The (perhaps) limited role of phonology in prediction during language comprehension

Yiling Huo

1st Chandler House Symposium

Generation of predictions might be a universal principle of the brain:

#### Prediction during language processing: Note: Now, prediction has also been found in language comprehension. For example, if I say to you... --- #### Prediction during language processing: "This morning I went to Starbucks to buy ..." Note: Now, prediction has also been found in language comprehension. For example, if I say to you...

Prediction during language processing:

"This morning I went to Starbucks to buy ... coffee"

#### What aspects of language are involved in prediction during language comprehension? Note: Now people have been researching what aspects of language are involved in the process of prediction during language comprehension? Turns out quite a lot of them: --- #### What aspects of language are involved in prediction during language comprehension? - Sentential and discourse context (Kutas & Hillyard, 1984; Federmeier & Kutas, 1999) - Event knowledge / Theta roles (Altmann & Kamide, 1999; Chow et al., 2016) - Morphosyntax (gender/case markers) (Kamide et al., 2003; Lew-Williams & Fernald, 2007, 2010) - ... Note: Now people have been researching what aspects of language are involved in the process of prediction during language comprehension? Turns out quite a lot of them:

"I eat rice"

Phonology in prediction?

DeLong et al. (2005)

My project is concerned with the role of phonology in prediction.

The most famous piece of work on phonology in prediction comes from DeLong et al 2005. In this study they showed people context constraining sentences such as '..', and here you would expect the ending to be 'a kite', and 'an airplane' would be unexpected. And indeed DeLong et al found that the unexpected airplane elicited a larger N400 than kite, which means it took more effort to process airplane than kite. And more importantly, they found that the preceding indefinite article an before airplane also elicited a larger N400 compared with the expected a. So not only were people predicting kite, when they get an indefinite article that's not compatible with kite, their brain notices it and takes more effort to process the article as well.

Phonology in prediction?

However: Nieuwland et al. (2018) (N=356)

Two questions to ask:

Can listeners use phonological information to generate predictions?
Can listeners detect prediction errors / revise their predictions using phonology?

Experiments 1 and 2

Tone sandhi in Mandarin Chinese

Lexical tone: pitch contour that encodes meaning.

ma1 "mother", ma2 "hemp", ma3 "horse", ma4 "scold"

ma1ma1-ma4-ma3 "Mom scolds the horse."

Tone sandhi in Mandarin Chinese

Tone sandhi: lexical tone can undergo change triggered by neighboring linguistic environments.

The yi sandhi: numeral yi1 "one"

yi4tian1, yi4tiao2, yi4ba3, yi2ge4

The T3 sandhi: numeral liang3 "two"

xiao3gou3 → xiao2gou3

Tone sandhi refers to the phenomenon that lexical tones can undergo change triggered by neighboring lisguistic environments. These changes does not affect meaning. For example, in Mandarin, if you have two tone-3 syllables such as xiao3gou3, the first syllable needs to be changed into something very similar to tone-2. This is called the T3 sandhi. We have another sandhi in Mandarin which I call the yi sandhi. This sandhi says that the numeral yi, which means one, should be realised in tone-4 whenever another syllable follows it in the same phrase, except when it's followed by another tone-4, in this case it's realised in tone-2. My experiments will use these two tone sandhi at the numeral position, so the yi sandhi obviously happens with the numeral yi, and the T3 sandhi happens with the numeral liang3, which means two.

Experiments 1 and 2

Experiment 1 Results

The yi sandhi ("yi"):

Experiment 1 Results

The T3 sandhi ("liang"):

Experiment 1 Results

Experiment 2 Results

The yi sandhi ("yi"):

Experiment 2 Results

The T3 sandhi ("liang"):

Experiment 2 Results

Experiments 1 and 2

Can listeners use tone sandhi to generate predictions?

Perhaps, but this effect is not always replicated.

Can listeners use tone sandhi to generate predictions?

Perhaps, but this effect is not always replicated.

Can listener use tone sandhi to detect prediction errors / revise their predictions?

Experiment 3

Experiment 3 Results

The yi sandhi ("yi"):

Experiment 3 Results

The T3 sandhi ("liang"):

Can listeners use tone sandhi to generate predictions?

Perhaps, but this effect is not always replicated.

Can listener use tone sandhi to detect prediction errors / revise their predictions?

No evidence for revising prediction using tone sandhi.

Take home message

Phonology's role in prediction may be limited.

But why?

We see that many aspects of language are involved in prediction. Why not phonology?

But why?

We see that many aspects of language are involved in prediction. Why not phonology?

Processing time hypothesis:

Evidence that English articles 'a/an' and Chinese tone sandhi are involved in prediction when SOA is long / speech rate is low (Ito et al., 2016; Liu et al., 2023).
Maybe the comprehender needs more processing time in order to get phonology involved in prediction.

But why?

We see that many aspects of language are involved in prediction. Why not phonology?

The 'Filter' hypothesis:

As language involves several levels of representation that require different levels of abstraction (phonology - semantics - syntax),
Maybe only information at a certain level of abstraction or above (e.g. semantics and above) is involved in prediction (under normal circumstances).

Thank you!

References

Altmann, G. T., & Kamide, Y. (1999). Incremental interpretation at verbs: Restricting the domain of subsequent reference. Cognition, 73(3), 247-264.

Chow, W. Y., Smith, C., Lau, E., & Phillips, C. (2016). A “bag-of-arguments” mechanism for initial verb predictions. Language, Cognition and Neuroscience, 31(5), 577-596.

DeLong, K. A., Urbach, T. P., & Kutas, M. (2005). Probabilistic word pre-activation during language comprehension inferred from electrical brain activity. Nature neuroscience, 8(8), 1117-1121.

Federmeier, K. D., & Kutas, M. (1999). A rose by any other name: Long-term memory structure and sentence processing. Journal of memory and Language, 41(4), 469-495.

Ito, A., Corley, M., Pickering, M. J., Martin, A. E., & Nieuwland, M. S. (2016). Predicting form and meaning: Evidence from brain potentials. Journal of Memory and Language, 86, 157-171.

Kamide, Y., Altmann, G. T., & Haywood, S. L. (2003). The time-course of prediction in incremental sentence processing: Evidence from anticipatory eye movements. Journal of Memory and language, 49(1), 133-156.

Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect word expectancy and semantic association. Nature, 307(5947), 161-163.

Lew-Williams, C., & Fernald, A. (2007). Young children learning Spanish make rapid use of grammatical gender in spoken word recognition. Psychological science, 18(3), 193-198.

Lew-Williams, C., & Fernald, A. (2010). Real-time processing of gender-marked articles by native and non-native Spanish speakers. Journal of memory and language, 63(4), 447-464.

Nieuwland, M. S., Politzer-Ahles, S., Heyselaar, E., Segaert, K., Darley, E., Kazanina, N., ... & Huettig, F. (2018). Large-scale replication study reveals a limit on probabilistic prediction in language comprehension. ELife, 7, e33468.

Shun, L., Chen, X., & Wang, S. (2023). The involvement of phonological information during spoken language prediction: evidence based on Chinese tone sandhi. OSF Preprints.