to Mary

(18b) John {gave/lent/*donated/*returned} Mary the book

There have been numerous attempts to explain the restriction on the

nonalternating verbs, usually by recourse to verbal semantics or putative
differences in theta-role assignment (e.g. Larson 1988; Pinker 1989;
Wechsler 1995). In his approach, which otherwise relies on
subcategorisation to account for the syntactic behaviour of verbs, Emonds
(1993:256) proposes a lexical rule which allows ‘Anglo-Saxon verbs’ to
acquire a feature licensing the prepositionless form of the indirect object.
Again, the rule does not differentiate between a Recipient of a dative verb
and a Beneficiary.
But here too, differentiation is required. Despite all the proposals
deriving dativisation from verbal properties of one kind or another, the
class of dative verbs disallowing an NIO is only partly predictable. As
Pinker (1989) details, monosyllabic verbs with native (initial) stress and
Germanic as opposed to Romance origin are more likely to allow an NIO
than verbs that lack one or more of these criteria, but there are still
‘negative exceptions’ which are resistant to capture by rule (guarantee me
a ticket).
In contrast, once we recognise the general condition on the
Beneficiary NIO, namely that it must be construed as an ‘intended
recipient’, there is no need to list idiosyncratic ‘exceptions’ for particular
‘benefactive’ verbs. For example, Pinker (1989:58) has choose as a
‘negative exception’ to dative shift (*Mary chose Linda a dress versus
Mary picked Linda out a dress). The sentence is undoubtedly
58 T. NISBET

ungrammatical on the (possibly most pragmatically likely) interpretation

that Mary chose a dress on Linda’s behalf, which is a Beneficiary type
incompatible with the NIO construction. On the other hand, one can
readily think of examples with choose which seem perfectly good, such as
Choose me a sandwich if you’re going to the baker’s, where the
interpretation of me as ‘intended recipient’ is less ambiguous, since the
sandwich is clearly to be chosen in order for the speaker to eat (‘have’) it.
In fact, there is good reason to doubt that the morphophonological
properties of verbs discussed by Pinker are criterial for the benefactive
NIO at all. Levin (1993) lists many verbs as ‘nonalternating’ for the
benefactive – which she suggests are due to the ‘Latinate constraint’ – but
I have found numerous native-speaker counterexamples such as the
following, which do not seem deviant, despite all of them being Latinate
and polysyllabic, and all bar the first example having non-initial stress as
well:

(19) I recently went to purchase my son a new uniform

I have asked my MP to obtain me a copy
Maybe they’ll invent you a better TV set
I got a builder to construct me a patio
We can manufacture you a garment made to your own
specifications
…the one piece the Reds have who can acquire them a new
pitcher

The hypothesis that conditions on the selection of Beneficiary versus

Recipient NIOs are cognitively dissimilar is supported by an experiment
on adult speakers by Gropen et al. (1989). They report that the
phonological criteria that influence acceptability of NIOs with dative verbs
were statistically insignificant with benefactives. This result is what we
would expect if Beneficiary NIOs are not arguments of the verbs they co-
occur with.

3.4 Only dative indirect objects receive verb-contingent semantic roles

A fourth piece of evidence mitigating against a uniform account for datives

and benefactives concerns the exact semantic role borne by the indirect
object. Hitherto I have referred to the indirect object role of dative verbs as
the ‘Recipient’ (by convention, theta theory more usually labels this role
‘Goal’). Whether or not individual researchers invest such labels with
cognitive content, it is clear that the actual role the indirect object of a
 BENEFACTIVES IN ENGLISH 59

dative verb is interpreted by depends on the meaning of the verb in

question. The entailments of the three sets of dative verbs (20a-c) each
differ from one another, in contrast to the entailments of the single set of
benefactives in (20d):

