University of Groningen
Experimental investigations into the semantics of distributive marking
Bosnić, Ana
DOI:
10.33612/diss.171644158
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2021
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Bosnić, A. (2021). Experimental investigations into the semantics of distributive marking: Data from Serbian, Korean and Dutch. University of Groningen. https://doi.org/10.33612/diss.171644158
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
6
1 Overview
We have come to the end of the book, and although we are stopping here, the story is far from over. We uncovered fascinating results, possible ongo-ing semantic changes, drastic discrepancies between children and adults in their interpretations of distributive markers and finally, we proposed sev-eral ways to account for these findings. All of this has sparked many more interesting research questions for further investigation, be it in Serbian or other languages with similar phenomena. This book discussed several experimental and theoretical topics that we outline and summarize below: Chapter 1 was a short introduction to the phenomenon of DistShare markers. We pointed out that experimental research in this domain is mainly lacking compared to investigations with DistKey markers (such as each and every). We discussed the notions of distributive numerals, distance-distributivity and dependent indefinites, and the reasons why we are not using this terminology, thus defining the scope of this thesis. Finally, we presented the roadmap of our research and summarized the main findings.
Chapter 2 tested children’s knowledge of DistShare and DistKey markers in Serbian, in comparison to Dutch, which only has DistKey markers (a.k.a. distributive universal quantifiers). The goal was to map the acqui-sition path of these markers using an act-out task that could give more information about children’s reasoning about the possible interpretations of the sentences with and without distributive markers. First, the results revealed that there is an intermediate response (the cumulative answer) between collective and distributive answers. In addition, we confirmed previous results that Serbian children are significantly late in acquiring both DistKey and DistShare markers. Dutch children, on the other hand, acquire their DistKey markers earlier, and pragmatic factors such as verb type and naturalness may affect Dutch and Serbian speakers in a different way – Dutch speakers (including adults) tend to give cumulative answers only when it is pragmatically more natural to do so, while Serbian speakers tend to rigidly follow the lexical marking. Lastly, we hypothesized that there is a third distributive competitor in Serbian (i.e., svaki+po that may be equivalent to (binominal) each) that potentially affects the acquisition of the distributive markers svaki and po independently. These results are relevant when looking at the range of interpretations of po, including different meanings of po (e.g., as a preposition or a verbal prefix), that may hinder the very late acquisition of po.
The experiments in Chapter 3 and Chapter 4 both focused on testing whether DistShare markers, such as po in Serbian (and -ssik in Korean), should be analyzed in terms of universal quantification over events (Zim-mermann 2002b, Balusu 2006) or in terms of event plurality markers (pluractionals) (Cable 2014, Knežević 2015). The experiments were all picture verification tasks that we used to establish the baseline interpre-tations of DistShare markers in Serbian (and Korean) independently of the theoretical claims we made on the basis of the data.
6
1 Overview
We have come to the end of the book, and although we are stopping here, the story is far from over. We uncovered fascinating results, possible ongo-ing semantic changes, drastic discrepancies between children and adults in their interpretations of distributive markers and finally, we proposed sev-eral ways to account for these findings. All of this has sparked many more interesting research questions for further investigation, be it in Serbian or other languages with similar phenomena. This book discussed several experimental and theoretical topics that we outline and summarize below: Chapter 1 was a short introduction to the phenomenon of DistShare markers. We pointed out that experimental research in this domain is mainly lacking compared to investigations with DistKey markers (such as each and every). We discussed the notions of distributive numerals, distance-distributivity and dependent indefinites, and the reasons why we are not using this terminology, thus defining the scope of this thesis. Finally, we presented the roadmap of our research and summarized the main findings.
Chapter 2 tested children’s knowledge of DistShare and DistKey markers in Serbian, in comparison to Dutch, which only has DistKey markers (a.k.a. distributive universal quantifiers). The goal was to map the acqui-sition path of these markers using an act-out task that could give more information about children’s reasoning about the possible interpretations of the sentences with and without distributive markers. First, the results revealed that there is an intermediate response (the cumulative answer) between collective and distributive answers. In addition, we confirmed previous results that Serbian children are significantly late in acquiring both DistKey and DistShare markers. Dutch children, on the other hand, acquire their DistKey markers earlier, and pragmatic factors such as verb type and naturalness may affect Dutch and Serbian speakers in a different way – Dutch speakers (including adults) tend to give cumulative answers only when it is pragmatically more natural to do so, while Serbian speakers tend to rigidly follow the lexical marking. Lastly, we hypothesized that there is a third distributive competitor in Serbian (i.e., svaki+po that may be equivalent to (binominal) each) that potentially affects the acquisition of the distributive markers svaki and po independently. These results are relevant when looking at the range of interpretations of po, including different meanings of po (e.g., as a preposition or a verbal prefix), that may hinder the very late acquisition of po.
The experiments in Chapter 3 and Chapter 4 both focused on testing whether DistShare markers, such as po in Serbian (and -ssik in Korean), should be analyzed in terms of universal quantification over events (Zim-mermann 2002b, Balusu 2006) or in terms of event plurality markers (pluractionals) (Cable 2014, Knežević 2015). The experiments were all picture verification tasks that we used to establish the baseline interpre-tations of DistShare markers in Serbian (and Korean) independently of the theoretical claims we made on the basis of the data.
6
In Chapter 3 we tested intransitive sentences in which po and -ssik at-tached to the only available argument – the subject (e.g., Po one monkey is
dancing; see Chapter 3, section 3 for details on the method and materials).
