Crosslinguistic differences in the interpretation of plural NPs

It is well known that English contrasts with (most) Romance languages on the interpretation of plural NPs. In English, generic readings are available to bare plurals, but not definite plurals, as illustrated in (1). Although (1a), with a bare plural (i.e., an articleless plural NP), is a statement about tigers in general, (1b), with a definite plural (i.e., a plural NP preceded by a definite article), cannot be a statement about tigers in general. The only reading available to (1b) is that some salient, specific tigers (e.g., the tigers in this zoo) eat meat, the same reading that is available to the demonstrative plural (i.e., a plural NP preceded by a demonstrative determiner) in (1c). We term the reading of the NP in (1a) the generic reading, and reading of the NP in (1b), the specific reading. The specific reading is not available to (1a), which cannot be a statement about a specific group of tigers: bare plurals in English cannot have specific reference.Footnote ¹

1. (1)

   

Unlike English, Spanish uses definite plurals for generic readings, as shown in (2). Preverbal bare plurals are ungrammatical in Spanish, as in (2a), whereas definite plurals have both specific and generic readings, as in (2b), in contrast to demonstrative plurals, which have only specific (nongeneric) readings available to them, as in (2c). The same is observed for other Romance languages, including Italian and French (Zamparelli, Reference Zamparelli, Beyssade, Bok-Bennema, Drijkoningen and Monachesi2002).

1. (2)

   

In the theoretical literature on cross-linguistic differences with generics, two major approaches can be identified: syntactic and semantic. On the syntactic approach (espoused by Longobardi Reference Longobardi1994, Reference Longobardi2001; Vergnaud & Zubizarreta, Reference Vergnaud and Zubizarreta1992, among others), the differences between English and Romance are placed in the syntax: Romance languages require overt determiners for syntactic reasons, whereas English allows null determiners. This syntactic requirement has consequences for other aspects of the language besides generic reference, such as article use in inalienable possession constructions (Vergnaud & Zubizarreta, Reference Vergnaud and Zubizarreta1992) and article use with proper names (Longobardi, Reference Longobardi1994, Reference Longobardi2001).

On an alternative semantic approach (Chierchia, Reference Chierchia1998; Dayal, Reference Dayal2004), definite articles in Romance languages encode a slightly different semantics than the English definite article. Definite articles in both English and Spanish lexicalize maximality: the tigers in (1b) denotes the maximal set of tigers in the discourse (e.g., the tigers just mentioned, the tigers visually present, etc.), and the same reading is available to los tigres in (2b). However, definite articles in Romance languages, in addition, encode kind-formation: los tigres in (2b) can denote either the maximal salient set of tigers (the definite/specific reading), or the entire category/kind of tigers (the generic/kind reading). This argument receives support from the fact that plural NPs behave the same with true kinds readings as with generic readings: with kind predicates such as be extinct or be common, English requires bare plurals ((#The) tigers are common in these woods), whereas Romance languages require definite plurals.Footnote ²

Interpretation of plural NPs in child language acquisition

Several recent studies have examined the interpretation of bare versus definite plurals in child language acquisition. Gelman and Raman (Reference Gelman and Raman2003) tested 2- to 4-year-old English-acquiring children on their ability to use determiners as a cue to NP interpretation. The children would be shown a picture of two tiny elephants, and asked either a specific question (“Are the elephants big or small?,” target answer “small”) or a generic question (“Are elephants big or small?,” target answer “big”). Gelman and Raman (Reference Gelman and Raman2003) found that the children appropriately differentiated between the specific and generic questions most of the time, interpreting bare plurals as generic and definite plurals as specific. However, the children did exhibit error rates in both directions: specific questions received generic responses between 5% and 10% of the time, whereas generic questions received specific responses between 16% and 27% of the time. As Gelman and Raman (Reference Gelman and Raman2003) state, the specific (nongeneric) reading was the more salient one in this task, because it was supported by a picture of the specific referents. The saliency of the specific reading in comparison to the generic reading may explain why the children were more accurate answering specific questions than answering generic questions.

In a study with older children (ages 4–7), Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) probed the interpretation of definite plurals in the L1 acquisition of both English and Spanish. In this study, the children would be told a story, accompanied by pictures, about two animals that were unusual for their kind—for example, tigers that preferred carrots to meat, or zebras that had spots instead of stripes. The children would then be asked a question such as “Do the zebras have spots?” On the specific interpretation, the target response is “yes”; on the generic interpretation (not available in English, but available in Spanish), the target response is “no.” The opposite question, “Do the zebras have stripes?” was also asked: here, the target response is “no” on the specific interpretation, and “yes” on the generic interpretation.

Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) found that both groups of children had a preference for the generic interpretation of definite plurals: the Spanish-acquiring children gave a response consistent with the generic interpretation about 85% of the time, whereas the English-acquiring children did so about 70% of the time, unlike a control group of native English-speaking adults, who allowed definite plurals to have generic readings less than 10% of the time (the English-acquiring children also correctly assigned the generic interpretation to bare plurals, answering questions like “Do zebras have spots?” correctly more than 90% of the time). Strikingly, the rates of generic interpretation for English definite plurals were much higher in the Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) study than in the Gelman and Raman (Reference Gelman and Raman2003) study, even though Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) tested older children. The task differences are most likely to blame. The task used by Gelman and Raman (Reference Gelman and Raman2003) was biased in favor of the specific reading of the plural, as noted above. In contrast, the task used by Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) was biased in favor of the generic reading of the plural (see the critique in Serratrice et al., Reference Serratrice, Sorace, Filiaci and Baldo2009). One potential methodological confound of Pérez-Leroux et al.'s (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) study was the fact that the animals in the pictures were introduced by name, as in “Zippy the zebra and Suzy the zebra are spotted.” As a result, it would be most natural to refer back to the zebras as “Zippy and Suzy.” Referring back to them as “the zebras” was infelicitous, and this could have led the children to misinterpret “the zebras” as referring to the entire kind of zebras, instead. Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) corrected this problem in a second experiment, in which first names were omitted (instead, the animals would be introduced simply as “two zebras”). In this experiment (with 3- to 5-year-old English-speaking children), the rate of generic responses went down to 40%, which, however, is still higher than the result obtained by Gelman and Raman (Reference Gelman and Raman2003).

Serratrice et al. (Reference Serratrice, Sorace, Filiaci and Baldo2009) examined the interpretation of both bare and definite plurals in English and Italian. They tested 6- to 10-year-old children, English–Italian and Spanish–Italian bilinguals as well as English-speaking and Italian-speaking monolinguals. Unlike the studies described above, which used meaning-focused tasks, Serratrice et al. (Reference Serratrice, Sorace, Filiaci and Baldo2009) used a form-focused, metalinguistic task, in which the children were asked to make judgments about the grammatical acceptability of the sentences (presented in the context of pictures). Sentences targeting the specific interpretation began with “Here,” as in “Here, the strawberries are red” (acceptable in both English and Italian) versus “Here, strawberries are red” (unacceptable in both English and Italian). Sentences targeting the generic interpretation began with “In general,” as in “In general, sharks are dangerous” (acceptable in English, but unacceptable in Italian) versus “In general, the sharks are dangerous” (acceptable in Italian, but unacceptable in English).

In the English version of the study, both English monolingual and English–Italian bilingual children were successful at accepting definite plurals with specific readings and bare plurals with generic readings. However, they also incorrectly overaccepted bare plurals with specific readings and definite plurals with generic readings, with the latter error being particularly common (more than 70% overacceptance for the younger children). Age rather than learning situation appeared to be the relevant factor: younger children (ages 6–7) made more errors than older children (ages 8–10), but monolingual and bilingual children within each age group performed quite similarly. All children were significantly worse than adults at rejecting ungrammatical sentences (although even adults were not at ceiling in their rejections). Serratrice et al. (Reference Serratrice, Sorace, Filiaci and Baldo2009) suggest that the children had difficulty integrating the semantics of the adjunct (here vs. in general) with the semantics of bare versus definite plurals.

In the Italian version of Serratrice et al.'s (Reference Serratrice, Sorace, Filiaci and Baldo2009) study, monolingual Italian adults, monolingual Italian children, and bilingual Spanish–Italian children were all at or near ceiling in accepting definite plurals with both generic and specific readings, and rejecting bare plurals with both: in the Italian version of the study, the test items with bare plurals were ungrammatical regardless of the adjunct, so no integration of adjunct and NP interpretation was necessary, and even younger children did quite well. Because Spanish and Italian pattern the same in this aspect of the grammar, the Spanish–Italian child bilinguals were as accurate as the Italian monolinguals. In contrast, the English–Italian bilinguals did quite poorly: although they correctly accepted the grammatical sentences with definite plurals, they had difficulty rejecting sentences with bare plurals. They had a particularly high rate of overacceptance of bare plurals with generic readings (which are grammatical in English), but they also overaccepted bare plurals with specific readings (which are ungrammatical in English as well as Italian). Thus, the English–Italian bilinguals appeared to be transferring the generic readings of bare plurals from English to Italian; according to Serratrice et al., (Reference Serratrice, Sorace, Filiaci and Baldo2009) the children's overacceptance of bare plurals with specific readings may have been due to difficulty integrating the semantics of the adjunct with that of the plural NP (as in the case of the English version of the study).

To sum up, the literature on child language acquisition of plural NP interpretation is somewhat inconclusive. Both Gelman and Raman (Reference Gelman and Raman2003) and Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) used meaning-focused tasks, yet obtained quite different results, most likely because of methodological differences. Serratrice et al.'s (Reference Serratrice, Sorace, Filiaci and Baldo2009) findings from a form-focused task indicate that children have difficulty with metalinguistic judgments that involve integrating the meaning of the NP form with that of the adjunct; this difficulty, which was attested for both bilingual and monolingual children in the English study, obscured potential effects of cross-linguistic influence from Italian to English. In the case of the Italian study, transfer effects were in evidence, with bilingual English–Italian children overaccepting bare plurals with a generic interpretation.

