Emotion Word Processing in Immersed Spanish-English/English-Spanish Bilinguals: An ERP Study

Round 1

Reviewer 1 Report (Previous Reviewer 1)

I would like to thank the authors for the immense work they have put into the revision of the paper, as it has been essentially re-written. In my opinion, the manuscript benefitted from the changes, but, unfortunately, I still have serious concerns regarding the statistical analyses and interpretation of the data that, of course, affects the results and discussion section.

Major points:
Let me start with the fact, that I don’t think the authors have paid enough attention to their behavioural results. I think these results themselves could make a good paper. The language-by-emotion interaction showing faster processing of positive vs. negative and neutral targets only in Spanish, the less dominant L1 for the majority of the bilinguals, is an exciting finding. Looking at the descriptive statistics it seems that negative Spanish targets led to the longest RTs and a slowdown of word processing, as it would be predicted by studies showing greater emotional resonance specifically for negative information in L1 but not in L2 (e.g. Iacozza et al. 2017, Wu and Thierry 2012, Sheikh and Titone 2016, Jończyk et al. 2016, 2019, Toivo and Scheepers 2019, etc.). I think that these results should be given more prominence as they seem to be much more interesting that ERP results, where the authors find basically language effects. This effect is even more interesting in light of the fact that the authors used a mixed design, which tends to attenuate if not abolish any language-related emotion effects, which I think is a huge issue for ERP data.
This takes me to my second point. The major part of the discussion is built upon results that are insignificant and thus should not be treated as such. Specifically, I relate to the following ERP findings:

(1) Early EPN:
“The main effect of valence fell just short of statistical significance, F(2,48) = 1.98, p =0.1” – I tend to disagree. This effect is just insignificant.  Hence, the next sentence referring to a comparison between EPN amplitudes to negative vs. neutral targets should not be included in the results and should not be further discussed. Also, the authors did not make specific planned comparisons regarding negative vs. neutral word pairs; rather, they planned to dive into the emotion effect, which includes a comparison between negative vs. neutral and positive vs. neutral. Such a comparison, when corrected for multiple comparisons, would end up being insignificant, which is why the main effect on the ANOVA is insignificant as well.

Also, the following sentence “An enhanced early EPN response to negative relative to neutral targets seemed largely driven by the results recorded for Spanish emotion words (negative: M = -1.45, CI [-4.06, 1.15] vs. neutral: M = 1.09, CI [-0.66, 2.84]), with English negative (M = -1.25, CI [-2.34, -0.17]) and neutral (M = -1.46, CI [-3.15, 0.24]) words eliciting comparable amplitudes (see Figure 2).” seems to be based solely on descriptive statistics, as no inferential statistics has been included to confirm this comparison. The critical difference cited here between Spanish negative and English negative EPN amplitude amounts to 0.20 microvolt, with comparable variance, which is far too small a difference to be meaningful. If any, the numerical difference is greater for neutral words across the two languages.

(2) Late EPN:
“In the planned comparisons analyses, we found a significant modulation of the EPN 200-300 ms amplitude as a function of AoA. Early AoA English learners showed more pronounced EPN amplitudes in response to English negative words (M = -1.92 µV, 95% CI [-3.31, -0.52]) compared to late AoA English learners (M = -0.41 µV, 95% CI [-1.99, 1.16]) t(19.9) = 1.5, p = 0.06, Cohen’s d = 0.6).” – the same argumentation applies here, and even more so given the low power of this comparison (as AoA is taken into consideration). There is no statistical support for running post-hocs to language-by-emotion interaction, given that the interaction is highly insignificant. As stated above, the planned comparisons here would need to include comparing negative vs. neutral and positive vs. neutral for both early and late AoA groups. Such a test would need to be corrected for multiple comparisons.

In essence, what I am trying to point out is that the authors have a very interesting behavioural effect and a relatively strong language effect on the ERPs, with no modulation by valence and no main effect of valence (as confirmed by inferential statistics). As a result, the majority of the discussion section is, unfortunately, unsupported by the data, because it is based on the emotion effects on ERPs that did not survive the statistical comparison.
Finally, I also think that the reversed LPC language effect is interesting and suggests a greater reanalysis of L1 Spanish words relative to L2 English words, which seems like a nice index of participants’ dominance in L2 English.

