Unpacking ERP Responses in Artificial Language Learning

Unpacking ERP Responses in
Artificial Language Learning
Pablo Bernabeu
in collaboration with
Christina Athanasiadi, Gabriella Silva,
Stella Pischinger, My Ngoc Giang Hoang,
Vincent deLuca, Jason Rothman,
Claudia Poch, Iva Ivanova,
Jorge González Alonso
LaBL

Morphosyntactic transfer
• Initial heuristics
• Cognitive economy
• Facilitative or non-facilitative

• Third language (L3) context
• transfer source(s) selected
• information guiding the selection
• time course of the selection

• Third language (L3) context
• transfer source(s) selected
• information guiding the selection
• time course of the selection
LESS Project
Linguistic Economy through
transfer Source Selectivity

Transfer in L3 acquisition
Sources

L2 by
default
L1
and/or
L2
Sources

L2 Status Factor Model (Bardel & Falk, 2012)
• L2 by default. Declarative memory for L2 and subsequent languages
L2 by
default
L1
and/or
L2
Sources

Cumulative Enhancement Model (Flynn et al., 2004)
• Property-by-property and only facilitative
L2 by
default
L1
and/or
L2
Sources

Linguistic Proximity Model (Westergaard et al., 2017)
• Property-by-property, with facilitative and non-facilitative outcomes
L2 by
default
L1
and/or
L2
Sources

Linguistic Proximity Model (Westergaard et al., 2017)
• Property-by-property, with facilitative and non-facilitative cases
Typological Primacy Model (Rothman, 2011)
• Full transfer from one language, based on overall structural similarity
L2 by
default
L1
and/or
L2
Sources

Morphosyntactic transfer in L3 acquisition

• Confounds when using natural languages: age of acquisition, frequency of use,
proficiency level, morphological salience, etc.

• Workaround: artificial languages
Much more artificial than other lab studies?
Consider Buzz Lightyear versus Woody
versus humans.

Much more artificial than other lab studies?
Consider Buzz Lightyear versus Woody
versus humans.
Compare experimental paradigms to
language in natural contexts.

• Consistency with natural language

• Acquired by statistical learning (Hudson Kam & Newport, 2005; Kidd, 2012; Monaghan et al., 2023)
• Similar processing signatures (Friederici et al., 2002; Uddén & Männel, 2018)

• Acquired by statistical learning (Hudson Kam & Newport, 2005; Kidd, 2012; Monaghan et al., 2023)
• Similar processing signatures (Friederici et al., 2002; Uddén & Männel, 2018)
• An extensive training in the artificial language could be necessary for standard
syntactic signatures (notably, P600) to appear in ERPs (González Alonso et al.,
2020; Pereira Soares et al., 2022).

González Alonso et al. (2020)
• Participants: L1 Spanish, L2 English
• Artificial language groups: Mini-Spanish (n = 26), Mini-English (n = 24)
• Grammatical property: gender agreement between nouns and predicative
adjectives in copular sentences. Example:
• Mini-Spanish: Jer mochil son carejur | Jer mochil son baratejur
• Mini-English: Jer bag are expensivejur | Jer bag are cheapejur
• Translation: The bags are expensive | The bags are cheap
• Session phases: vocabulary pre-training → grammatical training →
test (min. 80%) → ERP experiment → gender assignment task in Spanish

Je bag is cheapeju
Nouns Nouns
González Alonso
et al. (2020)
Artificial languages

González Alonso
et al. (2020)
Pre-training

González Alonso
et al. (2020)
Training

González Alonso
et al. (2020)
Test

González Alonso
et al. (2020)
Experiment

González Alonso
et al. (2020)
Gender assignment
task in Spanish

Hypotheses (based on Rothman et al., 2015) under the assumption that
transfer would happen before the ERP measurement.

Results
• Mini-Spanish group: broadly distributed 300–600 ms positivity,
most consistent with attention-related P300.
• No (N400)–P600
• Interpretation: allocation of attentional resources in preparation for
the selection of transfer source(s). In Mini-Spanish, larger focus on
word-final gender morphology, consistent with Spanish.