We took the core property that distinguishes universal quantifiers from event plurality markers to be the exhaustivity requirement of the DistKey argument. We pioneered an experimental design to test whether the DistKey argument needs to be exhausted or not, which, in case of intran-sitive sentences, is a covert spatio/temporal argument. We encouraged interpretations with a spatial argument as the DistKey by making the spatial arguments explicit in the pictures by showing cages and caves in a zoo. These cages/caves contained triplets of animals as shown in Figure 3, Chapter 3. Our hypothesis was the following: if the spatial arguments must be exhausted by one-monkey-dancing events then there is evidence that DistShare markers should be analyzed as universal quantifiers over events. If not, then it is only a matter of verbal/event plurality – as long as there are at least two events of one-monkey-dancing, the sentence is true under all conditions.
We actually found an exhaustivity requirement over non-atomic partici-pants (groups of participartici-pants/monkeys) and crucially not over designated spatial units. In fact, out of three potential DistKeys available with our visual stimuli (cages, triplets of monkeys in the cages and triplets of mon-keys), triplets of monkeys seemed to be the relevant DistKey. We take this conclusion to argue in favor of a universal quantification analysis by adapting Zimmermann’s (2002b) analysis of DistShare markers. We thus assume that DistShare markers can distribute over entities that are non-atomic, such as time and space, as well as entities that are bigger than atoms – that is, groups/pluralities of atomic individuals. This is further reflected in the parameter settings for distributivity operators, granularity and dimension, as proposed by Champollion (2016b). Specifically, the granularity parameter can be set to distribute over atoms (as it is the case with quantifiers such as each) and over non-atomic (non-count) dimensions such as time or space (and now non-atomic pluralities).
Chapter 4 was the logical next step for further testing DistShare markers and exhaustivity. Here we also focused on the question whether non-atomic (spatial) distribution is equally available in transitive sentences in which po marks the object. This is not a trivial question because, unlike in intransitive sentences where there is a single overt argument, transitive sentences have two, and the unmarked plural argument becomes an overt, linguistic competitor for the implicit DistKey.
In our three experiments, the subject of the sentence was an overt plurality that could serve as a potential DistKey over which DistShare argument is distributed. In the first two experiments, we tested the exhaustivity requirements of the DistKey in six possible conditions (pictures) that were also based on our conclusions from Chapter 3, i.e., that non-atomic individuals could serve as the DistKey that needs to be exhausted (e.g., for the test sentence Monkeys are holding po one umbrella; see Chapter 4, section 3 for details on the new experimental conditions).
The results from these experiments suggest there may be two populations of speakers, but it was unclear whether the observed effects are coming from the exhaustivity requirements of the marker or maximality effects of the plural definites. The third experiment was designed to test for homogeneity effects (building on Križ & Chemla 2015) across three types of negative transitive sentences: with the DistShare marker po, or with either the DistKey quantifier svaki (every) or a definite plural in subject position. Most importantly, we proposed that testing homogeneity effects is a good diagnostic test to tease apart exhaustivity requirements of universal quantifiers and maximality effects of definite (bare) plurals. Namely, this test has clear and distinct predictions for universal quantifiers and for definite plurals which were ultimately borne out by the claims from the first two experiments. Thus, the experiments revealed and then confirmed there are two populations of speakers – one that requires an exhaustivity requirement of the DistKey and one that does not. In other words, the first population seemingly interprets DistShare markers as universal distributive quantifiers over events, as initially predicted, and the second population interprets them (possibly) as pluractional markers. As a concluding remark, we speculate that po, as a universal quantifier, may be undergoing semantic weakening for some speakers, which then explains the coexistence two populations.
Summarizing the experimental findings presented in Chapters 2, 3 and 4, our research discovered intriguing facts about po and confirmed that DistShare markers cannot be uniformly semantically analyzed as either universal quantifiers or event plurality markers, because we found two distinct populations of speakers. It then follows that, for some people, po is a universal quantifier over events, and for others, po does not impose exhaustivity requirements on the DistKey. Due to the fact this non-ex-haustive po has not been found with intransitive sentences, we took the position that po may be losing its universal quantificational force and is being possibly reduced to a distributive marker of event plurality.
Chapter 5 focused on how to semantically analyze po for the subset of the population that does not treat po in transitive contexts as a universal quantifier, and thus does not require an exhausted DistKey. Specifically, we explored two possible accounts that do not analyze these markers as universal quantifiers – first we present the account we covered in earlier chapters, proposed by Knežević (2015) for the marker po, where she argues that po is a marker of event plurality, similar to pluractionals. The essence of Knežević’s proposal is that the only requirement needed for a sentence with
po to be felicitous is to have at least two events consisting of a po-marked
argument, distributed over spatio-temporal units. Crucially, no universal quantificational force is responsible for the interpretations of po sentences. The second proposal analyzes DistShare markers as group-forming devices, which also makes no use of universal quantification, but crucially, it also does not require a plurality of events. This is an account proposed by McKercher & Kim (1999), based on Gil’s (1990) NP-internal distributive readings, in which the marker distributes over the head noun it modifies
6
In Chapter 3 we tested intransitive sentences in which po and -ssik at-tached to the only available argument – the subject (e.g., Po one monkey is
dancing; see Chapter 3, section 3 for details on the method and materials).
