Predicting the next sentence (not word) in large language models: What model-brain alignment tells us about discourse comprehension.

Current large language models (LLMs) rely on word prediction as their backbone pretraining task. Although word prediction is an important mechanism underlying language processing, human language comprehension occurs at multiple levels, involving the integration of words and sentences to achieve a full understanding of discourse. This study models language comprehension by using the next sentence prediction (NSP) task to investigate mechanisms of discourse-level comprehension. We show that NSP pretraining enhanced a model's alignment with brain data especially in the right hemisphere and in the multiple demand network, highlighting the contributions of nonclassical language regions to high-level language understanding. Our results also suggest that NSP can enable the model to better capture human comprehension performance and to better encode contextual information. Our study demonstrates that the inclusion of diverse learning objectives in a model leads to more human-like representations, and investigating the neurocognitive plausibility of pretraining tasks in LLMs can shed light on outstanding questions in language neuroscience.

Syntactic structures in motion: investigating word order variations in verb-final (Korean) and verb-initial (Tongan) languages.

This study explored sentence processing in two typologically distinct languages: Korean, a verb-final language, and Tongan, a verb-initial language. The first experiment revealed that in Korean, sentences arranged in the scrambled OSV (Object, Subject, Verb) order were processed more slowly than those in the canonical SOV order, highlighting a scrambling effect. It also found that sentences with subject topicalization in the SOV order were processed as swiftly as those in the canonical form, whereas sentences with object topicalization in the OSV order were processed with speeds and accuracy comparable to scrambled sentences. However, since topicalization and scrambling in Korean use the same OSV order, independently distinguishing the effects of topicalization is challenging. In contrast, Tongan allows for a clear separation of word orders for topicalization and scrambling, facilitating an independent evaluation of topicalization effects. The second experiment, employing a maze task, confirmed that Tongan's canonical VSO order was processed more efficiently than the VOS scrambled order, thereby verifying a scrambling effect. The third experiment investigated the effects of both scrambling and topicalization in Tongan, finding that the canonical VSO order was processed most efficiently in terms of speed and accuracy, unlike the VOS scrambled and SVO topicalized orders. Notably, the OVS object-topicalized order was processed as efficiently as the VSO canonical order, while the SVO subject-topicalized order was slower than VSO but faster than VOS. By independently assessing the effects of topicalization apart from scrambling, this study demonstrates that both subject and object topicalization in Tongan facilitate sentence processing, contradicting the predictions based on movement-based anticipation.

Brain activity predicts future learning success in intensive second language listening training.

This study explores neural mechanisms underlying how prior knowledge gained from pre-listening transcript reading helps comprehend fast-rate speech in a second language (L2) and applies to L2 learning. Top-down predictive processing by prior knowledge may play an important role in L2 speech comprehension and improving listening skill. By manipulating the pre-listening transcript effect (pre-listening transcript reading [TR] vs. no transcript reading [NTR]) and type of languages (first language (L1) vs. L2), we measured brain activity in L2 learners, who performed fast-rate listening comprehension tasks during functional magnetic resonance imaging. Thereafter, we examined whether TR_L2-specific brain activity can predict individual learning success after an intensive listening training. The left angular and superior temporal gyri were key areas responsible for integrating prior knowledge to sensory input. Activity in these areas correlated significantly with gain scores on subsequent training, indicating that brain activity related to prior knowledge-sensory input integration predicts future learning success.

[Influence of seven absorption enhancers on nasal mucosa--assessment of toxicity].

OBJECTIVE: To study the influence of seven different absorption enhancers on nasal mucosa. METHODS: By testing last time of ciliary movement, velocity of ciliary movement, ciliary structural and specific cellular changes of nasal mucosa the influence of seven different absorption enhancers on nasal mucosa. RESULTS: The effect on lasting time of ciliary movement was 1%SDS>1%SDC>1%Brij35>5%Tween80>0.1%EDTA>5%HP-beta-CD>1%lecithin the effect on velocity of ciliary movement 1%Brij35>1%SDC>1%SDS>0.1%EDTA>1%lecithin>5%Tween80>5%HP-beta-CD,and the effect on ciliary structural and specific cellular changes of nasal mucosa 1%SDS approximately 1%SDC approximately 1%Brij35>5%Tween80>0.1%EDTA approximately 5% HP-beta-CD approximately 1%lecithin. CONCLUSION: The three methods have good correlation.

Nasal insulin delivery in the chitosan solution: in vitro and in vivo studies.

The effects of chitosan concentrations, osmolarity, medium and absorption enhancers in the chitosan solution on nasal insulin delivery were studied in vitro and in vivo. The penetration of insulin through the mucosa of rabbit nasal septum was investigated by measuring the transmucosal flux in vitro, while the nasal absorption of insulin in vivo was assessed by the efficiency in lowering the blood glucose levels in normal rats. It was demonstrated that increasing concentrations of chitosan up to 1.5% (w/v) caused an increase in the permeability of insulin across the nasal mucosa. Insulin given intranasally in hypo- or hyperosmotic formulation showed a higher hypoglycemic effect than insulin delivered in isoosmotic formulation. Insulin formulation in chitosan solution prepared with deionized water brought to a higher relative pharmacological bioavailability (Fr) value than that prepared with 50 mM pH 7.4 phosphate buffer. A formulation containing both 1% chitosan and 0.1% ethylenediaminetetraacetic acid (EDTA), 5% polysorbate 80 (Tween 80) or 1.2% beta-cyclodextrin (beta-CD) did not lead to a higher Fr than insulin formulated with 1% chitosan alone. The formulation containing both 5% hydroxypropyl-beta-cyclodextrin (HP-beta-CD) and 1% chitosan was more effective at reducing blood glucose levels than the formulation containing 5% HP-beta-CD or 1% chitosan alone. The studies indicated that chitosan concentrations, osmolarity, medium and absorption enhancers in chitosan solution have significant effect on the insulin nasal delivery. The results of in vitro experiments were in good agreement with that of in vivo studies.

[Preparation and properties of 5-fluorouracil-loaded chitosan microspheres for the intranasal administration].

OBJECTIVE: To study the preparation technique and release characteristic of 5-fluorouracil loaded chitosan microspheres for the intranasal administration. METHODS: Using the liquid paraffin as the oil phase,and span-80 as the emuifier; 5-fluorouracil-loaded chitosan microspheres were achieved by emulsion-chemical crosslink technique. The orthogonal experimental design was applied to optimize the preparation procedure. Dynamic dialysis method was used to determine the releasing characteristic of microspheres in vitro and it influencing factors. Swelling behavior was expressed by swelling ratio. The degree of mucoadhesion was investigated by determining the mucociliary transport rate(MTR) of the microparticle across a frog palate. RESULTS: Microspheres with a good shape and narrow size distribution were prepared. The average diameter was (43+/-4) microm. The drug loading was 38.5%+/-1.0%. The entrapment efficiency was 79.0%+/-1.8%. The drug release profile in vitro could be described by Higuichi equation Q=0.1035t(1/2)+0.0284 (r=0.9965). Chitosan had good mucoadhesive property and caused a significant reduction in MTR(P<0.01). CONCLUSION: The optimized preparation technique is stable and has a high entrapment efficiency. So it could be used to prepare 5-fluorouracil-loaded chitosan microspheres for the intranasal administration. Chitosan is a good material for nasal preparation and has prospective development in the pharmaceutical field.