Davidson’s analysis focuses on the meaning of such adverbials as slowly, with a knife, in the bathroom and at midnight, as in the sentence:

The adverbial phrases raise the issue of whether or not to treat them as arguments. Arguments are traditionally thought to be centrally connected to the meaning of the predicate. Adverbial phrases are only loosely associated with predicates. Arguments also combine with predicates in a different order. Adverbial phrases can be added to predicates in different orders. A final difference is that arguments are (usually) obligatory, whereas adverbial phrases are optional.

Davidson pointed out the entailment relation between the predicate with adverbials and the predicate without the adverbials. (1) entails:

Davidson proposed treating the adverbial phrases as separate propositions conjoined with the root predicate:

Davidson’s analysis raises the question of what argument to give the adverbial predicates slowly and with a knife. Davidson proposed making the event itself one of the arguments of action verbs:

b. ∃e (BUTTER (Jones, the toast, e) & SLOWLY(e) & WITH(e, a knife) & IN (e, the bathroom) & AT (e, midnight)

The variable e is existentially bound and restricted to ranging over events. Each adverbial predicate is expressed as a separate conjunct. The adverbials are independent of the root predicate and its arguments.

A Neodavidson analysis carries Davidson’s program further by treating arguments as conjuncts in the same way Davidson treated adverbial phrases. Hector-Neri Casteñeda (1967) suggest analyzing the sentence in (28a) as (29a):

29a. ∃e (FLY (e) & SUBJECT (I, e) & OBJECT (my spaceship, e) & TO (e, the Morning Star)

The passive construction allows the subject of a predicate to remain unexpressed. The Davidsonian analysis of a passive predicate translates the subject as an existentially bound variable:

A Neodavidsonian analysis of passives can express the subject as an existentially bound variable or omit it entirely:

Both of the semantic interpretations in (34) are possible for the passive. There is no necessity for the arguments in the logical representation to appear in the surface representation. Some mechanism is needed to insure that the required arguments actually surface in the utterance. The absence of the subject argument in (34c) is more serious. This translation suggests that the subject is not even present in the semantic representation.

Parsons (1990:96-9) argued that the Davidsonian analysis sometimes requires too many arguments. For example, the sentence ‘Brutus stabbed Caesar’ entails that ‘Brutus stabbed someone.’ In a Davidsonian analysis all arguments for the verb stab, including the object, are obligatory. An unexpressed object is replaced by an existentially bound variable as in (38d).

According to Parsons’ argument, the sentence ‘Brutus stabbed Caesar’ also entails the proposition ‘Brutus stabbed’ which lacks the implication that Brutus actually connected with anyone. The Davidsonian analysis doesn’t capture this possibility.

It is also possible to construct a Davidsonian analysis with too few arguments. The sentence ‘Brutus stabbed Caesar with a knife’ entails that ‘Brutus stabbed Caesar’. These sentences have the Davidsonian analyses shown in (39).

The problem is that all stabbing events, even the one in (39c), involve the use of a sharp instrument. The analysis in (39d) fails to capture this fact. Kearns (186) suggests that ‘the apparent obligatoriness of arguments is not really a semantic or logical matter.’ This solution solves the problem by ignoring it.

40. In a dream last night I was stabbed, although in fact nobody had stabbed me, and I wasn’t stabbed with anything.

Here, the dreamer is reporting an incoherent dream sequence in which the stabbing event is somehow separated from an assailant and a weapon. Parsons’ point is that although the dream event is incoherent, the language used to describe the dream is not incoherent—we can understand the anomalous event the dreamer is trying to report. A Davidsonian analysis of the dream sequence requires the use of a subject variable.

The subject variable in (41) produces a contradiction with the statement that ‘nobody had stabbed me.’ Parsons states ‘My dream may have been incoherent, but I am not, and what I am saying should not contain a self-contradictory logical form.’

Chomsky (1965) introduced the term selectional restrictions to describe semantic co-occurrence restrictions which are logically necessary. We can learn a lot about the meaning of verbs by analyzing their selectional restrictions. The verb die, for example, selects animate subjects.

Selectional restrictions on verbs are one of the main sources of semantic differences between languages. The breaking domain provides a good example of this variation (Pye, Loeb & Pao 1996). English speakers apply the word break to a wide array of objects—sticks, stones, bread, furniture, and machines. At the same time, English insists on marking the difference between breaking and tearing things. We break one and three dimensional objects, while we tear two dimensional objects.

