1 Introduction
Most of the discussions on object omission center around the factors that license omission or the syntactic representation of omitted objects, and there has been relatively little discussion of its precise nature – what it means for objects to be omitted. Hence, this squib attempts to take a closer look at the nature of object omission, by examining the referential behavior of omitted objects. Omitted objects are ‘semantically obligatory’ (Somers Reference Somers1984: 510), and yet object omission is generally distinguished from the phenomenon where an object is missing on the surface but ‘recoverable’ from the context. The difference between the two types of missing objects, however, has not been clearly explained, and this squib investigates this difference to gain a more precise view of object omission.
Allerton (Reference Allerton1975) and Fillmore (Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986) explain the difference in terms of definiteness: omitted objects are indefinite in reference, while recoverable objects are definite. For them, indefiniteness means that omitted objects are ‘unknown’ and insulated from their potential referents available in the context. However, this understanding of indefiniteness is at odds with observations that the referents of omitted objects are often clearly known from the context. This squib explores the commonality between omitted objects and indefinite NPs to propose that indefiniteness may be more precisely understood as referential indeterminacy, where omitted objects may have referents of varying specificity and likelihood but no fixed referent.
2 Two types of missing objects
There is widely shared agreement among linguists that when the direct object is missing in a well-formed sentence it can be one of two types. Fillmore (Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 95f., see also Reference Fillmore, Herbst and Götz-Votteler2007) calls these types DNC (definite null complements) and INC (indefinite null complements) and says:
With DNC the missing element must be retrieved from something given in the context; with INC the referent's identity is unknown or a matter of indifference. (Fillmore Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 96)
Allerton (Reference Allerton1975: 214–15) illustrates this by contrasting watch and read: in (1a) the hearer knows that George was watching the cricket, while in (1b) the hearer does not (have to) know what he is reading.
(1)
(a) I went to see the cricket. George was watching already.
(b) He's reading.
(Allerton Reference Allerton1975: 214)
Furthermore, Fillmore says that (2b) is not odd, while (2a) is odd because ‘one does not wonder about what one already knows’.
(2)
(a) They found out; ?I wonder what they found out.
(b) He was eating; I wonder what he was eating.
(Fillmore Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 96)
In this squib, the two types of missing objects will be called DNC-objects and INC-objects, and the term object omission will apply exclusively to INC-objects.
Allerton (Reference Allerton1975) and Fillmore (Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986) say that INC-objects are indefinite and DNC-objects definite, and by this they mean that the referents of INC-objects are ‘unknown’ to the hearer, and the referents of DNC-objects are known to the hearer based on the context. They present (3) to show that INC-objects remain unknown even when there is a likely referent in the context:
(3)
(a) A: What happened to my sandwich? B: *Fido ate.
(Fillmore Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 97)
(b) I see you've got today's Guardian. May I look? / *May I read?
(Allerton Reference Allerton1975: 215)
Fillmore's explanation of (3a) is that INC-objects are ‘obligatorily disjoint in reference with anything saliently present in the pragmatic context’; in other words, eat here rejects my sandwich as its object, which thus remains unknown. Allerton supports this with (3b), by comparing read with look, which does take a contextual referent.
For Fillmore and Allerton, however, the unknownness of INC-objects does not mean that they are known only as the abstract semantic role assigned to the object. Rather, they note that INC-objects are often understood in ‘semantic specialization’, for instance, as ‘a meal’ as in (4a), ‘alcoholic beverages’ as in (4b), and ‘breads or pastries’ as in (4c):
(4)
(a) We've already eaten.
(b) I've tried to stop drinking.
(c) I spent the afternoon baking.