Interpretation of plural NPs in adult L2 acquisition

In the case of adult L2 acquisition, early studies of English articles typically grouped generic NPs separately from nongeneric ones (Huebner, Reference Huebner1983; Master, Reference Master1987; Thomas, Reference Thomas1989), but did not otherwise examine generic interpretation in any detail, and did not specifically examine the interpretation of plural NPs. More recently, Slabakova (Reference Slabakova2006) examined the interpretation of bare plurals by L1-Italian L2-English and L1-English L2-Italian learners, using a truth–value judgment task (TVJT). English allows bare plural subject NPs such as white elephants to have both generic readings (the entire kind of white elephants) and existential readings (some white elephants), whereas Italian allows bare plurals to have only existential readings (generic readings are expressed with definite plurals, as discussed above). Thus, L1-Italian L2-English learners need to acquire a new interpretation of bare plurals: they have to learn that English bare plurals can be generic as well as existential. In contrast, L1-English L2-Italian learners have to unlearn an existing interpretation: they have to learn that Italian bare plurals cannot be generic, only existential. Slabakova (Reference Slabakova2006) found that learning was easier than unlearning. The L1-Italian L2-English learners were fully targetlike in allowing the generic reading for bare plurals; like the native English controls, they exhibited a preference for the generic reading over the existential reading. In contrast, the L1-English L2-Italian learners were not very successful at rejecting the generic readings of bare plurals. At the same time, unlearning was possible, and some L1-English L2-Italian learners were able to unlearn the generic reading of bare plurals. Slabakova (Reference Slabakova2006) argued that this unlearning was triggered by input from a different part of the grammar: according to Longobardi's (Reference Longobardi2001) proper name/bare noun parameter, the availability of bare plurals with generic readings (a semantic property) is directly related to noun–adjective order (a syntactic property). Slabakova (Reference Slabakova2006) predicted that the noun–adjective order in Italian (e.g., Roma antica, “ancient Rome”) would serve as a trigger informing L1-English L2-Italian learners that Italian has a different setting of the proper name/bare noun parameter than English does; once the learners had acquired this syntactic property, they should acquire the corresponding semantic property, and restrict bare plural NPs in Italian to existential readings. Slabakova's (Reference Slabakova2006) findings were consistent with this prediction, in that most of the learners who were successful on the interpretation of Italian bare plurals were also successful with noun–adjective order. Thus, Slabakova (Reference Slabakova2006) found evidence for L1 transfer as well as recovery from L1 transfer, and proposed a mechanism for this recovery; however, Slabakova's (Reference Slabakova2006) study considered only bare plurals, and did not examine how the acquisition of bare plurals might relate to the acquisition of definite plurals.Footnote ³

Objectives of the present study

The main goal of the present paper is to examine the interpretation and judgments of both definite plurals and bare plurals in L2 acquisition bidirectionally, with Spanish-speaking learners of English and English-speaking learners of Spanish. Our research questions are given in (3).

1. (3) Research questions:

   1. a. Is L1 transfer attested in learners’ interpretation and judgments of plural NPs in English and Spanish?

   2. b. Is targetlike attainment in the interpretation and judgments of plural NPs in L2 acquisition possible, in both English and Spanish?

   3. c. What presents the greatest learning challenge: learning about the (un)grammaticality of bare plurals in English versus Spanish, or learning the target interpretation of definite plurals?

Descriptively, the Spanish/English difference can be characterized as follows. First, there is a difference of grammaticality: in English, bare plural NPs are grammatical in preverbal subject position, whereas in Spanish, they are typically ungrammatical.Footnote ⁴ In contrast, definite plurals in preverbal subject position are grammatical in both languages. Second, there is a difference of interpretation: definite plurals have both generic and specific readings in Spanish, but only specific readings in English, where generic readings are expressed via bare plurals.

The extent to which these two differences—the (un)grammaticality of bare plurals and the interpretation of definite plurals—are viewed as related depends on the theoretical approach. In Chierchia's (1998) approach, the fact that bare plurals are ungrammatical in Spanish is directly tied to the fact that definite plurals have a generic reading: the definite article must be used precisely because bare NPs cannot appear in argument position. Furthermore, in Chierchia's view, the fact that English freely allows bare plurals in subject position is directly related to the fact that the generic reading is unavailable to definite plurals. According to Chierchia's avoid structure principle, an interpretation that can be encoded by a bare NP should be unavailable to a definite NP, because additional structure (in the form of a determiner phrase projection) should be avoided.

Other accounts (including Dayal, Reference Dayal2004; Longobardi, Reference Longobardi1994, Reference Longobardi2001) do not posit a direct link between the ungrammaticality of bare plurals and the generic readings of definite plurals. For example, Dayal's (Reference Dayal2004) approach allows for the coexistence of bare and definite plurals with generic readings within a single language; this account receives support from Brazilian Portuguese, in which both bare and definite plurals have generic readings (Schmitt & Munn, Reference Schmitt, Munn, Tamanji, Hirotani and Hall1999, Reference Schmitt and Munn2002; empirical support for this proposal is found in Ionin, Montrul, & Santos, Reference Ionin, Montrul, Kim and Philippov2011).

In light of these differences in the theoretical accounts, we do not assume a relationship between bare and definite plurals in learners’ Interlanguage grammars, but instead focus on the two learning tasks: learning whether preverbal bare plurals are (un)grammatical in the target language, and learning the interpretation of definite plurals in the target language. If L1 transfer is attested in this domain, then L1-English L2-Spanish learners should overaccept bare plurals as grammatical in Spanish, and in particular grammatical with generic interpretation (which is available to bare plurals in English); in addition, these learners should disallow definite plurals with generic interpretation, because this interpretation is not available to definite plurals in English. In the other direction, L1-Spanish L2-English learners should reject English bare plurals in preverbal position as ungrammatical, and should furthermore allow definite plurals to have generic as well as specific readings.