Pereira Soares et al. (2022)
Materials

Case morphology Adjective position
Language combination TPM LPM/SM TPM LPM/SM
L1 Italian—L1 German (L2 English)—L3 Mini-Latin No effect (N400)-P600 (N400)-P600 (N400)-P600
L1 German—L2 English—L3 Mini-Latin No effect (N400)-P600 No effect No effect
Hypotheses (based on Rothman et al., 2015) under the assumption that
transfer would happen before the ERP measurement.

Results
• From the abstract: […] “N200/N400 deflection for the HSs in case
morphology and a P600 effect for the German L2 group in adjectival
position. None of the current L3/Ln models predict the observed results,
which questions the appropriateness of this methodology.”

Our study
• Sites: Tromsø and Madrid
• Several groups per site
• Six sessions
• Three executive functions
• Three grammatical properties
• Several parts per session

Grammatical properties, examples, and presence in natural languages
Gender agreement Zer watch are yellowezur and …
in Spanish and Norwegian
Differential object marking Jessica provided fi ze key and …
in Spanish
Verb-object number agreement John cleanedevo fi zer fikey and…
in none of these languages

Wrap-up buffer: … and ze watch too
Sentence wrap-up effects: Real? Dogma? Real?
• Just and Carpenter (1980)
• Stowe et al. (2018)
• Desbordes et al. (2023)
• Meister et al. (2022)

Property Artificial lang.
Gender
agreement
Mini-Norwegian
Mini-English
Differential
object marking
Mini-Norwegian
Mini-English
Verb-object
number agreem.
Mini-Norwegian
Mini-English
Property Group Artificial lang.
Gender
agreement
L1 Eng, L2 Spa
Mini-Spanish
Mini-English
L1 Spa, L2 Eng
Mini-Spanish
Mini-English
Differential
object
marking
L1 Eng, L2 Spa
Mini-Spanish
Mini-English
L1 Spa, L2 Eng
Mini-Spanish
Mini-English
Verb-object
number
agreement
L1 Eng, L2 Spa
Mini-Spanish
Mini-English
L1 Spa, L2 Eng
Mini-Spanish
Mini-English

• Session 1. Individual differences (home-based session)
• Working memory (digit span), selective attention (Stroop) and implicit learning (serial reaction time)
• Language History Questionnaire (LHQ3; Li et al., 2020)
Sessions

+ 1 week: Session 2. Gender agreement
• Session begins with resting-state EEG (eyes-open, eyes-closed counterbalanced across participants)
Sessions

+ 1 week: Session 3. Differential object marking + Gender agreement
• Training only in the new property
• Experiment part contains both properties intermixed
Sessions

+ 1 week: Session 4. Verb-object agreement + Differential object marking + Gender agreement
• Same mechanism as in the previous session
Sessions

+ 1 week: Session 5. Retest of executive functions (home-based session)
Sessions

+ 4 months: Session 6. Retest of all grammatical properties (Morgan-Short et al., 2012)
• Session ends with control tests on the relevant properties in the natural languages
Sessions

Creation of the artificial languages
• Cognates avoided: lexicons tailored to Norwegian and Spanish sites.
• Content words (n, adj, adv, v): translated across the two artificial
languages in each site, and ideally across the three mini-languages.
• Must be picturable (González Alonso et al., 2020; Wendebourg et al., 2025)
• Adjectives paired by meaning, mostly by antonymy.
• Morphemes: same across the two artificial languages in each site,
and ideally across the three mini-languages.
• Design and materials described in González Alonso et al. (2025).

Stimulus creation: Phonological and semantic challenges
SPA_noun ENG_noun NOR_noun
habitación bedroom soverom
mochil bag bag
nuez walnut valnøtt
perch hanger henger
raiz root rot
taz cup kopp
ventan window vindu
aguacate avocado avokado
cuchil knife kniv
gor hat hatt
reloj watch klokke
zapat shoe sko
blue blaa
white hvit
first foerst
last sist
good bra
bad daarlig
easy lett
malet suitcase koffert
mes table bord
etiquet label merkelapp

• Modular framework formed of interoperable components
• Minimal components of each language contained in a base file
• Linguistic and visual stimuli finally presented are created by assembling minimal
components.
• Several controls exerted on the stimuli to prevent spurious effects. For instance, gender
and number are counterbalanced across experimental conditions. Similarly, words and
experimental conditions within the same set appear equally often.

