Elsevier

Cognition

Volume 198, May 2020, 104161
Cognition

Original articles
The puzzle of number agreement with disjunction

https://doi.org/10.1016/j.cognition.2019.104161Get rights and content

Abstract

In English, when two nouns in a disjunctive subject differ in number (e.g., the dogs or the cat), the verb tends to agree with the number of the nearer noun. This is exceptional, as a noun's linear proximity to the verb does not generally play a role in agreement. In the present study, we investigate a further puzzle about agreement with disjunction, namely, the existence of a pattern in which two singular disjuncts trigger plural agreement (e.g., The lawyer or the accountant are…). Two eyetracking studies in English show that plural agreement with a disjunction of singulars does not reliably disrupt readers' eye movements, in contrast to the immediate disruptive effect of other agreement violations. Three off-line rating studies in English show that plural agreement results in only a small decrement in acceptability, compared to other agreement violations, and that in some structural configurations there is no decrement at all. On the whole, the data do not support the hypothesis that plural agreement is licensed only when or has an inclusive reading; even when it has an exclusive reading, there is only a small penalty for plural agreement. Finally, we explored this issue in Italian, which has a richer system of inflectional morphology. Italian speakers showed a plural preference in a completion experiment, and singular and plural agreement did not differ in acceptability in a rating experiment. We conclude that agreement with disjunction is a grammatical lacuna or gap, in the sense that speakers' grammar simply does not prescribe a verb number following a disjunctive subject.

Introduction

Subject-verb agreement, in English and other languages, is notoriously prone to error in both production and comprehension. Speakers sometimes produce a verb that agrees in number with a noun intervening between the head of the subject phrase and the verb, as in (1a), a phenomenon known as agreement attraction (e.g., Bock & Miller, 1991). The presence of an intervening attractor can also make an agreement error less salient in comprehension (e.g., Wagers, Lau, & Phillips, 2009). Examples of error types in other structural configurations are in (1b; e.g., Bock & Miller, 1991; Staub, 2010) and (1c; Dillon, Staub, Levy, & Clifton Jr, 2017).

  • (1)

    a.*The key to the cabinets are on the table.

    b *The cabinets that the key open are in the laboratory.

    c *Which keys are the scientist using?

Agreement attraction is usually attributed to the processor rather than to the grammar. The use of a plural verb in (1) does not seem to reflect dialectal variation, or a gradient or indeterminate grammar. In all of these cases the verb should agree in number with the head of the syntactic subject (key in 1a and 1b, scientist in 1c); a speaker who produces one of these errors will generally agree, upon reflection, that the verb should be singular. Because these are regarded as processing errors, investigation of their causes has taken place within the psycholinguistic literature, which contains a variety of mechanistic accounts (Eberhard, Cutting, & Bock, 2005; Staub, 2009; Wagers et al., 2009).

However, other forms of variable agreement behavior may be attributed to optionality, indeterminacy, or variability in the grammar, rather than to errors in the cognitive processes underlying sentence production or comprehension. The influence of notional number on agreement gives rise to one class of examples. Subject nouns that are grammatically singular but denote a collective, such as family, sometimes trigger plural agreement in American English (e.g., Bock, Nicol, & Cutting, 1999; Humphreys & Bock, 2005), and generally trigger plural agreement in British English (Bock et al., 2006), and this variability has not usually been interpreted as reflecting processing error.

Agreement variability also arises when the syntactic subject is a coordinated phrase, headed by and or or, and this too has generally received an explanation in grammatical rather than processing terms. A widely observed tendency, cross-linguistically, is for the verb to agree in gender, number, or other features with the nearer conjunct or disjunct, a phenomenon known as closest conjunct agreement. This tendency is at odds with the general observation that nouns that are linearly close to the verb do not have a particular influence on agreement with non-coordinated subjects (i.e., the number attraction effect as in (1a) is not attributable to the proximity of the attractor and the verb; Franck, Vigliocco, & Nicol, 2002; Keung & Staub, 2018), and has important theoretical implications that go beyond the scope of this paper; see Nevins and Weisser (2018) for a recent review, and for a discussion of implications of closest conjunct agreement for the syntactic analysis of coordinate structures in terms of the structural relations inside the conjunction phrase and of the phases involved in agreement operations. A possibility is that closest conjunct agreement arises because the head of the subject phrase, the connective or or and, lacks a morphosyntactic number feature that can directly value the verb (Nevins & Weisser, 2018). However, this agreement pattern is often if not always optional, co-existing within a language, or within an individual speaker, with the use of other agreement paradigms; see Willer et al. (2018) for recent empirical work demonstrating this optionality within and across dialects.