(Fillmore Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 96–7)
The question, then, is how semantic specialization differs from the knownness of DNC-objects. Allerton (Reference Allerton1975: 218) answers this by saying: ‘the sentence: John's been drinking again may imply a particular KIND of object, but it does not refer to one established as DEFINITE contextually’. In other words, semantically specialized INC-objects always refer to a category, or ‘a particular KIND of object’, and never a specific instance of the category, while DNC-objects refer to specific individuals in the context. Indeed, meals, alcoholic beverages, breads and pastries are all categories, whereas the referents of DNC-objects above are a specific event, item or fact. For Allerton and Fillmore, then, the indefiniteness of INC-objects means that they are unknown as specific individuals, even when there is a specific individual in the context that could function as a possible referent; they may still be known as a bare semantic role (‘stuff’ - Fillmore Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 96) or a specialized category within the semantic role.
Indefiniteness as unknownness, however, is challenged by cases where INC-objects fairly clearly refer to a specific individual in the context. Groefsema (Reference Groefsema1995: 142, 44) introduces (5a, b) against Fillmore's comment on (3a) and says that here eat and drink do take the contextual referents (sandwiches, glass of beer) as their objects, although the amount eaten or drunk is ‘unspecified’:
(5)
(a) John brought the sandwiches and Ann ate.
(b) John picked up the glass of beer and drank.
(Groefsema Reference Groefsema1995: 142, 144)
(c) We needed a lot of pastries for the party. I have been baking all week.
(Condoravdi & Gawron Reference Condoravdi, Gawron, Kanazawa, Piñon and de Swart1996: 4)
(d) I started painting the inside walls of my house. This has been an off and on project for the past year and this was the first time in this calendar year that I have gotten the chance to paint. I came up with some beautiful designs and colors and I have been painting all week!
(e) Remove cookies from oven. Eat quickly (before they cool).
(Massam & Roberge Reference Massam and Roberge1989: 137)
(f) My dog got access to dog food and ate until his stomach look[ed] like a basketball.
(Ruppenhofer & Michaelis Reference Ruppenhofer, Michaelis, Bourns and Myers2014: 61)
Likewise, Condoravdi & Gawron, Massam & Roberge and Ruppenhofer & Michaelis affirm that the missing objects in (c), (e) and (f) respectively refer to the contextual referents pastries, cookies and dog food. Also, in (d) what I painted all week is most likely the walls.
Furthermore, DNC-objects do not always refer to specific individuals. For instance, (6) is a stand-alone message on an e-card, and the object of find out is most likely what children hide from mom, which is a category, rather than any specific fact:
(6) Has no one ever told you that MOM ALWAYS finds out?? You should never underestimate a mom's ability to find stuff out. Poor dear.
This shows that the indefiniteness of INC-objects as defined as unknownness is inadequate; it cannot account for the difference between INC-objects and DNC-objects, since INC-objects can refer to specific individuals, and DNC-objects categories. It is likely, then, that the difference between them lies not in what is known and unknown but in how their referents are known.
3 Indefiniteness as indeterminacy
3.1 The indefiniteness of indefinite NPs
Fillmore (Reference Fillmore, Nikiforidou, VanClay, Niepokuj and Feder1986: 97) draws the notion of indefiniteness from indefinite NPs. Hence, if contextual insulation does not help us understand the indefiniteness of INC-objects, a fresh look at behaviors of indefinite NPs may. Especially, let us see how indefinite NPs behave when there is an item in the context that is highly likely to be their referent.
(7)
(a) John brought a sandwich, and Ann ate a sandwich.
(b) You have killed a man.
(Steinbeck Reference Steinbeck1994 [1947])
(c) Wow, somebody did the homework. Thank you for the information!!!
In (7a) the indefinite article may be used for the second sandwich because it is different from the first sandwich. In this case, it serves to contextually insulate the second sandwich as Fillmore says. However, the indefinite article can be used even when it refers to the first sandwich, as a way of emphasizing the type of what was eaten. Irrespective of whether the first and second sandwiches are identical, the indefinite article indicates that the hearer is unable to pick out their referents (Lambrecht Reference Lambrecht1994: 77f, 87f.); hence, for the hearer the second sandwich may well refer to the first, but its referent cannot be fixed as such.