Other points:
“Concreteness was blocked, such that for each language, half the words were concrete and half were abstract.” – I think “blocked’ might be confusing as it suggests that you presented concrete and abstract words in separate experimental blocks, unless this is what you did? If not, maybe go for “matched”, “balanced”, or “controlled for”

“For the AoA of English, we had more compatible size groups in early (14) and late (12) bilingual categories. We therefore used planned comparisons (independent samples t-tests) to test the following prediction: AoA early English bilinguals should have a more pronounced EPN/LPC emotion effect for English emotion words than AoA late English bilinguals. While we are aware of the unequal sample size and the consequent low generalizability/power of this comparison analysis, we still wanted to observe a general trend. Where appropriate, to correct for unequal samples in planned comparisons, the unpaired t-test for unequal variance (the Welch’s test) is reported.” – this is tricky. With 14 vs 12 participants on EEG you get incredibly low power and a lot of variance. While I understand that you want to investigate a general trend, the results of this comparison are bound to be misleading, possibly resulting in type 1 error. I would recommend the authors to run a power analysis based on the collected data and based on the results conduct a separate experiment with the number of participants per each AoA group as indicated by the power simulation – provided that looking at AoA is of critical interest to the authors. For this manuscript, however, I would just say that the bilingual group varied on the AoA, with a relatively equal balance (half/half).

GRAPHS:

-    for behavioural results, it would be much more reader-friendly to have line graphs with error bars representing 95% CI. Note, that line graphs are also preferred for visualizing LMEMs. The authors may also want to use the sjPLOT package in R that produces very nice fixed term and interaction term plots for LMEMs.

-    For ERP plots, please plot a linear derivation of the electrodes of interest (rather than individual electrodes; you can place individual electrode plots in supplementary material) and – critically- present the full epoch for EPN and LPC (currently the plots are cut in half, with x axis running up to 500 ms for EPN and LPC starting from 500 ms)

-    one can see from the ERP plots that there is a lot of noise on some electrodes (e.g., alpha on O1, Oz; baseline issues on PO3, CP5, PO4, P3, P7 with significant amplitude changes already around 50ms post-stimulus onset or even in the pre-stimulus period).

In short, I still think that this study would contribute to the existing literature on language-emotion interactions in bilinguals. However, it just seems that the ERP data does not contribute much to this discussion due to low power and decent amount of noise (as judged from the ERP plots). I therefore think that the authors should reframe their discussion to focus more on emotion effects in the behavioural data and language effects (and lack of emotion effects) on the ERP data – possibly driven by the mixed design that would be expected to attenuate cross-language emotion effects.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report (New Reviewer)

The paper “Emotion word processing in immersed Spanish-English/English-Spanish bilinguals: An ERP study” is one of the current investigations about how emotions are processed by bilinguals in L1 and L2, and it contributes to the field when it chooses to analyze participants who are not only highly proficient in their L2, but for whom L2 often becomes their dominant language. To pursue this goal, it was conducted a lexical decision task to measure Spanish-English/English-Spanish bilinguals’ behavioral (RT) and electrophysiological (EPN, Early Posterior Negativity and LPC, Late Pos-itive Complex). The content is well described and contextualized with respect to previous and present theoretical background and empirical research on the topic, also all the cited references are relevant to the research.  The research design, questions, hypotheses and methods are clearly stated. Stimuli, procedures and data analyses are well described and lean on to the paper's argumentation. But, on methods, it is missing an objective proficiency evaluation to assure the high proficiency participant’s profile. Therefore, it lacks a discussion about the implications that the use of self-rating proficiency can have on the data and its following discussion. Hulstijn (2012) showed that more than 50% of papers published in the journal of Bilingualism: Language and Cognition used self-assessment as the main or only proficiency measure, but the studies used the self-assessment ratings as the independent variable, and the variability in those studies rarely could be explained in terms of proficiency. Hence, the data could not be explained in terms of the independent variable. Gollan et al. (2012) and Souza et al. (2015) show that self-assessment is not reliable when compared to objective proficiency measures. Gollan et al. (2012) shows that bilinguals are accurate at indicating which of their languages are the dominant one and in which contexts. However, bilinguals are not accurate when reporting the extension of their knowledge in their languages. The study also shows a weak correlation between subjective versus objective measures of proficiency. Souza et al. (2015) shows that there is internal consistency between objective measures of proficiency but not on subjective measures. This scenario indicates that it is problematic to use self-assessment as a reliable measure and it could be an effect in your data that should be discussed properly. It is not the case to invalidate the experiment or the found data, but this discussion is necessary and an objective test, such as a VLT (Nation, 1990), could be included on analyses. 