Unpacking ERP Responses in Artificial Language Learning

• Linguistic and visual stimuli finally presented are created by assembling minimal components.
• Several controls exerted on the stimuli to prevent spurious effects. For instance, gender and
number are counterbalanced across experimental conditions. Similarly, words and experimental
conditions within the same set appear equally often.
• All stimuli compiled through R scripts

All scripts are run
from a core script.

• Present framework facilitates reproducibility and inspection of stimuli, and allows extensions
• Parallel lists of stimuli used to enable some of the controls
• Open-source software OpenSesame used to present the stimuli and collect responses

• Present framework facilitates reproducibility and inspection of stimuli, and allows extensions
• Parallel lists of stimuli used to enable some of the controls
• Open-source software OpenSesame used to present the stimuli and collect responses
• Reproducible, testable, reusable materials available at https://osf.io/wbjyr

Stimuli ahead
• Example of stimuli presented next include different mini-languages.
• Mini-languages were distributed between groups. Each participant
saw one mini-language only.
• Semantic information, particularly using pictures, helps in artificial
language learning. It boosts performance, reduces perceived effort
and increases enjoyment (Wendebourg et al., 2025).

• Introduced, trained on and tested on in Session 2
• Maintained across subsequent sessions
Gender agreement

Locative filler during training

1. Correct
2. Gender agreement violation
3. Number agreement violation
4. Gender and number agreement violation
5. Semantic violation (i.e., opposite adjective)
Test on gender agreement - Session 2
Match image to one of five sentences

1. Correct
2. Gender agreement violation
3. Number agreement violation
4. Gender and number agreement violation
5. Semantic violation (i.e., opposite adjective)
If accuracy < 80%, training and test are repeated.
Test on gender agreement - Session 2

Differential object marking
• Introduced, trained on and tested on in Session 3
• Maintained across subsequent sessions

Training in differential object marking

Test on differential object marking - Session 3
1. Correct
2. DOM violation (i.e., object noun without DOM)
3. Article with number violation
4. Misplaced article
5. Noun with semantic violation (i.e., noun different from image)

Training in verb-object number agreement

Test on verb-object number agreement

Examples of conditions of interest
• zer watch are yellowejur and zer pencil too Gender agreement violation
• Jessica chose fi je street and fi ze roof too. Correct differential object marking
• Thomas chose je key but not je backpack. Violation of differential object
marking (missing fi)
∅ ∅

Examples of control conditions
• zer watch are yellowezu and zer pencil too Number violation (missing r)
• Jessica provided fi zekey and fi zewall too Article misplacement
• Ole bestilte fi jevegg og fi zetak ogsaa. Article misplacement
∅

Lab sessions
• Instructions written in the language corresponding
to each artificial language group (e.g., in Norwegian
for the Mini-Norwegian group).
• Sufficient written instructions to avoid linguistic
influence through spoken explanations.

Preliminary results
~ 23 participants per mini-language group by Session 5
~ 11 participants per mini-language group by Session 6
- Data collection will continue over the next few months.

Accuracy of grammaticality judgements

Event-related potentials
To be presented next…
• Midline medial region: generally representative of other regions in
the current data.
• Per grammatical property
• Per session

P600-like effect for misplaced articles
• Control effect is informative.
• Data is sensitive enough to detect such a large effect,
suggesting data and preprocessing are sound.
• This effect can be compared with the effects of interest,
even though smaller effects are expected for the latter.

Analysis of ERPs for gender agreement
• Data filtered to grammatical and ungrammatical conditions only.
• Categorical predictors numerically recoded in logical ways.
• Session: 0, 1, 2, 3
• Grammaticality: -0.5, 0.5
• Dependent and independent variables z-scored (Brauer & Curtin, 2018)
• Trial-by-trial mixed-effects model incl. baseline (Alday, 2019).
• Per time window (200–500, 300–600, 400–900 ms).
• Per macro-region (lateral or midline electrodes).

