PhiN 34/2005: 60

Bonaventure M. Sala (Yaounde/Cameroon)

Dangling Modifiers: What is wrong with them?*

Dangling Modifiers: What is wrong with them?
This paper revisits the concept of dangling modifiers, which overflows even the frontiers of Non-native Englishes. In it, we argue that dangling modifiers do more than just violate Equi-NP-deletion. We posit that dangling modifiers flout two rules in English that govern modifier topicalisation: The Co-indexation Condition and The Restrictive Relative Constraint. We further argue that the dangling modifier falls within a general problem with the interpretation of empty categories in Cameroon English. Its interpretation is based on context.

By deviating from native standards and having their own distinct, institutionalised forms, the New Englishes are progressively becoming autonomous error systems. One error that has been discovered, even in the productions of native speakers making their first steps in English, is the dangling modifier. This has occupied major columns in grammar sections of most college handbooks on writing. Yet error analysts have not investigated this "standard error" in great detail. The purpose of this paper is to revisit dangling modifiers as a global error as well as a feature of Cameroon English (henceforth, CamE) and other non-native Englishes (see Simo-Bobda 1994: 68 ff., Trudgill and Hannah 1985, Platt et al. 1984 and Arua 1998). We need to know the rules of English that this global phenomenon transgresses. This becomes the more important because focus in syntax today seems to have shifted from the rules of English grammar to the explication of ungrammatical sentences (see Radford 1997: 4).

College writing handbooks prescribe that to have a subjectless clause in a sentence initial position, we must ensure that its implied subject be the same as that of the main clause (see Hacker 1995: 71f., Strunk and White 1979: 13f., Warriner 1988: 323 ff. and Heffernan and Lincoln 1982: 229 ff.).

PhiN 34/2005: 61

Researchers have referred to dangling modifiers as modifiers that modify nothing in the sentence. Unless we are in the realm of prescriptive grammar, such a definition is biased because those who utter them intend that they modify something within the sentence. Though dangling modifiers are, by so doing, unacceptable in BrE, speakers of non-native varieties of English (henceforth, NVEs) use them variantly and understand them as modifying some particular constituent in the sentence. According to Simo-Bobda (1994: 68 ff.) dangling modifiers are common in CamE and are modifiers that are attached to their heads following a convention that is un-English. Simo-Bobda (ibid.) provides the following examples:

(1) a. Returning from work late in the evening alone, thieves caught him and took out all his money at gunpoint.
  b. Trapped in the wrecked car, we could remove him only by sawing the door open.
  c. To get to his friend's house, we explained to him that he should pass the traffic lights and turn left.
  d. Fifty of his articles were published while on the staff of the Department.
  e. As their principal, students knew that they owed him respect.
  f. At the age of twenty, do not expect your daughter to behave like a baby.
  g. Coming back from work yesterday, a car knocked him down.

Also notice that sentence (h.) below, which is very common in Cameroonian essays, also contains a dangling modifier:

  h. To conclude this essay, it is necessary to restate the points we have raised.

There is some grammar that governs the topicalisation of non-finite clauses in English. Though at first sight it may sound like another rule of proximity between constituents and their heads, it could have deep theoretical implications. To go straight to the point, there is some underlying knowledge about dangling modifiers that users of NVEs do not possess. The questions that this paper seeks to answer are the following: What is this knowledge about the focusing of modifiers? What is the internal logic of dangling modifiers? And how does context help interlocutors to understand each other?

It is necessary to begin with the grammar of modifier topicalisation in English. Modifier topicalisation in English (the focusing of modifiers) would present no problems if subjects were repeated or specified in all clauses of complex sentences. What is meant here is that, if (1g.) were written as (2) below, the problem of interpretation would not even arise:

PhiN 34/2005: 62

(2) [When he was coming back from work,] a car knocked him down.

In (2), the bracketed adverbial clause is complete because it has its own subject and verb. It is finite. It is therefore clear that the notion of dangling modifiers has nothing to do with finite clauses. I do not overlook the possibility that some finite adjectival clauses and adverbial clauses could have faulty headwords as in (3):

(3) Mr. Mbah and his dog, who always carries a walking stick, passed here.

