ka-i tealo anu-o tina-no

262 TABLE 107 PC1 AND PC2 AVERAGES FOR INVERSE FUNCTIONAL VOICE CLUSTER IN MORONENE Clause Type Occurrences Mean PC1 Mean PC2 NOM,V-Opro,S 13 -0.818 -0.463 V-Opro,S 13 -1.432 -0.130 In terms of PCA scores, the mean PC1 scores are between -0.8 and -1.5, indicating that the undergoer is significantly more topical than the actor. The mean PC2 scores are between -0.1 and -0.5, indicating that overall topicality is rather low. Syntactically these two clause types share the features of having a nominal subject and a verb with a suffixed pronominal object. The following are examples: 39. ka-i kaa-hira yo ica then-3sNOM eat-3pABS ART fish NOM,V-Opro,S „then the fish ate them‟ Icadiu 303 RRD Actor: New Und.: 1 TP Actor: 0 Und.: 0

40. ka-i tealo anu-o tina-no

then-3sNOM pass address-3sABS mother-3sPOS hi daa mo-turi COMPL be INTNF-sleep V-Opro,S „then her mother appeared and spoke to her when she was sleeping‟ Icadiu 215b RRD Actor: 1 Und.: 10 TP Actor: 4 Und.: 8 The topicality measures for this cluster of clause types are shown in Table 108. 263 TABLE 108 TOPICALITY MEASURES FOR INVERSE FUNCTIONAL VOICE CLUSTER IN MORONENE n=26 Anaphoric Continuity Cataphoric Importance Referential Distance Anaphoric Continuity Actor Undergoer Topic Persistence Actor Undergoer 1 High: 10 46 62 High: 3-10 50 50 2-3 Med.: 9-9.5 15 23 4-19 Low: 1-8 12 8 Low: 0-2 50 50 19 Reactiv. V. Low: 0.5 12 4 20 New V. Low: 0 15 4 Standard Deviation 4.27 2.72 Standard Dev. 2.82 2.78 Mean Anaphoric Continuity 7.02 8.85 Mean Topic Pers. 3.15 3.85 Mean Anaphoric Continuity Difference -1.83 Mean Topic Persistence Dif. -0.70 By both topicality measures the undergoer is more topical than the actor. With respect to topic persistence, both actors and undergoers have an equal proportion of high and low topic persistence, but the mean topic persistence difference has a small negative value, indicating the undergoer is slightly more topical. Actors retain considerable topicality by both measures. Hence this cluster matches the category of inverse functional voice. NOM,UVV,U Isolate The NOM,UVV,U clause type represents an isolate, located in the middle bottom part of Figure 7. In terms of PCA scores, the mean PC1 score is -0.946, indicating that the undergoer is more topical than the actor. The mean PC2 score is -1.218, indicating that overall topicality is quite low. The following is an example: 264 41. ka-i tibo die te-o-angka Colisi then-3sNOM sudden this RES-RED-lift Colisi NOM,UVV,U „then Colisi was suddenly lifted up‟ Colisi 118a RRD Actor: 1 Und.: 1 TP Actor: 7 Und.: 10 The topicality measures for this clause type are shown in Table 109. TABLE 109 TOPICALITY MEASURES FOR NOM,UVV,U IN MORONENE n=12 Anaphoric Continuity Cataphoric Importance Referential Distance Anaphoric Continuity Actor Undergoer Topic Persistence Actor Undergoer 1 High: 10 42 33 High: 3-10 42 42 2-3 Med.: 9-9.5 8 8 4-19 Low: 1-8 25 Low: 0-2 58 58 19 Reactiv. V. Low: 0.5 20 New V. Low: 0 50 33 Standard Deviation 5.18 4.54 Standard Dev. 3.69 3.81 Mean Anaphoric Continuity 4.96 5.79 Mean Topic Pers. 3.17 3.17 Mean Anaphoric Continuity Difference -0.83 Mean Topic Persistence Dif. 0.00 The topicality measures give a mixed message with regard to the relative topicality of the actor and the undergoer. Although more actors than undergoers have medium or high anaphoric continuity 50 versus 41, there are also more actors than undergoers with very low anaphoric continuity 50 versus 33. As measured by the mean anaphoric continuity difference, undergoers are slightly more topical than actors. As measured by cataphoric importance, actors and undergoers have exactly the same topicality. By both measures, significant numbers of both actors and undergoers have high topicality, and significant numbers have low topicality. The high standard deviations for every topicality measure indicate that the values have a wide spread. This clause type cannot be characterized according to any functional voice category. 265 NOM,UVV Isolate The NOM,UVV clause type represents an isolate, located in the bottom left hand part of Figure 7. In terms of PCA scores, the mean PC1 score is -2.275, indicating that the undergoer is much more topical than the actor. The mean PC2 score is -1.627, indicating that overall topicality is quite low. Syntactically this clause type is characterized by an undergoer voice verb and a pronominal undergoer subject. The following is an example: 42. ka-ndo po-weweu-hakono pua wulu, then-3pNOM AF-make-3sBEN blowpipe bamboo na-i o-oleo-hako ka-i te-wotu NEG-3sNOM RED-day-INST then-3sNOM RES-break NOM,UVV „then they made him a bamboo blowpipe, but it wasn‟t a day before it was broken‟ Maegani 187 RRD Actor: 1 Und.: 0 TP Actor: 1 Und.: 0 The topicality measures for this clause type are shown in Table 110. TABLE 110 TOPICALITY MEASURES FOR NOM,UVV IN MORONENE n=11 Anaphoric Continuity Cataphoric Importance Referential Distance Anaphoric Continuity Actor Undergoer Topic Persistence Actor Undergoer 1 High: 10 9 55 High: 3-10 9 55 2-3 Med.: 9-9.5 18 36 4-19 Low: 1-8 Low: 0-2 91 45 19 Reactiv. V. Low: 0.5 20 New V. Low: 0 73 9 Standard Deviation 4.44 2.96 Standard Dev. 1.85 3.54 Mean Anaphoric Continuity 2.59 8.91 Mean Topic Pers. 0.73 4.18 Mean Anaphoric Continuity Difference -6.32 Mean Topic Persistence Dif. -3.45 266 By both topicality measures the undergoer is much more topical than the actor. Undergoers with high cataphoric importance exceed actors by a large