class: center, middle, inverse, title-slide # L3 French Voice Onset Time at first exposure by Spanish-English bilinguals ## CASPSLaP, February 2022 ### Kyle Parrish ### Rutgers University --- <style type="text/css"> .remark-slide-content { font-size: 20px; padding: 20px 80px 20px 80px; } .remark-code, .remark-inline-code { background: #f0f0f0; } .remark-code { font-size: 24px; } .huge .remark-code { /*Change made here*/ font-size: 200% !important; } .tiny .remark-code { /*Change made here*/ font-size: 70% !important; } .med .remark-code { /*Change made here*/ font-size: 120% !important; } </style> # Overview .pull-left[ .big[ - Introduction - Background - The Present study - Conclusion ] ] --- class: center, middle background-color: black # Are L2 acquisition and L3 acquisition the same? --- # Introduction .big[ - If L2 and L3 learning are the same, there should be **no difference between monolingual and bilingual participants** when learning a new language. ] -- .big[ - Specifically, the **L2 should not impact L3 learning**. ] -- .big[ - The present study investigates this assertion by examining the production of Voice Onset Time in **bilinguals** and **monolinguals** in a language they do not yet know. ] --- # Introduction - L3 research suggests that L3 and L2 acquisition are distinct processes and investigates what factors impact how the L1 and/or L2 impact the acquisition of a new language. -- .pull-left[ .full-width[ .content-box-red[ - Some accounts suggest that **only one source language** plays a primary role in L3A. - **Typological Primacy Model** Closer psycho-typological language holistically influences the L3 (Rothman, 2015). - **The L2 Status Factor** The L2 has a special status and blocks access to the L1 (Bardel & Falk, 2007). ]]] -- .pull-right[ .full-width[ .content-box-red[ - Other accounts suggest that **both the L1 and L2** influence the L3. - **The Cumulative Enhancement Model** L3 learning is a cumulative process (Flynn et al., 2004). - **The Linguistic Proximity Model/Scalpel Model** L3 learning occurs on a structure-by-structure basis and is influenced by underlying structural similarity between either L1 or L2 and L3 (Westergaard et al. 2017; Slabakova, 2017). ]]] --- # L3 phonology - Work in L3 phonology is more scarce, but has seen similarly mixed results (L1 influence, L2 status, hybrid) -- .pull-left[ .full-width[ .content-box-red[ **Previous Studies in L3 VOT** - L3 VOT studies have been mixed - **L2 influence** (Llama et al., 2010; Tremblay, 2007; Wunder, 2010) - **Intermediate values** (Wrembel, 2011, 2014, 2015; Llama & Cardoso, 2018) - L2 influence has been reported in global accent ratings (Wrembel, 2010; Williams & Hammarberg, 1993) - No clear explanation for mixed findings. ]]] -- .pull-right[ .full-width[ .content-box-blue[ **My contribution** - **New language order** - Spanish L1, English L2, French "L3". - **First exposure** rather than more advanced stages. ]]] --- # The Present Study .big[ - The present study also was designed to examine the predictions of the models. ] -- .big[ .pull-left[ .content-box-red[ **RQ1**: Will bilingual participants produce words in a new language closer to their own productions in their first or second language? **RQ2**: Will differences in novel-language word production between bilingual and monolingual groups be seen? ]]] -- .big[ .pull-right[ .content-box-blue[ **H1**: L3 production will resemble L2 production (longer VOT) **H2**: Monolingual French production will resemble Spanish and Bilingual French production will resemble English ]]] --- # The Present Study .big[ .pull-left[ **Spanish L1 - English L2** speakers and **Spanish monolinguals** shadowed words in **French**, an unknown language. Spanish and English were also produced for comparison. ]] -- .pull-right[ .big[ **Voice onset time** of these productions was measured and compared. ]] --- class: center, middle # VOT cross-linguistic comparison .large[ English ≠ Spanish = French ] --- # Participants .big[ .pull-left[ 2 groups were recruited online: **Bilingual group**: 39 L1 Mexican Spanish - L2 English speakers. **Monolingual group**: 18 L1 Mexican Spanish monolingual speakers. ]] -- .pull-right[ .content-box-grey[ - **Location:** All participants were born and lived in Mexico at the time of the study. - **L2 Proficiency:** Self-rated (1-7 scale) + phonetic corroboration (Spanish-English t-test) - **Dominance and use:** Filtering location was