Linking meaning to language: linguistic universals and variation
Joshua K. Hartshorne, T. O’Donnell, Y. Sudo, Miki Uruwashi, J. Snedeker
CogSci
Abstract
Linking Meaning to Language: Linguistic Universals and Variation Joshua K. Hartshorne (jharts@wjh.harvard.edu) Department of Psychology, Harvard University 33 Kirkland St., Cambridge, MA 02138 Timothy J. O’Donnell (timo@wjh.harvard.edu) Department of Psychology, Harvard University 33 Kirkland St., Cambridge, MA 02138 Yasutada Sudo (ysudo@mit.edu) Department of Linguistics and Philosophy, Massachusetts Institute of Technology 77 Mass. Ave. 32-D808, Cambridge, MA 02139 Miki Uruwashi (mikiuruwashi@ruri.waseda.jp) Graduate School of Human Sciences, Waseda University 33 Kirkland St., Cambridge, MA 02138 Jesse Snedeker (snedeker@wjh.harvard.edu) Department of Psychology, Harvard University 33 Kirkland St., Cambridge, MA 02138 Abstract To use natural language, speakers must map the participants in events or states in the world onto grammatical roles. There remains considerable disagreement about the nature of these so-called linking rules (Levin & Rappaport Hovav, 2005). In order to probe the nature of linking rules, we investigate verbs of psychological state, which demonstrate complex linking patterns both within and between languages. We find that the typical duration of the psychological state guides the application of linking rules to novel verbs in both English and Japanese, consistent with a universal constraint. Nonetheless, there are marked differences in the baseline preferences for the individual linking rules across the two languages. We discuss these findings both in terms of theories of exceptionless linking rules and accounts on which linking rules are governed by probabilistic biases as well as cross- linguistic variation. Keywords: syntax; semantics; linking; UTAH; universal grammar; over-hypotheses. The Linking Problem To interpret Mary broke the vase, one must minimally identify the event described (breaking), the participants in that event (Mary, vase), and identify which participant played which role (Mary = breaker, not break-ee). This linking problem has received considerable attention both by theorists trying to correctly characterize the semantics- syntax links (see Levin & Rappaport Hovav, 2005, for review), and by developmental psychologists interested in how children discover these links (Bowerman, 1990; Pinker, A key issue is identifying the right level of generalization for the linking rules. Many data points suggest linking rules are highly regular. Regularity is seen both within verbs and across verbs. Not just Mary but all breakers are the subject and not object of break (John/the baby/the wind broke the vase/window/glass). Similarly, in English the object of a transitive change-of-state verb is systematically the entity that changes state while the subject effects that change (Mary broke/cleaned/opened the box). These intuitions generalize to novel words. If interpretable, The dax broke the blicket must mean that the dax is the breaker and the blicket is broken. Adults and children prefer an interpretation on which The bear pilked the horse means the bear did something to the horse, not vice versa (Marantz, 1982; see also Pinker, 1989). Moreover, these patterns are sufficiently regular across languages to suggest that some (Pinker, 1984) or all (Baker, 1988) linking rules are innate. However, there are numerous examples of apparent variation and exceptionality. An object moving from Mary’s possession to John’s can be described by Mary gave/lent/sent the package to John or John received/took/obtained the package from Mary. The same activity might be called Mary chasing John or John fleeing Mary. Many emotion verbs put the experiencer in subject position (John feared/hated/loved Mary), while others put the experiencer in object position (Mary frightened/angered/delighted John). Moreover, a relatively small number of languages appear to exhibit linking rules quite distinct from what is seen in languages like English (Dixon, 1994). In the present study, we investigate linking rule regularity and variation within and across two unrelated languages with respect to one such problematic case: psych verbs.