What is meant in the above statement is that the notion of dangling modifiers clearly relates to nonfinite clauses, that is, to clauses that have an empty INFL node and cannot, therefore, govern an overt subject as seen below:

Phrase marker for non-finite clauses

(4) coming back from work:

The tree diagram in (4) shows that the non-finite clause is just a VP with an empty INFL node and an unspecified subject. This means that the subject of the modifier clause has been deleted. Naturally, the deletion of elements containing lexical information must be compensated for by other elements in the clause. This is referred to as recoverability. Recoverability will be gained here by the proper interpretation of PRO in (4). In BrE, the deleted subject of subordinate clauses is recovered by being equivalent or co-indexed with the subject of the main clause. This is referred to as Equi-NP-deletion. In the literature on empty categories, PRO is the non-overt subject of subordinate clauses, in accordance with principles such as the theta criterion, the Extended Projection Principle (EPP) and the Projection Principle as seen below:

PhiN 34/2005: 63

(5) a. [IP2Thomasi wants [IP1PROi to eat some bread]]
  b. [IP2Thomas wants Pauli [IP1PROi to eat some bread]]

According to Haegeman (1984: 407), "the properties of a trace are recovered by virtue of the co-indexation with its antecedent". In (5a.), "Thomas" is co-indexed with PRO. In (5b.), "Paul" is co-indexed with PRO. A distinction is therefore made between subject and object control of PRO in (5a.) and in (5b.), respectively (Haegeman (1984: 257)). How is this phenomenon linked to the notion of dangling modifiers? Consider (6) below:

(6) *PROi coming back from work yesterday, a cari knocked him down.

As said above, PRO is the subject of the participial clause in (6). The sentence is ungrammatical because PRO cannot be or is wrongly co-indexed with "car", the subject of the matrix clause. (7) below is grammatical because PRO can be duly co-indexed with the subject of the matrix clause:

(7) PROi coming back from work yesterday, hei was knocked down by a car.

In (7), PRO can be co-indexed with "he" because PRO and "he" refer to the same entity. In the literature, the process whereby two NPs in different clauses are equivalent, with the result that the one is deleted and is controlled by the other, is referred to as Equi-NP-deletion. The ungrammaticality of (6) stems from the fact that there has been a non-Equi-NP- deletion. Equi-NP-deletion is a kind of deletion that responds to the characteristic of a language being economical. It helps speakers to rid themselves of the redundant repetition of the same NP in the same clause, especially in similar positions. It can be concluded, therefore, that the use of non-finite clauses as modifiers is closely linked to the notion of Equi-NP-deletion and that a dangling modifier is a non-Equi-NP deletion.

But why should only equivalent NPs be deleted? This question is important because it will explain better why the deletion of non-equivalent NPs is ungrammatical. I propose that only equivalent NPs can be deleted for purposes of interpretation. Consider the sentences in (8) below:

(8) a. Hei left the hall, [PROi shouting at the top of his voice]
  b. ?He scolded the boyi [PROi shouting at the top of his voice]

PhiN 34/2005: 64

In (8a.), the bracketed subordinate clause can be interpreted as referring to "he", so the PRO in it can be co-indexed with "he". But in (8b.), PRO cannot be co-indexed with "he" because of the intervention of a false head, "boy", which could also be interpretable with PRO. Hence, as far as modifiers are concerned, PRO in a non-finite subordinate clause can normaly only be interpreted if there is no intervening NP to block it. But in (8a.), "hall" is also an NP. However, PRO in (8a.) is not co-indexed with "hall" for grammatical and even contextual reasons1. It can be seen that context, after all, plays a role in the co-indexation of PRO. "Hall" has got nothing to do with the subject of "shouting". It cannot be its external argument. To solve the problem of (8b.), there has to be an obligatory topicalisation of the subordinate clause as in (9):

(9) PROi shouting at the top of his voice, hei scolded the boy.

In the grammar of topicalisation of non-finite modifiers, there is need to account for why, in (9), PRO cannot be co-indexed with "boy". In other words, one needs to explain why the object in the main clause cannot control PRO. I posit that the control of PRO, in topicalised non-finite clauses, has a domain of application and of interpretation. Notice that PRO can only be co-indexed with the subject NP of the matrix clause, with which it forms a governing domain. In the literature, it will be said that PRO must be C-commanded. This is why (10b.) is ungrammatical:

(10) a. The principal saw Thomasi [PRO i leading the strike.]
  b *PROi leading the strike, the principal saw Thomasi.

The problem with (10b.) is that the non-finite clause is dominated by a faulty headword. Hence, the modifier cannot be preposed because the object of the main clause controls it.

As mentioned above, the problem with dangling modifiers lies in the conditions of topicalisation. It has also been shown that one condition for the movement of the non-finite modifier clause is the equivalence of the NP. Let us call this "The Co-indexation Condition".

(11) The Co-indexation Condition.
Only a non-finite modifier with a PRO co-indexable with the subject NP of the matrix clause can be topicalised.