(20a) John gave/handed/lent me a book, {and I was glad to get

it/#but I never received it}

(20b) John refused/denied me his fortune, {#and I was glad to get

it/#but I never received it}

(20c) John owed/promised/bequeathed me a fortune,{and I was glad

to get it/but I never received it}

(20d) John bought/baked/iced me a cake, {and I was glad to get

it/but I never received it}

Only with the (20a) verbs must the indirect object be an actual recipient.
Accordingly, only the conjunct entailing actual reception by the indirect
object referent is compatible with the dative clause. In contrast, the (20b)
verbs are actually ‘verbs of refusal’, so the indirect object is necessarily
interpreted as a non-recipient and neither of the alternative conjuncts can
be entailed. The verbs of (20c) are verbs of ‘future having’ and therefore
whether the referent of the indirect object actually receives the theme a
fortune is ambiguous, and compatible with either continuation. The
Beneficiary indirect object, on the other hand, has the remarkably uniform
and precise role of ‘intended recipient’: whichever verb is selected in
(20d), it is not possible to interpret the Beneficiary NP with any other
sense, nor is there any necessary entailment that the cake was actually
received. This is so even though the verbs themselves are not ‘of a class’,
since their semantic relation to the direct object is diverse: ‘baking a cake’
creates one, ‘icing a cake’ affects one already baked, while ‘buying a cake’
acquires one both baked and iced.
Now, this lack of semantic uniformity across syntactically identical
arguments of different verbs is not especially surprising, if one accepts a
view of lexical semantic structure like that, say, of Rappaport & Levin
(1988), where the roles linked to argument slots directly reflect lexical
semantics, but the argument slots themselves, which by hypothesis
interface with syntax, do not. But if a constituent is not in an argument slot
of a verb in the first place, then we would expect semantic roles not to be
verb-contingent, which is what we find in the case of benefactives.
60 T. NISBET

This distinction between a unitary Beneficiary role and divergent

‘Recipient’ roles can be illustrated most clearly with the few verbs in
English that are compatible with both datives and benefactives.

(21a) #John brought some chocolates to Mary, but she wasn’t in

(21b) #John described the route to Mary while she was out

(22a) John brought some chocolates for Mary, but she wasn’t in

(22b) John described the route for Mary while she was out

When bring is dative, as in (21a), Mary must personally receive the

chocolates, but, as expected, when a benefactive overtly signalled by the
preposition for is present (22a), Mary may be absent when the chocolates
arrive, and in fact may never receive them at all. In (21b), Mary must be
around to hear the description, but in (22b), John could provide a
description on paper or on tape which she may (or may not) read/listen to
later.
Once again, this asymmetry between dative and benefactive indirect
objects is unexpected on the assumption that both are arguments: we
would expect either (1) dative and benefactive verbs to assign semantically
uniform roles in both cases, or (2) different benefactive verbs to assign
different kinds of Beneficiary role to their indirect objects, as dative verbs
in fact do.

3.5 Productivity

There is clear evidence that the NIO is in principle productive in English,

since novel verbs such as fax and text, presumably on the analogy of dative
verbs with similar meanings, readily accept both PIOs and NIOs:

(23a) Text your opinion to us

(23b) Text us your opinion

There can also be little doubt that the constituents (to) us and your opinion
in (23) represent arguments of text. However, whatever the conditions are
that permit the required lexical entries to be formed, dativisation is not
completely productive, since, as noted above, there is a sizeable class of
verbs that for most speakers do not permit the NIO. But if Beneficiaries are
 BENEFACTIVES IN ENGLISH 61

not listed in lexical entries, then we would expect them to be ‘free’ in

principle and thus manifest high productivity among existing verbs in the
lexicon. This prediction is borne out by the acceptability of the following
examples, which contain non-novel verbs with ‘novel’ co-occurrence of
Beneficiary NIOs:

(24) Why don’t you just tear him off a strip of your bread?
If you ask him he’ll milk you a pint
Wrap me up a kilo of salmon
I’ll root you out a fork if I can (‘find you a fork’)