margin: 55 - 9 = 46. The topicality of the actor is demoted, with all but 27 having very low anaphoric continuity, and only 9 having high cataphoric importance. Hence this clause type matches the category of passive functional voice. Theoretical Implications In this chapter I have clustered fifteen frequently occurring basic clause types in Moronene using principal component analysis and classified them in terms of functional voice categories. In doing this, I have tried not to impose the theoretical categories I have posited on data that does not match those categories. This is why I initially grouped them in small clusters according to how they appeared on the plot, even though it turned out that three of those small clusters fell into the category of active functional voice. As was the case with Hebrew, the active functional voice clusters are syntactically rather diverse. They all share the feature of having a pronominal or implicit subject, but they vary in the verb form and the form of the object. This is explicable in that active functional voice is the unmarked category in the functional voice system. As such, it does not need a distinctive syntactic marking as much as the other marked categories do. The results reported above are relevant to the problem of the classification of moN -V constructions in terms of syntactic and functional voice categories, as discussed in Chapter 7. It can be seen that moN -V clauses do not fall into one functional voice category. Some are classified as antipassive functional voice, namely, monV,O and NOM,monV,O; others are classified as active functional voice, namely NOM,monV and monV. This represents additional evidence that the moN - prefix does not function like a prototypical antipassive morpheme. 267 It is somewhat paradoxical that those clause types which have an explicit nominal object, and are therefore clearly transitive, fall into the antipassive category, which one expects to have low transitivity. In contrast to this, those clause types with implicit undergoers, and hence arguably syntactically intransitive, fall into the active functional voice category. In these latter two clause types, NOM,monV and monV, the undergoers are coded by zero anaphora. To explain this paradox, one needs to distinguish between the two common functions of zero anaphora, as mentioned in the discussion of implicit undergoer clauses in Chapter 4. Zero anaphora is typically used either for discourse- active referents which have been mentioned in the immediately previous clauses or else for unidentifiable unimportant referents. An examination of the referential distance figures for undergoers in NOM,monV and monV clauses reveals that 55 of the undergoers have a referential distance of one. Only 24 of the undergoers are new referents. This indicates that in a majority of cases, these two clause types are used when both actor and undergoer have relatively high topicality, thus falling into the category of active functional voice. At the end of Part II, I stressed that the analysis of Hebrew showed it was more insightful to analyze topicality and continuity in relation to clause types, rather than merely in relation to contrasting verb forms. The above discussion shows that what is true of Hebrew is also true of Moronene. Conclusion of Part III In Part III, I have analyzed functional voice categories in Moronene narrative. Some adjustments to the methodology were necessary to accommodate particular features of Moronene syntax. Some additional features are used to define clause types, namely, the choice of affix or clitic set used to mark the actor, the presence of an action focus prefix, and the use of undergoer voice verb forms. Unlike Hebrew, the selection of 268 clauses for analysis includes participial clauses. I restricted the analysis to narrative texts. I have used principal component analysis to cluster common clause types. Most of the clusters identified are similar to those found in Hebrew, namely antipassive functional voice, active functional voice, inverse functional voice, and passive functional voice. PART IV MATCHING CLAUSE TYPES IN TRANSLATION In Part IV, I compare the results of the separate analyses of Hebrew and Moronene as set out in Parts II and III. In Chapter 10, I match the clause type clusters identified for Hebrew in Chapter 3 with those identified for Moronene in Chapter 9. For each Hebrew clause type cluster, I identify Moronene matching clause types, that is, those which are most similar to the Hebrew cluster in their topicality patterns. Other common Moronene clause types are referred to as “non-matching clause types.” I then make a comparison of a Biblical Hebrew text and its translation into Moronene to see to what extent particular Hebrew clause types have been translated by Moronene matching clause types. This comparison is used to identify clauses in the existing translation that are potentially unnatural, and hence should be evaluated for possible revision. In Chapter 11, I describe how the analysis of functional voice can be applied to the task of improving the quality of a translation. I report the results of a revision process by Moronene translators based on their evaluation of the naturalness of non-matching clause types and rare clause types used in the Moronene translation. I discuss various categories of clauses that needed to be revised and those that were not revised. 270

CHAPTER 10 MATCHING HEBREW AND MORONENE CLAUSE TYPES