In English, the tendency for the verb to agree in number with the nearer noun is most easily observed when the subject is a disjunction. Prescriptive style guides (e.g. Fowler & Aaron, 2007) enforce the rule that if the subject is preverbal, the verb should agree in number with the second noun, giving rise to the patterns in (2):

  • (2)

    a. The cat or the dogs *is/are going to leave.

    b The dogs or the cat is/?are going to leave.

Using an elicited production task, Haskell and MacDonald (2005) confirmed that speakers do generally follow this rule, producing a singular verb 2% of the time in cases like (2a), compared to 72% of the time in cases like (2b). Keung and Staub (2018) replicated this pattern in a task requiring subjects to make a speeded choice of verb form after reading a subject phrase presented one word at a time; a singular verb was chosen on about 20% and 60% of trials in the two conditions, respectively. There is an asymmetry here, as in both data sets the ‘agree nearest’ rule is more consistently applied when the nearer noun is plural; in both Haskell and MacDonald (2005) and Keung and Staub (2018), plural verbs were more common after plural-or-singular subjects than were singular verbs after singular-or-plural subjects. We assume that this may be due to the markedness of the plural feature, which plays an important role in determining patterns of agreement attraction (Bock & Miller, 1991; Eberhard, 1997). Many theorists have proposed a ‘privative’ account of grammatical number, with singular being an unmarked or default value, and only plural marking involving the presence of a distinct feature. The same asymmetry is implied by the prescriptive advice in Fowler & Aaron, 2007, who suggest reordering a plural-or-singular disjunction in favor of singular-or-plural, on the grounds that there is no fully satisfactory agreement pattern in the former case.

The ‘agree nearest’ rule clearly implies that when the subject is a disjunction of singulars, the verb should be singular. Zwicky (2009) writes that this prescription is “utterly uncontroversial,” and he doubts that “anyone needs to be told what to do when confronted by disjunctive subjects of the same number.” Zwicky (2009) (see also McCawley, 1998) proposes that the general principle governing agreement with disjunction is that the verb must agree with each of the disjuncts, and that agreement with the nearer disjunct is deployed as a resolution principle only in cases where this basic principle cannot be applied, because the disjuncts differ in number. When both disjuncts have the same number, the grammar is unambiguous.

However, the starting point of the present study is the observation that variability in agreement with disjunctive subjects does not appear to be restricted to cases in which the two nouns differ in number. An informal search of the Corpus of Contemporary American English (COCA; Davies, 2008) confirms that singular agreement is more common than plural when the subject is a disjunction of singulars, but turns up several examples of plural agreement. Shown in (3) are three examples from COCA; several additional examples are reported in Ivlieva (2012):

  • (3)

    a. It would be implausible to maintain that either the assumption or the implication are wholly wrong, but I do not believe they are wholly right, either.

    b. That understanding is gained long before the patient or the surgeon have entered the operating room.

    c. I think the economy’s going to turn around in any case, regardless of what the administration or the president do.1

Plural agreement with a disjunction of singulars has also been attested in the laboratory. In the speeded choice task used by Keung and Staub (2018), subjects selected a plural verb about 20% of the time after a disjunction of singular noun phrases such as the maid or the butler. This pattern cannot be attributed simply to task demands or inaccurate responding; subjects selected a singular verb only about 5% of the time after a disjunction of plurals, and selected a plural verb only about 3% of the time when a singular subject was followed by a singular noun in a prepositional phrase, e.g., the helicopter for the flight. It appears that speakers do not actually treat a singular-or-singular subject phrase as unequivocally singular, for the purpose of computing verb agreement.