The rule in (11) is not strong enough because it does not specify the kind of clause that should be topicalised. If the rule were to be taken at face value, it would lead to ungrammatical sentences like (12b.) and (13b.):

(12) a. The boy running on the field is very intelligent.
  b. *[Running on the field,] the boy is very intelligent.
(13) a. The people sent to discuss with me arrived late.
  b. *[Sent to discuss with me] the people arrived late.

PhiN 34/2005: 65

The b-sentences above are ungrammatical though they respect the rule in (11) above. The bracketed clauses in them are adjectival clauses. A close observation shows that they are restrictive in nature. Our grammar, therefore, needs constraints to avoid such ungrammatical structures. Let this constraint be called "The Restrictive Relative Constraint".

(14) The Restrictive Relative Constraint.
Restrictive relatives cannot be topicalised.

The rule in (14) demands that only non-restrictive, non-finite relatives should be topicalised. This is logical because non-restrictive relatives do not define or restrict the sense of the noun they modify. This is verifiable from the fact that they are always set off by a comma, or a pair of commas, to show their loose attachment to their heads. They only supply additional information about the head. The rule in (14) enables us to distinguish between (15) and (16):

(15) a. The students who passed their exams went to the university.
  b. *Having passed their exams, the students went to the university.
(16) a. The students, who passed their exams, went to the University.
  b. Having passed their exams, the students went to the university.

(15b.) is ungrammatical because it implies that all the students passed their exams, which is not the meaning of (15a.). It means that the PRO in the non-finite clause in (15b.) cannot be co-indexed with "students" because (15a.) does not mean "the students passed their exams" but "some students pass their exams", that is, we are dealing with part of the whole and not the whole itself. (16b.) is grammatical because it shows that all the students passed their exams as shown in (16a.). It can be concluded that the rules in (11) and in (14) are mutually complementary. The PRO of a restrictive relative clause is not equivalent to the subject NP of the matrix clause, which is always greater. (15b.) is a glaring example. It is believed that if a learner combines the Co-indexation Condition in (11) with the Restrictive Relative Constraint in (14), he/she will topicalise the right non-finite clause and do so duly.

The data show that the problem with modifier topicalisation in CamE lies in the violation of the co-indexation condition, which leads to non-Equi-NP-deletion. What accounts for this problem? Evidence from PRO in complement clauses shows that PRO-interpretation poses a general problem in CamE. Consider the sentences below:

(17) a. He decided that he will go alone.
  b. He refused that he will not come.
  c. I cannot go to Bamenda that I want to see him.
  d. Do you think I can go to Bamenda that I want to see him?

PhiN 34/2005: 66

The sentences in (17) are typical CamE sentences. From observation, they are preferred to those in (18):

(18) a. He decided [PROi to go alone.]
  b. He refused [PROi to come.]
  c. Ii cannot go to Bamenda [PROi to see him.]
  d. Do you think Ii can go to Bamenda [PROi to see him?]

The preference of (17) over (18) is a signal of the fact that speakers of CamE avoid infinitival complements and choose that-clauses, which, as stated in Sala (2003: 142 ff.) are very recurrent in CamE2. I propose that the avoidance of infinitival complement clauses is linked to the problem of PRO interpretation. The same issue is also evident with WH-traces and NP-traces in CamE showing the various ways users grapple with the complexity of their interpretation (see Sala 2003: 260 ff.). PRO is another empty category whose complexity is fought against by the extensive use of that-clauses. Substrate evidence from local languages and Cameroon Pidgin shows that non-restrictive relatives and non-finite modifier clauses are alien categories in Cameroonian languages (see Sala 1999: 29 ff.) for a discussion of the issue concerning Lamnso). Their absence in indigenous languages explains why they are frequently misused.

Finally, the fact that both non-restrictive and non-finite modifier clauses are not attested in Cameroonian languages means that the rules in (11) and in (15) above are strange rules to Cameroonians who learn English. Hence, there is the violation of the co-indexation condition on modifier-clause topicalisation. How then do Cameroonians understand themselves after when violating the condition, given that the result is that a false head is wrongly attached to it? It is proposed that the context of use is the only thing that helps PRO in a dangling modifier to be interpreted. This is shown in (19) below:

(19) PROi coming back from work yesterday, a car knocked himi down.