Consider next how the learners might become targetlike, in terms of both (un)grammaticality of bare plurals and interpretation of definite plurals. With regard to learning about (un)grammaticality of bare plurals, both classroom learners of English and classroom learners of Spanish receive much explicit instruction in the matter. Spanish textbooks such as Iguina and Dozier (2008) and Alonso Raya et al. (Reference Alonso Raya, Castañeda, Martínez Gila, Miquel López, Ortega Olivares and Ruiz Campillo2008) (used in the Spanish program at the US university where our L2-Spanish learners were tested) emphasize the ungrammaticality of bare NPs in preverbal position, and the requirement for a definite article (although a textbook used in more advanced courses, King & Suñer (Reference King and Suñer1998), shows that the rule is not quite as simple as that, see also Note 4). With regard to learners of English in Argentina, where our L2-English learners were tested, more grammar-oriented textbooks, such as Cunningham and Moor (Reference Cunningham and Moor2007), discuss the “zero article,” and provide examples of grammatical English sentences with bare (article-less) plural NPs in subject position. Although the books focus primarily on (un)grammaticality, they also address interpretation, albeit to a limited degree. Cunningham and Moor note that the zero article is used to talk about “things in general,” whereas the Spanish textbooks discussed above give examples of plural NPs with definite articles used with generic as well as specific meanings.

Turning to theoretical considerations, there is reason to expect learning about (un)grammaticality to be easier than learning about interpretation. A number of recent proposals have argued that narrow syntactic phenomena (which involve (un)grammaticality) are easier for L2 learners to acquire, and less vulnerable to transfer, than phenomena at the interface between syntax and semantics/pragmatics, which involve interpretation (Slabakova, Reference Slabakova2008; Tsimpli & Sorace, Reference Tsimpli, Sorace, Caunt-Nulton, Kulatilake and Woo2006; White, Reference White, Ritchie and Bhatia2009). For example, Montrul (Reference Montrul2010), looking at the acquisition of clitics and Differential Object Marking in L2-Spanish learners and Heritage Spanish speakers (with English the L1/dominant language in both cases) found that a purely syntactic phenomenon (clitic placement) was acquired more successfully, and was less subject to transfer, than a phenomenon at the syntax/semantics interface (the semantic conditions under which the differential object marker a must be inserted before direct objects). Similarly, the Serratrice et al. (Reference Serratrice, Sorace, Filiaci and Baldo2009) study described above found that children, monolingual as well as bilingual, had an easier time learning about the (un)grammaticality of bare plurals in Italian, than about the target interpretation of definite versus bare plurals in English.

In light of these considerations, we hypothesize that L2 learners will also have more difficulty with interpretation than with (un)grammaticality when it comes to the interpretation of plural NPs. For L1-English L2-Spanish learners, we predict that they will be fairly successful in learning that preverbal bare plurals are ungrammatical in Spanish, and that definite plurals must be used instead of bare plurals, to express generic as well as specific readings. At the same time, L1 transfer may cause these learners to prefer the specific readings of definite plurals when given a choice between specific and generic readings. In the case of L1-Spanish L2-English learners, we predict that these learners will be successful at learning that English, unlike Spanish, allows bare plurals; however, we expect learners to have difficulty acquiring the target interpretation of bare versus definite plurals. In particular, L1 transfer should lead these learners to allow the generic interpretation of English definite plurals.

Study 1, reported below, used a meaning-focused task to investigate which reading—generic or specific—L2 learners of English and Spanish prefer for definite plurals. Study 2 used a form-focused, metalinguistic task to examine how L2 learners of English and Spanish judge sentences with definite versus bare plurals in generic versus specific contexts.

TVJT: Test categories and predictions

We probed learners’ interpretation of plural NPs by using a written TVJT, whose format was loosely modeled on the oral task used by Pérez-Leroux et al. (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) with child L1 learners. Each TVJT item consisted of a story accompanied by a picture and followed by a test sentence. The participants had to judge the test sentence as true or false in the context of the story. As in Pérez-Leroux et al.'s (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2004) task, each test story juxtaposed a specific reading with a generic reading, by describing animals that bear a characteristic unexpected for their species. A sample story from the English version of the task is given in (4), and the corresponding story from the Spanish version is given in (5). There was a total of eight target stories, and each appeared three times in the test, each time with a different target sentence.

In the English version, each story appeared once with a target sentence containing a bare plural (4a); once with a target sentence containing a definite plural (4b); and once with a target sentence containing a demonstrative plural (4c). The target interpretation is generic for bare plurals, and specific for definite plurals and demonstrative plurals. The demonstrative plural was a control item, because the specific reading is the only possible interpretation for it in Spanish as well as English. Target truth values were counterbalanced across items. For the item in (4), the sentences with bare and definite plurals (4a–b) are designed to be true, whereas the sentence with a demonstrative plural (4c) is designed to be false. This was the case for half of the target items, whereas for the other half, the sentences with bare and definite plurals were false, and the sentences with demonstrative plurals were true.

In the Spanish version, the setup was somewhat different. Because a bare plural in preverbal position is ungrammatical in Spanish, it was not possible to test target sentences analogous to (4a): participants cannot be asked to judge an ungrammatical sentence as true or false. Instead, each target item appeared once with a test sentence containing a definite plural, and designed to be false on the specific reading and true on the generic reading (5a); once with a test sentence containing a definite plural, and designed to be true on the specific reading and false on the generic reading (5b); and once with a test sentence containing a demonstrative plural (5c). Half of the demonstrative plural sentences were designed to be true on the specific reading, as in (5c), and half were designed to be false.