intended to find Spanish-dominant speakers - All participants reported not speaking or having studied a third language. ]] -- |Factor | Mean| SD| |:-----------|-----:|----:| |AoA | 11.90| 4.07| |Current age | 22.74| 3.43| |Proficiency | 5.26| 0.94| -- .footnote[ 37 participants did not produce distinct L1 and L2 sounds are their data was removed ] --- # Materials .big[ - L1 and L2 production - **Word reading task** ] -- .big[ - L3 production - **Shadowing task** ] -- .big[ - Proficiency and use - **Background questionnaire** ] --- # Procedure .content-box-red[ .big[ - All **data collection took place online** using a Pscyhopy experiment via Pavlovia.org - All tasks were given in a **single session**. - Task order was **counter-balanced**, and stimulus order within-tasks was **randomized**. ]] --- # Word Reading task .pull-left[ .content-box-blue[ .big[ - Spanish and English words were produced in isolation. - /p/, /t/ or /k/ initial words. - One or two syllables. ] ]] -- .pull-right[ .content-box-grey[ - **Word lists** - *Spanish:* tiro, tema, talla, quiso, queja, cama, piso, pena, pato - *English:* tipping, teller, tacky, penny , pass, parrot, kitten, kennel, cabbage ]] .footnote[ Word lists were adapted from the lists used in Llama and Cardoso (2018) ] --- background-image: url(./images/ept_example.png) background-size: contain --- # Shadowing Task .pull-left[ .content-box-grey[ .large[ - A French native speaker recorded each word - Participants repeated each word - *French:* tir, terre, tasse, quitte, quelle, pile, pere, patte ]]] -- .pull-right[ .large[ - **Tokens for shadowing task** - Produced by a native French speaker - Mean relative VOT = 0.064, sd = 0.042 ]] --- background-image: url(./images/shadowing_example.png) background-size: contain --- # Data segmentation procedure .big[ - Auto segmented in WebMaus and hand corrected for VOT ] -- .big[ - VOT was marked at the closest zero-crossing to the onset of visible aspiration and the zero-crossing before before the onset of the vowel. ] -- .big[ - Values were extracted from PRAAT textgrids using the `read_textgrid` function in R. ] --- background-image: url(./images/segmentation.png) background-size: contain --- # Statistical Analysis .large[ - All analyses were carried out in `R` ] -- .large[ - **Sample size justification**: Power analysis of pilot data ] -- .large[ - **Inclusion and subsetting**: T.tests ] -- .large[ - **Quantify the effect of language on VOT production**: Linear Mixed Effects Model `relative vot ~ language*group + (1 | word) + (1 | participant)` ] -- .large[ - **Determine equivalence**: Test of Equivalence (Lakens, 2017) ] --- class: center, middle # Results --- background-image: url(./images/comb.png) background-size: contain --- # Averages per segment full dataset .pull-left[ **Mean relative VOT per segment in bilinguals** |language |text | Relative VOT| SD| |:--------|:----|------------:|-----:| |english |k | 0.160| 0.055| |french |k | 0.139| 0.053| |spanish |k | 0.085| 0.034| |language |text | Relative VOT| SD| |:--------|:----|------------:|-----:| |english |t | 0.134| 0.046| |french |t | 0.073| 0.047| |spanish |t | 0.047| 0.022| |language |text | Relative VOT| SD| |:--------|:----|------------:|-----:| |english |p | 0.109| 0.058| |french |p | 0.057| 0.038| |spanish |p | 0.041| 0.020| ] -- .pull-right[ **Mean relative VOT per segment in monolinguals** |language |text | Relative VOT| SD| |:--------|:----|------------:|-----:| |french |t | 0.061| 0.029| |spanish |t | 0.051| 0.022| |language |text | Relative VOT| SD| |:--------|:----|------------:|-----:| |french |p | 0.044| 0.020| |spanish |p | 0.040| 0.015| ] --- # Linear Mixed Effects Model .pull-left[ This model was run to determine how language predicted VOT. `relative vot ~ language*group + (1 | word) + (1 | participant)` - Outcome was *relative VOT* (vot/total word duration). - Fixed effect predictors: *Language* (Spanish, English, French), *group* (monolingual or bilingual) and their interaction. - Random intercepts: *Word* and *participant*. ] -- .pull-right[ .tiny[ <table style="border-collapse:collapse; border:none;"> <tr> <th style="border-top: double; text-align:center; font-style:normal; font-weight:bold; padding:0.2cm; text-align:left; ">&nbsp;</th> <th colspan="3" style="border-top: double; text-align:center; font-style:normal; font-weight:bold; padding:0.2cm; ">relative