According to the paper's format, some text parts are in red, and it needs to be checked. Also there are some double semicolons and commas in text.

 

References

Hulstijn, Jan H. The construct of language proficiency in the study of bilingualism from a cognitive perspective. Bilingualism: Language and Cognition 15 (2). Cambridge University Press, p.422-433, 2012.

Gollan, T. H., Weissberger, G. H., Runnqvist, E., Montoya, R. I., & Cera, C. M. (2012). Self-ratings of Spoken Language Dominance: A Multi-Lingual Naming Test (MINT) and Preliminary Norms for Young and Aging Spanish-English Bilinguals. Bilingualism (Cambridge, England), 15(3), 594–615.

Nation, I. S. (1990). Teaching and Learning Vocabulary. New York: Heinle and Heinle.

 

SOUZA, R. A.; SILVA, J. S. . Exploring the measurement of vocabulary size to differentiate Brazilian Portuguese-English bilinguals? access to grammatical knowledge in the L2. Revista Linguística (Online), v. 11, 2015, p. 187-214.

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

This is a thorough and /or thoroughly revised paper on emotion and bilingual in Spanish and English. I am a Spanish/ English bilingual linguist and I read with interest.

I have noted a couple of typos with a comment but not more.

The bibliography is extensive however the remainder is not so original . Spanish/ English bilingual studies a high in number, the actual  so what factor is missing right till the end. There is no mention made or argument put forward as to who might be interested in this work and the purpose it serves. So what ? is the final conclusion after let us say plowing through infinite amount of  interesting but difficult to digest quantitative data. For this reason I have rated it average interest for the readership of Languages journal. From the humble opinion of this reviewer it is ' interesting' but there is no argument made as to its purpose or who would benefit from reading it, what the implications are or where further research on this topic would lead.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report (Previous Reviewer 1)

I would like to thank the authors for addressing my comments. I have no further suggestions except a small spelling mistake (I guess) in the following sentence:

Exploring L1 and L1 emotion word processing in such bilinguals might help throw new light on the interplay 288 of the various participant characteristics in modulating the emotion effect.  -> I think it should say L1 and L2 emotion word processing?

Author Response

We would like to once again express our deepest gratitude to the Reviewer for the very detailed guidance and feedback which has helped us immensely in gaining a new direction from the original draft and allowed improving the quality of the manuscript.

We thank the Reviewer for catching the spelling mistake in line 288. We have corrected it.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

This study addresses important and so far largely overlooked factors that may affect emotion word processing in bilingualism (i.e. the age of acquisition, language status, and language dominance) using behavioural and electrophysiological measures. To my knowledge, this is the first attempt to directly investigate the role of the abovementioned factors on bilingual emotion word processing and as such the study may constitute a relevant contribution to the field. Although I am overall enthusiastic about the overarching goal of the study and the statistical method used for data analysis, I think the study is largely underpowered for its design (2x3x2x2x3x2). Further, I have major reservations towards how the data (particularly EEG data) were analysed and/or preprocessed. These issues do not allow me to recommend this paper for publication in its present form. I provide my comments/suggestions for revision below.

 

Major comments

EEG recordings

- "EEG data was filtered offline with a band pass filter .10 Hz using MATLAB software (MATLAB R2018b)." - this is confusing. If bandpass filter was used, the authors should specify the parameters for highpass and lowpass. Also filter name, type and order should be reported (follow the recommendations by Keil et al. 2014, 10.1111/psyp.12147)

- "Ocular artifacts were corrected using the Independent Component Analysis (ICA) with Matlab" – the authors should report descriptive statistics (mean and SD) of the rejected ICs and provide a reference to ICA (Makeig, Bell, Jung, & Sejnowski , 1995).

- provide the mean number of epochs per condition that were included in the final analysis.

- note that it is recommended to use [-200 ms] rather than [-100 ms] baseline period for ERPs. Also, currently, there is no information about baseline removal.