Further analyses and hypotheses

+ 4 months: Session 6. Retest of all grammatical properties (Morgan-Short et al., 2012)
• Session ends with control tests on the relevant properties in the natural languages
Refresher on the Sessions

Executive functions and rs-EEG
• Primarily methodological analyses focussed on longitudinal stability.
• Analysed first as independent variables, and second as dependent variables.

Longitudinal role of executive functions and rs-EEG
• Independent-variable analysis: longitudinal stability (i.e., test-retest reliability) of the
variables (for similar analyses, see Fuhs et al., 2014; Samuels et al., 2016; Swanson, 2015).
• First analysis: longitudinal stability of the executive functions and rs-EEG on their own.
• Second analysis: longitudinal stability of the executive functions and rs-EEG as predictors of
language learning and morphosyntactic transfer over time.

Longitudinal effects on
executive functions and rs-EEG
• Dependent-variable analysis: whether cognitive enhancements in each executive function
(incl. rs-EEG) induced by language training are consistent with the baseline role of each
executive function (incl. rs-EEG).
• Cognitive enhancements analysed relative to the effect of each executive function (incl. rs-
EEG) as a predictor of language-learning performance following first exposure—i.e., in the
first test on the artificial language.
• Analysis intended to increase the methodological basis for the selection of measures to study
training-induced cognitive enhancements (see Grossmann et al., 2023; Kliesch et al., 2022;
Meltzer et al., 2023).
• Pre-post effects not analysed in absolute terms due to lack of control group (Sala & Gobet,
2017).

Language learning:
Analyses and hypotheses

Language learning
• Hypothesis 1: greater executive functions (overall, incl. rs-EEG) → better language learning
• Exploratory analysis: relative importance of the four executive function measures
• Hypotheses to be set a priori where possible
• Hypothesis 2: greater executive functions (overall, incl. rs-EEG) → greater longitudinal
improvements in language learning due to better accumulation of knowledge
• Hypothesis 3: longitudinal retests in the grammatical properties → better language learning

Morphosyntactic transfer:
Analyses and hypotheses

Standardised numeric predictions
• Effect sizes standardised between 0 and 10
• Visualisation of numerous comparisons across conditions and models
• Pave the way towards computational models with numeric predictions

Property Artificial lang. L2SFM CEM LPM TPM
L2 default prop. by prop. prop. by prop. full transfer
Gender
agreement
Mini-Norwegian 0 10 5-10 10
Mini-English 10 10 0-5 0-2
Differential
object marking
Mini-Norwegian 0-3 0-3 0-2 0-2
Mini-English 0-3 0-3 0-3 0-3
Verb-object
number agreem.
Mini-Norwegian 0-3 3 0-6 0-5
Mini-English 0-3 3 0-3 0-3
Standardised numeric predictions
• Effect sizes standardised between 0 and 10
• Visualisation of numerous comparisons across conditions and models
• Pave the way towards computational models with numeric predictions
Norway site predictions

Property Group Artificial lang. L2SFM CEM LPM TPM
L2 default prop. by prop. prop. by prop. full transfer
Gender
agreement
L1 Eng, L2 Spa
Mini-Spanish 10? 10 5-10 10
Mini-English 10? 10 5-7 0
L1 Spa, L2 Eng
Mini-Spanish 0 10 5-10 10
Mini-English 0 10 0-5 0
Differential
object
marking
L1 Eng, L2 Spa
Mini-Spanish 8? 8 5-8 8
Mini-English 8? 8 0-5 0-2
L1 Spa, L2 Eng
Mini-Spanish 0 8 5-8 8
Mini-English 0 8 0-5 0-2
Verb-object
number
agreement
L1 Eng, L2 Spa
Mini-Spanish 0-5? 0-3 0-4 0-3
Mini-English 0-5? 0-3 0-2 0-2
L1 Spa, L2 Eng
Mini-Spanish 0-3 0-3 0-4 0-3
Mini-English 0-3 0-3 0-3 0-2
Spain site predictions

Executive functions and morphosyntactic transfer
• Hypothesis 1: greater executive functions (overall, incl. rs-EEG) → greater precursors of
transfer (N2, P3)
• Hypothesis 2: greater executive functions (overall , incl. rs-EEG) → greater signatures of
transfer (P600)

Longitudinal effects in morphosyntactic transfer
• Hypothesis 1: any signatures of transfer (esp. P600) more likely to occur in later sessions.
• Hypothesis 2: any signatures of transfer (esp. P600) should persist in subsequent sessions.