vot</th> </tr> <tr> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; text-align:left; ">Predictors</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">Estimates</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">CI</td> <td style=" text-align:center; border-bottom:1px solid; font-style:italic; font-weight:normal; ">p</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">(Intercept)</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.13</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.13&nbsp;&ndash;&nbsp;0.14</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong>&lt;0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">group [monolingual]</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.00</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.01&nbsp;&ndash;&nbsp;0.01</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">0.994</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">language [french]</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.04</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.05&nbsp;&ndash;&nbsp;-0.04</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong>&lt;0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">language [spanish]</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.08</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.08&nbsp;&ndash;&nbsp;-0.07</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong>&lt;0.001</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; ">group [monolingual] *<br>language [french]</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.02</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; ">&#45;0.03&nbsp;&ndash;&nbsp;-0.01</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:center; "><strong>0.001</strong></td> </tr> <tr> <td colspan="4" style="font-weight:bold; text-align:left; padding-top:.8em;">Random Effects</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm;">&sigma;²</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left;" colspan="3">0.00</td> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm;">&tau;₀₀ _participant</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left;" colspan="3">0.00</td> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm;">ICC</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left;" colspan="3">0.16</td> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm;">N _participant</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left;" colspan="3">57</td> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm; border-top:1px solid;">Observations</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left; border-top:1px solid;" colspan="3">1248</td> </tr> <tr> <td style=" padding:0.2cm; text-align:left; vertical-align:top; text-align:left; padding-top:0.1cm; padding-bottom:0.1cm;">Marginal R² / Conditional R²</td> <td style=" padding:0.2cm; text-align:left; vertical-align:top; padding-top:0.1cm; padding-bottom:0.1cm; text-align:left;" colspan="3">0.289 / 0.404</td> </tr> </table> ]] --- # English L2 - French L3 TOST .pull-left[ <!-- --> ``` ## TOST results: ## t-value lower bound: 5.13 p-value lower bound: 0.000001 ## t-value upper bound: 1.60 p-value upper bound: 0.943 ## degrees of freedom : 75.72 ## ## Equivalence bounds (Cohen's d): ## low eqbound: -0.4 ## high eqbound: 0.4 ## ## Equivalence bounds (raw scores): ## low eqbound: -0.0234 ## high eqbound: 0.0234 ## ## TOST confidence interval: ## lower bound 90% CI: 0.022 ## upper bound 90% CI: 0.067 ## ## NHST confidence interval: ## lower bound 95% CI: 0.018 ## upper bound 95% CI: 0.071 ## ## Equivalence Test Result: ## The equivalence test was non-significant, t(75.72) = 1.595, p = 0.943, given equivalence bounds of -0.0234 and 0.0234 (on a raw scale) and an alpha of 0.05. ## Null Hypothesis Test Result: ## The null hypothesis test was significant, t(75.72) = 3.361, p = 0.00122, given an alpha of 0.05. ## Based on the equivalence test and the null-hypothesis test combined, we can conclude that the observed effect is statistically different from zero and statistically not equivalent to zero. ``` ] --- # Interim Discussion .large[ .pull-left[ - During hand-correction, I noticed many participants produced **target-like** VOT in L3 French, while others **produced long-lag stops**. - These two trends suggest that the overall observed group trends are the culmination of two specific trends, rather than most participants producing intermediate values. - As a result, the full bilingual dataset was subset post-hoc into **L1-influence** and **L2-influence** subsets. ] ] -- .large[ .pull-right[ - **The data were split based on the results of a t-test per participant** that compared that person's **L3 productions** compared to their **L2 English productions**. - **L1-influence subset**: L2 English - L3 French t-test (p < .05) - **L2-influence subset**: L2 English - L3 French (p > .05) ] ] --- class: center, middle # L1 subset results --- background-image: url(./images/l1_subset.png) background-size: contain --- # L1 subset: French-Spanish TOST .pull-left[ <!