Certain particles of colloquial Russian, such as -to, zhe and ved’, have been traditionally described as "intensifying" or "emphatic." Short examples with the particle -to serve as an illustration:

(1)Tonja-TO pozvonila.
Tonja-PART called.
"Tonja called." ("Tonja CALLED" as opposed to writing a letter or just showing up at the door, as was expected of her.)

(2)Pozvonila-TO Tonja.
Called-PART Tonja.
"Tonja called." ("TONJA called" as opposed to other people who were expected to call.)

It may appear strange that the element which -to cliticizes to is not the one that is being contrasted (cf. English only ); the reason that -to does not cliticize to the contrasted element is that -to contrasts full propositions and not elements within a proposition (Bitekhtin 1994). Its distribution is also determined by some other factors. The most important one is that -to never cliticizes to rhematic elements (Bonnot 1986), which are the last elements of the clause in (1-2).

Vallduví and Vilkuna (1998) (referred to as V&V henceforth) established a conceptual distinction between two notions that have been covered by the term "focus:" one is "rheme," a concept belonging to the domain of information packaging; and the other is what the authors labeled as "kontrast," the notion covering quantificational phenomena of a more formal semantic nature. It is the latter term that will be important for our purposes. V&V use the term "kontrast"

I propose to analyze such particles of colloquial Russian, as -to, zhe and ved’, as lexical means of marking kontrast and label them as "kontrastive markers," or "k-markers." Even though these particles differ in distribution and scope of interpretation, all of them possess set-evoking properties. For example, the particle -to , as in (1-2), can be analyzed as kontrastive in the following way: -to instructs the hearer to interpret the proposition that contains -to as being part of a set of alternatives that are present in the context in the form of implications.

Colloquial Russian k-markers differ from each other in at least two major ways. First, k-markers differ with respect to types of sets that they evoke: the membership is limited to terms, propositions, or larger discourse segments. Second, k-markers differ in the way they evoke sets: some instruct the hearer to evoke a set that is known to the hearer but not activated in the discourse at the moment, while others give instructions to the hearer to construct a set out of activated, or salient, elements in discourse that might (or might not) be new to the hearer (see Gundel et al. 1993 and Givón 1992 among others). Also, it is important to track the discourse status of the k-marked elements, i.e., whether they are discourse-old/new and hearer-old/new (Prince 1992).

Each k-marker will be classified with respect to the following criteria: the type of set it evokes (terms, propositions, etc.) and with respect to the saliency and informational status of the set members.