We took the core property that distinguishes universal quantifiers from event plurality markers to be the exhaustivity requirement of the DistKey argument. We pioneered an experimental design to test whether the DistKey argument needs to be exhausted or not, which, in case of intran-sitive sentences, is a covert spatio/temporal argument. We encouraged interpretations with a spatial argument as the DistKey by making the spatial arguments explicit in the pictures by showing cages and caves in a zoo. These cages/caves contained triplets of animals as shown in Figure 3, Chapter 3. Our hypothesis was the following: if the spatial arguments must be exhausted by one-monkey-dancing events then there is evidence that DistShare markers should be analyzed as universal quantifiers over events. If not, then it is only a matter of verbal/event plurality – as long as there are at least two events of one-monkey-dancing, the sentence is true under all conditions.
We actually found an exhaustivity requirement over non-atomic partici-pants (groups of participartici-pants/monkeys) and crucially not over designated spatial units. In fact, out of three potential DistKeys available with our visual stimuli (cages, triplets of monkeys in the cages and triplets of mon-keys), triplets of monkeys seemed to be the relevant DistKey. We take this conclusion to argue in favor of a universal quantification analysis by adapting Zimmermann’s (2002b) analysis of DistShare markers. We thus assume that DistShare markers can distribute over entities that are non-atomic, such as time and space, as well as entities that are bigger than atoms – that is, groups/pluralities of atomic individuals. This is further reflected in the parameter settings for distributivity operators, granularity and dimension, as proposed by Champollion (2016b). Specifically, the granularity parameter can be set to distribute over atoms (as it is the case with quantifiers such as each) and over non-atomic (non-count) dimensions such as time or space (and now non-atomic pluralities).
Chapter 4 was the logical next step for further testing DistShare markers and exhaustivity. Here we also focused on the question whether non-atomic (spatial) distribution is equally available in transitive sentences in which po marks the object. This is not a trivial question because, unlike in intransitive sentences where there is a single overt argument, transitive sentences have two, and the unmarked plural argument becomes an overt, linguistic competitor for the implicit DistKey.
In our three experiments, the subject of the sentence was an overt plurality that could serve as a potential DistKey over which DistShare argument is distributed. In the first two experiments, we tested the exhaustivity requirements of the DistKey in six possible conditions (pictures) that were also based on our conclusions from Chapter 3, i.e., that non-atomic individuals could serve as the DistKey that needs to be exhausted (e.g., for the test sentence Monkeys are holding po one umbrella; see Chapter 4, section 3 for details on the new experimental conditions).
The results from these experiments suggest there may be two populations of speakers, but it was unclear whether the observed effects are coming from the exhaustivity requirements of the marker or maximality effects of the plural definites. The third experiment was designed to test for homogeneity effects (building on Križ & Chemla 2015) across three types of negative transitive sentences: with the DistShare marker po, or with either the DistKey quantifier svaki (every) or a definite plural in subject position. Most importantly, we proposed that testing homogeneity effects is a good diagnostic test to tease apart exhaustivity requirements of universal quantifiers and maximality effects of definite (bare) plurals. Namely, this test has clear and distinct predictions for universal quantifiers and for definite plurals which were ultimately borne out by the claims from the first two experiments. Thus, the experiments revealed and then confirmed there are two populations of speakers – one that requires an exhaustivity requirement of the DistKey and one that does not. In other words, the first population seemingly interprets DistShare markers as universal distributive quantifiers over events, as initially predicted, and the second population interprets them (possibly) as pluractional markers. As a concluding remark, we speculate that po, as a universal quantifier, may be undergoing semantic weakening for some speakers, which then explains the coexistence two populations.
Summarizing the experimental findings presented in Chapters 2, 3 and 4, our research discovered intriguing facts about po and confirmed that DistShare markers cannot be uniformly semantically analyzed as either universal quantifiers or event plurality markers, because we found two distinct populations of speakers. It then follows that, for some people, po is a universal quantifier over events, and for others, po does not impose exhaustivity requirements on the DistKey. Due to the fact this non-ex-haustive po has not been found with intransitive sentences, we took the position that po may be losing its universal quantificational force and is being possibly reduced to a distributive marker of event plurality.
Chapter 5 focused on how to semantically analyze po for the subset of the population that does not treat po in transitive contexts as a universal quantifier, and thus does not require an exhausted DistKey. Specifically, we explored two possible accounts that do not analyze these markers as universal quantifiers – first we present the account we covered in earlier chapters, proposed by Knežević (2015) for the marker po, where she argues that po is a marker of event plurality, similar to pluractionals. The essence of Knežević’s proposal is that the only requirement needed for a sentence with
po to be felicitous is to have at least two events consisting of a po-marked
argument, distributed over spatio-temporal units. Crucially, no universal quantificational force is responsible for the interpretations of po sentences. The second proposal analyzes DistShare markers as group-forming devices, which also makes no use of universal quantification, but crucially, it also does not require a plurality of events. This is an account proposed by McKercher & Kim (1999), based on Gil’s (1990) NP-internal distributive readings, in which the marker distributes over the head noun it modifies
6
and results in a plurality of these entities, but not in a plurality of events (i.e., the events are not distributed). The two accounts differ in the fact that the latter account allows scenarios that involve a single event, in which distributivity only happens within the DP/NP that the marker attaches to and has no impact on the plurality of events. As expected, the event plurality account cannot compute these readings.