-- --> ``` ## TOST results: ## t-value lower bound: 1.77 p-value lower bound: 0.044 ## t-value upper bound: -0.496 p-value upper bound: 0.312 ## degrees of freedom : 26.6 ## ## Equivalence bounds (Cohen's d): ## low eqbound: -0.4 ## high eqbound: 0.4 ## ## Equivalence bounds (raw scores): ## low eqbound: -0.0159 ## high eqbound: 0.0159 ## ## TOST confidence interval: ## lower bound 90% CI: -0.015 ## upper bound 90% CI: 0.033 ## ## NHST confidence interval: ## lower bound 95% CI: -0.02 ## upper bound 95% CI: 0.038 ## ## Equivalence Test Result: ## The equivalence test was non-significant, t(26.6) = -0.496, p = 0.312, given equivalence bounds of -0.0159 and 0.0159 (on a raw scale) and an alpha of 0.05. ## Null Hypothesis Test Result: ## The null hypothesis test was non-significant, t(26.6) = 0.635, p = 0.531, given an alpha of 0.05. ## Based on the equivalence test and the null-hypothesis test combined, we can conclude that the observed effect is statistically not different from zero and statistically not equivalent to zero. ``` ] --- # Individual Differences - L1 subset <!-- --> --- class: center, middle # L2 subset results --- background-image: url(./images/l2_subset.png) background-size: contain --- # L2 subset: French-English TOST .pull-left[ <!-- --> ``` ## TOST results: ## t-value lower bound: -0.0271 p-value lower bound: 0.511 ## t-value upper bound: -2.74 p-value upper bound: 0.004 ## degrees of freedom : 43.29 ## ## Equivalence bounds (Cohen's d): ## low eqbound: -0.4 ## high eqbound: 0.4 ## ## Equivalence bounds (raw scores): ## low eqbound: -0.024 ## high eqbound: 0.024 ## ## TOST confidence interval: ## lower bound 90% CI: -0.054 ## upper bound 90% CI: 0.005 ## ## NHST confidence interval: ## lower bound 95% CI: -0.06 ## upper bound 95% CI: 0.011 ## ## Equivalence Test Result: ## The equivalence test was non-significant, t(43.29) = -0.0271, p = 0.511, given equivalence bounds of -0.024 and 0.024 (on a raw scale) and an alpha of 0.05. ## Null Hypothesis Test Result: ## The null hypothesis test was non-significant, t(43.29) = -1.384, p = 0.174, given an alpha of 0.05. ## Based on the equivalence test and the null-hypothesis test combined, we can conclude that the observed effect is statistically not different from zero and statistically not equivalent to zero. ``` ] --- # Individual differences - L2 subset <!-- --> --- class: center, middle # Discussion and Conclusion --- # Discussion and Conclusions .big[ .pull-left[ .content-box-red[ **RQ1**: Will bilingual participants produce words in a new language closer to their own productions in their first or second language? **RQ2**: Will differences in novel-language word production between bilingual and monolingual groups be seen? ]]] -- .big[ .pull-right[ .content-box-blue[ **H1**: L3 production will resemble L2 production (longer VOT) **H2**: Monolingual French production will resemble Spanish and Bilingual French production will resemble English ]]] --- # Discussion and Conclusions .pull-left[ .full-width[ .content-box-red[ **Is L3A another case of L2A?** - Monolinguals and bilinguals were different. - If participants are trilinguals performing a task in a new or learned language, one should keep in mind that the CLI from both previously learned languages might be a possible influence. **Does the L1 or L2 impact the L3?** - Both seem to at the group level. - Individual trends exist in the data. ]]] -- .pull-right[ .full-width[ .content-box-red[ **Conclusion and Future directions** - Overall, L3 French fell between L1 Spanish and L2 English relative VOT values. - There was variation in the results - My research contributes to L3A in that it shows that individual differences may play an important role is predicting CLI - Future and ongoing projects aim to explore which factors might be associated with this variation - PSTM - Mood/energy ]]] --- class: center, middle # Thank you!