On the origin of the term ‘implicit argument’ (Tom Roeper)

 On the origin of the term ‘implicit argument’

Tom Roeper, University of Massachusetts, Amherst 

Sept 21,2024       

In 1981 I wrote the paper “Implicit arguments and the Head-Complement relation” which was eventually published in LI in 1987 where I introduced the term ‘implicit argument’. I wrote a follow-up called ‘Copying implicit arguments’ that was published earlier in WCCFL in 1985.  

Originally Luigi Burzio and Rita Manzini had observed that passives allowed control in the sentences.  

1.

The boat was sunk to collect the insurance.

I observed that these had been called “agentless passives” before but clearly the agent was somehow present. Then I added several more constructions which allowed the same kind of control:

2.

a. Nominalizations: the destruction of the city to PRO prove a point

b.  Compounds: meat-eating to gain weight.

c. -able: goods are easily exportable to PRO improve profits.

d. small clauses: the game was played [pro drunk]

I said they shared the property of having an ‘implicit argument’ which I thought should have an explicit designation as such. Others thought that it was an implicit thematic role and that control referred directly to a verb’s theta-grid. I responded that an ‘implicit thematic role’ seemed an apt claim for middles, which as Jay Keyser and I showed (1984), required an Agent for interpretation, but blocked both by-phrases and control:

3.

a. *the city destroys easily by the army

b. *the city destroys easily to prove a point

That claim was captured precisely by assuming the presence of the thematic grid, but no mechanism to control PRO. The absence of a control-mechanism with middles further enhanced the idea that there must be a representation capable of control for implicit arguments. (See Landau (2010) for more arguments that implicit arguments need to be syntactically explicit.)

At the time I was a Visiting Scholar at MIT invited under a Sloan grant (1981), and I was immediately challenged by several people who said “arguments are obligatorily expressed and only thematic roles in adjuncts are optional” so ‘implicit argument’ is theoretically incoherent.  And it was suggested that the by-phrase option showed that it really was an adjunct. Fortunately, I had an ally in Noam Chomsky who liked the idea and asked me to put the paper on reserve for the course he was teaching (see reference in Chomsky (1986)). He suggested that (pc) it revealed an important aspect of the Head-complement relation: arguments needed to be c-commanded (hence the title: “Implicit arguments and the Head-complement relation”).

Edwin Williams (my colleague at the time) thought the whole domain would have a significant impact on linguistic theory. But for nominalizations (see (2a) above) he suggested that the whole clause was the subject of the nominalization with control which indeed can be in this example:

4.

the destruction of the city proved the point.

However, I pointed out to him - which he accepted in subsequent publications - that other examples showed that it could be just the Agent that controlled, as in (5a), because the whole clause would be ungrammatical (5b):

5.

a. the use of drugs to go to sleep

b. *the use of drugs went to sleep.

And in fact, it could be an instrument too, which also points to the notion of an argument and not a VP adjunct:

6.

On the assembly line, the wheel was automatically rotated to put in screws.

Howard Lasnik (pc) pointed out that non-agents can control as well:

7.

Jesus died to save our souls

(7) indicates, I believe, that a Theme can be a controller as well, which is not surprising.

The implicit argument paper argued that Agent was projected onto a subject PRO position of the nominalization, which was, crucially, blocked if an object was preposed:

8.

a. the destruction of the city to prove a point

b. ?* the city’s destruction to prove a point

In (8b) again the question arises of whether the whole DP, not just the Agent, is what “proves the point”, but the presence of a preposed object clearly decreases grammaticality and if it occupies the PRO position, we have an explanation.     

Luigi Rizzi at the time asked me if I thought that the concept of implicit argument could extend to a case that he discovered in one of his classes when he said:

9. this leads to conclude

And a student pointed out that in English one must say:

10. This leads us to conclude

Rizzi observed that the object could be left empty in Italian, but apparently not in English.   

This raised the question of what the conditions on implicitness are. My view at the time was that the existence of an empty Pro/pro position meant that the argument was not implicit because it had a clear overt syntactic position, even if empty (pro), exactly what is not the case in passive or -able where the object fills the subject position:

11.

a. This leads pro/us to conclude.

b. The truck was driven/ driveable.

It appears that the determiner specifier, normally occupied by an article can also host such a PRO in examples like (2a). Chomsky (pc.) suggested that the determiner is above PRO:

12. [DP the [NP [VP PRO destroy]]]

While when an alternate thematic role, like Theme, is projected onto PRO, it blocks an AGENT in ‘the [city’s destruction]’.  This leads to the judgement that (8a) and (8b) involve a significant contrast and (8b) ‘the city’s destruction to prove a point’ is therefore ungrammatical because there is no room for the Agent PRO in the specifier of ‘destruction’. However, the potential for Agent control still seems to be there for most people even if the syntax seems “ungrammatical”, thus the idea of covering PRO as something which makes it “ungrammatical” or less grammatical but not uninterpretable  seems to correspond to the subtlety of the judgements involved.    

The idea that a syntactic position is projected is further supported in complex nominalizations which contain a passive marker: -ed or -able.  If (15a,b) is all right with either Agent or Theme in the subject position, then why is there a contrast between (13) and (14)?

