"In this book, Richard Kern explores how technology matters to language and the ways in which we use it. Kern reveals how material, social and individual resources interact in the design of textual meaning, and how that interaction plays out across contexts of communication, different situations of technological mediation, and different moments in time."
Date: Mon, 23 May 2005 15:08:50 +0200 From: Susanna Bartsch Subject: The Ups and Downs of Child Language
AUTHOR: Gualmini, Andrea TITLE: The Ups and Downs of Child Language SUBTITLE: Experimental Studies on Children's Knowledge of Entailment Relationships and Polarity Phenomena SERIES: Outstanding Dissertations in Linguistics PUBLISHER: Routledge YEAR: 2005
Susanna Bartsch, unaffiliated
Gualmini's book (x + 198 pages) is one of the volumes of the series ''Outstanding Dissertations in Linguistics'' published by Routledge since 1992 and comprehending at the moment more than 90 titles. The volume under review is structured in 6 chapters preceded by a table of contents, acknowledgements, and a 2-page long introduction; an appendix, the reference section, an index of names, as well as an index of topics complete the volume. Every chapter ends with a conclusion in which the results not only are summed up, but also in some cases completed with new information.
Introduction Gualmini starts his dissertation with the assertion: ''Entailment relations among sentences are relevant for several distributional and interpretive phenomena across _natural languages_'' (p. 3; my emphasis, SB). And further: ''[...] linguistic research has [...] uncovered several interesting quirks that arise because of the particular way _natural languages_ make reference to entailment relations'' (p. 3; my emphasis, SB). And: ''From the perspective of the acquisitionist, entailment relations provide an interesting case study in that a fairly complex theoretical apparatus is needed to account for seemingly simple facts'' (p. 3). In the introduction, Gualmini presents short summaries of each of the six chapters of the dissertation (p. 4).
Chapter 1 - Language and Acquisition (pp. 5-37) In this chapter, Gualmini introduces the discussion which runs through the whole volume under review, i.e., ''entailment relations, polarity items and inferences across natural languages'' (pp. 6f.) and how these linguistic phenomena can be approached when using one of three models of language acquisition: the ''Conservative Learning Model'', the ''Rich Input Model'', and the ''Continuity Assumption'' (pp. 24f.).
Gualmini is mainly concerned with ''downward entailments'' (henceforth: DE): DE ''describes the environments in which a noun phrase can be substituted with one that picks out a subset of the denotation of the original noun phrase without affecting the truth of the original sentence'' (p. 9):
(7a) John graduated BEFORE HE WROTE HIS FIRST PAPER licenses the inference: John graduated BEFORE HE WROTE HIS FIRST GOOD PAPER. (p. 8)
(Note: Gualmini's examples are reproduced in the present review with their original numbers. I use capital letters for terms, which appear underlined in the original text.)
Gualmini discusses and, in part, challenges some of the features of DE as maintained in previous literature. According to him:
(i) a ''downward entailing operator OP-DE'' (p. 12; the original notation is with capitals OP plus low-placed capitals DE), as _without_ and _before_, is not automatically operative (pp. 16f., 20);
(ii) the view that ''negative polarity items (NPIs)'', as _any_ and _ever_, are used grammatically in downward entailing environments, whereas the occurrence of ''positive polarity items'' (PPIs), as _every_, and _already_, in such environments often leads to ungrammaticalities (pp. 9f., 22f.) is challenged, since there are cases of grammatical sentences containing PPIs in downward entailing contexts (p. 22), a claim further developed in Chapter 6;
(iii) but DE does not occur in negative environments only (pp.20f.);
(iv) sentences containing an OP-DE and the disjunction operator _or_ license conjunctive inferences, according to the scheme OP-DE(A or B) entails OP-DE(A) and OP-DE(B) that corresponds to ''one of De Morgan's laws of propositional logic'' (pp. 12f.);
(v) the decisive force for the occurrence of DE is the structure dependent notion of c-command (p. 16), a claim further developed in Chapter 3.
