Resisting attraction

We propose that attraction errors in agreement production (e.g., the key to the cabinets are missing) are related to two components of executive control: working memory and inhibitory control. We tested 138 children aged 10 to 12, an age when children are expected to produce high rates of errors. To increase the potential of individual variation in executive control skills, participants came from monolingual, bilingual, and bidialectal language backgrounds. Attraction errors were elicited with a picture description task in Dutch and executive control was measured with a digit span task, Corsi blocks task, switching task, and attentional networks task.

Overall, higher rates of attraction errors were negatively associated with higher verbal working memory and, independently, with higher inhibitory control. To our knowledge,
this is the first demonstration of the role of both working memory and inhibitory control in attraction errors in production. Implications for memory- and grammar-based models are discussed.

Read the article here: Veenstra, A., Antoniou, K., Katsos, N., & Kissine, M. (2018, April 19). Resisting Attraction: Individual Differences in Executive Control Are Associated With Subject–Verb Agreement Errors in Production. Journal of Experimental Psychology: Learning, Memory, and Cognition. Advance online publication. http://dx.doi.org/10.1037/xlm0000516

Advertisement

Executive control and agreement errors

When I was running earlier subject-verb agreement experiments, I noticed that some people make a lot more agreement errors than others. Although the overall mean error rates were similar to the existing literature, I became intrigued by the large individual variation in agreement production.

Some studies had already shown (some more convincingly than others) that working memory might have something to do with it. We also hypothesized that inhibitory control might prove useful in producing correct agreement. Remember the sentence “the key to the cabinets are missing”? Whereas the number of the head noun has to te kept active until the point in time where the agreement link with the verb is made, the number from the local noun has to be inhibited so that the verb does not accidentally agree with the local noun.

We tested this hypothesis with two groups of children: monolingual Dutch speakers and bilingual French-Dutch speakers. Some studies have suggested that bilinguals have better executive control skills than monolinguals. If these skills are needed for agreement, bilinguals might be better at agreement production as well.

All children completed a agreement production task (which we first introduced here), a verbal and a non-verbal working memory task, a switch task, and the Attentional Networks Task. A bit to our surprise, the monolingual and bilingual children did not differ in their executive control skills or in their agreement error rates. However, we did find overall correlations of verbal working memory with agreement error rates: children with a higher working memory made fewer errors. Although there was a trend for inhibitory control in the hypothesized direction, this effect did not reach significance. More children are being tested to see whether this is a power issue.

Read the original article:

Veenstra, A., Antoniou, K., Katsos, N., & Kissine, M. (2017). The role of executive control in agreement attraction in monolingual and bilingual children. Proceedings of the 41st annual Boston University Conference on Language Development, 706-717.

Notional and grammatical number in subject-verb agreement

In an earlier paper, we have argued that for subject-verb agreement, not only the grammatical number properties of the subject phrase are important, but also the notional number properties. Speakers make agreement errors when a grammatically singular head noun is combined with a grammatically plural local noun: #The bowl with the stripes are broken (versus The bowl with the stripe is broken). But more errors are also made when a grammatically singular head noun is combined with a notionally plural subject phrase: #The bowl with the spoon are broken (versus The bowl with the stripe is broken). This notional number mismatch effect appears to be additive to the grammatical number mismatch effect (attraction). In this paper, we investigated whether the two factors work independently from each other, by observing the change in agreement error rates when the notional number is made more salient.

One group of participants heard subject phrases with nouns that matched and mismatched in grammatical number, and subject phrases in which the notional number matched and mismatched with the head noun. Participants had to press a button for the singular or plural verb phrase with which they wished to continue the sentence. Another group of participants heard the same subject phrases, while being presented with a drawing of that subject phrase. This drawing made the notional number of the phrase very clear.

There were effects both of grammatical and notional number mismatch, which did not interact. Moreover, the notional number effect was stronger in the group that saw the drawings, whereas the grammatical number effect was identical across both groups. This suggests that notional and grammatical number information each have their independent influences on the agreement process.

Read more:

Veenstra, A., & Acheson, D. J. (2016). Semantic integration and subject-verb agreement: Independent effects of notional and grammatical number. Studies of the Belgian Linguistics Circle, 10:5.

The acquisition of clitics and pronouns

This study develops a single elicitation method to test the acquisition of third-person pronominal objects in 5-year-olds for 16 languages. This methodology allows us to compare the acquisition of pronominals in languages that lack object clitics (“pronoun languages”) with languages that employ clitics in the relevant context (“clitic languages”), thus establishing a robust cross-linguistic baseline in the domain of clitic and pronoun production for 5-year-olds. High rates of pronominal production are found in our results, indicating that children have the relevant pragmatic knowledge required to select a pronominal in the discourse setting involved in the experiment as well as the relevant morphosyntactic knowledge involved in the production of pronominals. It is legitimate to conclude from our data that a child who at age 5 is not able to produce any or few pronominals is a child at risk for language impairment. In this way, pronominal production can be taken as a developmental marker, provided that one takes into account certain cross-linguistic differences discussed in the article.