In fact, the linguistic literature has noted the use of plural agreement with a disjunction of singulars. Peterson (1986) reports that speakers differ in whether they accept singular or plural agreement, and also that “some speakers [accept] both at different times” (Peterson, 2004, p. 670; see also Eggert, 2002; Morgan, 1985). Here we investigate this phenomenon in detail. One possibility is that naturally occurring instances of plural agreement with disjunction like those in (3) are simply errors, on a par with attraction errors, and that the 20% plural responding obtained by Keung and Staub (2018) is due to the demands of a speeded, and unnatural, laboratory task. If so, then plural agreement with a disjunction of singulars will be unacceptable for most English speakers, and will be processed as an agreement error, at least most of the time. However, there are other salient possibilities, which do not presuppose that plural agreement with disjunction is ruled out by speakers' grammar. One possibility is that there is systematic individual variation in agreement preference with disjoined subjects, with some speakers requiring singular, more-or-less categorically, and other speakers requiring plural. But it is also possible that both singular and plural are somewhat degraded, for a given speaker, or that neither is degraded, i.e., both are acceptable. These last two potential patterns are especially interesting, as they would suggest that agreement with disjunction is a grammatical lacuna or gap, in the sense that a speaker's grammar simply does not prescribe a determinate verb number when the subject is a disjunction of singulars.

A further goal of the present work is to evaluate a specific hypothesis about a factor that may modulate the acceptability or processing of plural agreement with disjunction. According to this hypothesis, the notional number of a disjunctive subject determines the use and acceptability of plural agreement. This hypothesis builds on the fact that disjunction is notionally singular if interpreted exclusively, but may be notionally plural if interpreted inclusively.

To introduce this hypothesis, it is necessary to provide minimal background regarding the semantics and pragmatics of disjunction. These issues are discussed in detail from a theoretical perspective in Chierchia (2013), and have recently been explored experimentally in a variety of studies with children and adults (see Sauerland & Yatsushiro, 2018, for an alternative account and an overview of experimental studies). Natural language or is, in a typical or positive context, interpreted exclusively. For example, the speaker of (4a) would be interpreted as asserting that Chuck will meet with Brian or Lyn, but not both. This interpretation is often attributed to a Gricean (Grice, 1991) process of pragmatic inference, in which the logical, inclusive meaning of or is strengthened to an exclusive interpretation. However, a disjunction that occurs in certain contexts, such as in the antecedent of a conditional (4b) or in the scope of negation (4c), receives an inclusive interpretation according to which both disjuncts might be true at the same time.

  • (4)

    a. Chuck will meet with Brian or Lyn after lunch.

    b. If Chuck meets with Brian or Lyn after lunch, he’ll miss the colloquium.

    c. Chuck won’t meet with Brian or Lyn after lunch.

In (4b), the speaker is naturally understood as asserting that Chuck will miss the colloquium if he meets with Brian, with Lyn, or with both. In (4c), the speaker is understood as asserting (at least in English; cf. Szabolcsi & Haddican, 2004, for evidence of cross-linguistic variation) that Chuck will not meet with Brian, he will not meet with Lyn, and he will not meet with both. These contexts, in which the pragmatic strengthening of or is cancelled or suspended, are known as downward entailing contexts, for reasons that go beyond the scope of the present paper.

It has been pointed out by several previous authors that a disjunction of singulars may trigger a plural verb when the disjunction occurs in a downward entailing context, and is therefore interpreted inclusively rather than exclusively. Ivlieva (2012) reports the following intuitive contrast in Russian, in which plural agreement is acceptable only when a disjunctive subject is embedded under negation, in which case singular may be degraded:

  • (5)

    a. Petja ili Vasja prišël[+sing] /*prišli[+plu] [Petja or Vasja came].

    b. Ja ne dumaju, čto Petja ili Vasja ?prišël[+sing]/prišli[+plu] [I don’t think that Petja or Vasja came]

Morgan (1985) conjectures that a similar contrast exists in English.

  • (6)

    a. John or Bill is/are going to win the race.

    b. I don’t think that John or Bill are/?is going to win the race.

Peterson (1986) reports data from a small survey that confirms these intuitions, with a majority of respondents preferring singular in (6a) and plural in (6b).

To recap, plural agreement with a disjunction of singulars may be used by speakers, and may be acceptable to comprehenders, primarily when the disjunction has an inclusive reading. This would suggest that agreement with disjunction is sensitive to the subject's notional number, and that when a disjunction receives an inclusive reading, it is interpreted as notionally plural, at least for the purpose of computing agreement. One possible variant of this hypothesis, which we also test in the present study, would hold that the verb's number is actually used by comprehenders as a cue to the interpretation of a disjunctive subject as exclusive or inclusive.