Next, Gualmini discusses which model of language acquisition is the most adequate for explaining ''how children achieve th[is] intricate pattern of linguistic behavior'' (p. 18). Tomasello's (2000) ''Conservative Learning Model'' and Pullum's & Scholz' (2002) ''Rich Input Model'' are out of question because of their notion of qualitative distinction between children's and adults' linguistic competence (pp. 30f.), their refusal of the Poverty-of-Stimulus hypothesis (p. 34), as well as their relying on what Gualmini calls ''syntax-blind mechanisms'' (Chapter 2, p. 50). Gualmini endorses the ''Continuity Hypothesis'' (MacNamara 1982, Pinker 1984, Crain & Thornton 1998, Crain & Pietroski 2001) according to which ''observed mismatches between children and adults are compatible with U[niversal] G [rammar]'' (p. 34). Since ''[t]he property of downward entailment is closely related to the meaning of certain linguistic expressions across natural languages'' (p. 35), and the acquisition of such meanings ''is largely determined in advance by Universal Grammar'', ''we expect children's knowledge of downward entailment to be essentially adult-like from the earliest stages of language development, [...] we expect children to be like adults in the classification of a linguistic context as DE or non-DE, since natural languages do not differ in this respect'' (p. 36).
Chapter 2 - Entailment and Polarity Phenomena in Child Language (pp. 39-61) In this chapter, Gualmini reviews the experimental research on child acquisition of DE (pp. 40f.), before he presents his own Experiment I, a study of children's conjunctive interpretation of ''the disjunction operator _or_ in the scope of the quantified expression _None of the Ns_'' (pp. 48f.).
Gualmini points out to substantial and methodological shortcomings in the previous research and proposes solutions for them (pp. 44f.):
(i) the focus on items limited to DE environments, as _any_, should be expanded to accommodate different structural configurations, as sentences with a non-downward entailing environment; and
(ii) the methodology (Elicited Production tasks) is concerned mainly with performance (production) and should be substituted/completed with Truth Value Judgment tasks (developed by Crain & McKee 1985, as Gualmini in Chapter 3, p. 73 informs us), focusing, therefore, on comprehension/interpretation (competence).
An example for such an experimental improvement is Boster's & Crain's (1993) study on children's interpretation of disjunction in the scope of _every_ employing the Prediction Mode of the Truth Value Judgment task, in which the children are presented with a short story and, at the end, with a target sentence commenting on the final outcome of the story (pp. 46f.). (Note: All experiments employing the Truth Value Judgment task discussed in the volume under review have this basic design.)
Nevertheless, Gualmini points out that Boster's & Crain's results might conduct to a misleading conclusion, namely, that children never assign a conjunctive interpretation to the disjunction operator, therefore never assuming that _or_ could appear in downward entailing contexts (pp. 46f.). In his Experiment I, using the Truth Value Judgment task, Gualmini found out that children do assign an adult-like conjunctive interpretation to the disjunction operator when it appears in a downward entailing environment, e.g., when it falls within the nuclear scope of an OP-DE, as the negative quantified expression _None of the Ns_ (pp. 47f.). Gualmini's subjects (30 children from age 3;10 to 5;10) interpreted (25) in 87,5% of the time as equivalent to (26a) (pp. 52f.):
(25) [...] none of the pirates found the jewel or the necklace.
(26a) None of the pirates found the jewel AND none of the pirates found the necklace.
Such findings suggest that ''syntax-blind mechanisms'' (p. 50) as noun substitution as proposed in Tomasello's (2000) Conservative Learning model ''do[...] not achieve descriptive adequacy-not to mention explanatory adequacy-as it pertains to children's interpretation of sentences containing disjunction''. Noun substitution, as ''a very simple mechanism'', might guide their _production_ of such sentences, but for their _interpretation_ ''something more elaborated'' is needed (pp. 50, 53, 60). A better explanation is provided by the Continuity Hypothesis, since children's interpretations ''conform to the scheme [OP-DE(A or B) entails OP-DE(A) and OP-DE(B)] from the early stages of language acquisition, because that interpretation is UG-compatible'' (p. 50).
