Implicit Arguments for Dummies

In this post, I summarize my recent MIT Press monograph for non-linguists. 

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

The purpose of this blog post is to make the subject matter of my monograph as accessible as possible for a non-linguist audience. By non-linguist, I mean not only non-linguist academics, such as computer scientists, philosophers and cognitive scientists, but also the educated lay person. Any person interested in following recent developments in the science of human language might appreciate this post.

By using the title “Implicit Arguments for Dummies”, I do not mean to denigrate my readers, but only to highlight that my goal is to write something as accessible as what is found in that popular series of books.

Since I am trying to keep this presentation as simple as possible, I do not cite any references on the topic. See my monograph for detailed background references.

I would love to hear feedback on my attempt from both linguists and non-linguists.

Syntax and Semantics

When we hear a sentence, we try to calculate the meaning of that sentence from the words in it and how they are put together. For example, we all know that there is a huge difference between:

1.

A dog bit the postman. 

2.

The postman bit a dog.

While the two sentences contain the same words, they are put together in a different way, so they yield completely different meanings. In (1) the agent (the doer of the action) is a dog, who bites the postman. In (2), the agent is the postman, who bites a dog. 

The study of the rules that determine how words are put together is called ‘syntax’. The study of the meaning of the words and the sentences they form is called ‘semantics’. In the example in (1), ‘bite’ is a verb, and ‘a dog’ and ‘the postman’ are noun phrases. We say that ‘bite’ takes two arguments. 

How do the arguments of a verb combine with the verb? The primary syntactic operation is Merge, which takes two phrases and combines them into a larger phrase. To form the phrase ‘bite a dog’, you take the verb ‘bite’ and combine it together with the noun phrase ‘a dog’. This operation is written as follows:

3.

Merge(bites, [a dog]) = [bites a dog]

Paraphrase: The operation Merge takes the word ‘bites’ and combines it with the noun phrase ‘a dog’ to yield the bigger verb phrase ‘bites a dog’.

In other words, Merge takes two phrases and snaps them together like Lego building blocks to create a larger structure. I call this the Lego block theory of syntactic structure.

My MIT Press monograph is all about the arguments of verbs. This area of research is referred to as ‘argument structure’. 

Implicit Agents

In some sentences, crucial words seem to be missing, making it hard to see how the meaning is calculated. A classical case is the passive:

4.

Mistakes were made.

In this example, the verb ‘make’ takes on the participle form which is ‘made’. In addition, there is a helping verb ‘were’ which is added before the participle. This construction is called the passive.

If you hear somebody say (4), you immediately wonder, who made the mistakes? There is no explicit mention of the agent in the sentence. In fact, the speaker has purposefully omitted the agent. But even so, there is an ‘implicit’ agent. When (4) is uttered, we know for certain that somebody made mistakes. 

The use of the term ‘implicit’ is common in English. For example, we speak of an ‘implicit bias’, which is an unconscious bias that I might have when making a hiring decision, for example. An implicit bias is somehow hidden, and influences my decision without me even knowing about it. My use of the term ‘implicit agent’ does not have such negative connotations, but represents a hidden agent, not pronounced but still influencing my understanding the sentence.

We can easily continue the discussion started in (4) by making explicit who the agent really is. In (5), there is no longer an implicit agent, rather now there is an explicit agent (expressed by the by-phrase). 

5.

Yes, in fact, the mistakes were made by John.

Invisible Agents versus Missing Agents

But what is the theoretical status of implicit arguments? How are they represented in the minds of the language users? There are two general ideas hypotheses that need to be tested.

The first hypothesis is that an implicit argument is like the invisible man. The invisible man is somebody who is physically present. He can lift things up and push things over. But he is completely invisible.

On analogy with the invisible man, an implicit argument (for example, the implicit agent in (4)) would be syntactically present but invisible. In other words, in (4) there is a phrase [BY SOMEBODY], that forms a part of the sentence but is not pronounced. The capital letters indicate that the phrase is not pronounced. 

On the invisible argument theory, the passive sentence in (4) would have the following structure (see the monograph for a more technically accurate presentation).

6.

Mistakes were made BY SOMEBODY.

The sentence in (6) is formed by the usual syntactic rules of English. The operation Merge takes the phrase ‘mistakes were made’ and combines it with the non-pronounced phrase ‘BY SOMEBODY to produce the sentence ‘Mistakes were made BY SOMEBODY.’

The other hypothesis is that an implicit argument is not present in the sentence at all, but we just infer it. When we hear (4), we infer that somebody made mistakes, but there is no invisible agent in the sentence representing the maker of the mistakes. 

A possible analogy here is a missing puzzle piece. If we see a puzzle depicting a house and it has a missing piece, we assume that the piece is somewhere, maybe lost, but we do not assume that there is an invisible piece in the puzzle. From the rest of the puzzle, we might be able to infer quite a bit about the missing piece. Perhaps we can tell that the missing piece would depict a door with a tree next to it. Similarly, in the case of the passive sentence in (4), we can infer quite a bit about the missing agent without necessarily assuming that there is an invisible agent phrase present in the sentence.

