Home » Articles » LS35. Ellipsis in coordination

LS35. Ellipsis in coordination

ellipsis cowIn previous posts we’ve looked at how AGENT, THEME and GOAL may encompass more than one noun and also at multiple pairings of THEME + GOAL.  We now look at coordination of more complex groupings.  These are treated as multiple clauses with material needing to be projected from one to another.

Projecting forwards

A sentence such as (125) from LS34 may be thought of as a series of clauses of which the later ones have some material brought forward from the first and that material therefore need not be repeated in the phonological stream.

(125) John gave Mary bananas Odette damsons and…

In (125) it is subject and verb that are shared.  (122) from LS34 is similar but neater, having only two clauses:

(122) John gave Mary apples and Nancy bananas

It is possible to share just subject:

(126) John gave Mary apples and sold Nancy bananas

It is also possible to share just verb:

(127) John gave Mary apples and Ken Nancy bananas

Sentence (127) shows that ellipsis of the verb may occur with an explicit subject at the left of the clause.  But consider:

(128) John gave Mary apples and Ken bananas

Sentence (128) does not have the meaning John gave Mary apples and Ken gave Mary bananas.  Explicit Ken is to the left of omitted Mary.  Evidently ellipsis of nouns in later clauses occurs only to the left of explicit material.

The problem is accounting for the propositions delivered to cognition for the later clause.  In every case, one or more propositions include one concept from the explicit material and another concept brought forward from the first clause.

Projecting backwards

As well as material from an earlier clause being projected forwards, it is possible for material in a later clause to be projected backwards.  In (129), John is also the subject of sold and Mary is also the indirect-object of gave.

(129) John gave apples and sold bananas to Mary

Sentence (130) shows apples being projected backwards as the direct-object of gave.

(130) John gave (to) Mary and sold (to) Nancy apples

The simplest explanation is that if cognition has VERB1 / AGENT / NOUN1, then delivering VERB2 / COORD / VERB1 entails VERB2 / AGENT / NOUN1 unless overridden by another noun local to verb2.  And similarly if cognition has VERB2 / COORD / VERB1, then delivering VERB2 / THEME (or GOAL) / NOUN1 entails VERB1 / THEME (or GOAL) / NOUN1 unless overridden.

This is feasible if it is assumed that COORD is a ‘reversible’ REL, i.e. the two concepts it links can be either QUO and SIC or SIC and QUO.  Sentence (129) delivers propositions arranged thus:

SELL / AGENT / JOHN and GIVE / GOAL / MARY are enabled via the COORD link.  But SELL / THEME / APPLE is overridden by SELL / THEME / BANANA, and GIVE / THEME / BANANA is overridden by GIVE / THEME / APPLE.

The diagram makes it look as if SELL forms part of the realisation of GIVE and vice versa.  Both realisations are activated at almost the same instant, so this may not be an issue.  But if cognition can’t cope with that, then the simple solution would be for the GIVE and SELL nodes shown here to be temporary tokens each linked to the appropriate permanent concept.  Those links would use the same distinctive relation (TYPE perhaps) and the two links would override each other, in the same way as do the THEME links.


The analysis for (129) is as follows.

(129) John gave apples and sold bananas to Mary

SELL / AGENT / JOHN and GIVE / GOAL / MARY do not appear in the table.  They take effect by the delivery to cognition of SELL / COORD / GIVE combined with GIVE / AGENT / JOHN and with SELL / GOAL / MARY.

Type of coordination

There are two new ideas in the analysis of (129).  The first is that the junction apples__and delivers directly a proposition showing the type of coordination currently in play (differentiating from or, for example).  Note that and is the dependent here.  In pre-verb contexts it is always parent in (noun)__and junctions.  The apple concept plays no part in the proposition; (any word)__and is the fastest way to derive the proposition and apples happens to be the closest word.

The type of coordination can change within a string of items:

(131) John gave Mary apples and Nancy bananas or Odette cherries

This approach suggests that the scope of coordination is determined by there being only one conjunction effective at one time in left-to-right processing.  In (131), brackets that might be explicit in a formal language are notionally around Mary apples and Nancy bananas, not around Nancy bananas or Odette cherries.  This is a big topic but will not be pursued further in LanguidSlog.

There is no and__Nancy junction with and as parent.  In any case, Nancy as dependent in that junction would rule out gave__Nancy, which is essential for the analysis.

One interesting consequence is that, even without and, coordination can occur – although of an unspecified type.  Presumably COORD / HASP / AND is the mind’s default setting.  This aligns with native-speaker intuition that, with and nowhere, John gave Mary apples Nancy bananas Odette cherries is OK.

At clause-end

The other new idea in the analysis of (129) concerns what else happens when the first clause ends.  The apples__and junction marks the end of the clause.  Clause-end (the heavy horizontal line in the table) forces consolidation of APPLE on THEME – in the same way that sentence-end would.

At that point some or all of AGENT, THEME and GOAL propositions have been delivered.  In the following clause some or all of AGENT, THEME and GOAL propositions will be (re)constituted and delivered.

Propositions 1, 2 and 3 are therefore cleared of activation and SIC concepts in readiness for that.  In general, the QUO and REL concepts are retained because the same verb may be used in the following clause.  This is reasonable because the QUO concept is from the verb which will not have acted as dependent in any junction.  Therefore the verb will not have had its activation depleted (an assumption already made in LanguidSlog).

For (129) this is not shown because the start of the clause is immediately signalled by a new verb, needing its own set of arguments.  In (129) another ditransitive, sold, has the same set as would have been brought forward from the gave clause.

Depletion of the activation on the complement material for the first clause presumably means it disappears from the sentence being processed.  For (122), this would leave John gave Nancy bananas.  There is no searching for (noun)__(noun) junctions between complements and no need to account for Mary__Nancy and apples__bananas (both of which must be available) not being actioned.

How the material disappears needs further thought.  There can be an adjunct (preposition phrase or subordinate clause) following the first clause’s second noun and qualifying it.  The last noun cannot disappear before the junction connecting the adjunct has been processed, but the adjunct material also needs to disappear to avoid spurious junctions being actioned later.

More to come

At this point a means by which material can be projected from one clause to another has been established.  Next week the discussion will focus on how the material that is explicitly coordinated is used to form the correct propositions.  And after that, at least two more LanguidSlog posts will be needed to finish outlining coordination.

Mr Nice-Guy