1. (4) Sample test story: English study

   Everyone knows that a zebra always has stripes. But not in our zoo! Our zoo has two zebras, and they are really unusual: they have spots instead of stripes! That's really strange.

   • 

2. (5) Sample test story: Spanish study

   El zoológico de Buenos Aires tiene dos cebras nuevas. Estas cebras no son comunes: tienen manchas en vez de rayas. ¡Son muy extrañas!

   1. a. Las cebras tienen rayas. [The zebras have stripes.] TRUE generic, FALSE specific

   2. b. Las cebras tienen manchas. [The zebras have spots.] FALSE generic, TRUE specific

   3. c. Estas cebras tienen rayas. [These zebras have stripes.] FALSE

Thus, there was a total of 24 test items: in the English test version, 8 contained bare plurals, 8 contained definite plurals, and 8 contained demonstrative plurals; in the Spanish test version, 16 items contained definite plurals, and 8 contained demonstrative plurals. There were also 36 filler items in each test version, for 60 items total. The fillers consisted of stories juxtaposing either two different aspectual interpretations or two different temporal interpretations. All filler items had proper personal names or determinter phrases with possessive pronouns in subject position, to avoid a focus on the presence versus absence of articles.

The TVJT is designed to focus learners’ attention on the meaning of the sentence, rather than the (un)grammaticality of the plural NP. At the same time, the learners’ responses of TRUE versus FALSE provide information about whether the learners are interpreting the plural NP specifically or generically, because the stories are set up to juxtapose the specific and generic reading: if one is true, the other is false, and vice versa. In coding the data, we therefore scored each response as either generic or specific: for example, for (4a) and (5a), a response of TRUE would be scored as generic (because it is true that zebras as a kind have stripes) but a response of FALSE would be scored as specific (because it is false that the specific zebras in the story have stripes). For (4b) and (5b), a response of FALSE would be scored as generic (because it is false that zebras as a kind have spots), whereas a response of TRUE would be scored as specific (because it is true that the zebras in the story have spots). Recall that target truth values were counterbalanced across NP types. The dependent measure was the proportion of generic responses, out of all responses (generic + specific), for each category.

For the English version of the TVJT, we can make the following predictions. For native English speakers, we expect all sentences with bare plurals to receive generic responses, and all sentences with definite plurals and with demonstrative plurals to receive specific responses, regardless of target truth value. In contrast, L1-Spanish L2-English learners who transfer the interpretation of Spanish definite plurals onto English may give generic as well as specific responses to English definite plurals. Because Spanish definite plurals can have both generic and specific readings, and because all our test items made the target sentence true on either the generic or the specific reading, the learners may always judge sentences with definite plurals as TRUE, resulting in 50% generic responses (given to items where the generic reading is true) and 50% specific responses (given to items where the specific reading is true).

The design of the Spanish TVJT was somewhat different from that of the English TVJT: recall that there were no items with bare plurals in the Spanish TVJT, and therefore twice as many items with definite plurals. Native Spanish speakers are expected to provide only specific responses for demonstrative plurals, but to provide both generic and specific responses for definite plurals. A possible response pattern is that the native speakers would always judge sentences with definite plurals as TRUE, because they are always true on either the generic reading or the specific readings, and both readings are available to definite plurals. In that case, we would expect a 50% rate of generic responses to definite plurals for native Spanish speakers. In contrast, L1-English L2-Spanish learners who transfer the interpretation of definite plurals from English are expected to give specific responses only, regardless of target truth value.

Results of the Spanish TVJT

Figure 3 reports the means and standard deviations for the proportion of generic responses for both categories across proficiency groups in the Spanish TVJT. All groups correctly interpreted demonstrative plurals as specific, but differed on definite plurals. The native Spanish group overwhelmingly preferred the generic reading of definite plurals, but the low group preferred the specific reading. The intermediate and high groups were in between, allowing both generic and specific readings for definite plurals. As shown by the breakdown in Figure 4, the target truth value did not make a difference: the native Spanish speakers opted for the generic reading even if it made the sentence false; rather than responding TRUE to all sentences with definite plurals, the native speakers gave mostly TRUE responses if the generic reading was true (and the specific reading false), and mostly FALSE responses if the generic reading was false (and the specific reading true). This is highly reminiscent of the performance pattern of lower-proficiency L1-Spanish L2-English learners in the English TVJT (see Figure 2): there is a strong preference on the part of Spanish speakers for the generic reading of definite plurals, both in Spanish and (for lower proficiency L2-English learners) in English. The low L2-Spanish group showed the opposite pattern: their strong preference was for the specific reading of definite plurals, regardless of whether it made the sentence true or false. Again, this is consistent with performance in English (compare the native English group in Figure 2, and the Low L2-Spanish group in Figure 4). The intermediate and high L2-Spanish groups pattern in between, opting for the specific reading of definite plurals about half the time, and for the generic reading about half the time. As shown by Figure 4, this split is not a result of these learners always saying TRUE: the learners gave the generic response about 50% of the time when the generic reading was true, but also when the generic reading was false. An individual subjects analysis suggests that the 50% performance of the intermediate and high groups is in fact the result of a largely bimodal subject distribution. Of the 28 learners in these two groups, 14 consistently opted for the generic reading (choosing it at least 75% of the time), nine consistently opted for the specific reading (choosing it at least 75% of the time), and only five went back and forth between the two readings. (In contrast, the majority of the participants in the low group—10 out of 13—consistently opted for the specific reading, and the majority of native speakers—14 out of 17—consistently opted for the generic reading.)

