"Buenos dias", "buenas noches" -- this was the first words in a foreign language I heard in my life, as a three-year old boy growing up in developing post-war Western Germany, where the first gastarbeiters had arrived from Spain. Fascinated by the strange sounds, I tried to get to know some more languages, the only opportunity being TV courses of English and French -- there was no foreign language education for pre-teen school children in Germany yet in those days. Read more
To find some answers Tim Machan explores the language's present and past, and looks ahead to its futures among the one and a half billion people who speak it. His search is fascinating and important, for definitions of English have influenced education and law in many countries and helped shape the identities of those who live in them.
This volume provides a new perspective on the evolution of the special language of medicine, based on the electronic corpus of Early Modern English Medical Texts, containing over two million words of medical writing from 1500 to 1700.
SUMMARY This volume investigates a curious non-adult-like performance pattern displayed by children in their interpretation of quantifier-scope interaction in double-object constructions. Focusing on children acquiring Dutch, the author presents a series of experiments that establish the non-target-like pattern, dubbed the ‘Reverse-pattern’, and tests a number of lexically-based hypotheses about the source of the non-target-like behaviour. With evidence that the non-target-like performance is restricted to sentences containing distributive universal quantifiers, the author provides an account of the Reverse-pattern that attributes children’s non-target-like scope interpretations to a non-adult-like ability to revise an initial interpretation strategy.
Chapter 1 lays the groundwork for the discussion. The object of study is the Frozen Scope Constraint (FSC), exemplified in double-object (DO) constructions such as (1). Unlike the 'to'-dative construction in (2), (1) is scopally unambiguous: only the surface scope interpretation, on which the indirect object scopes over the direct object, is available.
(1) The car salesman showed a man every car. (a>>every; *every>>a) (2) The car salesman showed every car to a man. (a>>every; every>>a)
Van der Ziel reviews two alternative accounts of the FSC: Bruening’s (2001) syntactic account, and Goldberg’s (2006) information-structural account. Van der Ziel ultimately adopts Goldberg’s account, according to which the indirect object in the DO construction tends to take wide scope due its (secondary) topic status. On the assumption that DPs higher on the topicality scale typically scope above DPs lower on the topicality scale, van der Ziel suggests that when a speaker chooses to produce a DO construction over the ‘to’-dative, this signals that the new information to be conveyed concerns the theme of the action (the direct object) and not the recipient (the indirect object). Goldberg’s account predicts that scope freezing should be acquired as soon as children have acquired the secondary topic status of the indirect object.
Chapter 2 carefully establishes the non-target-like ‘Reverse-pattern’ in children’s interpretation of DO constructions such as (3).
(3) The bear gave a hedgehog every piece of cake.
In (3), the FSC blocks the inverse, distributive reading whereby the universal scopes over the indefinite. Van der Ziel presents a series of experiments using the truth value judgment task, the results of which suggest that English- and Dutch-speaking preschool children lack knowledge of the FSC. More puzzling, they not only allow the reading blocked by the FSC, they reject the one interpretation that adults allow. The critical test sentences contain an indefinite and a universally quantified NP (as in (3)), and are presented in two kinds of contexts: a distributive ‘ALL>ONE’ context, in which each piece of cake is given to a different hedgehog, and a non-(prototypically-)distributive ‘ONE>ALL’ context, in which a single hedgehog receives all the pieces of cake. Given the FSC, the sentence in (3) should only be true in a ONE>ALL context. Testing a variety of structures, including DO constructions, ‘to’-datives, scrambled ‘to’-datives, and simple transitive sentences, van der Ziel finds the Reverse-pattern surfacing at a rate of 38-52%. She then argues that the Reverse-pattern cannot be attributed to non-target-like scope assignment, incomplete acquisition of DO datives, or a lack of exhaustive pairing.