Chapter 3 - The Structure of Child Language (pp. 63-88) In this chapter, Gualmini reviews previous studies on the role of (innate) language-specific structural constraints in first language acquisition in general (pp. 63-77). Then he discusses the general relationships between structure dependence and polarity phenomena in terms of the notion of c- command (77-81), before presenting Experiment II and III in which he investigates the role of structure dependent principles in children's interpretation of downward entailments (pp. 81-87).
In his review on the role of structure dependence in child language acquisition, Gualmini again concludes that some of these studies, focusing on language production (using Elicited Production tasks), present the deficiencies discussed in Chapter 3 (pp. 63-72), that other studies, focusing on comprehension/interpretation (using True Value Judgment tasks) do not (pp. 73-77). On the other hand, these studies provide evidence for the claim that ''structure dependence is a general constraint on all principles of grammar'' (p. 70, 72) and a ''structure-blind rule [allusion to the Conservative Learning Model] is unlikely to allow the child to converge on the adult grammar in absence of negative evidence'' (p. 65); the same applies for ''domain general cue[s]'' (pp. 80f.).
Gualmini then further develops the claim outlined in Chapter 1 that the common features of DE are constrained by the structural notion of c- command. In the case of the conjunctive interpretation of disjunction, the OP-DE must c-command disjunction (p. 78). In a study by Crain, Gardner, Gualmini & Rabbin (2002), employing the Truth Value Judgment task, the subjects assigned a conjunctive interpretation to (47)-with negation preceding and c-commanding disjunction-in 92% of the time (pp. 79f.):
(47) The girl who stayed up late will NOT get a dime OR a jewel.
To rule out the hypothesis that domain-general cues, as linear precedence or distance between negation and disjunction, could play a role in children's interpretations, and to verify the role of c-command, Gualmini designed Experiment II and III, employing the Prediction Mode of the Truth Value Judgment task.
In Experiment II, 30 children (3;08-6;05) assigned in 85% of the time a conjunctive interpretation to sentences as (58), in which _or_ was preceded and c-commanded by negation and both operators were a distance apart from each other, without being troubled by the distance between both operators (pp. 82f.):
(58) I said that Winnie the Pooh would NOT let Eeyore eat the cookie OR the cake.
In Experiment III, 35 children (3;05-6;05) did not assign, in 80% of the time, the conjunctive interpretation to sentences as (68), in which negation preceded, but not c-commanded, disjunction, and with a short distance between both operators (pp. 84f.):
(68) The Karate Man will give the Pooh Bear he could NOT lift the honey OR the donut.
The over-all conclusion is that domain-general cues do not play any role in children's interpretations of DE, whereas c-command is a sufficient criterion for both children's and adult's interpretations (pp. 78f., 83, 86). Gualmini's findings ultimately corroborate the Continuity Hypothesis and refute ''the shallow linguistic representations'' of the Conservative Learning Model and the Rich Input Model (p. 88).
Chapter 4 - Asymmetries of Child Language (pp. 89-116) This chapter is dedicated to the investigation of the universal quantifier _every_ in child language. Gualmini reviews studies attempting to explain children's systematic non-adult interpretation of sentences containing _every_ in certain cases (pp. 89f.), before presenting Experiment IV on sentences containing disjunction and _every_ (pp. 105f.).
Inhelder & Piaget (1964) firstly observed a non-adult ''symmetrical response'' in children presented with sentences containing _every_, as:
(1) Every boy is riding an elephant,
in a context with the ''extra-object condition''-in this case, three boys are riding an elephant and a fourth elephant is not being ridden (pp. 89f.). The children rejected (1) as a description of the context.