13.

a. ?The lecture’s preparedness by the professor

b. the grammar’s learnability by a child

14.

a. *the professor’s preparedness of the lecture

b. *the child’s learnability of grammar

15.

a. the enemy’s destruction of the city

b. The city’s destruction by the enemy

Why is (14) out if (15) is possible?  Why are both Agent and Theme possible in the possessive position for -tion but not for -ability or -edness?  I argued that while -tion and -ity are not making thematic projections, -ed and -able are projecting theta-roles even though they are interior affixes. The difference clearly stems from the presence of passive markers -ed and -able in ‘preparedness’ and ‘learnability’. That is, the Theme projection to subject in: ‘grammar is learnable’ is still present when one has a nominalizer [Theme [learn] able] ity]. This fits Chomsky’s (1970) original argumentation about argument inheritance in nominalizations.

Under Burzio’s account the affix had the effect of ‘dethematizing’ the subject position. However,  we argue that the position has a Theme projection when -able is added.  The argument now is that there continues to be a Theme projection in subject position even if it is invisible and this can explain a variety of nominalizations: (16c) manifests this Theme and (16b) violates it with an Agent:

16.

a. The learnability of grammar by children

b. *children’s learnability of grammar

c. ?grammar’s learnability by children

Again, if -ed and -able project an object Theme into the subject position—it is precisely that role which blocks the Agent, but allows (16c) (See Roeper (2006, 2021) for more extensive arguments).  This then supports the notion that some kind of PRO-like position must also be syntactically present onto which the Agent or Theme can be projected.

Borer (2013) and Collins (2024) argue as well that a syntactic pro is projected into the Spec position of vP, as did Van Hout and Roeper (1999). Roeper (2024) points out (see Collins 2024 as well) that this introduces the possibility that an implicit argument could control another implicit argument such as in (17) where the implicit argument of ‘prepared’ could control the same argument in ‘well-seasoned’ because we construe the seasoning as part of the preparing.

17.

The food was prepared well-seasoned

Now predictably, bare nominalizations also seem to allow Agents in Spec or subject position:

18.

a. Mary silenced her children with an angry glance

b. Mary’s glance silenced the children

c. Mary’s glance was enough to silence the children.

Here a Theme projection is blocked in (19a):

19.

a. * The children’s glance (by Mary) silenced them

b. A glance at the children by Mary silenced them

Bare nominals are often simply results (the harvest was good = food or amount harvested was good), but that fact does not eliminate potential Agent readings in other bare nominals like ‘glance’, although such facts have not been a part of the literature thus far but I think they should be.    

It is noteworthy that plurals can elicit the result reading and block agent as well in the of-phrase:: (20a) ‘loss’ means fewer bankers  and not what bankers lose, while (20b) allows bankers to be the subject (agent).

20.

a. The loss of bankers (object)

b. The losses of bankers (agent)

c. The conquest of the Indians (Indians = object)

d. The conquests of the Indians (Indians = agent preferentially, but object possibly)

The ‘loss’ is an event, but ‘losses’ pluralizes the object result reading of ‘loss’ and forces bankers to be Agents if any thematic role is to be projected. This contrast can be captured if we assume the continuing presence of the verb-complement structure. This fairly subtle syntactic interaction deserves a precise analysis and reveals again that the projection of arguments within nominalizations resembles the precise syntax most syntax discussions seek.

A relevant shift through articles can change the thematic role as well:

21.

a. John was in control of the army. 

            John1 is in [NP [VP PRO1 control (of the) army]] (‘army’ is the object)

b. John was in the control of the army. (‘army’ is the agent)

How does (21a) force an Agent reading? If we assume a PRO in (21a) [PRO control] then we need a theory that can co-index in (21b) the of-PP to the Agent thematic role. It is not clear what that theory should be.

In the ‘copying implicit arguments’ paper, the argument was made that where a structure failed to project arguments in a canonical position, then it could be co-indexed with a PP where the Prep expressed the relevant relation (by, to, of) which was in effect a copy of that argument, but the argument itself originated on the theta-grid and was projected onto c-commanded syntactic positions, and for subjects linked to a pro in Spec of little vP. 

In sum, the term ‘implicit argument’ has referred primarily to the Agent projected onto PRO (or pro). The PRO position has been argued to be in the Spec of little vP which has the introduction of the Agent as a defining property (as Borer, Collins and originally Chomsky and Kratzer argued).  

The term implicit argument does not refer (in my usage) to arguments linked to empty object or indirect object syntactic positions although a number of scholars assume that it does and they use it in this broader way where it engages in further syntactic processes, as Rizzi’s indirect object case also involves control [it leads pro1-indirect object1 [PRO1 to conclude]]. I would suggest that a term - if any term is needed - like ‘unexpressed thematic roles’ could cover the whole domain (including implicit arguments and middles) while implicit arguments in my usage covers precisely  -passive, -able, -ed as they are projected onto PRO in passives, compounds, -able, and small clauses. Nonetheleless our goal as always, I think, is to promote terminology that aims at technical precision and transparency so that the theory advances with increased theoretical subtlety without engendering confusion.

I hope this discussion clarifies some of the issues and enables more useful terminology.

Selected References

Collins, Chris. 2024. Principles of Argument Structure: A Merge-Based Approach. MIT Press, Cambridge.

Roeper, Thomas. 1987. Implicit Arguments and the Head-Complement Relation. Linguistic Inquiry 18, 267-310.