In (7b) let us suppose that the hearer knowingly killed only one man, and the speaker knows this as well. Then the interlocutors share the knowledge of only one referent, but even in this case a man can be used to emphasize the fact that the hearer killed a human being rather than a plant or a mosquito. Again, for the hearer what the speaker refers to with a man is not necessarily the one the hearer knows – the speaker may wrongly believe that the hearer killed another man, or the hearer may have killed another man without knowing. In (7c) the obvious referent of somebody is the hearer, who provided the information. I suggest that the use of the indefinite somebody here is a move that diverts the attention from the actor to the act (homework doing) (Lambrecht Reference Lambrecht1994: 137f., 168). Here again, the one who did the homework does not have to be the one the speaker is thanking.
In sum, the reference of indefinite NPs is indeterminate: for the hearer, even when there is an obvious referent for the NP, its status as the correct referent can be denied. On the other hand, the reference of definite NPs is determinate: the referent obvious to the hearer cannot be denied to be the correct referent. Thus, if the boldface NPs in (7) are changed to a definite NP (the sandwich, the man, he), they can only refer to a single entity whose knowledge is shared by the interlocutors. Indeed, Farkas (Reference Farkas2002: 226) understands definiteness in terms of ‘determinacy of reference’. According to her, the referents of definite NPs are ‘determined’ and fixed as ‘a singleton set’, while the referents of indefinite NPs are ‘non-determined’ and ‘variable’ (2002: 223).
3.2 The indefiniteness of INC-objects
I propose that the indefiniteness of INC-objects should also be understood as indeterminacy. Let us return to (5) and see if indeterminacy can be found here. In (5a, b) what is ‘unspecified’ is not just the amount. What Ann ate and what John drank are not necessarily limited to the sandwiches or the glass of beer. Furthermore, there is even a possibility that Ann did not in fact eat John's sandwiches, or John did not drink the beer he picked up, but they ate or drank something else. Therefore, it is still possible to say after (5a, b) ‘I wonder what Ann ate’ and ‘I wonder what John drank’ respectively. Similarly, in (5c, d) there is a legitimate possibility of surprise for the hearer, where the actual referent of the missing object is not the contextually obvious one. For instance, (5c) may be followed by ‘So can you bake pastries for the party because I am too tired?’ and (5d) by ‘But it was for my mom's urgent job, and I will get back to my walls now that it is done.’ On the other hand, in (5e, f) the contexts do not seem flexible enough to reasonably allow the actual referent to differ from the obvious one. But this only shows an accidental feature of the context, not the nature of INC-objects. The point here is that INC-objects themselves do not fix the actual referent as the obvious one; hence, whenever the context is flexible enough as in (5a–d), it is possible to dissociate the actual referent from the obvious one without sounding odd.
In contrast, DNC-objects require a clear and fixed referent. So in all contexts if the actual referent is different from what the hearer expects on the basis of the context, the object cannot be deleted; therefore, if an object is deleted as a DNC-object, the actual referent cannot differ from the obvious referent no matter how flexible the context may be. The referents of DNC-objects are determinate.
This difference between INC- and DNC-objects seems to be illustrated and confirmed by (8):
(8)
(a) There was an adult show, and George was watching. George had no permission to watch the show.
(b) There was a cake, and George was eating. The cake was for the wedding the next day.
In (8a) there is no ambiguity about what George watched: he behaved badly and watched the show. By contrast, while (8b) suggests that George behaved badly and ate the cake, it is open to a different reading – that he behaved well and did not touch the cake.Footnote 1 These two sentences are maximally flexible to allow the deviation of the actual from the obvious: neither of them has anything in the context that indicates whether George behaved well or badly. Nevertheless, the well-behaved interpretation is possible only in (b); this difference can only be attributed to the inherent difference between INC-objects and DNC-objects in referential determinacy.