Two kinds of explanations for this phenomenon were offered. The ''Partial Competence View'' explains it in terms of non-target ''Event Quantification'', ''Weak Quantification'', or ''Weak Mapping'' of _every_ (pp. 90f.). Gualmini rejects this explanation since it presupposes qualitative differences between children and adults, in which ''child language violates important linguistic universals uncovered by research in formal semantics'' (pp. 95f.). Gualmini endorses the Full Competence View, according to which children's non-adult responses are attributed to the methodology used: Thus, Crain and associates (1996) found out an improvement in children's performance in the Truth Value Judgment task (pp. 93f.), when felicity conditions are better achieved.
According to Gualmini, _every_ has two arguments: The first argument (''restrictor'') is the NP occurring with _every_; the second argument (''nuclear scope'') is the VP (pp. 96f.). Moreover, _every_ is asymmetrical, in that it is downward entailing in its first argument, but upward entailing on its second argument (p. 101f.). Previous studies, as Gualmini, Meroni, & Crain (2003) and Boster & Crain (1993), dealing respectively with sentences containing disjunction in the restrictor and in the nuclear scope of _every_, show that children are aware of this asymmetry.
In his Experiment IV, using the Truth Value Judgment task in its Description Mode and investigating children's interpretations of sentences containing disjunction in the second argument of _every_, Gualmini found out that his subjects, 23 children (3;10-5;09), have an adult-like knowledge of the asymmetry between the two arguments of _every_, as well as of the meaning of _or_ (p. 108), since they do not interpret the following sentences as being equivalent:
(22) [...] Every kid took a tiger or a dinosaur. (23) Every kind took a tiger AND every kid took a dinosaur. (p. 107)
According to Gualmini, ''[t]his finding is not surprising if one adopts the Full Competence View [...], but it is unanticipated on the [Partial Competence] view'' (p. 109). Such findings provide arguments against the view of domain-general learning mechanisms (p. 115) and for the view that ''[I]n absence of guidance from Universal Grammar, it is difficult to see how children could successfully master this asymmetry'' (p. 109). Thus, children do not assume that determiners ''could be downward entailing on its second argument, but upward entailing on its first argument'' (p. 113), since this is a possibility that is not exploited by ''natural language determiners'' (p. 114). The findings also provide arguments for the Continuity Hypothesis, since differences between children's and adults' interpretations never violate ''core principles of Universal Grammar'', when experiments achieve good felicity conditions (pp. 114f.)
Chapter 5 - Structure and Beyond (pp. 117-144) The starting point of this chapter is, again, a non adult-like linguistic behavior observed in children, namely their non-target interpretation of _some_ (within the scope of negation), in opposition to their adult-like interpretation of _any_. Gualmini reviews previous research on this phenomenon (pp. 118f.), relating it with a discussion on the question of felicity conditions of experiments (pp. 126ff.), and presents his Experiment V on children's interpretation of sentences containing _some_ within the scope of negation (pp. 132ff.).
Some researchers observed that, to target sentences related to stories the children were presented with, as (1), children consistently did not assign the adult-like interpretation (2), but the non-adult like interpretation (3) (pp. 117f.):
(1) The detective didn't find some guys. (2) There are some guys that the detective didn't find. (3) It is not the case that the detective found some guys.
To account for this phenomenon, Musolino (1998) and Musolino, Crain, & Thornton (2000) proposed the ''Observation of Isomorphism'' (''isomorphic'' interpretation=non-adult like; ''non-isomorphic'' interpretation=adult- like), according to which children's interpretations are determined by syntactic structure if there are mismatches between syntactic and semantic scope (p. 119), which suggests ''that children's grammar might be constrained by linguistic structure to a greater extent than adult's grammar'' (p. 121). Gualmini rejects this idea of distinct competences in child and adult, and argues that the observed differences between children's and adults' responses are rather the logical consequence of the use of infelicitous experimental material (p. 123). The question of felicity of experimental material is crucial in the case of negative contexts, since negation is usually employed to indicate divergences between expectations and facts (pp. 129ff.). Gualmini argues that, in some of the reviewed studies, the experiments were not designed in a way to evoke in the children explicit expectations that were not fulfilled in the end of the stories presented (pp. 130f.).