Thank you
Questions and feedback very welcome.

References (1/5)
Alday, P. M. (2019). How much baseline correction do we need in ERP research? Extended GLM model can replace baseline
correction while lifting its limits. Psychophysiology, 56(12), e13451. https://doi.org/10.1111/psyp.13451
Bardel, C., & Falk, Y. (2012). The L2 status factor and the declarative/procedural distinction. In J. Cabrelli, S. Flynn, & J.
Rothman (Eds.), Third Language Acquisition in Adulthood (pp. 61–78). John Benjamins Publishing Company.
https://doi.org/10.1075/sibil.46.06bar
Desbordes, T., Lakretz, Y., Chanoine, V., Oquab, M., Badier, J.-M., Trébuchon, A., Carron, R., Bénar, C.-G., Dehaene, S., & King,
J.-R. (2023). Dimensionality and ramping: Signatures of sentence integration in the dynamics of brains and deep language
models. Journal of Neuroscience, 43(29), 5350–5364. https://doi.org/10.1523/JNEUROSCI.1163-22.2023
Flynn, S., Foley, C., & Vinnitskaya, I. (2004). The cumulative-enhancement model for language acquisition: Comparing adults’
and children’s patterns of development in first, second and third language acquisition of relative clauses. International
Journal of Multilingualism, 1(1), 3–16. https://doi.org/10.1080/14790710408668175
Friederici, A. D., Steinhauer, K., & Pfeifer, E. (2002). Brain signatures of artificial language processing: Evidence challenging the
critical period hypothesis. Proceedings of the National Academy of Sciences, 99(1), 529–534.
https://doi.org/10.1073/pnas.012611199
Fuhs, M. W., Nesbitt, K. T., Farran, D. C., & Dong, N. (2014). Longitudinal associations between executive functioning and
academic skills across content areas. Developmental Psychology, 50(6), 1698–1709. https://doi.org/10.1037/a0036633

References (2/5)
González Alonso, J., Bernabeu, P., Silva, G., DeLuca, V., Poch, C., Ivanova, I., & Rothman, J. (2025). Starting from the very
beginning: Unraveling Third Language (L3) Development with Longitudinal Data from Artificial Language Learning and
EEG. International Journal of Multilingualism, 22(1), 119–142. https://doi.org/10.1080/14790718.2024.2415993
Grossmann, J. A., Aschenbrenner, S., Teichmann, B., & Meyer, P. (2023). Foreign language learning can improve response
inhibition in individuals with lower baseline cognition: Results from a randomized controlled superiority trial. Frontiers in
Aging Neuroscience, 15. https://doi.org/10.3389/fnagi.2023.1123185
Hudson Kam, C. L., & Newport, E. L. (2005). Regularizing unpredictable variation: The roles of adult and child learners in
language formation and change. Language Learning and Development, 1(2), 151–195.
https://doi.org/10.1080/15475441.2005.9684215
Just, M. A., & Carpenter, P. A. (1980). A theory of reading: From eye fixations to comprehension. Psychological Review, 87(4),
329–354. https://doi.org/10.1037/0033-295X.87.4.329
Kliesch, M., Pfenninger, S. E., Wieling, M., Stark, E., & Meyer, M. (2022). Cognitive benefits of learning additional languages in
old adulthood? Insights from an intensive longitudinal intervention study. Applied Linguistics, 43(4), 653–676.
https://doi.org/10.1093/applin/amab077
Li, P., Zhang, F., Yu, A., & Zhao, X. (2020). Language History Questionnaire (LHQ3): An enhanced tool for assessing multilingual
experience. Bilingualism: Language and Cognition, 23(5), 938–944. https://doi.org/10.1017/S1366728918001153