In the subcategorisation-based framework Emonds has pursued (see

especially Emonds 2000 for further elaboration), there is then a learnability
problem. On his assumptions, the acquisition of argument structures
involves the building of subcategorisation frames on the basis of positive
evidence. Then, in order to explain the acceptability of the examples in
(24), the theory would have to require adults to have readily-available
lexical entries for each of these verbs including the ‘optional indirect
objects’; and, crucially, they would have had to have heard examples of
both the monotransitive and ditransitive uses of exactly these verbs during
the critical acquisition period. This is implausible.
To sum up, the following facts seriously undermine an ‘optional
argument’ account for benefactives: (1) there is no evidence for an
‘implied argument’ with simple transitive use of the verbs concerned; (2)
acceptability of the Beneficiary NIO is dependent on the interpretation of
the transitive clause, not on properties of the verb alone; (3) the occurrence
of the benefactive NIO is fully predictable whereas that of the dative NIO
is not; (4) indirect object roles of datives are contingent on the verb of
which they are an argument, but the role of Beneficiary never is; (5) the
benefactives manifest free productivity over existing verbs whereas
productivity of datives is limited to the entry of novel verbs into the
lexicon. The conclusion is that benefactives and datives are nonequivalent
and that the lexical and/or syntactic principles that license them must be
different.

4. Can benefactives be generated by a lexical rule?

In this section I consider the alternative hypothesis that benefactive clauses

are related to their monotransitive variants by a lexical rule of some sort.
Pinker’s (1989) account of argument structure alternations includes a
62 T. NISBET

lexical rule account of datives, which subsumes benefactives as a subclass.

But Pinker (1989:71) assumes that ‘every distinct set of grammatical
functions is associated with a fully-formed argument structure for that verb’
[my italics]; in other words, his rules map existing argument structures
onto other argument structures, but do not add or subtract arguments, so
that for the benefactive subclass, his proposed rule converts an argument
structure linked to the representation ‘X ACTS ON Y FOR THE BENEFIT OF Z’
to one with the meaning ‘X CAUSES Z TO HAVE Y’. As this again begs the
question as to why we should assume the benefactive ‘argument structure’
is fully-formed in the first place, I will not go on further with his account
here.

4.1 Larson’s (1990) lexical rule

I will instead consider a ‘productive rule’ hypothesis: that the grammar of

English contains a lexical rule that operates on a specified subclass of
monotransitive verbs to add a Beneficiary argument, generating a
ditransitive verb. In principle this means that two separate lexical entries
for a verb can exist: one licenses monotransitive clauses, the other
ditransitive clauses with a third Beneficiary argument. It is easy to see that
this will get round the problem discussed in section 3.1 above: if no
Beneficiary is implied in a given clause, then the verb is in its
monotransitive version and the Beneficiary ‘slot’ is simply not available.
Larson (1990), in response to Jackendoff (1990a), suggests just such
a rule for benefactives. He proposes that lexical entries of transitive verbs
may be ‘augmented’ with a Beneficiary argument, on the condition that the
verb ‘denotes an event of creation or preparation’ (Larson 1990:616). The
ditransitive verb that is the output of this rule is then able to undergo the
syntactic derivation of the ‘oblique’ (=PIO) and ‘double object’ (=NIO)
structures that are proposed for datives in Larson’s (1988) paper.
Larson claims that the semantic constraint on the class of verbs able
to undergo the rule is sufficient to explain the ‘illusion’ of a semantic
constraint on the double object construction itself (namely, as discussed
above in section 3.2, that the first object be construed as the ‘intended
recipient’ of the second object). However, this claim is falsified by the
acceptability of Beneficiaries with verbs that are not ‘verbs of creation or
preparation’ and by instances of both acceptability and nonacceptability
with the same verb. I give further examples below:

(25) {Pour/Save} me some wine!