In his experiment V employing the Truth Value Judgment task, Gualmini designed a context that falsified the non-adult (isomorphic) interpretation and verified the adult (non-isomorphic) interpretation, and manipulated children's expectations about the end of a story (p. 132) about a firefighter looking for dwarves, evoking the expectation that the firefighter would find all the dwarves, but not the expectation that he would miss all of them (pp. 134). Group I, 15 children (4;01-5;06), were then presented with the target sentence:
(22) [...] The firefighter DIDN'T find SOME dwarves,
whereas Group II, 15 children (4;02-5;8) were presented with the target sentence:
(23) [...] The firefighter DIDN'T miss some dwarves.
(22) is felicitous since it indicates a discrepancy between expectations and facts, whereas (23) is infelicitous because it does not indicate a similar discrepancy. This pattern of felicity conditions corresponds to the findings: The children accepted (22) in 90% of the time, whereas they accepted (23) only in 50% of the time. Though the level of acceptance of (23) is not low (50%), the children who accepted (23) could not give explanation for their answers, whereas the children who accepted (22) consistently gave the right explanation for their answers (p. 134).
Gualmini conclusions, which are closely related to each other, are:
(i) children's non-adult interpretations of _some_ in negative sentences do not require a grammatical explanation, as proposed by Musolino and associates; the assumption about an imperfect competence with _some_ is not necessary (pp. 136f.), since there are no relevant differences between children's and adults' competence and performance (pp. 143f.).
(ii) experimental designs must consider felicity conditions (pp. 137f.), since, in the real life, ''there really is no null context'' (p. 144); the findings show, moreover, the relevant role children's pragmatic competence plays in their interpretations (pp. 143f.).
Chapter 6 - The Structure of Universal Asymmetries and Beyond (pp. 145-163) In this chapter, Gualmini reviews previous research on the relations between DE and PPI (Positive Polarity Items) (pp. 145f.), before he presents his Experiments VI and VII, focusing on the first and the second argument of the universal quantifier and PPI _every_, respectively (pp. 153ff.).
Szabolcsi 2002a has recently challenged the view that PPIs, as _some_, _someone_, _every_, in downward entailing environments always lead to ungrammatical sentences, pointing out ''the rich typology of polarity items witnessed across natural languages'' (p. 147) and arguing for the investigation of PPIs and NPIs as one single item class. Instead of speaking of resistance to and licensing by OP-DEs, Szabolcsi argues that the notion of licensing is sufficient and that polarity items are better classified as ''simple polarity items'', as _any_ or _some_, and ''complex polarity items'', as [_not...someone_] (p. 149). The complex item [_not...someone_] can occur in the scope of an antiadditive operator (e.g. _not_) or other OP-DE, whereas a simple item, as _someone_, cannot (pp. 147, 149), unless both the additive operator and the simple item ''occur within the scope of a downward entailing operator'', as _at most_ (p. 150). Thus, a sentence like (2) is ungrammatical if _someone_ is interpreted in the scope of negation (p. 146), whereas (9) is grammatical on the same interpretation:
(2) John didn't call someone is not equivalent to: John didn't call anyone (p. 146)
(9) At most five boys think that John didn't call someone is equivalent to: At most five boys think that John didn't call anyone. (p. 149)
According to Gualmini, to acquire this intricate linguistic pattern, the child cannot rely on input evidence only, since this pattern is related to ''a generalization at considerable distance from the input'' (p. 150).
Next, Gualmini observes that the asymmetry of the quantifier _every_ is not changed in sentences containing a complex polarity item as [_not...somewhat_]. The first argument licenses the occurrence of the complex item (12), whereas the second does not (13) (pp. 150f.):
(12) Every boy who doesN'T like pizza SOMEWHAT ordered lasagna.
(13) *Every boy doesN'T like pizza SOMEWHAT.
Gualmini explains the difference between these two sentences in terms of ''availability of a particular interpretation'' (12) and ''grammaticality of the entire sentence'' (13).