Object	English	Japanese	Mandarin	Spanish	K’iche’
pen cap	take off	tor-u	na xia	quitar	-esa:j
apples	pick	tor-u	zhai	arrancar	-ch’up
cherries	pick	tzum-u	zhai	cortar	-mak
string	break	kir-u	duàn	cortar	-t’oqopi:j
stick	break	or-u	duàn	quebrar	-q’upi:j
paper	tear	yabur-u	xi (kai)	romper	-rach’aqi:j
cracker	break	war-u	bo (kai)	romper	-pi’i:j
peanut	break	war-u	bo	pelar	-paq’i:j

The Neodavidsonian analysis does not capture the selectional restrictions on verbs. We have also seen the need to treat collective predicates like surround differently from distributive predicates such as break.

Some linguists have used a semantic feature analysis to account for the selectional restrictions on verbs. Jackendoff (1990) proposed the following lexical conceptual structure for the verb drink.

[_Event CAUSE([_Thing ]_i, [_Event GO ([_Thing LIQUID]_j,[_Path TO ([_PlaceIN ([_Thing MOUTH OF ([_Thing ]_i,)])])])])]

Jackendoff conceptualizes the act of drinking in terms of a causal event resulting in the conveyance of a liquid to someone's mouth. Such analyses are only as good as the semantic features they employ. We have already seen the spatial terms such as to and in are suspect.

Break verbs add a further concern. While it might be possible to posit general restrictions for the English verbs break and tear, such restrictions do not apply to the breaking verbs in languages like Japanese or K’iche’. The Mayan languages as a whole have incorporated features of a narrow range of objects into the meaning of their verbs (Pye 1996). The K’iche’ verb -t’oqopi:j is used for one dimensional objects, the verb -rach’aqi:j for two dimensional objects, the verb -pi’i:j for soft things, and the verb -q’upi:j for hard things. These distinctions are understandable to English speakers, but the K’iche’ breaking verbs also include verbs like -jixi:j ‘to tear leaves along the veins’, -joyopi:j ‘to break a banana from a bunch of bananas’, -chiko:j ‘to break something by throwing it’, and -qasa:j ‘to lower; to break something with a downward motion’. K’iche’ even has the verb -jochopi:j ‘to break off a banana by failing to support a bunch of bananas’.

The Neodavidsonian predicates SUBJECT and OBJECT are somewhat confusing since they apply the traditional names for grammatical relations to the level of semantic interpretation. Linguists such as Charles Fillmore introduced the notion of semantic or thematic roles into linguistic theory. Thematic roles or thematic relations are broad semantic categories of arguments. They include the roles of:

45. agent, patient, theme, source, recipient, goal, instrument, benefactive, experiencer, stimulus

The thematic roles lack explicit definitions. The descriptions in Kearns are merely suggestive. Fillmore noted a preferential mapping between thematic roles and grammatical relations. Agents usually surface as subjects, patients and themes surface as objects, and instruments and goals surface as adverbial phrases. If an agent is absent, a theme or instrument may advance to subject position:

The use of thematic roles assumes that the same semantic relation applies across a wide range of predicates. For example, Jones might be assigned as the agent in all of the following sentences:

Ideally, a theory of thematic roles would predict the mapping relation between semantic arguments and grammatical relations. The theory becomes circular when the grammatical relations serve as the basis for determining the thematic roles. A precise delineation of thematic roles remains to be determined.

Parsons’ stab example demonstrates the ability of many English verbs to appear in sentences with different types of arguments. This variation is referred to as argument alternation or verbal diathesis. The causative alternation, shown in the example below, represents one of the better known examples of argument alternation.

The example in (1) illustrates the lexical form of the causative alternation where the same form of the verb appears in the intransitive (1a) and transitive (1b) sentences. Many English verbs such as fall and throw do not participate in the causative alternation.

The Neodavidsonian approach cannot explain these differences; it provides the same analysis for all three verbs.

Fall can appear in a causative sentence, but English grammar requires the use of a periphrastic construction (4a). Likewise, throw occurs in intransitive sentences by means of the passive alternation (4b).

Many linguists have noted the semantic contrast between the lexical and periphrastic causative constructions (Levin & Rappaport 1994). The lexical construction indicates the direct involvement of the agent in the event, while the periphrastic causative indicates an indirect connection between the agent and the event.

The examples with fall and throw demonstrate that while the lexical causative usually involves direct causation, the absence of a lexical alternation is not always associated with indirect causation. For some verbs the periphrastic construction is associated with direct causation. Indeed, the periphrastic construction is the main construction used to express direct causation in languages like French and Thai.