I will refer to these two theories as the ‘invisible agent’ theory and the ‘missing agent’ theory. I summarize these two theories below.

7.

The Analysis of Implicit Agents

a.

Hypothesis 1: Invisible Agents (analogy: the invisible man)

Implicit agents are invisible (not pronounced), but syntactically present in the sentence.

b.

Hypothesis 2: Missing Agents (analogy: a missing puzzle piece)

Implicit agents are inferred, but missing syntactically in the sentence.

As it turns out, deciding between these two approaches has posed a huge challenge to linguistic theory. The issue is rather controversial. Therefore, I took up the challenge of systematically investigating it in my monograph. My conclusion is that the ‘invisible agent’ theory is the correct. In the section below, I present some of the evidence for my conclusion.

Testing the Predictions

To summarize, we have introduced the concept of ‘implicit agent’, illustrated in (4). We have also sketched two theories of implicit agents: the ‘invisible agent’ theory and the ‘missing agent theory’. Since linguistics is a science, we should be able to test these hypotheses, and choose the one that best handles the data.

We need to find tests that can detect the presence of arguments in a sentence. In the monograph, I developed five different tests, but here I will focus on one, the distribution of reflexive pronouns. 

A reflexive pronoun is a pronoun that refers back to another noun phrase in the same sentence, called its antecedent. In English, reflexive pronouns have the form X-self (himself, herself, yourself, myself, etc.).

8.

John takes care of himself.

In this example, ‘himself’ refers back to ‘John’, and so ‘John’ is the antecedent. If there is no antecedent, the sentence is unacceptable (indicated with a *):

9.

*I take care of himself.

The reason that this sentence is unacceptable is that ‘himself’ needs an antecedent, but the pronoun ‘I’ cannot be its antecedent. We can sharpen up this explanation by noting that ‘himself’ is a third person singular pronoun, but the subject of the sentence in (9) is first person singular. They do not match, and so ‘himself’ does not find an antecedent in the sentence.

The general principle accounting for (8) and (9) is the following (called Principle A by syntacticians):

10.

A reflexive pronoun has an antecedent in the sentence.

It is important to note at this point that (10) is a syntactic principle. It says that if a sentence contains a reflexive pronoun, then there needs to be a noun phrase in the sentence that can act as the antecedent for the reflexive pronoun.

But now, let’s consider how reflexive pronouns work in the passive. Here is an example from the internet, which I also find acceptable:

11.

Some things are better kept to yourself.

On first impression, (11) seems like a straightforward counterexample to the principle in (10), which is otherwise a solid generalization about English. The problem is that there is a reflexive pronoun ‘yourself’, but it does not appear to have any antecedent in the sentence. Clearly, ‘some things’ is not the antecedent of ‘yourself’ because they do not match. The noun phrase ‘some things’ is third person plural, and the reflexive pronoun ‘yourself’ is second person singular. But then, what is the antecedent?

But let’s unpack (11) a little. The reflexive pronoun ‘yourself’ refers back to the keeper. Therefore, the implicit agent is the antecedent of the reflexive pronoun. 

In this way, (11) is entirely parallel to the active:

12.

You had better keep some things to yourself.

In this example, the antecedent of the reflexive pronoun ‘yourself’ is the subject of the sentence. Another example from the internet which illustrates the same point is given below:

13.

Rumor has it that Mike Tyson bought over 200 cars throughout his career, totally $4,5 million. Many were bought for himself and others as gifts for his friends and family.

Focus narrowly on the clause below:

14.

Many were bought for himself and others…

We need to ask what the antecedent of ‘himself’ is in this sentence? How does this sentence conform to the principle in (10)?  But there is an implicit agent in (14), none other than Mike Tyson, and that implicit agent can serve as the antecedent of the reflexive pronoun.

From these examples, we have established the following generalization:

15.

The implicit agent in a passive can be the antecedent of a reflexive pronoun.

In this way, implicit agents behave exactly like overt agents in active clauses. We can account for the generalization in (15) under the ‘invisible agent’ theory. The reason an implicit agent can be the antecedent of a reflexive pronoun is because it is actually present in the sentence (even though it is not pronounced).

It is far less clear how to account for (15) under the ‘missing agent’ theory. If the agent is really not present (but only inferred) then how could it be the antecedent of a reflexive pronoun?

Overall, the conclusion is the following:

16.

The implicit agent in the passive is syntactically present, but not pronounced.

Conclusion

I have tried to explain in layman’s terms one of the main discoveries of my recent MIT Press monograph: Implicit arguments are syntactically present but not pronounced. 

This result shows that the meaning of a sentence cannot always be calculated simply on the basis of the pronounced words and phrases in the sentence. Sometimes there are words and phrases that are present in the sentence but not pronounced, and it is also necessary to incorporate these into the calculation of meaning. 

The theory of implicit arguments that I have been developed could be important for other related disciplines. The existence of implicit arguments raises questions for the study of language acquisition. For example, a child learning a language who hears a sentence such as (4) will have to recover the meaning of the implicit agent in some way. How do they do this? What cues do they use? What kinds of representations form in their brains? 

It is to be hoped that my monograph will help pave the way for such further research.