Semantic roles: an attempt at clarification In my lecturette #3, i said '... in an agentless passive clause like "The door was opened", 'door' behaves in some respects like the subject (it precedes the verb and controls verb-agreement) and in some respects like the direct object (most obviously in the semantic sense, in that it is the patient or theme of the action).' One of our participants said in response to this: 'This is a very common claim, and it utterly escapes me. In English you can say "the door opened" or even "the door opened itself"; in French it seems to me that this is the standard idiom. In what sense is the door not the agent? It does, after all, open ... 'I can tell the difference between the English pattern and the ergative pattern, and between verbs of a close semantic class i can see when the arguments are aligning similarly and when not. But between languages or between verb classes i've *never* been able to see such patterns. In the "real world" do things rotate or are they rotated? Surely all such pat- terns are linguistically imposed!' Perhaps they are, in which case we are dealing with facts about linguis- tic patterms, not about 'real-world' phenomena, which is fine as long as we recognize the distinction; after all it's primarily (human) language we're claiming to study here, not physics. We're dealing here not only with the relationship between human language and phenomenal (i.e., 'real world') reality but that between human lan- guage and human cognition, i.e. how we as human beings perceive phenome- nal reality. I think any linguist, if pressed, would admit that human language is a part and a manifestation of the human cognitive system, and therefore that many characteristics, including idiosyncratic ones, of the human cognitive system are also characteristics of human language. A further question would be, Are there any (possible idiosyncratic) charac- teristics of human language that are not shared by general human cogni- tion? On this question linguists differ; Chomskyans would say 'yes' while some others (e.g., if i'm not mistaken, Dick Hudson) would say 'no'. If the Chomskyan position on this is correct -- and i think this is probably an empirical question -- then there will be some characteris- tics of the behaviour of human language -- or even of specific human lan- guages -- that have no equivalent in general human cognition and there- fore are not direct reflections of either phenomenal reality or of how we humans perceive it. Getting down to the brass tacks of terminology, we have to bear in mind that there is no essential isomorphism between semantic roles such as Agent, Patient, Theme, etc. and grammatical relations such as Subject and Object. There are some correspondence defaults, e.g., Agents are typi- cally represented by Subjects, but conceptually they are distinct. To address the specific question raised above, the mere fact that in the sentences The door opened. The door was opened. The door opened itself. 'door' is the Subject does not make it in any sense an Agent. Only in a particular interpretation of the third (for instance, if the door is claimed to belong to a haunted house or something like the Beast's Castle in Disney's Beauty and the Beast, in which the articles of furniture are endowed with volitionality) is the door INSTIGATING or WILLING and ac- tion, which is to the best of my knowledge (cf. caveat below) an essen- tial part of the technical definition of 'Agent'. In all of them, how- ever, the door is the 'thing in motion', which is part of the definition of 'Theme'. And at least in the second and third it is the 'affected entity', which makes it a Patient, as well. Note that it is possible for a single entity to serve more than one semantic role in a given sentence. (A parallel would be the range of interpretations possible for the sentence 'Bill rolled down the hill'. As the 'thing in motion', Bill is definitely the Theme of this sentence, but can only be identified as the Agent under the interpretation that he deliberately cast himself down the hill.) Even in French, in which as noted above the expression 'la porte s'est ouverte' would be the standard way of reporting the event of the door being opened without any obvious or explicit Agent, there is generally no assumption that the door is tiself the Agent; again, you would need a peculiar context to elicit that implication. Another of our participants notes that the impossibility of 'door' con- trolling the subject position in a purpose clause as in (1) below (com- pare both (2) and (3)) is evidence that, unlike 'boy' in (3), it is not an Agent. This sort of test is pretty reliable, as far as i know, at least in English and probably a lot of other Modern Western-European languages as well. (1) *The door opened to let in a draft. (2) The door opened, letting in a draft. (3) The boy opened the door to let in a draft. (Note: this is a semantic distinction, not a grammatical one. It is per- fectly possible to choose to ascribe volitionality to the door in ques- tion, and then to speak about the door in a manner consistent with and implementing that choice. In which case one would presumably use senten- ces like (1). Such sentences would be rejected by the general public on the grounds not of ungrammaticality but of semantic anomaly. I believe, though i am not sure, that their status is equivalent to that of senten- ces like (4-6) below. Likewise, i can assume that hobbits or unicorns have objective, phenomenological reality, and write for pages on the ba- sis of that assumption, without once offending against my internalized grammar. (4) Sam frightens sincerity. (5) My toothbrush is pregnant. (6) 3 is angry. A caveat: I do not claim to be a semanticist, and most of my understan- ding of these issues is derived from my reading of Ray Jackendoff's Semantic Structures (Current Studies in Linguistics 18, MIT Press, 1990) and Robert D. Van Valin's (ed.) Advances in Role and Reference Grammar (Current Issues in Linguistic Theory 81, Benjamins, 1993). I personally find the discussion of the taxonomy of semantic roles in these works interesting and illuminating, though certainly not uncontroversial. And these issues are well worth discussing; i am convinced that the develop- ment of syntactic theory will ultimately require a clearer understanding than we have at present of the nature of semantic roles and of the inter- face between syntax and semantics. These issues are bound to recur again and again in our further discussions. Best, Steven --------------------- Dr. Steven Schaufele 712 West Washington Urbana, IL 61801 217-344-8240 fcosws@prairienet.org **** O syntagmata linguarum liberemini humanarum! *** *** Nihil vestris privari nisi obicibus potestis! ***