We contribute to this argumentation by first discussing different cri-teria for individuating and counting events (e.g., Donazzan & Müller 2015, Lasersohn 1995 and Tunstall 1998) in order to directly apply these criteria to the counterexamples of the group-forming story (e.g., John is
weighing po three apples), showing that the situations cannot be considered
collective unless there is also multiplication/distribution of the implicit instrument (the scales in this example), which is not predicted by this proposal. Furthermore, we introduce our own crucial borderline example to demonstrate that what is (seemingly) a single event scenario should actually be considered as involving multiple events after applying the differentiation criteria to count the events. To illustrate, we showed that the sentence with a singular subject argument that is engaging in a single event of pulling three carts each with three suitcases (the sentence being:
John is pulling po three suitcases) involves three subevents, differentiated
by the three carts with three suitcases along the spatial dimension. That is, these three subevents do not share the same spatial trace, but they do share the individual/agent dimension and the temporal dimension. We claim that if there is at least one dimension over which the events are not overlapping, this counts as a minimal differentiation between these events and they should be counted as separate (sub)events. In conclusion, we argue in favor of the pluractional analysis of non-exhaustive po by showing that we are indeed dealing with a plurality of events in all these cases, and not just plurality of individuals/groups.
2 Unsolved puzzles for future research
Where do we go from here?
We strongly believe that the results that emerged in this book create excit-ing new ways of lookexcit-ing at DistShare markers and language change. Apart from numerous possibilities of continuing and expanding this research, pointed out in almost all of our chapters, there are other linguistic puzzles related to DistShare markers that did not find their place in the book, either because they exceeded the scope of our research, or because they came up in the very last stages of the research. Nevertheless, we provide a few snippets of these puzzles here, as a way to illustrate just how many more exciting topics still lie ahead.
2.1 Svaki+po
It became clear throughout our research that the svaki+po construction is something that is definitely worth exploring further. While we have some
evidence for the interpretations of this construction from Knežević & Demirdache (2018), more questions arose about the specific contribution of po in these sentences and we only offered speculations.
First, in Chapter 2 we hinted that po in the svaki+po construction may not be a “regular” distributive po – i.e., that it is neither a quantifier, nor a plu-ractional. Instead, this use of po may only be there to highlight distributivity, to emphasize it.1 The contribution of po here would merely be emphatic
in this case, which is an idea that has precedents. Kim (2012) talks about Korean -ssik being used as an emphatic marker, and a conversation with a Telugu speaker revealed that RedNum too is used primarily for emphasis. Perhaps it is not impossible to assume that po in svaki+po does the same. Thus, for example, the sentence Svaki (every) boy is pushing po three toy
cars emphasizes the numeral and the fact there is mapping between each
individual boy and three toy cars, preventing a scenario where one boy is pushing only two cars, or a scenario in which distribution is non-atomic on the boys. On the other hand, po in svaki+po seems to strongly enforce atomic readings, as shown in Knežević & Demirdache (2018), while inde-pendently, these markers are not necessarily atomic. Being emphatic does not entail that po should force atomicity, so the actual contribution of po in svaki+po still remains an open question in need of answers.
Recall that in Chapters 4 and 5 we discussed that svaki+po construction might influence the semantic change of po. In Chapter 4 we proposed an explanation based on semantic weakening story, with the idea being that
po starts off as a universal quantifier, but it gets weakened to a pluractional,
by discussing two reasons: the svaki+po construction and the fact that we only found the non-exhausted po with transitive sentences. First, given that the svaki+po construction is very common in adult language and acquired earlier by children than these markers independently (compare Knežević & Demirdache (2018) and Chapter 2), it could be that speakers resort to svaki+po for exhaustive interpretations and begin to reanalyze the contribution of po – if svaki is not present, there is no longer an exhaus-tivity requirement on a DistKey. This then means that po is treated as a weaker, non-exhaustive distributive marker. Second, the experiments with intransitive sentences did not reveal two populations. In fact, we only found the universal quantification pattern of responses across all participants, in which the implicit DistKey must be exhausted. We could then argue that non-exhaustive po only appears as a pragmatically weaker form of a universal quantifier in transitive situations where the intended exhaustivity requirement can be achieved with a better construction – svaki+po.
However, at this stage of investigation, we cannot be certain of the direction of change, so it is also possible that po is undergoing semantic strengthening, starting off as an existential plural undergoing pragmatic strengthening to yield a universal meaning – such a process of (“double strengthening”) has been argued for by Bar-Lev & Margulis (2014), for the
1 If the emphasis is on distributivity, and distributivity is a byproduct of event plurality, it could also be that emphatic uses of certain markers such as po still fall under event plurality. However, this is yet to be tested.
6
and results in a plurality of these entities, but not in a plurality of events (i.e., the events are not distributed). The two accounts differ in the fact that the latter account allows scenarios that involve a single event, in which distributivity only happens within the DP/NP that the marker attaches to and has no impact on the plurality of events. As expected, the event plurality account cannot compute these readings.
We contribute to this argumentation by first discussing different cri-teria for individuating and counting events (e.g., Donazzan & Müller 2015, Lasersohn 1995 and Tunstall 1998) in order to directly apply these criteria to the counterexamples of the group-forming story (e.g., John is
weighing po three apples), showing that the situations cannot be considered
collective unless there is also multiplication/distribution of the implicit instrument (the scales in this example), which is not predicted by this proposal. Furthermore, we introduce our own crucial borderline example to demonstrate that what is (seemingly) a single event scenario should actually be considered as involving multiple events after applying the differentiation criteria to count the events. To illustrate, we showed that the sentence with a singular subject argument that is engaging in a single event of pulling three carts each with three suitcases (the sentence being:
John is pulling po three suitcases) involves three subevents, differentiated
by the three carts with three suitcases along the spatial dimension. That is, these three subevents do not share the same spatial trace, but they do share the individual/agent dimension and the temporal dimension. We claim that if there is at least one dimension over which the events are not overlapping, this counts as a minimal differentiation between these events and they should be counted as separate (sub)events. In conclusion, we argue in favor of the pluractional analysis of non-exhaustive po by showing that we are indeed dealing with a plurality of events in all these cases, and not just plurality of individuals/groups.