Figure 3.

Spanish Study 1, truth-value judgment task (TVJT) results: mean proportion “generic” responses, by category. Error bars ±1 SD.

Figure 4.

Spanish Study 1, truth-value judgment task (TVJT) results: mean proportion “generic” responses on the definite plural category, by target truth value. Error bars ±1 SD.

We conducted a mixed ANOVA on the group results, with NP type (definite vs. demonstrative) and target truth value (TRUE vs. FALSE) as the within-subjects variables, and group as the between-subjects variable. There was a significant effect of determiner type, F (1, 56) = 110, p < .001, but no effect of target truth value, F (1, 56) = 0.20, p = .66, which indicates that there was no truth bias in the participants’ responses. There was a significant effect of group, F (3, 56) = 5.2, p < .01, which was due to the native group being significantly different from the low group (there were otherwise no significant group differences). Group interacted significantly with NP type, F (3, 56) = 9.6, p < .001, and there was also a significant interaction between NP type and target truth value, F (1, 56) = 4.3, p < .05, as well as a three-way significant interaction between NP type, target truth value, and group, F (3, 56) = 6.2, p < .01. The interaction between NP type, target truth value, and group had the following source: the low and intermediate groups were more likely to treat definite plurals as generic if the generic reading was true, and more likely to treat definite plurals as specific if the generic reading was false; that is, they had a slight tendency to opt for the reading that made the definite plural sentence true. The high and native groups did not exhibit any truth bias, and none of the groups exhibited a truth bias for demonstrative plurals, which were interpreted as specific regardless of the target truth value.

The interaction between group and NP type was due to the fact that all groups performed similarly on demonstrative plurals, but differed on definite plurals, as revealed by a follow-up one-way ANOVA on definite plurals, F (3, 56) = 8.1, p < .001. Post hoc tests (Tukey) revealed significant differences between the low group and each of the intermediate and native groups on definite plurals, and no other significant differences between any pair of groups.

AJT: Task categories and predictions

The AJT consisted of 40 items, and each item in turn consisted of a paragraph-long story followed by five different test sentences. The participants were instructed to rate the acceptability of each test sentence in the context of the preceding story, using a scale from 1 (unacceptable) to 4 (acceptable). The instructions specifically stated that two or more sentences could receive the same rating, that is, responses did not need to be ranked; the availability of four scale points and five target sentences necessitated that at least two sentences receive the same rating. It took participants about 45 min to complete this task.

The use of a 4-point rating scale, instead of a binary yes/no scale, allowed us to probe participants’ responses to fairly subtle distinctions. Although it would have been ideal to present each story with only one target sentence underneath it, repeating each story five times with different sentences, we opted for the presentation of multiple test sentences after the same story instead. We made this decision because presenting each story five times would have made the test too long and taxing on participants’ attention, and would have caused participants to become bored with reading the same stories multiple times.

Of the 40 test items, one-half (20 items) contrasted the interpretation of nominal expressions (the other 20 items were fillers, and contrasted types of aspectual interpretation rather than nominals). The 20 test items that focused on nominals were further broken down into five categories, four items per category. Here, we focus on two of the categories, which are directly related to the categories tested in Study 1. The specific-plural category tested the acceptability of different NP types when the context set up a specific reading of a plural NP, and the generic category tested the acceptability of different NP types when the context set up a generic reading. The other three test categories were specific-singular (analogous to specific-plural, but for specific readings of singular NPs), mass-noun, and kind-reference. The kind-reference category examined judgments of different NPs types in combination with a kind predicate (such as be extinct). As mentioned earlier, kind readings, like generic readings, are expressed by bare plurals in English but by definite plurals in Spanish; for reasons of space, we do not discuss the kind-reference category here, but we note that the results were very similar across the generic and kind-reference categories, for both native speakers and L2 learners.

For each test category, each item was a story followed by five different target sentences, which contrasted in the form of the NP subject: bare plural (chairs), definite plural (the chairs), indefinite singular (a chair), definite singular (the chair), and bare singular (chair). We discuss here only the first two NP types, bare plurals and definite plurals, because our focus is on the interpretation of plural NPs. Singular NPs have the same characteristics in English and Spanish, in both generic and specific contexts, and in fact, both the L2-English and the L2-Spanish learners in the present study were largely accurate with singular NPs. See Ionin et al. (Reference Ionin, Montrul and Santos2011) for the results of the native English and native Spanish controls on all five NP types, and Ionin, Montrul, Kim and Philippov (Reference Pérez-Leroux, Munn, Schmitt and DeIrish2011) on the L2-acquisition of these five NP types by L2-English learners whose L1s (Russian and Korean) lack articles.