Having laid out the experimental evidence for the Reverse-pattern, van der Ziel devotes the next two chapters to lexical factors that may give rise to the Reverse-pattern. Chapter 3 develops and tests hypotheses that relate the Reverse-pattern to incomplete acquisition of the indefinite article. On the one hand, a modified version of Su’s (2001) Lexical Factor Hypothesis posits that children exhibiting the Reverse-pattern have not fully acquired the meaning of the indefinite article ‘a’/‘an’; this predicts that the Reverse-pattern should be restricted to this particular indefinite. On the other hand, Krämer’s (2000) Non-Integration Account and the Singleton Restriction Hypothesis (SRH) (building on Schwarzschild’s (2002) theory of indefinites) posit that children have not fully acquired the meaning of singular indefinites more generally; the latter, for example, posits that children have difficulty restricting the domain of indefinites to a singleton set when the discourse context provides multiple salient individuals that could plausibly be members of the domain. Given these predictions, van der Ziel presents a series of experiments showing first that the Reverse-pattern is not restricted to a particular indefinite, but rather is observed with a variety of singular indefinites (e.g., ‘some’, numeral ‘one’ or ‘één’ in Dutch) and second, that children are able to interpret indefinites in a target-like way when the sentences in question do not contain universal quantification. In light of these findings, van der Ziel proposes that the source of the Reverse-pattern lies not with the indefinite, but rather the universal quantifier.
Given evidence that the Reverse-pattern is not driven by a deficiency in the interpretation of the indefinite, Chapter 4 explores an alternative lexical explanation: that the problem lies instead in children’s interpretation of universal quantification. Van der Ziel formulates the Distributivity Hypothesis (building on aspects of Drozd and van Loosbroek (2006)). According to this hypothesis, 5-year-olds know that universal quantifiers like ‘every’ are obligatorily distributive, but reject sentences in contexts that do not unambiguously support a distributive interpretation. Consider (3) again. In an unambiguously distributive ALL>ONE context, where there is a one-to-one correspondence between hedgehogs and pieces of cake, children have no problem verifying that the property of being given to a hedgehog holds for each individual piece of cake; if unconstrained by the FSC, they will accept the sentence in such contexts. In a ONE>ALL context however, where there is no one-to-one correspondence between hedgehogs and pieces of cake, children have difficulty verifying whether every piece of cake is given to a hedgehog, and so reject the sentence, giving rise to the Reverse-pattern. Thus the problem lies in a mismatch between the perceived (lack of) distributivity in the context and the obligatory distributivity of the universal quantifier. The experiments presented in Chapter 4 reveal that the Reverse-pattern surfaces with sentences containing obligatorily distributive universal quantifiers such as ‘ieder’ (every), but not cardinal quantifiers or non-obligatorily distributive universal quantifiers such as the collective ‘alle’ (all), providing preliminary evidence for the Distributivity Hypothesis. Children have acquired the distributivity feature as part of the lexical feature specification of distributive universal quantifiers, but have difficulty evaluating whether the context unambiguously satisfies this distributivity requirement.
Given the restriction of the Reverse-pattern to test sentences containing distributive universal quantifiers, Chapter 5 further explores the role of distributivity, i.e. the requirement that whatever applies to a larger set also applies to each individual element of the set. While Tunstall’s (1998) Event-Distributivity Condition states that sentences containing an ‘every’-NP can only be true of event structures that are at least partially distributive, van der Ziel presents results from a picture selection task showing that adults tend to prefer fully distributive event structures over partially distributive ones. Thus she hypothesizes that the lexical meaning of a distributive universal quantifier encodes a requirement for fully distributive or prototypically distributive event structures, and moreover that adults but not children can accommodate deviations from this basic interpretation:
(4) The Weak Prototypical Distributivity Hypothesis: Distributive universal quantifiers force a prototypically distributive event structure, i.e. an event structure in which every element in the restrictor set of the universal quantifier is associated with a distinct subevent. However, adults can deviate from this prototypical interpretation, whereas children cannot. (p. 148)
(5) The Strong Prototypical Distributivity Hypothesis: Distributive universal quantifiers force a prototypically distributive event structure, i.e. an event structure in which every element in the restrictor set of the universal quantifier is associated with a distinct subevent. In addition, these subevents need to be unique. However, adults can deviate from this prototypical interpretation, whereas children cannot. (p. 149)
Although the results of a truth value judgment task do not fully support either hypothesis as an explanation for children’s performance, van der Ziel observes crucially that all contexts in which children accepted the test sentences involved a full (though not necessarily one-to-one) linking of the set of possible recipients and a set of objects, a property dubbed ‘Full Set Linking.’ This property was not attested in any of the contexts which lead children to reject the test sentences.