It is difficult to discern which argument set is basic for verbs in languages with the lexical form of the causative alternation. Levin & Rappoport (1994) suggest using the selectional restrictions for English verbs to establish their basic transitivity. Thus, they suggest that the English verb break is basically transitive since it has more transitive than intransitive uses.

The intransitive form seems to be basic for languages like French and Thai that rely on the periphrastic alternation. Many languages support an anticausative alternation that intransitivizes a transitive verb stem. Mandink, a Niger-Congo language of West Africa, uses a verb affix to effect this change. The transitive form of kati ‘stick_break’, for example, has a corresponding intransitive form kati-ta. K’iche’ Maya uses an antipassive affix to make this change on a subset of transitive verbs (Mondloch 1981). The antipassive morpheme in K’iche’ typically suppresses the internal argument and changes a transitive verb stem to an intransitive stem. The following table provides an idea of the cross-linguistic variation that exists in this domain.

	Alternation Type	Intransitive Form	Transitive Form
English	Lexical	break	break
Spanish	Anticausative	romper-se	romper
Mandarin	Compound	dwàn	nòng dwàn
Thai	Periphrastic	hak	t^haam hak
Garifuna	Causative Affix	halagua	halagua-na
Mandink	Anticausative	kati-ta	kati
Korean	Idiosyncratic	bureojida	bureoddeurid
K’iche’	Antipassive	-q’upi-n	-q’upi

The Neodavidsonian treatment can be applied to the arguments of perception verbs. Perception verbs typically take complement clauses which describe the stimulus:

Traditional logic would analyze the stimulus clauses as propositions with an interpretation as true or false. The Neodavidsonian analysis allows the content of the stimulus clauses to be decomposed into a series of conjoined propositions:

∃e∃e'(SEE (e) & EXPERIENCER(j, e) & STIMULUS(e',e) & SHAKE(e') & AGENT(l,e') & PATIENT(the bottle,e'))

This analysis commits the speaker to the entailment that Lina actually shook the bottle since (56) asserts the existence of the event e', spelled out in (57):

If Jones is reporting a dream or hallucination then the existence of event e’ must be restricted to the appropriate context.

Lina can be analyzed as an argument of the perception verb in (56), which leads to the further entailment that ‘Jones saw Lina’. Unfortunately, the analysis in (56) does not show that Lina is a stimulus for the event e. The analysis only shows that the event e' is the stimulus. It would be possible to add a meaning postulate to capture this inference. Meaning postulates were introduced by Carnap (1952) to describe ‘nonlogical elements’ that might play a role in a formal analysis. In the present case, we might apply the meaning postulate:

While meaning postulates offer a quick fix to this problem, they invariably lead to other difficulties such as the distinction between logical and nonlogical elements. Kearns examples in (59) raise questions about the status of our meaning postulate.

Adding tense and quantified NPs to a Neodavidsonian analysis leads to questions about the scope of the event variable relative to these other elements. We might assume that a tense operator always has wide scope with respect to an event variable since the tense operator marks a boundary of the event. Past and future events do not exist at the time of speech.

‘At a past time there was an event and the event was a leaving event and Jones was its agent.’

Combining tense operators with quantified NPs produces scope ambiguities like the following:

Explicitly representing tense as quantification over times allows the relationship between an event and its time of occurrence to be expressed by a predicate like ‘AT(e,t)’, as in (63).

[The x: PRESIDENT(x)] ∃t ∃e(t < t* & VISIT(e) & AGENT(x,e) & PATIENT(h,e) & AT(e,t))

Davidson, D. 1967. The logical form of action sentences. In N. Rescher (ed.), The Logic of Decision and Action, pp. 81-120. Pittsburgh: University of Pittsburgh Press.

Levin, Beth & Rappaport Hovav, M. 1994. The acquisition of the lexicon, A preliminary analysis of causative verbs in English, ed. by Lila Gleitman and Barbara Landau, 35-77. Cambridge, MA: MIT Press.

Pye, C. 1996. K’iche’ Maya verbs of breaking and cutting. Kansas Working Papers in Linguistics 21.87-98.

Pye, C., Loeb, D. F. & Pao, Y. Y. 1996. The Acquisition of Breaking and Cutting. In E. V. Clark (Ed.), The Proceedings of the Twenty-seventh Annual Child Language Research Forum, 27.227-236. Stanford: Center for the Study of Language and Information.