2 Unsolved puzzles for future research
Where do we go from here?
We strongly believe that the results that emerged in this book create excit-ing new ways of lookexcit-ing at DistShare markers and language change. Apart from numerous possibilities of continuing and expanding this research, pointed out in almost all of our chapters, there are other linguistic puzzles related to DistShare markers that did not find their place in the book, either because they exceeded the scope of our research, or because they came up in the very last stages of the research. Nevertheless, we provide a few snippets of these puzzles here, as a way to illustrate just how many more exciting topics still lie ahead.
2.1 Svaki+po
It became clear throughout our research that the svaki+po construction is something that is definitely worth exploring further. While we have some
evidence for the interpretations of this construction from Knežević & Demirdache (2018), more questions arose about the specific contribution of po in these sentences and we only offered speculations.
First, in Chapter 2 we hinted that po in the svaki+po construction may not be a “regular” distributive po – i.e., that it is neither a quantifier, nor a plu-ractional. Instead, this use of po may only be there to highlight distributivity, to emphasize it.1 The contribution of po here would merely be emphatic
in this case, which is an idea that has precedents. Kim (2012) talks about Korean -ssik being used as an emphatic marker, and a conversation with a Telugu speaker revealed that RedNum too is used primarily for emphasis. Perhaps it is not impossible to assume that po in svaki+po does the same. Thus, for example, the sentence Svaki (every) boy is pushing po three toy
cars emphasizes the numeral and the fact there is mapping between each
individual boy and three toy cars, preventing a scenario where one boy is pushing only two cars, or a scenario in which distribution is non-atomic on the boys. On the other hand, po in svaki+po seems to strongly enforce atomic readings, as shown in Knežević & Demirdache (2018), while inde-pendently, these markers are not necessarily atomic. Being emphatic does not entail that po should force atomicity, so the actual contribution of po in svaki+po still remains an open question in need of answers.
Recall that in Chapters 4 and 5 we discussed that svaki+po construction might influence the semantic change of po. In Chapter 4 we proposed an explanation based on semantic weakening story, with the idea being that
po starts off as a universal quantifier, but it gets weakened to a pluractional,
by discussing two reasons: the svaki+po construction and the fact that we only found the non-exhausted po with transitive sentences. First, given that the svaki+po construction is very common in adult language and acquired earlier by children than these markers independently (compare Knežević & Demirdache (2018) and Chapter 2), it could be that speakers resort to svaki+po for exhaustive interpretations and begin to reanalyze the contribution of po – if svaki is not present, there is no longer an exhaus-tivity requirement on a DistKey. This then means that po is treated as a weaker, non-exhaustive distributive marker. Second, the experiments with intransitive sentences did not reveal two populations. In fact, we only found the universal quantification pattern of responses across all participants, in which the implicit DistKey must be exhausted. We could then argue that non-exhaustive po only appears as a pragmatically weaker form of a universal quantifier in transitive situations where the intended exhaustivity requirement can be achieved with a better construction – svaki+po.
However, at this stage of investigation, we cannot be certain of the direction of change, so it is also possible that po is undergoing semantic strengthening, starting off as an existential plural undergoing pragmatic strengthening to yield a universal meaning – such a process of (“double strengthening”) has been argued for by Bar-Lev & Margulis (2014), for the
1 If the emphasis is on distributivity, and distributivity is a byproduct of event plurality, it could also be that emphatic uses of certain markers such as po still fall under event plurality. However, this is yet to be tested.
6
universal quantifier kol in Hebrew, and by Magri (2014) for the definite article. In fact, an argument for strengthening can be made using the same
svaki+po construction. Imagine, for example, that po is a marker of event
plurality with weak requirements (i.e., no exhaustivity requirements). However, svaki+po sentences, which must be exhaustive (because of the presence of svaki), being more commonly used, cause that the exhaustivity requirements of svaki get transferred to both distributive elements, and
po then assumes the role of an exhaustive distributive marker even on its
own. This could then be imputed to semantic strengthening. The topic of semantic change is a much more intricate topic for further (and extensive) research. For instance, investigating diachronic or sociolinguistic change (e.g., testing whether there is an effect of age or regional differences in the uses of po) may be a useful starting point of determining the direction
po is changing.
Lastly, researchers should also focus their attention on the cross-linguistic aspect of this phenomenon. Other Slavic languages have a similar svaki+po construction, and it would be worth checking if other languages with similar DistShare markers (Korean, for example) could reveal something peculiar about possible co-occurrences of these markers with universal quantifiers.