 BENEFACTIVES IN ENGLISH 63

(26a) John killed Mary a chicken

(26b) *John killed Mary a spider

(27a) John wrote Mary a reference

(27b) *John wrote Mary her essay

(28) (*)Mary burned John a steak (Green 1974)

Neither pour nor save in (25) credibly create or ‘prepare’ any wine;3
kill is a verb of destruction not creation, and yet allows the Beneficiary in
(26a) but not (26b) because only ‘killing a chicken’ can be pragmatically
construed as creating something a person might want to ‘have’ (i.e. in
order to eat it). On the other hand, write is quite plausibly a verb of
creation, since in the examples both a reference and an essay come about,
yet (27b) is unquestionably bad because the most plausible interpretation is
that John wrote the essay on Mary’s behalf, which, as we have already
seen, is not a legitimate reading of the NIO construction in English.
Finally, (28) is either good or bad, depending on whether Mary
deliberately burnt the steak because John likes it that way (good), or on
whether Mary accidentally burned it to John’s disadvantage (a malefactive
sense, which is bad).
Now, Larson (1990:616) in fact attaches a rider to the statement of
his rule: the ‘result’ must be that ‘the theme is for the benefit of the
beneficiary’. But this is pure stipulation, added in order to account for the
effects of the construction he claims are illusory. Moreover, a lexical rule
which can ‘see into’ syntax to check that the direct object has the required
interpretation is stretching the definition of ‘lexical’ into the realms of the
miraculous.
Jackendoff (1990b:196) suggests an extension of the benefactive-
taking class to verbs involving acts of ‘performance’ (sing) and ‘making
available’ (buy). But amending Larson’s rule accordingly does not make it
fare much better: the Beneficiary NIO can still be good or bad depending
on the interpretation of the verb plus its direct object. While one can say
Enrico sang Helen a song (one of Jackendoff’s examples), if the meaning
is that the song was intended for Helen personally to hear (metaphorically,
3
Even the verb that Larson uses as an example –bake– is not necessarily a verb of
‘creation’. While ‘baking a cake’ results in a cake, ‘baking a potato’ does not create a
potato. What Pustejovsky (1995) calls the ‘permeability of word senses’ is true of many
other verbs as well (see section 4.2).
64 T. NISBET

to ‘receive’), it is not possible to say *Enrico sang the Queen a song

(meaning the song was performed for her official entertainment, with her
in the audience), since ‘intended reception’ is no longer a possible
interpretation in such a context. Although Jackendoff admits that his
characterisation is ‘informal’, it does not seem promising to attempt to
account for benefactives by extending the range of verbs to more and more
speculative and rather ill-defined classes.

4.2 Levin & Rappaport Hovav’s lexical rules

A more fleshed-out lexical rule account of argument structure alternations

is represented by the work of B. Levin and M. Rappaport Hovav. In their
framework, verbal meanings are based on templates they call ‘lexical
conceptual structures’ (LCS) which represent the ‘syntactically relevant
aspects of verb meaning’ (Levin & Rappaport Hovav 1995:21). They then
posit lexical rules to account for ‘variable behaviour verbs’: these lexical
rules map the LCS of one class of verb onto the LCS of another class. For
example, the input of a rule may be a verb of ‘manner of motion’ (run) and
have as output a verb of ‘directed motion’ (run into the room) (Levin &
Rappaport Hovav 1995:197). The aim of this approach is to provide a
principled account for apparently regular patterns of verbs occurring with
more than one argument structure and thus avoid stipulating (unmotivated)
multiple lexical entries for one verb.
Suppose then we adopt their approach and try to capture the relation
between a transitive verb and its putative ‘benefactive’ variant by positing
a hypothetical lexical rule (29). (Here I am adapting the notation of
Rappaport Hovav & Levin (1998) and borrowing a broadly ‘causative’
structure to capture the representation of a transitive verb, which is the
input to the rule. The variables {x, y, z} represent argument slots, the
capitalised elements the components of the LCS, and the item in angled
brackets the ‘constant’, i.e. the idiosyncratic part of the verb meaning.)