Gualmini then presents his Experiments VI and VII dealing with children's interpretation of sentences containing the complex item [_not...some N_] in the first and second argument of _every_, respectively, and employing the Truth Value Judgment task. In Experiment VI, 15 children (3;10-5;08) were presented with stories and target sentences containing [_not...some N_] in the first argument of _every_, i.e., in a downward entailing environment, as (29) (pp. 153f.):
(29) This was a story about five farmers and one Indian. Every farmer had to clean three animals and I know what happened. EVERY FARMER WHO DIDN'T CLEAN SOME ANIMAL HAS A BROOM.
From the two possible interpretations of such target sentences containing [_not...some N_] in downward entailing contexts, isomorphic (30) and non- isomorphic (31), the children consistently (in 88% of the time) favored the former. Interestingly, this was also the preferred interpretation (92,5% of the time) of a group of adult controls.
(30) EVERY farmer who didN'T clean ANY animal has a broom.
(31) EVERY farmer for whom there is SOME animal that he didN'T clean has a broom.
In Experiment VII, 15 children (4;01-5;05) were presented with stories and target sentences containing [_not...some N_] in the second argument of _every_, i.e., in an upward entailing environment, as (38) (pp. 157f.):
(38) This was a story about three farmers and one Indian. Every farmer had to clean three animals and I know what happened. EVERY FARMER DIDN'T CLEAN SOME ANIMAL.
From the two possible interpretations of such target sentences containing [_not...some N_] in upward entailing contexts, isomorphic (33) and non- isomorphic (35), the children consistently (in 78% of the time) favored the latter. Again, this was also the preferred interpretation (82,4% of the time) of a group of adult controls.
(33) EVERY farmer didN'T clean ANY animal.
(35) For EVERY farmer, there is an animal that he did NOT clean.
Gualmini's explanation for these findings is that children's interpretation of the scope ambiguity between _not_ and _some_ is constrained by abstract properties of the context, i.e., _some_ receives a narrow scope in downward entailing contexts, and a wide scope in upward entailing contexts (p. 161).
Gualmini's over-all conclusion to the whole dissertation is that ''the Conservative Learning model and the Rich Input model could not even begin to account for children's behavior. By contrast, we have encountered no evidence against the Continuity Assumption as yet, even when occasional mismatches between children's and adults' behaviors have surfaced. The next step is to accept Continuity Assumption as the null assumption for child language research. Even in the domain of semantic competence, there is no reason to assume that child language differs from adult language in ways that would exceed the boundary conditions imposed by Universal Grammar'' (p. 163).
Appendix (pp. 165-179) The appendix contains the experimental materials of all trials of all experiments (I-VII) by Gualmini, i.e., here the reader can find all the stories and target sentences designed for the Truth Value Judgment tasks (except for the trials presented in the respective chapters). Additionally, Gualmini also presents here the individual subject results for every trial.
Firstly, I would like to make some editorial remarks on the dissertation's edition and on the two review copies I received from the publisher. The edition is very accurate in that, for instance, it seems to contain no misprints whatever. Another positive aspect is the use of footnotes instead of unpractical notes at the end of the volume. In the index of topics (pp. 197-198), however, there are several relevant terms missing, as ''Conservative Learning Model'' and ''Rich Input Model'' (''Continuity Hypothesis is, on the other hand, not missing...), ''downward entailment operators'', ''noun substitution'', ''Full Competence View'' and ''Partial Competence View'', ''Prediction Mode'' and ''Description Mode'' (of the Truth Value Judgment task), ''determiners'' and ''quantifiers'', ''restrictor'' and ''nuclear scope'', ''domain-specific learning mechanisms'' and ''domain- general learning mechanisms'', amongst others. As for the review copies, in the first one I got, pages 189-198 were missing, i.e., a part of the references and both indexes. The second one was bound the wrong way round, so that one had to read it backwards.