2.2 Homogeneity and non-atomic readings
Apart from svaki+po issues, the feature of homogeneity proved to be an invaluable diagnostic tool for distinguishing between exhaustivity require-ments imposed by po and maximality effects stemming from the inter-pretations of definite plural expressions. By testing homogeneity effects, we confidently managed to tease apart two populations of speakers found in the experiments in Chapter 4. Homogeneity effects arise in statements with plural definites in which a statement can be true, false or undefined (i.e., the statement is neither true, nor false, so it creates a so-called “gap” scenario). Checking for homogeneity effects in negative sentences with
po allowed us to test whether po is a universal quantifier or not, given the
distinct predictions for universal quantifiers and definite plural expressions. Specifically, universal quantifiers in negative statements should be judged as true when paired with the non-exhausted scenarios, and, conversely, statements with definite plural expressions should either be judged as false or undefined when paired with non-exhausted scenarios. Thus, the results for the negative po sentences clearly showed that one population judges them as true (akin to the universal quantifiers) and the other as both false and undefined (akin to definite plurals.
While we focused only on individual (atomic) distribution in negative contexts, we still do not know what happens in event-distributive (non-atomic) “gap” scenarios. It is still unclear whether different patterns of responses (and two populations) could be identified in the critical condi-tions in Chapter 4 – the ones based on the conclusions of Chapter 3 that show non-atomic groups of individuals that define the relevant (spatial) DistKey over which (exhaustive) distribution takes place.
Furthermore, in Chapter 3, in which there were only cases of non-atomic distribution with intransitive sentences, we did not find evidence of two populations. This could mean that we may not have homogeneity effects with intransitive sentences because all participants would treat
po as a universal quantifier only. An idea would be to test intransitive
sentences in negative contexts, but the challenge there is that the sole argument in the sentence would need to act as a DistShare argument in the sentence with po and as a DistKey argument with svaki. In addition, for both intransitive and transitive sentences, the “gap” scenario could also be the exhausted scenario in which the groups, but not individuals, are exhausted. However, given that we simply do not have experimental data on non-atomic/spatial distribution (in either type of sentences), we cannot infer anything else about the current results. Nevertheless, what is especially interesting is that testing for homogeneity effects in other situations, namely intransitive sentences and non-atomic distributive readings, could potentially contribute to determining the directionality of the semantic change that has resulted in two populations. For instance, if there are no homogeneity effects in non-atomic readings and no evidence for two populations, then the only occurrence of non-exhaustive po would be found in individual distribution, which could not be imputed to the consistent diachronic semantic change in the population, but possibly only to a pragmatic weakening of po. If, however, we find evidence that po in non-atomic readings shows homogeneity effects, then non-exhaustive po is a real and a more unified phenomenon in the language, present across different conditions. Although this would not reveal the direction of the change, it would warrant further research into diachronic change of po.
Going even further, experiments on definite and bare plurals (as well as distributive marking) in positive statements would uncover insightful crosslinguistic facts about homogeneity effects of plural nouns in general. Our intention is to pursue the research on homogeneity further and try to establish baselines about Serbian and hopefully apply our generalizations to other Slavic languages or languages without definite/indefinite articles.
2.3 Temporal distribution and covariation
Temporal distribution has been mentioned back-to-back whenever event-distributive readings were discussed. For instance, in Chapter 3 (section 4) we said that our conclusions that po is a universal distributive quantifier that exhaustively distributes over non-atomic DistKey should apply to temporally-distributive readings, and that we should be able to observe similar exhaustive requirements. It is important to note that tem-poral distribution might be complex in some aspects and easier in others. On the one hand, there may be difficulties for setting up experimental conditions that clearly demonstrate temporal distribution. On the other hand, temporal readings usually seem intuitively easier to understand than spatial readings in terms of imagining multiple events (as demonstrated in Chapter 5). However, since we focused exclusively on spatially-distrib-utive readings, which were also rarely researched, temporally-distribspatially-distrib-utive
6
universal quantifier kol in Hebrew, and by Magri (2014) for the definite article. In fact, an argument for strengthening can be made using the same
svaki+po construction. Imagine, for example, that po is a marker of event
plurality with weak requirements (i.e., no exhaustivity requirements). However, svaki+po sentences, which must be exhaustive (because of the presence of svaki), being more commonly used, cause that the exhaustivity requirements of svaki get transferred to both distributive elements, and
po then assumes the role of an exhaustive distributive marker even on its
own. This could then be imputed to semantic strengthening. The topic of semantic change is a much more intricate topic for further (and extensive) research. For instance, investigating diachronic or sociolinguistic change (e.g., testing whether there is an effect of age or regional differences in the uses of po) may be a useful starting point of determining the direction
po is changing.
Lastly, researchers should also focus their attention on the cross-linguistic aspect of this phenomenon. Other Slavic languages have a similar svaki+po construction, and it would be worth checking if other languages with similar DistShare markers (Korean, for example) could reveal something peculiar about possible co-occurrences of these markers with universal quantifiers.
2.2 Homogeneity and non-atomic readings
Apart from svaki+po issues, the feature of homogeneity proved to be an invaluable diagnostic tool for distinguishing between exhaustivity require-ments imposed by po and maximality effects stemming from the inter-pretations of definite plural expressions. By testing homogeneity effects, we confidently managed to tease apart two populations of speakers found in the experiments in Chapter 4. Homogeneity effects arise in statements with plural definites in which a statement can be true, false or undefined (i.e., the statement is neither true, nor false, so it creates a so-called “gap” scenario). Checking for homogeneity effects in negative sentences with
po allowed us to test whether po is a universal quantifier or not, given the
distinct predictions for universal quantifiers and definite plural expressions. Specifically, universal quantifiers in negative statements should be judged as true when paired with the non-exhausted scenarios, and, conversely, statements with definite plural expressions should either be judged as false or undefined when paired with non-exhausted scenarios. Thus, the results for the negative po sentences clearly showed that one population judges them as true (akin to the universal quantifiers) and the other as both false and undefined (akin to definite plurals.
