Topicalization


Topicalization is a mechanism of syntax that establishes an expression as the sentence or clause topic by having it appear at the front of the sentence or clause. Topicalization often results in a discontinuity and is thus one of a number of established discontinuity types. Topicalization is also used as a constituency test; an expression that can be topicalized is deemed a constituent. The topicalization of arguments in English is rare, whereas circumstantial adjuncts are often topicalized. Most languages allow topicalization, and in some languages, topicalization occurs much more frequently and/or in a much less marked manner than in English. Topicalization in English has also received attention in the pragmatics literature.

Examples

Typical cases of topicalization are illustrated with the following examples:
Assuming that the a-sentences represent canonical word order, the b-sentences contain instances of topicalization. The constituent in bold is fronted in order to establish it as topic. The first two examples, which use topicalized adjuncts, are typical, whereas the two latter examples with topicalized object arguments are comparatively rare. The appearance of the demonstrative determiners that and those is important, since without them, topicalization of an argument seems less acceptable, e.g. ?A pizza I won't eat.
Topicalization can occur across long distances, e.g.

Further examples

Topicalization is similar to wh-movement insofar as the constituents that can be wh-fronted can also be topicalized, e.g.
And topicalization is similar to wh-fronting insofar as the islands and barriers to wh-fronting are also islands and barriers to topicalization, e.g.
These examples illustrate the similar behavior of topicalization and wh-fronting. Further data, which will not be produced here, could show, however, that topicalization is unlike the other two major discontinuity types, i.e. scrambling and extraposition.

Theoretical analyses

The theoretical analysis of topicalization can vary greatly depending in part on the theory of sentence structure that one adopts. If one assumes the layered structures associated with many phrase structure grammars, all instances of topicalization will involve a discontinuity. If, in contrast, less layered structures are assumed as for example in dependency grammar, then many instances of topicalization do not involve a discontinuity, but rather just inversion. This point is illustrated here first using flatter structures that lack a finite VP-constituent. Both constituency- and dependency-based analyses are given. The example itself is a piece of Yoda wisdom, and is certainly of questionable acceptability in this regard. It is, however, perfectly understandable:
The upper two trees show the analysis using flat constituency-based structures that lack a finite VP constituent, and the lower two trees are dependency-based, whereby dependency inherently rejects the existence of finite VP-constituents. The noteworthy aspect of these examples is that topicalization does not result in a discontinuity, since there are no crossing lines in the trees. What this means is that such cases can be analyzed purely in terms of inversion. The topicalized expression simply "inverts" to the other side of its head.
Instead of the flat trees just examined, most constituency grammars posit more layered structures that include a finite VP constituent. These more layered structures are likely to address topicalization in terms of movement or copying, as illustrated with the following two trees:
Tree a. shows the canonical word order again, and tree b. illustrates what is known as the movement or copying analysis. The topicalized expression is first generated in its canonical position but is then copied to the front of the sentence, the original then being deleted.
The movement analysis of discontinuities is one possible way to address those instances of topicalization that cannot be explained in terms of inversion. An alternative explanation is feature passing. One assumes that the topicalized expression is not moved or copied to the clause-initial position, but rather it is "base" generated there. Instead of movement, there is feature passing, however. A link of a sort is established between the topicalized expression and its governor. The link is the path along which information about the topicalized expression is passed to the governor of that expression. A piece of Yoda wisdom is again used for illustration, the full sentence being Careful you must be when sensing the future, Anakin:
The nodes in red mark the path of feature passing. Features about the topicalized expression are passed rightward through the tree structure to the governor of that expression. This path is present in both analyses, i.e. in the constituency-based a-analysis on the left and in the dependency-based b-analysis on the right. Since topicalization can occur over long distances, feature passing must also occur over long distances. The final example shows a dependency-based analysis of a sentence where the feature passing path is quite long:

Information about the topicalized such nonsense is passed along the path marked in red down to the governor of the topicalized expression spouting. The words corresponding to the nodes in red form a catena. A theory of topicalization is then built up in part by examining the nature of these catenae for feature passing.