References (3/5)
Meltzer, J. A., Kates Rose, M., Le, A. Y., Spencer, K. A., Goldstein, L., Gubanova, A., Lai, A. C., Yossofzai, M., Armstrong, S. E. M.,
& Bialystok, E. (2023). Improvement in executive function for older adults through smartphone apps: A randomized
clinical trial comparing language learning and brain training. Aging, Neuropsychology, and Cognition, 30(2), 150–171.
Meister, C., Pimentel, T., Clark, T., Cotterell, R., & Levy, R. (2022). Analyzing wrap-up effects through an information-theoretic
lens. In S. Muresan, P. Nakov, & A. Villavicencio (Eds.), Proceedings of the 60th Annual Meeting of the Association for
Computational Linguistics (Volume 2: Short Papers) (pp. 20–28). Association for Computational Linguistics.
https://doi.org/10.18653/v1/2022.acl-short.3
Monaghan, P., Donnelly, S., Alcock, K., Bidgood, A., Cain, K., Durrant, S., Frost, R. L. A., Jago, L. S., Peter, M. S., Pine, J. M.,
Turnbull, H., & Rowland, C. F. (2023). Learning to generalise but not segment an artificial language at 17 months predicts
children’s language skills 3 years later. Cognitive Psychology, 147, 101607.
https://doi.org/10.1016/j.cogpsych.2023.101607
Morgan-Short, K., Finger, I., Grey, S., & Ullman, M. T. (2012). Second language processing shows increased native-like neural
responses after months of no exposure. PLOS ONE, 7(3), e32974. https://doi.org/10.1371/journal.pone.0032974
Pereira Soares, S. M., Kupisch, T., & Rothman, J. (2022). Testing potential transfer effects in heritage and adult L2 bilinguals
acquiring a mini grammar as an additional language: An ERP approach. Brain Sciences, 12(5), 669.
https://doi.org/10.3390/brainsci12050669

References (4/5)
Rothman, J. (2011). L3 syntactic transfer selectivity and typological determinacy: The typological primacy model. Second
Language Research, 27(1), 107–127. https://doi.org/10.1177/0267658310386439
Rothman, J., Alemán Bañón, J., & González Alonso, J. (2015). Neurolinguistic measures of typological effects in multilingual
transfer: Introducing an ERP methodology. Frontiers in Psychology, 6. https://doi.org/10.3389/fpsyg.2015.01087
Sala, G., & Gobet, F. (2017). Does far transfer exist? Negative evidence from chess, music, and working memory training.
Current Directions in Psychological Science, 26(6), 515–520. https://doi.org/10.1177/0963721417712760
Samuels, W. E., Tournaki, N., Blackman, S., & Zilinski, C. (2016). Executive functioning predicts academic achievement in
middle school: A four-year longitudinal study. The Journal of Educational Research, 109(5), 478–490.
Stowe, L. A., Kaan, E., Sabourin, L., & Taylor, R. C. (2018). The sentence wrap-up dogma. Cognition, 176, 232–247.
https://doi.org/10.1016/j.cognition.2018.03.011
Swanson, H. L. (2015). Growth in working memory and inhibition predicts literacy in English language learners: A cross-
sectional and longitudinal study. Memory, 23(5), 748–773. https://doi.org/10.1080/09658211.2014.927504
Uddén, J., & Männel, C. (2018). Artificial grammar learning and its neurobiology in relation to language processing and
development. In S.-A. Rueschemeyer & M. G. Gaskell (Eds.), The Oxford Handbook of Psycholinguistics (p. 0). Oxford
University Press. https://doi.org/10.1093/oxfordhb/9780198786825.013.33

Wendebourg, K., Öttl, B., Meurers, D., & Kaup, B. (2025). Semantic information boosts the acquisition of a novel grammatical
system in different presentation formats. Language and Cognition, 17, e30. https://doi.org/10.1017/langcog.2023.47
Westergaard, M., Mitrofanova, N., Mykhaylyk, R., & Rodina, Y. (2017). Crosslinguistic influence in the acquisition of a third
language: The Linguistic Proximity Model. International Journal of Bilingualism, 21(6), 666–682.
https://doi.org/10.1177/1367006916648859
References (5/5)

Unpacking ERP Responses in Artificial Language Learning

More Related Content

Similar to Unpacking ERP Responses in Artificial Language Learning (20)

Recently uploaded (20)

Unpacking ERP Responses in Artificial Language Learning