The Kensington rune-stone: Authentic and important By Robert A. Hall, Jr (review)

BOOK NOTICES 437 ious tests for subjecthood give mixed results? In Korean , there are constructions in which nomináis in nominative case do not behave as subjects, and nomin áis in some nonnominative case, e.g. dative, behave like subjects. Two approaches have been taken to this mismatch ofcase and grammatical relation. First, one can take the syntactic tests for subjecthood to be crucial and then write the case rules in some way to account for cases other than the expected ones. This approach, taken in various government binding and relational grammar treatments of Korean, has led to the notion that a case such as dative does not refer to a surface relation but rather to an inherent or semantic property of the NP. Second, one can take the case rules to be crucial and then rewrite the syntactic rules to refer to some other notion than subject. This is the approach adopted by O in this work. Arguing from the viewpoint of learnability, O sets forth two principles that govern his analyses. The Uniqueness Criterion states that each case suffix encodes a single corresponding linguistic relation. The Monostratality Criterion states that the relations encoded by the case markers of Korean are manifested at a single level of syntactic representation (surface structure). Thus, O assigns case a heavy functional load and then builds an analysis of Korean constructions to fit this view of case. O's analysis of Korean is cast in a simplified version of categorial grammar. Sentences are built by combining categories (TV, IV, NP, S) in accordance with their inherent combinatorial properties. There are three components to his grammar. First are the Korean case conventions: The genitive marks a nominal that combines with an N category. The accusative marks a nominal (or a QP) that combines with a (T)TV category. The nominative marks a nominal (or a QP) that combines with an IV category. The dative postposition marks a nonterm bearing a verbdetermined thematic role. Second are the mapping rules, which match thematic relations to categoriallybased tree structures. The mapping grid provides a representation of the relative 'combinatorial order' to be employed when a verb takes more than one term. Third are the conversion rules—passivization, recipient conversion, locative conversion, etc.— which make a specific term NP into a nonterm or vice-versa. These conversion rules operate as the NP is entered into the categorial tree; they do not convert one tree into another tree. Thus, a 'monostratal' syntactic analysis is possible. Furthermore, the availability ofconversion rules that turn a term into a nonterm distinguish this approach from government-binding analyses, where theta-theoretic principles block nonterm landing sites for movement rules. What O accomplishes with this relatively straightforward grammar is an account of the behavior of nomináis with respect to syntactic tests without reference to grammatical relations. Relative ranking between nomináis is achieved by height in categorial trees—not via GRs—so what is elsewhere referred to as subject in Korean would be equated with the highest of the NPs/PPs in the categorial tree. Thus, a dative-marked PP in a psych construction, which by the case conventions cannot be the subject, nevertheless behaves like a subject since it is the highest constituent in the tree. As with similar relational grammar accounts, word order does not come into play in O' s analysis since the relative height of a nominal in a categorial tree is not correlated with its position in the word order. O gives a discontinuousconstituents analysis of scrambling phenomena in Korean. O has done an excellent job at presenting consensus judgments by checking data with many speakers. Nevertheless, the reader should be warned that disagreement in judgment exists, though often nothing crucial rests on this fact. For example, some speakers do allow genitive-marked possessors to trigger honorific agreement. Also, speakers have different judgments concerning which NPs are eligible for extraction in WH-questions, relative clauses, clefts, and topics. So O's data do not always provide noncontroversial tests for an NP's status for all speakers. There are also a few typos to note: p. 104 (17) should read casin-uy, p. 124 (22) and p. 278 (footnote...