We now discuss each of our target categories in turn. Below, we present a sample item for each category, followed by the two relevant continuations, bare plurals and definite plurals. The target sentence variant (the one expected to be rated high by native speakers) is highlighted here; it was not highlighted in the actual test instrument. In the specific-plural category, exemplified in (6) and (7), each item consisted of a story that mentioned two NPs, one singular and the other plural.Footnote ⁶ Each continuation sentence included second mention of the target NP. As illustrated by (6), the specific-plural category always discussed unusual/noncharacteristic properties of the relevant individual(s), such as birds that bark, or chairs that are bright orange in color. This was done on purpose, to ensure that the variant with the bare plural, in English, would not be interpreted generically, for example, (6a) is not acceptable on the generic reading, because chairs are not, as a rule, orange.

1. (6) Specific-plural category (English). My friend Gilbert bought some furniture for his kitchen: two chairs and one table. Gilbert likes colorful furniture: for instance . . .

   • 

2. (7) Specific-plural category (Spanish). Mi amigo Gilberto compró muebles para la cocina: dos sillas y una mesa. A Gilberto le gustan los muebles de colores brillantes. Por ejemplo . . .

   • 

The generic category in (8)–(9) tests the availability to generic readings to definite versus bare plurals. The stories in the generic category are set up to ensure that no specific entities are under discussion: for example, no specific toy animals are mentioned in (8)–(9). The target sentences in (8)–(9) are generic sentences that discuss the typical properties of the referent of the subject NP; for example, toy animals are typically/usually/as a rule good children's gifts.

1. (8) Generic category (English). It's my niece's birthday this Saturday—she is going to be 3 years old. I'm not sure what to get her. Maybe I'll just get her some toy, like a stuffed dog or bear. I can't go wrong with that. We all know that . . .

   • 

2. (9) Generic category (Spanish). Este sábado es el cumpleaños de mi sobrina y cumple 3 años. No sé qué comprarle. A lo mejor le compro un juguete, como un osito de peluche o un perrito. Creo que eso le va a gustar. Todos sabemos que . . .

   • 

L1 transfer should lead L1-Spanish L2-English learners to rate definite plurals high and bare plurals low in the generic category in (8), whereas L1-English L2-Spanish learners should rate definite plurals low and bare plurals high in (9). L1 transfer should not affect the specific-plural category in (6)–(7), where English and Spanish behave the same. If interpretation is more of a learning challenge than (un)grammaticality, as we hypothesized, then the L2-Spanish learners are expected to be fairly targetlike: as long as they have learned that bare plurals are ungrammatical in Spanish, and definite plurals are grammatical, they should rate (9a) as well as (7a) low, and rate (9b) as well as (7b) high. In contrast, L2-English learners would have learned that both bare and definite plurals are grammatical in English, so their task is more difficult: they need to decide which form, definite plurals and/or bare plurals, is acceptable in the context in (6) vs. (8).

Results of the English AJT

Figure 5 presents the mean ratings for bare and definite plurals in the specific-plural and generic categories in the English AJT. As this figure shows, all three groups rated definite plurals above bare plurals in the specific-plural category, and rated bare plurals above definite plurals in the generic category. However, the contrasts were numerically much weaker in the low/intermediate group than the high and native groups.

Figure 5.

English Study 2, acceptability judgment task (AJT) results: mean ratings on a scale from 1 to 4. Error bars ±1 SD.

We conducted a mixed ANOVA on the results, with category (specific-plural vs. generic) and NP type (bare plural vs. definite plural) as the within-subject variables, and group as the between-subjects variable. There was no significant effect of category, F (1, 51) < 1, and no significant effect of NP type, F (1, 51) = 2.2, p = .14, but category interacted significantly with NP type, F (1, 51) = 335, p < .001, a result of the fact that definite plurals were rated higher in the specific-plural than the generic category, whereas the reverse was the case for bare plurals. Group did not have a significant effect, F (2, 51) = 1.58, p = .22, but there was a significant three-way interaction between group, category, and NP type, F (2, 51) = 38.9, p < .001. There were no other significant interactions.

In order to isolate the source of the three-way interaction, we conducted two separate mixed ANOVAs comparing the three groups on the two NP types in the specific-plural category and in the generic category. To avoid inflating the Type I error rate by conducting two separate analyses, we set the α level at 0.025 (Bonferroni correction, 0.05 divided by 2, the number of contrasts). In the specific-plural category, there was a significant effect of NP type, F (1, 51) = 98, p < .001, due to definite plurals being rated above bare plurals. There was also a marginally significant effect of group, F (2, 51) = 3.5, p = .038 (marginal at the Bonferroni-corrected α level of 0.025), with the low/intermediate group performing differently from the other two groups. There was also a significant NP by group interaction, F (2, 51) = 17, p < .001. Interaction plots revealed that while all three groups rated definite plurals above bare plurals in the specific-plural category, the low/intermediate group made the smallest distinction between bare and definite plurals, the native group made the largest distinction, and the high group was in between: as group proficiency increased (from low/intermediate to high to native), ratings of definite plurals went up while ratings of bare plurals went down.

In the generic category, there was a significant effect of NP type, F (1, 51) = 203, p < .001, but no significant effect of group, F (2, 51) < 1. Group interacted significantly with NP type, F (2, 51) = 18, p < .001; interaction plots revealed that although all three groups rated bare plurals above definite plurals in the generic category, this contrast was stronger in the high and native groups than in the low/intermediate group: the low/intermediate group rated definite plurals higher, and bare plurals lower, than did the other two groups (which did not differ from each other).

Results of the Spanish AJT

Figure 6 presents the mean ratings for bare and definite plurals in the specific-plural and generic conditions in the Spanish AJT. As this figure shows, both learner groups were remarkably similar to the native group, with all three groups rating definite plurals much higher than bare plurals in both specific-plural and generic categories.