The co-indexation of PRO with "him" is just a contextual phenomenon and an ungrammatically determined one. The problem with the sentence in (19), according to exonormative standards, lies in the fact that the word that is modified in the main clause is not the subject of the sentence. Consequently, what is meant by dangling modifiers modifying nothing in the sentence is simply that they do not modify the subject NP. However, speakers of NVEs do not feel that it was the car that was coming back from work. The context seems to set this point right. A car does not go to work though it may be driven to and from work. Even if we replaced "car" by "mad man" or "policeman" as in (20), speakers of NVEs would still understand that it is neither the mad man nor the policeman that was coming back from work. Arua (1998: 145) arrives at the same conclusion in his discussion of dangling modifiers in Swazi English.

PhiN 34/2005: 67

(20) Coming back from work, a madman/policeman knocked him down.

(20) is, therefore, ambiguous because it could mean either (21a.) or (21b.).

(21) a. A madman, who was coming back from work, knocked him down.
  b. A madman knocked him down when he was coming back from work.

The context in (20) is that a madman cannot go to work, so only "him" could be coming back from work. This context makes (21b.) a more likely interpretation of (20) than (21a.). As already shown in (8a.) above, the co-indexation of an NP with PRO has something to do with context. This context dependency of PRO-co-indexation may have been overgeneralised in NVEs.


This paper has seen the notion of dangling modifiers, which stems from the lack of knowledge governing the topicalisation of non-finite clauses in British English. The grammar includes two rules, which I have called The Co-indexation Condition and The Restrictive Relative Constraint. According to The Co-indexation Condition, only non-finite modifiers with a PRO co-indexable with the subject NP of their matrix clauses can be topicalised. According to The Restrictive Relative Constraint, restrictive relatives cannot be topicalised. I have linked non-finite relatives to the concept of Equi-NP-deletion and dangling modifiers are seen to be cases of Non-Equi-NP-deletion. This involves PRO-interpretation. The difficulty of PRO-interpretation in dangling modifiers is due to the general problem of the interpretation of empty categories in CamE. However, dangling modifiers, far from being modifiers that modify nothing in the sentence, are interpreted as modifying specific constituents in the sentence by users of NVEs. I have concluded that contextual and not grammatical factors determine such an interpretation. To account for the prevalence of dangling modifiers, evidence from the languages in Cameroon show that non-finite modifier topicalisation and, therefore, the two rules governing their grammar are unknown to Cameroonians. This explains the context dependency of the interpretation of topicalised non-finite modifiers in CamE.

PhiN 34/2005: 68


Arad, Maya (1996): "A Minimalist View of the syntax-Lexical Semantics interface", in: Working Papers in Linguistics 8, Department of Phonetics and Linguistics, University college London.

Arua, A. F. (1998): "Some Syntactic Features of Swazi English", in: World Englishes 17/2, 139–151.

Hacker, Diana (1995): A Writer's Reference. New York: Bedford Books.

Haegeman, Liliane (1992): Introduction to Government and Binding Theory. Oxford: Blackwell Publishers.

Heffernan, James A. W. and John E. Lincoln (1982): Writing: A College Handbook. New York: W.W. Norton and Company, Inc.

Platt, John J. and Heidi Weber and Mian Lian Ho (1984): The New Englishes. London: Routledge and Kegan Paul plc.

Radford, Andrew (1997): Syntactic Theory and the Structure of English: A Minimalist Approach. Cambridge: CUP.

Sala, Bonaventure M.(1999): "Aspects of Clefting in Lamnso". Unpublished M.A. Dissertation, Department of Linguistics and African Languages, University of Yaounde I.

Sala, Bonaventure M. (2003): "Aspects of the Cameroon English Sentence". Unpublished Ph. D. Dissertation, Department of English, University of Yaounde I.

Simo-Bobda, Augustin (1994): Watch Your English: A Collection of Remedial Lessons on English Usage. Yaounde: Atelier Material Audio-visuel.

Strunk, William and E. B. White (1979): The Elements of Style. New York: Macmillan.

Trudgill, Peter and Jean Hannah (1985): International English: A Guide to Varieties of Standard English. London: Arnold.

Warriner, John E. (1988): English Composition and Grammar. Orlando: Harcourt Brace Jovanovich Publishers.

PhiN 34/2005: 69


* This paper is culled almost in its entirety from a Ph.D dissertation presented to the Department of English in the Faculty of Arts, Letters and Social Sciences in 2003, 312–326.

1 I have in mind the fact that "hall", from a syntactic perspective, is not an internal argument of "left". But if we consider Arad's (1996) analysis of telicity, we see that "hall" and "boy" are both telic events, by way of having "measurers" of their events.

2 In Sala (2003: 408) we hypothesise the concept of selection and avoidance as a researchable domain to account for the discrepancy between the demand and supply of English in Cameroon.