Martin (1992: 427):
The problem of overlapping strings and chains can be illustrated with respect to the domicile string in [6:33] which overlaps with two distinct reference chains, one identifying the castle and one identifying the cottage:
domicile string: castle–castle–home–woods–cottage–inside–home–inside domicile chains: a castle–the castle–the woods this little cottage–there–here
The strategy adopted here to overcome this difficulty will be to analyse cohesive harmony from the point of view of reference chains as far as people, places and things are concerned and from the point of view of lexical strings for actions and qualities. The strategy it must be noted has strong effects on the kind of string/chain interaction recognised for a text.
 This nicely illustrates how the 'problem' of overlapping lexical strings and reference chains arises directly from Martin's misunderstanding of the cohesive system of reference.
As previously explained, the overlapping of Martin's lexical strings and reference chains arises from Martin's discourse semantic system of identification — 'the semantics of reference' — confusing two distinct types of cohesion, reference and lexical cohesion, largely as a consequence of confusing the referent with the system of reference. See previous critiques here, here, here, here and here.
 In SFL theory, reference is a cohesive system for marking the textual status of an element as identifiable (Halliday & Matthiessen 2004: 550). There are two types: co-reference (personal or demonstative) and comparative (general or specific). Reference is a grammatical cohesive system, rather than a lexical cohesive system. Here, because of the confusion explained in , Martin has included lexically cohesive relations in his reference chain, which thus accounts for the string/chain overlap.
 On the basis of what was said above, this amounts to arbitrarily splitting up lexical cohesion into two kinds of cohesive links, and mistaking one of them for grammatical reference.
 The consequences for this arbitrary division are less dire when it is understood that strings and chains are actually both largely concerned with lexical cohesion.