Chapter 8 Number disagreement between verb and subject in Shakespearean English

In modern English, the grammatical concord in number between the verb and its subject prescribes the plural verb form when:

  1. the nominal subject is in plural;

  2. the nominal subject contains an enumeration (e.g. mother, father, and children);

  3. the nominal subject is a collective noun (e.g. police, committee)2.

In Shakespeare’s times, the grammatical concord in the verb and its nominal subject was not fully established yet, allowing for singular in cases unacceptable for modern English.

8.1 Extraction Strategy in Universal Dependencies

We want to find cases where a verb is in singular whereas its subject would suggest plural in the modern English. Then we will manually examine the concordances and try to find patterns.

Constraints on the verb

  1. The verb must be a verb and not a nominal part of a copula predicate3 (upos = 'VERB').

  2. The verb must be in the singular (feats/Number = 'Sing') and in the third person (feats/Person='3').

Constraints on the subject

The subjects of interest can take several forms.

  1. The subject must be a nominal subject, not a clausal subject. When we disregard coordinations, then the constraint is simply deprel = 'nsubj' . For the singular in verb to be interesting, this nominal subject should be in plural (feats/Number = 'Plur').

  2. When we want to consider second and further members of coordinations (e.g. a man and a woman), we must look for a node that is not a direct child but grandchild (descendant one step down from child) and has a parent with deprel = 'nsubj'.

PMLTQ query

Let’s build the query from the perspective of the verb and call the verb node $verb3. Its definition and attributes are going to look like this:

tok $verb3 := [upos='VERB', feats/Number='Sing', feats/Person='3']

To keep the queries simple, let’s create a separate query for each subject option.

The first option is going to be a nominal subject in the plural:

tok $nsubjPlur := [feats/Number = 'Plur', deprel='nobj'].

Node $nsubjPlur is a child of $verb3, and therefore we embed it in the $verb3 node:

tok $verb3 := [upos='VERB', feats/Number='Sing', feats/Person='3',

tok $nsubjPlur := [feats/Number = 'Plur', deprel='nsubj']

]

The second option is going to be coordinated subjects. A very simple way to draw a subtree that would capture all members of the coordination under the subject would be explicitly mentioning the subject as a node $nsubj with deprel='nsubj'and let it have a descendant $coordsubjwith deprel='conj'. The nesting looks like this:

tok $verb3 := [upos='VERB', feats/Number='Sing', feats/Person='3',

tok $nsubj := [deprel='nsubj',

descendant tok $coordsubj := [deprel='conj']]

]

When the coordination (enumeration) has more than two elements, it should produce two hits: one with the subject and the first coordination member and one with the subject and the second coordination member.

Good hits:

So indeed all disquiet, horror, and perturbation follows her.

My shame and guilt confounds me.

Your patience and your virtue well deserves it.

There is Douglas and Lord Mortimer.

Small cheer and great welcome makes a merry feast.

Time and the hour runs through the roughest day.

A French song and a fiddle has no fellow.

There is no end, no limit, measure, bound,

There is reasons and causes for it.

There is tears for his love, joy for his fortune, honor for his valor, and death for his ambition.

This query could be further tweaked because it produces noise:

I hope it is some pardon or reprieve. A verb in singular with subject in coordination with or is not interesting.

Fear, and not love, begets his penitence. And neither is a coordination with negation, which is a sort of apposition, as shown here.

There is no ancient gentlemen but gard’ners, ditchers, and grave-makers.

To tighten the constraints, we forbid negation on the subject and the conjunctions or and but as children and siblings on its descendants:

tok $verb3 := [upos='VERB', feats/Number='Sing', feats/Person='3',

tok $nsubj := [deprel='nsubj',

0x tok [lemma = "no"],

descendant tok $coordsubj := [deprel='conj',

0x tok [lemma ~ "(or|but)", deprel = 'cc'],

0x sibling tok [lemma ~ "(or|but)", deprel = 'cc']

]]

]

Xie, Yihui. 2015. Dynamic Documents with R and Knitr. 2nd ed. Boca Raton, Florida: Chapman; Hall/CRC. http://yihui.org/knitr/.
———. 2021. Bookdown: Authoring Books and Technical Documents with r Markdown. https://CRAN.R-project.org/package=bookdown.

  1. There is a grey area in mass nouns: when they are modified by plural nouns denoting individuals, the verb is often in plural, agreeing with the modifier nouns rather than with the mass noun, e.g. a handful of leaflets (this phenomenon is also known as transparent head).↩︎

  2. This is to prevent noise from a frequently observed tagging error in the UDPipe version used at the moment: the nominal part of a copula predicate correctly governs a copula verb with the proper dependency relation, but is still tagged as a verb and hence even has the Number feature, so that it makes a match.↩︎