While we focused only on individual (atomic) distribution in negative contexts, we still do not know what happens in event-distributive (non-atomic) “gap” scenarios. It is still unclear whether different patterns of responses (and two populations) could be identified in the critical condi-tions in Chapter 4 – the ones based on the conclusions of Chapter 3 that show non-atomic groups of individuals that define the relevant (spatial) DistKey over which (exhaustive) distribution takes place.
Furthermore, in Chapter 3, in which there were only cases of non-atomic distribution with intransitive sentences, we did not find evidence of two populations. This could mean that we may not have homogeneity effects with intransitive sentences because all participants would treat
po as a universal quantifier only. An idea would be to test intransitive
sentences in negative contexts, but the challenge there is that the sole argument in the sentence would need to act as a DistShare argument in the sentence with po and as a DistKey argument with svaki. In addition, for both intransitive and transitive sentences, the “gap” scenario could also be the exhausted scenario in which the groups, but not individuals, are exhausted. However, given that we simply do not have experimental data on non-atomic/spatial distribution (in either type of sentences), we cannot infer anything else about the current results. Nevertheless, what is especially interesting is that testing for homogeneity effects in other situations, namely intransitive sentences and non-atomic distributive readings, could potentially contribute to determining the directionality of the semantic change that has resulted in two populations. For instance, if there are no homogeneity effects in non-atomic readings and no evidence for two populations, then the only occurrence of non-exhaustive po would be found in individual distribution, which could not be imputed to the consistent diachronic semantic change in the population, but possibly only to a pragmatic weakening of po. If, however, we find evidence that po in non-atomic readings shows homogeneity effects, then non-exhaustive po is a real and a more unified phenomenon in the language, present across different conditions. Although this would not reveal the direction of the change, it would warrant further research into diachronic change of po.
Going even further, experiments on definite and bare plurals (as well as distributive marking) in positive statements would uncover insightful crosslinguistic facts about homogeneity effects of plural nouns in general. Our intention is to pursue the research on homogeneity further and try to establish baselines about Serbian and hopefully apply our generalizations to other Slavic languages or languages without definite/indefinite articles.
2.3 Temporal distribution and covariation
Temporal distribution has been mentioned back-to-back whenever event-distributive readings were discussed. For instance, in Chapter 3 (section 4) we said that our conclusions that po is a universal distributive quantifier that exhaustively distributes over non-atomic DistKey should apply to temporally-distributive readings, and that we should be able to observe similar exhaustive requirements. It is important to note that tem-poral distribution might be complex in some aspects and easier in others. On the one hand, there may be difficulties for setting up experimental conditions that clearly demonstrate temporal distribution. On the other hand, temporal readings usually seem intuitively easier to understand than spatial readings in terms of imagining multiple events (as demonstrated in Chapter 5). However, since we focused exclusively on spatially-distrib-utive readings, which were also rarely researched, temporally-distribspatially-distrib-utive
6
readings in terms of exhaustivity requirements remained experimentally unverified.
In Chapter 5 we also discussed the differences between the semantic proposals of Knežević (2015) and Cable (2014), and we stressed that the covariation requirement is the crucial ingredient in Cable’s semantics that Knežević does not include. An argument against covariation comes out more prominently in temporally-distributive readings, where it can be argued that it is actually possible to re-use the distributive share arguments (e.g., in the sentence: The girls are bathing po two dogs, it may be possible to bathe the same two dogs over the course of a week).2 In addition, since
there could be intermediate cases (i.e., maybe some dogs can be different, but not all of them on every occasion), the facts here are not definitively established. Thus, focusing more on temporal distribution could also give answers to covariation questions about how much covariation is sufficient for different DistShare markers, and experimental research in this direction is highly encouraged.
2.4 NP-internal distributive and cumulative readings
Last but not least, we would like to highlight two more topics that emerged from this work and that would really merit investigating further. First, let us remind a reader that even though the entirety of Chapter 5 was focused on our borderline case with minimal spatial separation, “the pulling” example, that most accurately depicts an NP-internal distributive reading in an (alleged) single-event scenario (Figures 3 and 4, Chapter 5, section 4), this example (and all others described in Chapter 5, including weighing of the apples, mashing vs. boiling potatoes, and once-only predicates such as to
kill) has not been experimentally confirmed. We indeed have judgments
from some informants, who confirmed these readings, but in order to see the range of acceptance of similar borderline cases, we need a carefully designed experimental study that especially focuses on the amount of context needed to more confidently accept these readings.
Second, the topic of cumulative readings, although not discussed in detail in this book, has been an important finding in our act-out experiment in Chapter 2. Specifically, cumulative (1-to-1) answers are found as an intermediate stage of acquisition of distributive markers in Serbian for both distributively marked (which is an incorrect answer) and unmarked sentences (which is a possible correct answer). These answers were some-thing in between a collective and a distributive answer – while the total number of objects acted upon was interpreted non-scopally (say, to total of three), the objects were distributed 1-to-1 between the three agents, so that each agent got one object. On the other hand, Dutch speakers gave cumulative answers only when they are pragmatically plausible (e.g., when it made more sense to put one hamburger for each boy to eat, rather than arranging a collective action of eating three hamburgers). It
2 There may be other plausible situations in which covariation is not necessary (as discussed in Chapter 5). For instance, in the sentence: During the games, the referee raised po two yellow cards, the cards could be the same and raised several times (see Chapter 5, section 2.1.1).