Now, indeterminacy seems to explain why the uses of eat and read in (3) are odd despite the fact INC-objects can take contextual referents: here eat and read are required to have a determinate object, but this is not possible in their objectless use. To explain why eat and read need a determinate object: in (3a) Fido ate is an answer to a question about my sandwich, so it has to specifically address my sandwich and nothing else; and in (3b) May I read? asks about an item owned by the hearer, so it has to be specifically about that item and nothing else, for the hearer cannot answer the question if the referent is unclear. Cummins & Roberge (Reference Cummins, Roberge, Auger, Clements and Vance2004: 128) support this explanation when they say: ‘when a referential [= determinate: my note] interpretation is forced, a null object is impossible’. By contrast, in (5) none of the sentences with an INC-object is required to be about any specific referent despite the presence of a likely contextual referent.
Indeterminacy also explains why eat and apply behave differently in negation as shown in (9) even when both can refer to the preceding contextual referent:
(9)
(a) There was a piece of bread on the table but John didn't eat.
(b) There was a good job available here but Fred didn't apply.
(Condoravdi & Gawron Reference Condoravdi, Gawron, Kanazawa, Piñon and de Swart1996: 3)
According to Condoravdi & Gawron (Reference Condoravdi, Gawron, Kanazawa, Piñon and de Swart1996: 3), (9a) ‘implies that John didn't eat anything’, whereas (9b) ‘implies simply that Fred didn't apply for the good job that was available here (although he may have applied for other jobs)’. This may be explained by assuming that negation only takes scope over explicit or determinate elements: apply takes job as its determinate complement, and thus not negates job as well as the verb. On the other hand, eat has no determinate object despite the previous mention of bread, and thus not negates only the verb, with the result that John did not engage in eating at all.
Just like the contextual referents discussed above, semantic specialization is contextually established, not only through immediate contexts such as the form, time and place of utterance, but also through general contexts such as world knowledge and speaker's life pattern, although its contents are always categories rather than specific individuals. For instance, when someone says at lunchtime I have already eaten, the most likely object is lunch. García Velasco & Portero Muñoz explain that when we understand eat as ‘eat a meal’:
it is our world knowledge, the fact that we eat several times on the day, which leads us to the right interpretation of the understood object. (García Velasco & Portero Muñoz Reference García Velasco and Muñoz2002: 5)
Furthermore, Levin (Reference Levin1993: 33) notes that Mike ate can mean he ate ‘something one typically eats’, and Rice (Reference Rice, Axmaker, Jaisser and Singmaster1988: 204) says that in Each afternoon, John reads, the missing object is likely to be books. Indeterminacy makes no exception for semantic specialization, however. In the lunchtime example above, lunch is a highly likely, but not a necessary object of eat, and there is a possibility that the speaker in fact ate something simple and small rather than a full lunch. After all, the object of eat is a matter of indifference here because the speech is not really about what the speaker ate but about the fact that he has already performed an eating activity and does not wish to repeat it.
Our discussion in this squib suggests that the indefiniteness of the INC-object does not lie in being unknown but in being indeterminate over its possible referents: none of these referents can be fixed as its actual referent, and thus any of them can be its actual referent. The context may suggest a likely referent for an INC-object with varying degrees of specificity (individual or category) and likelihood. However, this is an accidental feature of the context, and the underlying, inherent indeterminacy remains; in other words, when the context is flexible enough, the likely referent can always be denied to be the actual referent, so it can never remove the indeterminacy although it may reduce or limit it. And when the context suggests no likely referent, the INC-object has full indeterminacy over all possible referents within the semantic role (i.e. it is known as the object semantic role).
4 Conclusion
I have proposed that the difference between DNC- and INC-objects, or definiteness and indefiniteness, should be understood not in terms of known and unknown but in terms of whether they relate determinately or indeterminately to their potential referents. Object omission, then, is a phenomenon where an object is missing and understood indeterminately, with varying degrees of specificity and likelihood based on immediate and general contexts as well as the object's semantic role.
The next step in the study of object omission is to investigate its lexical semantic basis. Although this is a highly complex and controversial issue (Condoravdi & Gawron 1996: 5), I believe indeterminacy provides a useful basis for the task by serving as an independent means of distinguishing between INC- and DNC-objects.
