Programm StuTS 70

Exploring the interactions between visual input, agreement attraction and numerical cognition: an (online) study in Dutch

While humans are unique among species in their cognitive capacity for learned mathematical skill (counting, performing arithmetic operations), our innate ability for intuitively grasping and evaluating quantities and magnitude characterize most of our daily experience.
Our Approximate Number System (ANS) is thought to be involved in the estimation of ‘how much’ there is of a given element in ecological scenarios; for example, when visually assessing how much rice there is in a plate, based on approximated grain size and surface distribution. Research suggests that display properties of the visual stimuli influence magnitude and quantity estimation. When elements of an image array appear more spread-out, numerosity perception tends to increase at a certain magnitude threshold (Krueger, 1972).
How this latter subsystem interacts and interferes with other higher cognitive functions, like language, is not completely understood as of now. However, it is the case that such contact occurs routinely, as the information the ANS carries is vital for expressing numerosity linguistically.

In the context of psycholinguistics, the encoding and processing of numerical information tends to be reflected in linguistic comprehension and production performance. Language-specific number marking, lexical numerals and quantifier restrictions all serve to encode quantity and index numerical information within the clause.
Phenomena like number agreement attraction (e.g. erroneous number marking of VP for a given NP head subject, as in ‘The key to the cabinets are on the table’) is typically induced in English (and other languages) when a local noun (here, ‘the cabinets’) seizes agreement by error. This effect can be of particular interest, as such errors have been shown to be modulated by more than just grammatical number, including non-linguistic numerical information during the processing stage. Linguistic encoding of number and quantity has two main streams of information: grammatical number (language-specific rules for agreement and overt marking) and notional number (conceptual-semantic numerical properties of referents, based on typicality). These two types of information guide agreement (errors) and appear to interact with each other during processing (Eberhard, Cutting & Bock, 2005), while also revealing interactions with nonlinguistic information available at this stage (e.g. visual input). As such, the exact computational nature of these interactive processes during sentence production remains unclear.
How does the ANS interact with grammatical and notional number during language processing? Can the involvement of these factors be modeled to account for differences observed in error rates for number agreement? Finally, can these interactions inform a finer-grained picture of the processing steps and the relationship between numerical cognition and linguistic number, and beyond, between perceptual input and language production in general?

To address these questions, a study was conducted in English (Brehm, to appear) in which the effect of notional typicality, grammatical specification, and visual array was evaluated for number agreement attraction during a sentence completion task. This study replicated previous results (Eberhard, 1997), showing that when conceptual number aligned with a specified determiner or quantifier, protective effects on agreement were observed. Moreover, this study showed novel effects of visual array on agreement: the biggest contrast reported occurred when far-spread arrays elicited more plural completions with ‘each’ and ‘every’, which are notionally plural but grammatically singular, suggesting that perceptual input may have a more significant effect in cases of conflicting, or distributive number (following Humphreys & Bock, 2004 and Eberhard, 1999).
In order to replicate and further extend these results, we conducted a similar study for Dutch, as it is phylogenetically close enough to English to predict similar effects, but features certain lexical specificities which make it a good probe for more general linguistic properties. In the context of the pandemic, the study was fully carried online, using the Gorilla Experiment Builder platform.

Within a sentence completion task, subjects were presented with a set of NPs for which they had to provide a (number agreeing) VP head. They were sequentially presented with (1) a singular subject NP and its plural PP modifier (2) a cue target adjective serving as complement for the verb head they were asked to provide in each case. While subjects were listening to the subject NP, they were presented with a visual array stimuli, followed by the adjective target on screen. Importantly, subjects had to orally produce these completions while being recorded.
As in Brehm’s original study, we manipulated notional and grammatical number in linguistic stimuli, and set three array type conditions (far, close or mid/filler) in visual primes.
Quantifiers and determiners of the head noun varied in grammatical number (S/P) being specified or underspecified, as well as in notional number (typicality of referent number given a determiner or quantifier). Every subject was presented with every combination of each condition.
Specifically for this study, we exploited determiner gender/number contrasts of Dutch by using diminutive forms of the noun (as in Antón-Méndez & Hartsuiker, 2010), both as a control for the neutral ‘het’ (maximally specific singular determiner) compared to ‘de’ (underspecified in number), and as a test for a putative effect of the diminutive in singular agreement protection, which is a hypothesis that shall be further explained in the course of this talk.

The data are currently under transcription and analysis, but we will provide a general discussion for the implications of upcoming results.
On a more methodological note, and given the growing body of online research in linguistics since 2020, we will provide a more in-depth overview of Gorilla Experiment Builder as it applies to psycholinguistic experimental needs.
This study was conducted under the supervision of Laurel Ellen Brehm and Antje Meyer at the Max Planck Institute for Psycholinguistics in Nijmegen, and launched in July 2021 for data collection.

Info

Tag: 20.11.2021
Anfangszeit: 14:00
Dauer: 00:30
Raum: 🥝

Sprache: en

Links:

• iCalendar

Feedback

Uns interessiert Ihre Meinung! Wie fanden Sie diese Veranstaltung?