Heili og tungumál

In Málstofur by Eiríkur Smári Sigurðarson

Hvernig vinnur heilinn úr orðum og setningum sem honum berast? Þessari spurningu er ekki hægt að svara með innsæinu einu saman heldur þarf að framkvæma tilraunir til að ákvarða hvaða merkingarlegu og setningarlegu þættir skipta máli. Taugafræðileg vinnsluferli málfræðiupplýsinga í heilbrigðum íslenskumælandi einstaklingum hafa lítið verið til umfjöllunar þótt stórstígar framfarir hafa orðið síðustu ár á sviði kennilegra málvísinda. Verkefninu Heili og tungumál: taugamálfræði og samspil setningagerðar og merkingar í íslensku, sem fékk þriggja ára RANNÍS-styrk 2016, er ætlað að bæta úr þessum skorti. Þar er rannsakað fyrirbæri í íslensku sem gerir kleift að aðskilja vinnslu upplýsinga í beygingum og setningargerð annars vegar og merkingu hins vegar. Í málstofunni verða þrír fyrirlestrar: sá fyrsti segir frá einni af fyrstu heilaritunartilraunum verkefnisins um merkingu í íslensku; næsti segir frá tilraun í hljóðfræði íslenskunnar sem er undirbúningur fyrir heilaritunartilraun í haust; þriðji kynnir rannsóknir um þýsku sem nýdoktorinn í verkefninu stóð fyrir í doktorsnámi sínu.

Fyrirlestrarnir verða á ensku.

Þórhallur Eyþórsson kynnir fyrirlesara og stýrir umræðum.

Laugardagur 11. mars

Hvar
Stofu 218 í Aðalbyggingu Háskóla Íslands
Hvenær
Kl. 15.00-16.30

Málstofustjóri:
Matthew Whelpton


Fyrirlesarar og titlar erinda

11. mars kl. 15.00-16.30

Collaborators: Patrick Kelley, Drew Trotter, Karthik Durvasula, Joan Maling, and Alan Beretta

Nouns which refer to substances (oil) are traditionally classed as mass nouns (She poured oil on her salad); however, such mass nouns can easily be used like a count noun (She poured two oils on her salad), on the condition that they receive a sort interpretation (she used two different sorts of oil). Such coerced interpretations should carry a processing cost of the kind reported in the literature (e.g., Kuperberg et al. 2010; Paczynski et al. 2014). However, Whelpton et al. (2014) found no effect for sort coercion in Icelandic.

Our hypothesis is that this relates to the markedness of the contrasting sorts involved in the coercion interpretation (several oils might be olive vs walnut vs coconut). The more marked these contrasting subsets are, the more processing work is required to isolate them. To test this hypothesis, we took the original stimulus set from Whelpton et al. (2014) and added adjectival modifiers to the relevant noun phrases.

Coercion: Konan þáði aðra dálítið þykka sósu á endanum

The woman accepted another rather thick sauce in the end

Neutral: Konan þáði aðra svakalega bleika dúkku á endanum

The woman accepted another outrageously pink doll in the end

In the coercion condition, we must identify subsets of rather thick sauce which constitute coherent sorts: perhaps a rather thick French cream sauce, a rather thick Indian lentil sauce etc. These are more marked than for the unmodified noun. At 450-750 ms, we found an increased negativity for the Coercion condition relative to the Neutral condition that was most prominent in anterior-central regions.

In addition we ran a second ERP study with the same stimuli but without the modifiers. We found no effect, replicating the null finding in Whelpton et al. This suggests that sort coercion is a subject to a pragmatic constraint related to the markedness of the contrasting sorts that one must retrieve in the coercion interpretation.

Collaborators: Lilja Björk Stefánsdóttir, Bjarni Barkarson, Tatiana Kantorovich, Joe Jalbert, Matthew Whelpton, Alan Beretta

Prosody (changes in pitch, loudness, length) can serve as a cue for syntactic structure in sentences. To provide a basis for future studies, we investigate the differences between words in a single noun context (1a) and within a compound (1b).

1.    a. Gætir þú keypt fyrir mig annan penna í Kosti

b. Gætir þú keypt fyrir mig annan pennastokk í Kosti

While representational theories of Icelandic stress predict few differences between these two contexts, we find phonetic differences between these two contexts that can serve as cues for the presence (or the lack) of a compound. We examine three major acoustic variables (pitch, loudness, length) in prosody, and find differences between single nouns and compounds within all three variables.

Collaborators: Tyler Roberts, Alan Beretta

Nouns are syntactically and semantically important in noun phrases, so when one appears in a sentence, it assumed to be the main word, or head, in that phrase. Encountering compounds (words that contain two or more nouns) can require one to reanalyze which is the head within the context, which can be a costly operation in processing. We test this hypothesis in German, where gender agreement can provide clues about upcoming structure. We test the following three conditions:

1) a. Der Finger -nagel Full Gender Match (FGM)
The.masc finger.masc – nail.masc

b. Die Glass -tür Gender Agreement (GA)
The.fem glass.neut -door.fem

c. *Das Tier -artz Gender Violation (GV)
The.neu animal.neut – doctor.masc

In two different experiments, using electroencephalography (EEG), we examine the cost of head reanalysis. In the first experiment, we find that subjects experience a syntactic reanalysis cost in cases where they cannot predict the compound (FGM) compared to when they can (GA). There is also a cost for ungrammatical phrases (GV). In the second experiment, where the compounding structure becomes predictable, we find a semantic reanalysis cost for cases where there are no cues for compounding (FGM) compared to cases that do have those cues (GA).

Deila færslunni