Read the original article:

Varlokosta, S., Belletti, A., Costa, J., Friedmann, N., Gavarro, A., Grohmann, K. K., Guasti, M. T., Tuller, L., Lobo, M., Anđelković, D., Argemí, N., Avram, L., Berends, S., Brunetto, V., Delage, H., Ezeizabarrena, M-J., Fattal, . I., Hamann, E., van Hout, A., Jensen de Lopez, K., Katsos, N., Kologranic, L., Krstić, N., Kuvac Kraljevic, J., Miękisz, A., Nerantzini, M., Queraltó, C., Radic, Z., Ruiz, S., Sauerland, U., Sevcenco, A., Smoczynska, M., Theodorou, E., van der Lely, H., Veenstra, A., Weston, J., Yachini, M. & Yatsushiro, K. (2016). A cross-linguistic study of the acquisition of clitic and pronoun production. Language Acquisition, 23(1), 1-26.

Parallel planning and agreement

We know that speakers make agreement errors when the subject phrase consists of a singular head noun followed by a plural local noun (e.g., the key to the cabinets are missing). It has been suggested that more agreement errors are made when these two mismatching nouns are planned simultaneously before speech onset, compared to when they are planned one-by-one: the singular and plural number features would be active simultaneously and lead to more interference, and thus, to more errors.

We tried to find out whether this statement is true. In a picture description task with eye-tracking, we asked participants to name two sets of pictures: a head noun and a local noun, and complete the sentence with a color adjective. This lead to sentences such as “the apple next to the cars is green.” To make sure agreement errors were made, the head and local nouns differed in number. To make sure that the two nouns were planned in parallel, the pictures were located at such a short distance, that the second picture could be processed while focusing on the first. Of course, there was also a condition in which the pictures were located at a large distance where simultaneous processing was much more difficult. The next step was to make sure that this manipulation indeed affected the planning strategy. We used pictures that were semantically related and unrelated. We predicted to find semantic interference (when speech onsets and gaze durations take longer because it is harder to name two similar pictures, compared to unrelated pictures) in the close condition, and not in the far condition. Then, you can compare the agreement error rates between the parallel and sequential planning conditions.

First of all, agreement errors were made for items in which the head and local noun differed in number (both for singular and plural head nouns). Second, we found semantic interference in the close condition, and not in the far condition. This suggests that we were succesful in creating a parallel and a sequential planning condition. However, the error rates did not differ between those conditions. Parallel planning of mismatching nouns does not seem to increase error rates.

Read the original article:

Veenstra, A., Meyer, A. S., & Acheson, D. J. (2015). Effects of parallel planning on agreement production. Acta psychologica, 162, 29-39.

 

Keep it simple!

In many linguistic experiments, there is a rich variation in the words used. Agreement studies are no exception. Ofcourse, our daily language also contains a rich variation of words and structures. But if we want to study grammar, it is useful to limit the lexical variation to a minimum, so that we can get to the heart of the grammar.

In agreement studies, researchers have found that certain factors make it difficult to generate correct agreement. Remember the key to the cabinets are missing? We wanted to see if these patterns can still be found when the lexical variation in the items is reduced. Our sentences only contained four different words: circle, rectangle, triangle, and star.

We reproduced the standard finding (attraction when a singular head noun is combined with a plural local noun: the circle next to the stars are red), but also found a new finding (attraction when a plural head noun is combined with a singular local noun: the circles next to the star is red).

On top of that, we also designed a very cool new paradigm to test agreement: picture description. In contrast to more traditional agreement tasks, picture description does not involve reading or listening. I can’t wait to try this out in new populations!

Read the original article:

Veenstra, A., Acheson, D. J., & Meyer, A. S. (2014). Keeping it simple: Studying grammatical encoding with lexically-reduced item sets. Frontiers in Psychology (5): 783.

 

Semantic integration

Researchers have long believed that subject-verb agreement is only affected by grammatical factors. One example is attraction, here the verb does not agree with number of  the head noun, but with the number of an intervening local noun: the key to the cabinets are missing. So instead of singular, the verb becomes plural, which we call an agreement error.

More and more studies are finding semantic factors in addition to the grammatical factors. One is brought about by semantic integration (not to be confused with semantic integration as often seen in EEG reading studies). The semantic integration I am discussing here concerns the conceptual link between two words in a phrase. For example, in the phrase the bowl with the stripes, the bowl and the stripes are conceptually tightly linked: the stripes are part of the bowl. In the bowl with the spoons the bowl and the spoons are less tightly linked. The tightness of this link influences the agreement process.

However, the exact effect of semantic integration is under debate. One account claims that a tight link makes the grammatical attraction effect stronger. Another account claims that a tight link makes the phrase “feel” singular, whereas a weak link makes the phrase “feel” plural. This means that a sort of semantic attraction can occur, which increases agreement error rates.

We looked at the effect of semantic integration in Dutch, and found support for the second, notional number account. So in addition to the grammatical attraction effect, we found that agreement was more difficult when a singular head noun was combined with a “plural feel”.

Read more:

Veenstra, A., Acheson, D. J., Bock, K., & Meyer, A. S. (2014). Effects of semantic integration on the production of subject-verb agreement: evidence from Dutch. Language, Cognition and Neuroscience, 29(3), 355-380.