(29) [[x ACT] CAUSE [BECOME [y <STATE>]]] →

[[x ACT] CAUSE [z HAVE y] BY [[x ACT] CAUSE [BECOME
[y <STATE>]]]

In Rappaport Hovav & Levin’s terms, the effect of the rule is to augment
the original meaning of the verb by ‘subordinating’ the input event (the
transitive verb) to another one (the ‘benefactive event’). This seems
intuitively quite plausible for benefactives, where the Beneficiary NP
implies a potential secondary ‘giving’ event. So if the verb buy undergoes
 BENEFACTIVES IN ENGLISH 65

the rule, the following representation obtains:

(30) [[x ACT] CAUSE [BECOME [y <BOUGHT>]]] →

[[x ACT] CAUSE [z HAVE y] BY [[x ACT] CAUSE [BECOME
[y <BOUGHT>]]]

The output of the rule thus allows a ditransitive verb buy to enter a
syntactic derivation and generate John bought Mary a car. The rule would
also license the generation of the benefactive clauses that Larson’s rule
does not, such as save me some wine, John killed Mary a chicken and
Mary burned John a steak.
However, (30) would also license examples such as the following:

(31) *The boy ate the baby a hamburger

*The soldiers destroyed the citizens the city
*Mary amused the guests their child

Naturally, none of these examples can be understood to mean that

somebody causes somebody else to have something by eating, destroying
or amusing it, because of the obvious semantic and pragmatic
contradictions. However, a lexical rule must be ‘blind’ to the semantic
interpretation of surface sentences. Although it might be countered that the
above sentences are too ‘crazy’ to get into syntax in the first place, the
point is that the lexical rule will generate the derived ‘benefactive’ entry
for any verb whose LCS meets its input conditions. This entails a
potentially limitless number of ‘redundant’ verbs sitting about ‘in the
lexicon’.
A further complication for the lexical rule hypothesis is the
interaction of the benefactive rule with other lexical rules. Research on the
lexicon (Pustejovsky 1995) has revealed rampant ‘complementary
polysemy’ among the ‘same’ words. For example, many English verbs are
compatible with a range of interpretations in morphologically simplex
form: cut can be used to mean ‘put a cut in’ ‘cut through’ ‘cut into pieces’
‘cut out’, etc. On the hypothesis that ‘the syntactic properties of verbs are
determined by their meaning’ (Levin & Rappaport Hovav 1995:1, my
italics), every separate syntactic frame a verb occurs in is associated with a
separate meaning for that verb. Thus, as discussed in Rappaport et al.
(1993), a ‘contact-effect’ verb such as cut (its ‘basic’ sense as in cut the
cloth) would undergo a rule yielding a representation ‘obtain by means of
cutting’ in order to license cut a slice off the loaf. We can now see that the
sentence John cut Mary some string would therefore require the verb cut to
66 T. NISBET

undergo this ‘obtain-by-means-of X’ rule first, and it is the output of this

rule which would have to serve as the input to a benefactive rule, contra
the representation in (30). But then, for some verbs, e.g. kill as in kill Mary
a chicken, it looks very much as if the first rule would only exist to provide
the required input to the second, since John killed a chicken does not mean
‘John obtained a chicken by killing one’, which seems to be on
theoretically very dubious ground. In short, it looks as if these lexical rules
can only account for the distribution of grammatical benefactives by
generating enormous numbers of verbs that do not (or could not) exist.
I conclude then that lexical rules are either too constrained and
stipulative (Larson 1990), or too unconstrained, as on the Rappaport
Hovav & Levin approach. Lexical rules, being ‘presyntactic’, cannot ‘look
forward’ into surface syntax to see the semantic conditions that are
necessary to rule in the ‘good’ cases and rule out the ‘bad’ ones we see in
the data.