Chapter 6 presents van der Ziel’s overarching account of the Reverse-pattern, driven by the crucial observation of Full Set Linking: all contexts which yielded acceptances (but no contexts which yielded rejections) were ones where the set of recipients and the set of objects were exhaustively linked (though not necessarily in a one-to-one correspondence). The question arises as to why Full Set Linking comes into play only when children are evaluating sentences containing distributive universal quantifiers. Van der Ziel assumes that both children and adults adhere to a weak version of Prototypical Distributivity, wherein each element in the restrictor set of the distributive universal needs to be associated with a distinct subevent. Both groups also use a short-cut strategy to verify prototypical distributivity: they evaluate the end state. When this shortcut fails, adults can revise their strategy and subsequently verify whether each element in the restrictor set is associated with a distinct event. Children on the other hand continue to evaluate the end state, and thus induce the requirement for Full Set Linking. Moreover, unlike children, adults can accommodate situations that deviate from prototypical distributivity; for example, when a reading is blocked by the FSC, adults can reject the sentences even if prototypical distributivity holds. This takes us full circle back to the question of whether children have knowledge of the FSC. Given the persistence of the Reverse-pattern errors through the age of 5 years, van der Ziel endorses the information structure account of the FSC, suggesting that children have not yet acquired the link between the indirect object position and its topic status. Adults strongly adhere to the FSC because they strongly associate the indirect object with topicality (and thus wide scope). Children who have not acquired the obligatory topic status of indirect objects are unconstrained by the FSC, and adhere to the Full Set Linking requirement, thereby yielding the Reverse-pattern.
EVALUATION This volume presents a very detailed, well thought-out investigation into what appears to be a very challenging set of data. It grapples with a performance pattern that appears to involve layers of complexity, touching on facets of syntax (i.e. movement and scope), semantics (i.e. interpretation strategies), information structure (e.g., topicality), and processing (i.e. the ability to revise one’s interpretation strategy), and ultimately motivates a similarly multi-layered account.
The investigation is well situated within the context of previous studies on English, Chinese, and Dutch (Su & Crain, 2000; Su, 2001; Philip, 2005; Hendriks, Koops van’t Jagt, and Hoeks, 2012; Philip & Coopmans, 1995), and the introductory chapters clearly set the stage for an analysis of the unusual pattern of results dubbed the Reverse-pattern. Presenting her own experimental evidence for the existence of the Reverse-pattern, van der Ziel systematically works through a number of possible hypotheses regarding the source of the Reverse-pattern, deftly leading the reader through the predictions of these hypotheses, and the carefully designed experiments that put these predictions to the test. Building on the information gathered in each subsequent experiment, van der Ziel arrives at a plausible, multi-layered explanation for the Reverse-pattern, notably one which succeeds in incorporating the multiple different factors that seem to be at play.
One factor that makes the larger puzzle (but also the task of working through the sometimes dense and complex experimental findings) all the more challenging is the overall non-uniformity of the data. The dissertation is predicated on a single pattern of behaviour that itself only surfaces at a rate of about 38-52%. Given the variability in the data, it is very useful to see the individual results, and van der Ziel very helpfully provides classifications that group the children based on their patterns of responses (target-like, Reverse-pattern, ambiguity, unclear, etc.). It is not entirely clear what to make of the sometimes relatively large number of children who fall into the ‘mixed/unclear’ category (in Experiment II, for example). Children who fail to be classified according to any of the expected response patterns present a curiosity, as they must have nevertheless passed all control items sufficiently, in order to be included in the data analysis.
The volume should also generate some lively discussion, as it leaves open some very interesting questions. I lay out a few of them here. One question concerns the sets of non-target-like adults who were tested across the various experiments. The 10 adult controls in Experiment I (Chapter 2) for example, performed at 70-80% accuracy. Two of the ten adults displayed the Reverse-pattern and one showed mixed behaviour. What is one to make of this data? Van der Ziel notes (footnote 59) that her use of the term ‘targetlike performance’ is not to be equated with ‘adultlike performance’, as the adults she tests do not always perform according to the predictions of the linguistic theory. The difficulty of this underscores the challenging nature of the problem to be solved. Given the subset of adults who consistently do not conform to the experimental predictions, a compelling linguistic theory must be able to account not only for the child data, but also for these adults’ performance. The question then arises whether van der Ziel’s overarching account can capture the non-target-like adult data. Presumably these adults were on task if they passed the control trials sufficiently to be included in the data analysis, and presumably they were consistent enough in their responses in order to be classified into a category of response pattern. The natural explanation, if one adopts van der Ziel’s final account, is that these adults stuck to their initial verification strategy (checking for prototypical distributivity) -- and for any number of reasons (fatigue, lack of attention), did not end up revising their interpretation strategy. In this context, an interesting consideration is what would happen if we were to increase the processing load of the experimental condition; if the reason that children cannot revise the interpretation strategy is due to the processing costs of revision for example, we might expect to see adults turn into children under heavy load conditions.
