Topics are usually pragmatically defined as what sentences are about. Usually, what is being talked about in a sentence are people, that is, what they do, how they behave, and why. Hence, the philosophical saying 'Small people talk about people, average people talk about events, and great people talk about ideas' has been challenged by linguists, who instead say: 'All people talk about people'. Corpus studies of spoken language have indeed shown that the subject of most sentences is human, or at least animate, and psycholinguistic research has revealed that in many languages there is a preference both in production and comprehension for constructions that start with an animate noun phrase (see van Bergen 2011). This principle has also been proposed in functional typology to account for crosslinguistic, canonical word-order patterns. Topics are what sentences are about, which explains why they often occur sentence-initially. Topics are often the subject of a sentence, or the other way around, the grammatical subject of a sentence is often the topic. Since topics are usually subjects and subjects are often agents, it is not surprising that topics are often agents, too. Brunetti (2009) argues for Spanish and Italian that the argument with the most proto-agent properties gets selected as the sentence-initial topic, and she shows that this approach can explain why topics are often, but not always, subjects.
Subjects are often prominent or salient in the discourse: agentive, animate, and definite, but of course objects can be prominent and salient in the discourse, animate, and definite, too. Often, when a language shows differential object marking (DOM)—the phenomenon that some objects in a language trigger case marking or agreement on the verb while others do not—the marked objects are in fact these prominent or salient, animate, and/or definite objects. Dalrymple and Nikolaeva link this to the notion of topic in information structure. In their book they show that marked objects in DOM languages are often topical, while unmarked objects are nontopical. I believe this is an important insight. In their view a sentence is not simply divided into a topic and a comment, by which the object automatically ends up as part of the comment when the subject is the topic. D&N note that more than one referent can be under discussion at the time of an utterance, and hence, an utterance can have more than one topic, at least two. The two topics are not equally important, however, and that is why D&N dub them 'primary' and 'secondary' topics. Clearly, subjects are more often topical than objects, and even in contexts where objects are topical, subjects do not have to be less topical. D&N note that, although there is no one-to-one mapping between grammatical functions and information-structure roles, there are strong crosslinguistic tendencies. They propose that while subjects can be viewed as grammaticalized primary topics, grammatically marked objects can be viewed as grammaticalized secondary topics. The difference between primary and secondary topics is very important for their line of argumentation. If one assumes there is only one topic per sentence, the marking of objects could not easily be understood as topic marking, due to the presence of an often more topical subject. When an object is marked as a secondary topic, however, it immediately follows that there must be a primary topic (often the subject) in the sentence as well.
I turn now to a discussion of the book's main claim on the basis of Chs. 6-10. Ch. 6 discusses the main evidence for the relevance of topicality in the grammatical marking of nonsubjects. Chs. [End Page 179] 7 and 8 present languages in which topical objects receive special (case or agreement) marking while nontopical objects remain unmarked. In Ch. 9 this general approach is extended to ditransi-tive constructions in languages with differential object marking, while Ch. 10 sketches a historical approach to differential object marking as grammaticalized secondary topic marking.
Ch. 6, 'Topical marking on nonsubjects...