Secondly, I would like to make some observations on the structure and the style of Gualmini's dissertation. Both structure and style are to my view exceptionally good. Gualmini's language and concerns seem to me fairly clear, even though it took me some time to get into it. The dissertation is, all in all, less hermetic than other generative texts commonly are. That is due in part to Gualmini's frequent recapping 'pauses' that are very helpful. I have here only two criticisms. Some terms are introduced without explanations. Thus, as far as I can judge, Gualmini does not explain anywhere in his book what polarity items really are and why they are designated as such. More importantly, there is no true over-all conclusion to the dissertation as a whole, apart from the last paragraph of the conclusion to Chapter 6 quoted above.
Finally, I would like to make some comments on more substantial aspects of Gualmini's dissertation.
First of all, one of Gualmini's greatest merits is his effort for methodological improvements in experimental settings. In my view, his preference for Truth Value Judgment tasks over Elicited Production tasks is not only understandable, but also essential for an experimental study on child language acquisition, since, as he points out, the latter focuses on performance (production), whereas the former focuses on competence (comprehension, interpretation, processing), and this is valid not only for generative studies. I am personally more concerned with naturalistic data in child language acquisition research, since the controlled situation of experiments may produce data that are not wholly reliable because of their more or less artificial nature. However, it is not difficult to recognize that, for some research questions, naturalistic data may turn out to be too elusive. This is the case of Gualmini's concerns on children's comprehension of entailment relationships. Moreover, Gualmini designed his experiments, at least partially, in such a way that aspects of every day's language use, i.e., _pragmatic_ aspects, could be taken into account. See, for instance, his discussion on felicity conditions in experiments investigating children's interpretations of negative sentences and his statement that ''there really is no null context'' (p. 144). A methodological desideratum would be, in my view, that experimental and naturalistic data could be used complementarily.
A second merit of Gualmini's dissertation is his tireless endeavor to interpret his findings in terms of both _explanations_, corresponding to Chomsky's demand for explanatory adequacy, and contributions to theory formation. Thus, all over the book, he indefatigably relates his findings to Continuity Assumption and Universal Grammar. He presents his findings as arguments for the notion that there are no differences between children's and adults' grammars that violate principles of UG, and also for the claim that children cannot acquire intricate patterns of linguistic behavior on the basis of input data only.
There are, however, some crucial inconveniences in the dissertation that are strongly intertwined with each other.
The first inconvenience is related to the way Gualmini explains the processing of entailment relations by the human brain. As exposed above, the structural notion of c-command is, in Gualmini's view, the decisive force that triggers DE. But he is concerned with the _semantics_ of DE what explains the absence of structure trees so common in generative texts. However, his 'explanations' focus on formal semantics, relying on laws of propositional logic. The question here is: Who says that the human brain processes entailment relations in natural languages the way Gualmini argues it does? This question turns out to be very crucial, since it challenges the very nature and raison d'être of Gualmini's whole enterprise.
The second inconvenience, related to the first one, is the explanation of the acquisition of DE in terms of putative universal principles of logic and the rejection of cognitive-functional models of language acquisition. Where are the arguments against the perspective that children can acquire DE on the basis of the data they are exposed to in concert with domain- general mechanisms, such as skills of ''pattern-finding'' and ''intention- reading'' (Tomasello 2003)? The fact is that Gualmini rejects cognitive- functional models of language acquisition without presenting arguments that are really consistent. It would be interesting to know if there are studies on the acquisition of entailment relations done in a more cognitive-functional spirit. Gualmini mentions none. He only provides speculations, and this not very convincingly. He speculates what, in his view, would be predicted by the Conservative Learning Model and the Rich Input Model, e.g., in terms of ''syntax-blind'' mechanisms, as noun substitution. And he speculates what, in her view, an explanation for DE in terms of domain-general cues would look like, namely in terms of linearity and relative proximity/distance between operators within sentences (Experiment II and III). This latter speculation is, in my view, particularly awkward, since it presupposes that there are only two possible explanations for children's interpretations of DE: structure dependence in terms of c-command or structure-blind cues in terms of linearity and relative proximity of operators. Such a presupposition casts doubt on his conclusion that his findings provide evidence for (i) the correctness of the assumption that c-command and structure dependence are crucial in children's interpretation of DE; (ii) the irrelevance of domain- general cues in the comprehension of DE; and (iii) his claim that cognitive-functional approaches proposes ''shallow linguistic representations'' only.