5. Conclusion

In this paper I have shown that benefactives cannot be explained by the

same theoretical devices as for datives, since there is no evidence that
verbs are subcategorised for Beneficiary NPs. There can therefore be no
‘optional’ indirect object in the lexical entry or subcategorisation frame of
‘benefactive verbs’, nor can any theoretical proposal be upheld that relies
on the verb ‘assigning’ a Beneficiary role. I have also demonstrated that a
competing lexical hypothesis, whereby argument structure phenomena
which appear to show regularities across classes of verbs are handled by a
lexical rule, is unsustainable for benefactives, since the factors that license
benefactives cannot be found in either the hypothesised ‘input’ or ‘output’
structures of verbal meanings. Alternative solutions must be sought which
do not depend on assumptions that Beneficiaries either already have
argument status, or can be promoted to it.

References
Allerton, D. J. (1978). Generating indirect objects in English. Journal of Linguistics 14.
21-33.
Emonds, J. (1993). Projecting indirect objects. The Linguistic Review 10. 211-263.
Emonds, J. (2000). Lexicon and grammar: the English Syntacticon. Berlin: Mouton de
Gruyter.
Fillmore, C. (1965). Indirect object constructions in English and the ordering of
 BENEFACTIVES IN ENGLISH 67

transformations. The Hague: Mouton.

Goldberg, A. (1995). Constructions: a Construction Grammar approach to argument
structure. Chicago: University of Chicago Press.
Green, G. (1974). Semantics and syntactic regularity. Bloomington: Indiana University
Press.
Gropen, J., Pinker, S., Hollander, M., Goldberg, R. & Wilson, R. (1989). The
learnability and acquisition of the dative alternation in English. Language 65.
203-257.
Hoffman, M. (1995). The structure and surface form of benefactives and other
prepositional grammatical relations. In Burgess, C., Dziwirek, K. & Gerdts, D.
(eds.) Grammatical relations: theoretical approaches to empirical questions.
Stanford, CA: CSLI Publications. 117-129.
Jackendoff, R. (1990a). On Larson’s treatment of the double object construction.
Linguistic Inquiry 21. 427-455.
Jackendoff, R. (1990b). Semantic structures. Cambridge, MA: MIT Press.
Langacker, R. (1987). Foundations of Cognitive Grammar, vol. I: theoretical
prerequisites. Stanford, CA: Stanford University Press.
Larson, R. (1988). On the double object construction. Linguistic Inquiry 19. 33-91.
Larson, R. (1990). Double objects revisited: reply to Jackendoff. Linguistic Inquiry 21.
589-682.
Levin, B. (1993). English verb classes and alternations: a preliminary investigation.
Chicago: University of Chicago Press.
Levin, B. & Rappaport Hovav, M. (1995). Unaccusativity: at the syntax-lexical
semantics interface. Cambridge, MA: MIT Press.
Marantz, A. (1984). On the nature of grammatical relations. Cambridge, MA: MIT
Press.
Oerhle, R. (1976). The grammatical status of the English dative alternation. PhD
dissertation, MIT.
Pinker, S. (1989). Learnability and cognition: the acquisition of argument structure.
Cambridge, MA: MIT Press.
Pustejovsky, J. (1995). The generative lexicon. Cambridge, MA: MIT Press.
Rappaport, M., Laughren, M. & Levin, B. (1993). Levels of lexical representation. In
Pustejovsky, J. (ed.) Semantics and the lexicon. Dordrecht: Kluwer. 37-54.
Rappaport, M. & Levin, B. (1988). What to do with θ-roles. In Wilkins, W. (ed.)
Thematic relations. San Diego, CA: Academic Press. 7-36.
Rappaport Hovav, M. & Levin, B. (1998). Building verb meanings. In Butt, M. &
Geuder, W. (eds.) The projection of arguments. Stanford, CA: CSLI
Publications. 97-134.
Wechsler, S. (1995). The semantic basis of argument structure. Stanford, CA: CSLI
Publications.

View publication stats