"This process of substitution...functions as a pure play of traces...[it] operates within the order of the pure signifier which no reality, no absolutely external reference...can come to limit, bound, or control; this substitutionŠcan go on infinitely in the element of the linguistic permutation of substitutes, of substitutes for substitutes...(Dissemination 89)."
Derrida speaks of the way in which words naturally refer to or "cite" one another. A simple example has everything to do with etymology. The word "text" is derived from the Latin verb "texere," meaning "to weave." Of course, the object "text" is quite different from the act of weaving; yet a metaphorical association between these two concepts can be facilitated by their material resemblance.
Words are capable of performing a double-operation: they are materially differentiated in order to create differences in meaning, but they can also cite each other by their material kinship, resulting in the activation of additional or inflected meanings. The power of these two operations is never equal; in every case, it seems that one can be found to dominate over the other (in "differance," Derrida purposely allows the trace to prevail). Traces take the form of an endless referential chain connecting to other words along with their respective meanings and descendents. But one must realize that difference works precisely to cut the chain. Therefore--and this is Derrida's most important discovery in Dissemination--the two coexisting functions of differance also counteract one another other. If meaning is constituted by differance, then meaning is necessarily haunted by internal conflict. It is the battleground between the forces of scission, and the forces of citation. Locked in eternal tension, one is never to vanquish the other.
Hence the second interpretation of differance as an act of "deferral": every word postpones the immediate grasp of the object by tracing its meaning to other words. This somewhat accounts for the logic of connotation. All words, are embedded in a greater signifying network. Meaning is never positively present in writing; it isn't just "there" to be interpreted. It cannot be understood except by difference and by trace, i.e. by differance. And speech is no exception: first of all, speech is intelligible only because it is uses phonic (rather than graphic) differences to establish meaning; secondly, it may in fact be even more difficult to differentiate words that sound alike than those that are spelled similarly can, allowing the utterance to blend into other utterances resembling or preexisting it. For Derrida, this happens to be the root of the indeterminate and unsystematic characteristic of language in general--writing and speech included. Since both writing and speech share the same qualities, the former cannot be conceived as subordinate or derivative to the latter. They should instead be viewed as two special forms in the linguistic order.