Finally, yet a third issue that we address here is the extent to which indeterminate or variable agreement with disjunction is present cross-linguistically. It is possible that the apparent indeterminacy of agreement with disjunction in English is due, at least in part, to the fact that English has relatively impoverished inflectional morphology, and subject-verb number agreement in English does not play an important role in determining either a syntactic analysis of the sentence or the identity of the subject. We examine Italian, which has richer system of agreement morphology than does English, with all verbs in Italian being inflected for number. Number agreement production in Italian has been investigated in the psycholinguistic literature (e.g., Franck, Lassi, Frauenfelder, & Rizzi, 2006; Vigliocco, Butterworth, & Semenza, 1995), although no study has investigated coordinated subjects. Agreement morphology arguably plays a critical role in syntactic and referential processing in Italian. Unlike in English, the subject is not obligatorily expressed in Italian, as it is a pro-drop language, and as a result the verb's number is in some circumstances informative about the identity of the subject. Because of these features, Italian speakers might be more sensitive to subject-verb number agreement, and might have more definitive number agreement preferences.2

Interestingly, informal canvassing of Italian native speakers reveals both no knowledge of any prescriptive rule for agreement with disjunction, and variability both within and between speakers in their preferences. This variability can be observed in written translations. Consider two different translations of the clause in (7a) below, from Alice's Adventures in Wonderland (Carroll, 1865). Because all verb forms are marked for number, the option of an unmarked form, like the English had, is not available in Italian. In one translation (7b) the verb is singular, while in the other (7c) it is plural3:

  • (7)

    a. (…) as if a dish or kettle had been broken to pieces.

    b. (…) come se un piatto o una caldaia andasse[+sing] in pezzi.

    c. (…) come se un piatto o una caldaia andassero[+plu] in pezzi.

  • (8)

    a. The European Parliament or the Council can seek the opinion of the Board during the entire procedure.

    b. Il Parlamento europeo o il Consiglio può[+sing] chiedere il parere della Commissione durante tutta la procedura.

    c. Il Parlamento europeo o il Consiglio possono[+plu] chiedere il parere della Commissione durante tutta la procedura.

We report seven experiments addressing these issues. We note before proceeding that there are a number of previous attempts to address the issue of subject-verb agreement with disjunctive subjects in rating or judgment studies, including Peterson (1986), Eggert (2002), Morgan and Green (2005), and Garley (2008). These studies have made use of very small samples, mostly of linguistics students. We regard the present study as the first to investigate these questions using modern psycholinguistic methods. Two of the present experiments (Experiments 1 and 3) are English eyetracking-during-reading experiments that investigate the processing of agreement with a disjunction of singulars during online comprehension. Three (Experiments 2, 4, and 5) are English rating experiments that investigate whether speakers' off-line judgments show sensitivity to agreement with a disjunction of singulars. In the last two experiments we turn to Italian; we present one sentence continuation study (Experiment 6), and one rating study (Experiment 7).

To anticipate our empirical conclusions, we find that in English, there is no reliable penalty for plural agreement with a disjunction of singulars in on-line reading measures, and only a small penalty in off-line ratings. We also find that both ratings and eye movements seem to be affected only very little, if at all, by factors that might encourage or discourage an inclusive reading of a disjunctive subject. Finally, we find that in Italian, which lacks a prescriptive rule in this regard, there is a preference for plural agreement following a disjunction of singulars in speakers' completions, but no clear preference for singular or plural agreement is expressed in ratings.

Section snippets

Experiment 1

Experiment 1 is an English eyetracking-during-reading study examining readers' sensitivity to number agreement when the matrix subject is a disjunction of singular definite noun phrases. We created a total of 12 experimental conditions, by manipulating subject type (disjunctive; conjunctive; singular; plural) and verb number (singular; plural; unmarked), as in (9):

  • (9)

    a. The lawyer or the accountant is coming/are coming/will come to the meeting.

    b. The lawyer and the accountant is coming/are

Experiment 2

Experiment 2 used the same materials as Experiment 1, but in a rating study, to assess whether off-line judgments show a sensitivity to agreement with disjunction.

