The extracts below are from Martin, J.R. 2013 Systemic Functional Grammar: A Next Step Into The Theory — Axial Relations
Saussure's concept of the sign is regularly mistaken for a conception he was in fact struggling against — namely the common sense idea that a symbol stands for a concept. …
Saussure does his best, in other words, to establish the sign as the bond between signifié [signified] and signifiant [signifier]. One important implication of this is that unless one is introducing Saussure's concept of the sign, as we are doing here, there is no need in linguistics or semiotics to refer to either the signifié or signifiant as distinct entities. What matters in linguistics and semiotics is the sign.
If we return to our initial image of the sign for a moment, we can see that what Hjelmslev is developing is the idea that the line we drew binding signifié with signifiant needs to be theorised with enough brea[d]th and depth to accommodate the complexity of the relations involved. It might help in this regard to use the analogy of a coin rather than a piece of paper: the coin binds heads with tails, but with a little more room than a piece of paper has between one side and the other. Linguists and semioticians work in this space, developing ways of formalising the network of relations binding 'heads and tails' (signifié and signifiant). … We've expanded this workspace to highlight what we mean in the two-dimensional diagram below. …
 This is precisely what Saussure does mean by a sign; see the quotes in the previous post. As Saussure (1915: 67) explicitly states:
I call the combination of a concept and a sound-image a sign
 Martin misconstrues Saussure's sign, two levels of symbolic abstraction (in SFL: stratification), as the "bond" between the levels (in SFL: realisation). As Saussure (1915: 67) explicitly states:
I propose to retain the word sign [signe] to designate the whole
The question that arises from Martin's misinterpretation is: What is the epistemological status of signifié (e.g. 'go') and signifiant (e.g. 'green') if they are external to linguistics and semiotics? What is it that is being "bonded" by linguistics and semiotics?