Figure 6.

Spanish Study 2, acceptability judgment task (AJT) results: mean ratings on a scale from 1 to 4. Error bars ±1 SD.

We conducted a mixed ANOVA on the results, with category (specific-plural vs. generic) and NP type (bare plural vs. definite plural) as the within-subject variables, and group as the between-subjects variable. There was no significant effect of category, F (1, 44) = 1.4, p = .24, but there was a significant effect of NP type, F (1, 44) = 431, p < .001, with definite plurals rated higher than bare plurals. Category interacted significantly with NP type, F (1, 44) = 32, p < .001, a result of the fact that the distinction between definite and bare plurals was bigger in the specific-plural than the generic category. Group did not have a significant effect, F (2, 44) < 1, but there was a significant three-way interaction between group, category and NP type, F (2, 44) = 4.5, p < .05. There were no other significant interactions.

In order to isolate the source of the three-way interaction, we conducted two separate mixed ANOVAs comparing the three groups on the two NP types in the specific-plural category and in the generic category. To avoid inflating the Type I error rate by conducting two separate analyses, we set the alpha level at 0.025 (Bonferroni correction, 0.05 divided by 2, the number of contrasts). In the specific-plural category, there was a significant effect of NP type, F (1, 44) = 699, p < .001, with definite plurals rated above bare plurals. There was no significant effect of group, F (2, 44) = 1.1, p = .35, and group did not interact with NP type, F (2, 44) < 1. Thus, in the specific-plural category, the two learner groups performed just like the Native group.

In the generic category, there was a significant effect of NP type, F (1, 44) = 160, p < .001, with definite plurals rated above bare plurals. There was no significant effect of group, F (2, 44) < 1, but group interacted with NP type, F (2, 44) = 3.6, p = .036, marginal at the Bonferroni-corrected α level of 0.025. Interaction plots revealed that although all three groups rated definite plurals above bare plurals, the magnitude of the contrast increased with proficiency: as proficiency went up (from low/intermediate to high to native), ratings of definite plurals went up, while ratings of bare plurals went down.

Discussion of Study 2 findings

Study 2 found more targetlike performance in L2-Spanish than in L2-English. In the Spanish study, the learners were fully targetlike in the specific-plural category (rating definite plurals above bare plurals to the same extent as native speakers), but less targetlike on the generic category. Like native speakers, learners at both proficiency levels rated definite plurals well above bare plurals in the generic category, which indicates that they are aware that Spanish, unlike English, uses definite plurals for generic interpretation. At the same time, the learners rated bare plurals higher, and definite plurals lower, in the generic category than did native speakers. This difference was particularly pronounced for the low/intermediate group. These results are fully consistent with L1-transfer from English, where bare plurals rather than definite plurals are used for generic readings. At the same time, the fact that even the low/intermediate group rated definite plurals well above bare plurals indicates that they were in the process of recovering from L1-transfer. In the specific-plural category, where Spanish and English behave alike, learners were fully targetlike even at the lowest proficiency levels.

The results from the English study present a rather different picture. The findings from the generic category are fully consistent with L1-transfer: the L2-English learners, especially those in the low/intermediate group, rated definite plurals too high and bare plurals too low, in comparison with native speakers (although even the low/intermediate group correctly rated bare plurals above definite plurals, like the natives). However, the results on the specific-plural category are not expected under L1-transfer: even though English behaves just like Spanish with regard to specific readings, the learners rated bare plurals unexpectedly high, and definite plurals unexpectedly low. A possible explanation is that the learners had learned the basic grammatical fact about English—that in English, unlike Spanish, bare plurals are grammatical in subject position—but had not yet fully acquired the interpretation of bare versus definite plurals in English, in particular which are required for generic versus specific interpretation.Footnote ⁷

The results of Study 2 are highly reminiscent of Serratrice et al.'s (Reference Serratrice, Sorace, Filiaci and Baldo2009) findings, despite the differences in test populations and task formats. Both our study and Serratrice et al.'s (Reference Serratrice, Sorace, Filiaci and Baldo2009) study find that learners have difficulty making metalinguistic judgments about bare and definite plurals in English; although problems judging plural NPs in generic contexts might be attributable to transfer from Italian/Spanish, transfer cannot account for performance in the specific contexts, where English behaves just like Italian/Spanish. Furthermore, recall that Serratrice et al. (Reference Serratrice, Sorace, Filiaci and Baldo2009) found similar difficulties among monolingual children, where transfer cannot possibly play a role. Thus, there is reason to think that different types of learners (monolingual children, bilingual children, adult L2 learners) have difficulty deciding which contexts in English are compatible with definite plurals and which, with bare plurals. As children get older and/or get more input, or as adults become more proficient, performance becomes more targetlike. At the same time, our results are not entirely parallel to Serratrice et al. (Reference Serratrice, Sorace, Filiaci and Baldo2009): whereas the children in their study made primarily errors of overacceptance (i.e., they correctly accepted grammatical sentences but also overaccepted ungrammatical ones), the adult learners in our study made errors of overrejection as well as overacceptance. There are many possible reasons for this difference, given the methodological differences between the studies; one possible explanation is that child learners, more than adult learners, exhibit a “yes” bias, which leads them to (over)accept all test sentences.