The third inconvenience is also related to the experiments' design. As stated above, Gualmini's efforts to take pragmatic aspects in account in the design of his experiments is by all means praiseworthy. Nevertheless, some of the target sentences children were presented to (as well as some of the example sentences included for expository reasons) did not sound 'natural' to me, which was confirmed by two adult English native speakers (no linguists!) to whom I showed the trials with the stories and the target sentences. For instance, we had in Chapter 5 sentence (1) in the adult-like interpretation (2) and the non-adult like interpretation (3):
(1) The detective didn't find SOME guys. (2) There are some guys that the detective didn't find. (3) It is not the case that the detective found some guys.
My informants told me that, in order to reach interpretation (2), they never would utter sentence (1), unless they would insert _of_:
(1a) The detective didn't find SOME OF the guys.
My informants would complement sentence (2) of Chapter 6 as in (2a)
(2) John didn't call SOMEONE. (2a) John didn't call SOMEONE OF THE PEOPLE HE WANTED TO CALL.
They would also replace sentence (29) of Chapter 6 with (29a), and (38) with (38a):
(29) Every farmer who didn't clean SOME animal has a broom. (29a) Every farmer who didn't clean ALL animals has a broom.
(38) EVERY farmer didn't clean SOME animal. (38a) NOT EVERY farmer cleaned ALL animals.
And sentence (12) of Chapter 6 they classified as mere nonsense:
(12) Every boy who doesN'T like pizza SOMEWHAT ordered lasagna.
Of course, I had only 2 informants, whereas Gualmini had control groups of 8-36 adults in his experiments. These were often ''undergraduates'' (linguistics students?) and Gualmini does not inform us as to what they knew about the purposes of the study (in the case they were linguistics students), nor on their possible estranged reactions (in the case they were not linguistic students).
Gualmini's dissertation contains some very important virtues, above all related to both methodological aspects (consideration of pragmatic factors in the experimental designs, the preference for Truth Value Judgment tasks over Elicited Production tasks), and theoretical aspects (relations between findings/conclusions and theory formation), as well. The value of these merits is, nevertheless, diminished by the inconveniences exposed above that cast doubt on the general pretensions and the conclusions of the dissertation. Specifically, Gualmini's aim of investigating linguistic phenomena present in ''natural languages'', as he repeatedly states (e.g., pp. 3, 6, 12, 16, 23, 77, 110, 121, 150, 162, etc.), in terms of universal principles turns out to remain unattained because the assumption that DE is processed by the human brain in terms of propositional logic remains mere speculation; and also because, apart from a few allusions to studies dealing with 4 or 5 other languages, Gualmini's findings are essentially based on data from children learning English as first language, employing sentences that even adult English native speakers partially find 'ungrammatical' or 'unnatural'. The over-all conclusion that first language acquisition researchers should ''accept Continuity Assumption as the null assumption for child language research'' (p. 163), whereas ''the shallow linguistic representations'' of the Conservative Learning Model and the Rich Input Model (p. 88) ''could not even begin to account for children's behavior'' (p. 163) must be questioned for the same reasons.
All the crucial problems present in Gualmini's dissertation are not surprising, however, since they are some sort of unfailing characteristics of generative studies. At the end, the reader might even get the impression that, as common in generative studies, hypotheses were posited axiomatically and that the whole experimental enterprise were designed in such a way to provide evidences for them and not to check them up, to scrutinize them, as it should actually be indispensable in genuine scientific work.
Tomasello, M. (2003): Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA & London: Harvard University Press.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
The reviewer is currently working on her M.A. thesis on the acquisition of argument constructions in a bilingual child within a cognitive-functional framework. Her research interests include first language acquisition, multilingualism, cognitive science, developmental psychology, as well as history of linguistics.