Home » Articles » NG57. Network Grammar: so far, so good

NG57. Network Grammar: so far, so good

Guy FawkesYes, this is number 57 of the 57 pieces promised in NG1.  The number was arbitrary, chosen for rhetorical effect and because I needed a target.  I knew roughly the ground to be covered but only a few weeks’ worth had been drafted at the start.  Perhaps too many weeks have been spent on some topics, and certainly too few on some others – for example, on production.

Nevertheless it’s fair to say that the important principles have been covered and also that the set of principles is small and has not grown in proportion to the amount of syntax covered.  There’s a lot more syntax to cover – and NG must be tried on other languages.  Perhaps more principles will be required then.  We’ll see.


The principles of Network Grammar are:

  • a mental network holds knowledge, innate and acquired
  • each node in the network is the locus of a unique concept
  • a concept is realised by progression along the paths fanning out from the node
  • a proposition comprises three concepts
  • language knowledge is held as special types of proposition – words and rules
  • words and rules are connected by category nodes, of which there are probably thousands to enable a many-to-many relationship between phonological words and meanings, and to allow common words also to have idiomatic meanings
  • a junction of two not-necessarily-adjacent words creates a proposition consisting of the two meanings joined by a syntactic/semantic relation
  • the words in a sentence are processed in left to right sequence, with junctions being identified by scanning backwards (right to left) from each word
  • each junction has one word as parent, the other as dependent
  • an incoming word brings activation the one-and-only time in the sentence it forms a dependent; a word may also act zero, one or more times as a parent in the sentence
  • a junction may form more than one incomplete proposition (therefore not immediately delivered to cognition), allowing completion as a result of a later junction in the left-to-right processing of the sentence; this is important for the correct allocation of words to argument roles
  • there is the possibility of a word’s category changing because of its participation in a proposition, thereby affecting the possibilities for its participation in later propositions; important for the correct allocation of words to adjunct roles.

Network Grammar has shown how these principles are enough to account for much of English syntax, exploding the myths of orthodox theory. The still-sceptical reader might try applying them to a topic I’ve not yet explored – for example, ‘extraction islands’ (see Wikipedia on ‘wh-movement’).

You should see that NG easily explains why, for example, a ‘WH-expression can occur at the front of a sentence regardless of how far away its canonical location is’.  Wikipedia then gives:

(245) Who does Carl believe that Bob knows that Mary likes ___ ?

The verbs believe, know and like all allow either an NP or a clause to follow.  The junctions who__believes creates an incomplete propositions that dies when the clause following the verb is encountered; who__knows likewise; but who__likes gets completed because there is no following clause.

Wikipedia then shows a variety of ungrammatical sentences.  One example is (247), an interrogative formed by ‘moving’ (allegedly) the WH-word to the front from the position the corresponding NP would take in the declarative (246).

(246) Susan asked why Sam was waiting for Fred

(247) * Who did Susan ask why Sam was waiting for ___ ?

The verb ask can allow an NP and a clause to follow.  The junction who__ask creates an incomplete proposition but this one is completed when the following clause is encountered.  The sequence is perfectly grammatical until the preposition for which is left without a complement.

OK, this is only a start on a big topic.  So see if you can find any examples where this approach doesn’t work.  Perhaps you will – and perhaps I’ll have to find another principle or two.


Some theoreticians have responded while LanguidSlog developed.  I’m grateful for their input.  However all have ultimately been sceptical despite my answers to their challenges on NG.  And no one has answered my challenges on the assumptions made in orthodox theories (although there are some lengthy exchanges of comments – attached to NG2 and NG4 for example).

Professionals must be cautious because they have careers to pursue and the linguistics community has often been cruel to anyone proposing anything unorthodox.  Noam Chomsky once wrote:

“The smart way to keep people passive and obedient is to strictly limit the spectrum of acceptable opinion but allow a very lively debate within that spectrum.”

He was writing about politics and presumably didn’t intend the irony when you apply the same words to linguistics.  Is there anyone out there brave enough to move linguistics towards a believable theory that will allow a rapprochement with other disciplines?


The next thing planned is to refine the existing material into a book.  I’ll also be promoting the ideas by mailshots to selected academics – in neuroscience, cognitive science, psychology, computing as well as in theoretical linguistics.

The aim of LanguidSlog has not been to present a slam dunk, but to show that further exploration of NG is justified.  With other things on the bucket list, I’m reluctant to continue alone without commitment from someone in academia.

Does anyone want to do that?

Look, your man is the giant from whose shoulders progress would be made, but I can’t resist a jokey envoi.  Noam or . . .