Experiment 3

In Experiment 1, readers' eye movements showed only delayed sensitivity to verb number following a subject that is a disjunction of singulars. Experiment 3 provides an opportunity to replicate this finding, and also explores the potential role of the subject's syntactic position and its interaction with semantic-pragmatic factors. The processing of agreement with disjunction may be modulated when the disjunction occurs in the antecedent of a conditional, or when it is embedded under an attitude

Experiment 4

Experiment 4 was a rating experiment using the same manipulations as in Experiment 3. Because disjunctive subjects appeared in all of the critical items, we were concerned that in the absence of an unusually large number of fillers these would be highly salient in a rating experiment. Thus, we used only eight of the 24 items from Experiment 3, with each participant rating only a single item in each of the eight conditions.

Experiment 5

Experiment 5 tested the acceptability of plural agreement when the predicate is designed to rule out an inclusive reading of or, because it can be predicated of only one entity, such as is/are going to become the next CEO of the company. To the extent that plural agreement enforces an inclusive reading, it should be incompatible with such predicates, leading to reduced acceptability. We created items such as (11), in which the predicate either did (11a) or did not (11b) allow an interpretation

Experiment 6

In the English rating studies reported above, we found only a weak preference for a singular verb when the subject is a disjunction of singulars, which was eliminated altogether in some structural configurations. In eyetracking-during-reading experiments, neither verb number resulted in notable disruption in the course of incremental processing. The most general conclusion from these experiments is that English speakers are highly tolerant of both singular and plural agreement with a

Experiment 7

In Experiment 7 Italian speakers provided ratings of singular and plural agreement with disjunction, as well as with conjoined, singular, and plural subjects, in a design mirroring that of Experiment 2 conducted in English. Unmarked verb conditions were not included, as these forms are absent in Italian.

General discussion

In this paper we presented two eyetracking-during-reading studies in English, four judgment studies in English and Italian, and one completion study in Italian to assess the processing and acceptability of singular and plural agreement following a subject that is a disjunction of singular definite noun phrases. In the eyetracking studies we found that plural agreement with a disjunction of singulars did not disrupt readers' eye movements, with the exception of a downstream effect that appeared

Conclusion

The present eyetracking, rating, and completion studies may be regarded as providing psycholinguistic validation of a claim about agreement with disjunction that is not new, going back to Peterson (2004) and Reis (1974), but which lacked empirical support. This is the claim that agreement with disjunction is an unusual, perhaps unique, grammatical lacuna: Neither singular nor plural agreement is ungrammatical, and the choice of verb number does not have clear interpretive consequences.

Authors' contribution

The two authors contributed equally to this work.

References (53)

  • G. Vigliocco et al.

    Constructing subject-verb agreement in speech: The role of semantic and morphological factors

    Journal of Memory and Language

    (1995)
  • M.W. Wagers et al.

    Agreement attraction in comprehension: Representations and processes

    Journal of Memory and Language

    (2009)
  • M. Baroni et al.

    The WaCky wide web: A collection of very large linguistically processed web-crawled corpora

    Language Resources and Evaluation

    (2009)
  • J.K. Bock et al.

    The ties that bind: Creating number agreement in speech

    Journal of Memory and Language

    (1999)
  • K. Bock et al.

    Number agreement in British and American English: Disagreeing to agree collectively

    Language

    (2006)
  • D. Braze et al.

    Readers' eye movements distinguish anomalies of form and content

    Journal of Psycholinguistic Research

    (2002)
  • L. Carroll

    Alice's adventures in wonderland

    (1865)
  • G. Chierchia

    Logic in grammar: Polarity, free choice, and intervention

    (2013)
  • R.H.B. Christensen

    Ordinal - Regression models for ordinal data

  • C. Clifton et al.

    Evaluation of the epistemic state of the speaker/author

    The Quarterly Journal of Experimental Psychology

    (2017)
  • M. Davies

    The corpus of contemporary American English (COCA): 560 million words, 1990–present

  • B. Dillon et al.

    Which noun phrases is the verb supposed to agree with?: Object agreement in American English

    Language

    (2017)
  • A. Drummond

    Ibex farm

  • K.M. Eberhard et al.

    Making syntax of sense: Number agreement in sentence production

    Psychological Review

    (2005)
  • R. Eggert

    Disconcordance: The syntax, semantics, and pragmatics of or-agreement

    (2002)
  • H.R. Fowler et al.

    The little, brown handbook

    (2007)
  • View full text