Another question that might be investigated in future research is whether we can shift participants’ biases for different verification strategies. Figure 1 on page 32 provides the illustration accompanying the test sentence ‘Snow White gave a lady every flower’. It’s worth noting that none of the three ladies that appear in the image is more salient than the other two. If children did have difficulty restricting the domain of an indefinite to a singleton individual unless that individual was very salient, we might expect the domain for the indefinite ‘a lady’ to include all three ladies; in this case, the child might then be biased towards a (prototypically distributive) verification strategy that checks for each lady whether she was involved in a giving event. So while it may be that children appear to require a kind of exhaustive pairing, perhaps other factors drive this apparent need; for example, non-saliency of the individuals in the possible domain might bias towards a distributive verification strategy. Another possibility for future research involves testing the Full Set Linking requirement experimentally, that is, monitoring participants’ verification strategies by tracking their eye movements as the stimuli are presented. Given the two matching sets (of recipients and objects), what are the participants attending to? Given van der Ziel's proposal, we might expect children to attend to whether the two sets are matched up; adults on the other hand might initially check the two sets, only to revise and specifically attend to the individuals in the restrictor set of the universal.
Yet another question pertains to the developmental story. What has to mature in order for the children to behave in a target-like way? Interestingly, van der Ziel notes that no effects of age were found across the 4-, 5-, and 6-year-olds. The lack of an age effect is especially interesting, if what has to mature are the mechanisms for verification, and the processes involved in revising initial interpretations (cf. Conroy et al. (2009) for a study that finds a U-shaped developmental trajectory in children’s scope interpretations; these authors suggest that children go through an intermediate stage where they have acquired adult-like parsing preferences but are not adept at revising their interpretations).
Finally, consider the implications of interactions among the relevant factors that van der Ziel raises in her goal of deriving the Reverse-pattern. Van der Ziel’s discussion takes us through the many layers of the problem at hand: syntax/information structure, distributivity/lexical requirements, interpretation strategies, and revision abilities. She ultimately concludes that the Reverse-pattern finds its source in the combination of two factors: (i) the Reverse-pattern children have not acquired the FSC; and (ii) these children continue to use the short-cut strategy for evaluating quantifier interpretation. In light of the multiple relevant factors however, it’s interesting to consider whether possible interactions among these different factors can predict the other response patterns, namely the Ambiguity pattern (yes-responses in both ONE>ALL and ALL>ONE contexts) and the target-like pattern (yes-responses in the ONE>ALL contexts and no-responses in the ALL>ONE contexts). It seems to me that there are at least three forces in tension with each other, having to do with the different ingredients (raised throughout the book) that the child must acquire before she can be fully target-like: the FSC (information-structural in nature), the distributivity requirement (lexical in nature), and revision abilities (involving processes that generate and select among possible interpretations). In van der Ziel’s story, the distributivity requirement is lexically obligatory but violable, while the FSC is generally less violable (or in terms of constraint ranking, appears to rank higher than the distributivity requirement for adults). Finally, it appears that children’s revision abilities require time to develop to an adult-like capacity. Assuming these three factors, we seem to be able to predict exactly those three response patterns observed by van der Ziel: (1) target-like children include any children who have acquired the FSC, whether they have acquired the distributivity requirement or adult-like revision abilities; the FSC, being less violable than distributivity, will always lead to a yes-response in the ONE>ALL context and a no-response in the ALL>ONE context. (2) Reverse-pattern children include children who have not acquired the FSC or adult-like revision abilities, but have acquired the distributivity requirement; unconstrained by the FSC and without the ability to revise interpretation strategies, they will always check for prototypical distributivity and respond accordingly. Finally, the children who fall into the “Ambiguity” category include children who have not acquired the FSC, but have acquired the distributivity requirement and adult-like revision abilities; they are unconstrained by the FSC and prefer prototypical distributivity, but can deviate like adults. The remaining two logically possible groups are: (i) children who have not acquired the FSC or the distributivity requirement, but have adult-like revision abilities, and (ii) children who have not acquired the FSC, the distributivity requirement, or adult-like revision abilities. The predictions for these last two groups are unclear. In short however, it seems that some interaction of the factors raised by van der Ziel can successfully capture the observed response patterns.