would certainly be interesting to explore these cases further and see why cumulative 1-to-1 answers are dominant with Serbian children, even if they are completely incorrect with distributive markers. Furthermore, cumulativity as a phenomenon is also intriguing to check from a different perspective. In particular, the analysis of cumulative partitioning is also one of the differences between Knežević’s (2015) and Balusu’s (2006) accounts in terms of how the unmarked argument is distributed over the partitioned events. For instance, in a sentence with an object marked as a distributive share, Balusu’s semantics require the distribution of the whole event (e.g., Figure 1b, Chapter 3), including the agents, while Knežević leaves them out (e.g., Figure 1c, Chapter 3). That means that for Knežević, the unmarked argument can cumulatively act on the marked argument, resulting in many different scenarios, as discussed in Chapters 4 and 5. It would then be useful to tease apart these constraints experimentally and (ideally) compare Serbian and Telugu in this respect.
With all of these intriguing puzzles, we leave the reader here. Throughout this book we have taken every opportunity to highlight the benefits of experimental analysis as a way to confirm existing theoretical accounts, and bring out topics that offer a fertile ground for further experimental research. We dove into the largely unexplored territory of DistShare mark-ers and got exciting results, but, more importantly, we discovered many paths and possibilities to explore this territory further. We can only hope that with this research we sparked some interest in carrying out, hand in hand, experimental and theoretical investigation of different aspects of quantification which could, perhaps, help complete our story.
6
readings in terms of exhaustivity requirements remained experimentally unverified.
In Chapter 5 we also discussed the differences between the semantic proposals of Knežević (2015) and Cable (2014), and we stressed that the covariation requirement is the crucial ingredient in Cable’s semantics that Knežević does not include. An argument against covariation comes out more prominently in temporally-distributive readings, where it can be argued that it is actually possible to re-use the distributive share arguments (e.g., in the sentence: The girls are bathing po two dogs, it may be possible to bathe the same two dogs over the course of a week).2 In addition, since
there could be intermediate cases (i.e., maybe some dogs can be different, but not all of them on every occasion), the facts here are not definitively established. Thus, focusing more on temporal distribution could also give answers to covariation questions about how much covariation is sufficient for different DistShare markers, and experimental research in this direction is highly encouraged.
2.4 NP-internal distributive and cumulative readings
Last but not least, we would like to highlight two more topics that emerged from this work and that would really merit investigating further. First, let us remind a reader that even though the entirety of Chapter 5 was focused on our borderline case with minimal spatial separation, “the pulling” example, that most accurately depicts an NP-internal distributive reading in an (alleged) single-event scenario (Figures 3 and 4, Chapter 5, section 4), this example (and all others described in Chapter 5, including weighing of the apples, mashing vs. boiling potatoes, and once-only predicates such as to
kill) has not been experimentally confirmed. We indeed have judgments
from some informants, who confirmed these readings, but in order to see the range of acceptance of similar borderline cases, we need a carefully designed experimental study that especially focuses on the amount of context needed to more confidently accept these readings.
Second, the topic of cumulative readings, although not discussed in detail in this book, has been an important finding in our act-out experiment in Chapter 2. Specifically, cumulative (1-to-1) answers are found as an intermediate stage of acquisition of distributive markers in Serbian for both distributively marked (which is an incorrect answer) and unmarked sentences (which is a possible correct answer). These answers were some-thing in between a collective and a distributive answer – while the total number of objects acted upon was interpreted non-scopally (say, to total of three), the objects were distributed 1-to-1 between the three agents, so that each agent got one object. On the other hand, Dutch speakers gave cumulative answers only when they are pragmatically plausible (e.g., when it made more sense to put one hamburger for each boy to eat, rather than arranging a collective action of eating three hamburgers). It
2 There may be other plausible situations in which covariation is not necessary (as discussed in Chapter 5). For instance, in the sentence: During the games, the referee raised po two yellow cards, the cards could be the same and raised several times (see Chapter 5, section 2.1.1).
would certainly be interesting to explore these cases further and see why cumulative 1-to-1 answers are dominant with Serbian children, even if they are completely incorrect with distributive markers. Furthermore, cumulativity as a phenomenon is also intriguing to check from a different perspective. In particular, the analysis of cumulative partitioning is also one of the differences between Knežević’s (2015) and Balusu’s (2006) accounts in terms of how the unmarked argument is distributed over the partitioned events. For instance, in a sentence with an object marked as a distributive share, Balusu’s semantics require the distribution of the whole event (e.g., Figure 1b, Chapter 3), including the agents, while Knežević leaves them out (e.g., Figure 1c, Chapter 3). That means that for Knežević, the unmarked argument can cumulatively act on the marked argument, resulting in many different scenarios, as discussed in Chapters 4 and 5. It would then be useful to tease apart these constraints experimentally and (ideally) compare Serbian and Telugu in this respect.
With all of these intriguing puzzles, we leave the reader here. Throughout this book we have taken every opportunity to highlight the benefits of experimental analysis as a way to confirm existing theoretical accounts, and bring out topics that offer a fertile ground for further experimental research. We dove into the largely unexplored territory of DistShare mark-ers and got exciting results, but, more importantly, we discovered many paths and possibilities to explore this territory further. We can only hope that with this research we sparked some interest in carrying out, hand in hand, experimental and theoretical investigation of different aspects of quantification which could, perhaps, help complete our story.