Overall, I found van der Ziel’s argumentation to be compelling and well supported by her experimental findings. This volume is clearly organized and presents a strong narrative. The research questions are well motivated, the hypotheses are well thought out, and the experiments are well-designed to test the hypotheses. The book should primarily be of interest to linguists and psychologists interested in child language acquisition, particularly of syntax/semantics, scope, quantification, and information structure. In addition, the data (particularly the adult data) presented should be informative for theoretical linguists interested in theories of quantification, distributivity, and event structure. Finally, the dissertation contains an array of eight carefully designed experiments; clear descriptions of all the methodologies are provided, and appendices include the full experimental stimuli, allowing for easy reconstruction of the designs. The extensive experimental component makes it a useful methodological guide to the strategies and issues that are relevant for researchers conducting experimental investigations on scope ambiguity, as well as researchers conducting child language studies more generally.
REFERENCES Bruening, B. 2001. QR obeys superiority: Frozen scope and ACD. Linguistic Inquiry 32.2:233-273.
Conroy, A., J. Lidz, and J. Musolino. 2009. The fleeting isomorphism effect. Language Acquisition 16:106-117.
Drozd, K.F. and E. van Loosbroek. 2006. The effect of context on children’s interpretations of universally quantified sentences. In V. van Geenhoven, ed., Semantics Meets Acquisition. Dordrecht: Kluwer, 115-140.
Goldberg, A.E. 2006. Constructions at Work: The Nature of Generalization in Language. Oxford: Oxford University Press.
Hendriks, P., R. Koops van ‘t Jagt and J. Hoeks. 2012. Restricting quantifier scope in Dutch: Evidence from child language comprehension and production. In B. Stolterfoht and S. Featherston, eds., Empirical Approaches to Linguistic Theory: Studies of Meaning and Structure. Studies in Generative Grammar. Berlin: De Gruyter Mouton, 147-167.
Krämer, I. 2000. Interpreting indefinites: An experimental study of children’s language comprehension. Doctoral dissertation, Utrecht University.
Philip, W.C.H. 2005. Pragmatic control of specificity and scope: Evidence from Dutch L1A. In E. Maier, C. Bary, and J. Huitink, eds., Proceedings of SUB9. Nijmegen: NCS, 271-285.
Philip, W.C.H. and P.H.A. Coopmans. 1995. Symmetrical interpretation and scope ambiguity in the acquisition of the universal quantification in Dutch and English. In J. Don, B. Schouten, and W. Zonneveld, eds., OTS Yearbook 1994, 85-134.
Schwarzschild, R. 2002. Singleton indefinites. Journal of Semantics 19.3:289-314.
Su, Y.-C. 2001. Scope and specificity in child language: A cross-linguistic study on English and Chinese. Doctoral dissertation, University of Maryland.
Su, Y.-C. and S. Crain. 2000. Children’s scope taking in double-object constructions. CLS 36: the Panels, 485-498.
Tunstall, S. 1998. The interpretation of quantifiers: Semantics and processing. Doctoral dissertation, University of Massachusetts, Amherst.
ABOUT THE REVIEWER:
Lyn Tieu is currently a PhD candidate in the Department of Linguistics at the University of Connecticut. As of January 2014, she will join the LINGUAE (CNRS) group in Paris as a post-doctoral researcher. Her research interests involve child language acquisition and the development of linguistic phenomena that lie at the syntax-semantics-pragmatics interface. Her dissertation focuses in particular on the acquisition of the negative polarity item ‘any’, investigating children's sensitivity to NPI licensing conditions, as well as their knowledge of domain widening and exhaustification of alternatives. She is also more generally interested in formal semantics, experimental syntax/semantics/pragmatics, theoretical syntax, and bilingual first language acquisition.