- the manuscript does not contain any ERP plots which are more informative to EEG/ERP community than barplots. Please, include ERP plots for the relevant comparisons [I am aware of the fact that the plots are to be found in supplementary materials, but (1) there is no link provided in the PDF generated for review that would allow me to view these plots, (2) as mentioned above, I would put the priority on ERP plots and treat bar plots as supplementary]

Statistical analyses

- please, report the final structure for each of the GLMM and LMM models. Was each of the models significantly better than the minimal model?

- report how factors were coded in the models.

- I think that running separate models for language status, dominance, and AOA was a good idea. However, I still have reservations towards the reliability of the results due to a small sample and possibly very low power. As reported in the manuscript, there are serious imbalances within the group for language status, language dominance and AOA. Although mixed effect models are great to tap into covariates in the data, this should be still supported by a representative sample. I do not see how comparing, for instance, 6 English dominant participants to 17 Spanish dominant participants would constitute a reliable statistical comparison. Further, I am confused by how AOA was entered into the analysis given that ouf of 26 participants, 22 were early speakers of Spanish while 12 were early speakers of English. It would only make sense if language was factored in but then there are only 4 late learners of Spanish.

- the abovementioned argument seems even more serious in the context of EEG data where much more power is needed to suppress noise and variance in EEG. For EEG I would recommend using a 2 (English, Spanish) x 3 (positive, negative, neutral) design. Concreteness is well matched so unless the authors have specific questions about concreteness (currently concreteness is not considered a main question in this study), I would recommend collapsing across this factor to get more power.

- time windows and topography for EEG analysis cannot be established based on visual inspection - for a discussion and solutions see Luck & Gaspelin 2017(http://dx.doi.org/10.1111/psyp.12639). I would recommend running mass univariate analysis (Groppe, Urbach & Kutas, 2011; https://doi.org/10.1111/j.1469-8986.2011.01273.x) which is an open-source plugin in MATLAB and works well with both eeglab and erplab. Alternatively, the authors may select the exact time windows and topographies for EPN & LPC based on available recommendations and/or published manuscripts (e.g. Citron 2012, Opitz & Degner, 2012; Conrad et al. 2011).

- I am not convinced that LMEs were run according to the recommendations for EEG data. The manuscript seems to suggest that LMEs were run on aggregated data (which is typically what one gets when analysing ERPs). However, for LMEs such data should be restructured to contain stimuli-related information for each trial (as for single-trial analysis; for a recent turorial see, for example, Heise, Mon & Bowman, 10.1016/j.dcn.2022.101070)

- all descriptive statistics should be supplemented with 95% confidence intervals (preferably) or at least with standard deviations or standard error of the mean.

- it is slightly confusing to read F-statistics reports for mixed effects models (or it's just me). Why don't the authors use beta?

 

Minor comments

- the first question that the authors ask in this study is "Are emotion effects comparable for both L1 and L2 or will L2 exhibit attenuation (smaller) effects? I had the impression, however, that the evidence showing emotional attenuation in L2 is not addressed sufficiently in the introduction. There are studies that directly address this question and that have not been discussed, e.g. Toivo & Scheepers, 2019 (http://dx.doi.org/10.1371/journal.pone.0210450); Iacozza et al. 2017 (https://doi.org/10.1371/journal.pone.0186027); Sheikh & Titone 2015 (http://dx.doi.org/10.1080/02699931.2015.1018144); Bromberek-Dyzman et al. 2021 (https://doi.org/10.1177%2F1367006920987306); Wu and Thierry 2012 (doi: 10.1523/JNEUROSCI.6119-11.2012); Jończyk et al. 2021 (doi: 10.1093/scan/nsz066); Ferre et al. 2022 (doi: 10.1017/S1366728922000189); or Garcia-Palacios et al. 2018 (doi: 10.1038/s41598-018-19352-8)

- "we tentatively expected to find no qualitative differences between L1 vs. L2 emotion effects" - please rephrase. One cannot hyothesise a null effect. On a related note, a lot of prior studies did report such differences (see above).

- did participants see English target words and their translation equivalents in the same experimental session?

- valence and arousal ratings for Spanish and English words should be compared statistically. It may be interesting in and of itself to see how words were rated on valence and arousal in L1 and L2.

- were reponse sides counterbalanced across participants?

Reviewer 2 Report

 

This contribution aims at exploring individual and joint contributions of AOA, language dominance and language status in emotion processing, which is a relevant and timely issue. To do so, authors implemented a lexical decision task. They measured Spanish-English/English-Spanish bilinguals’ reaction times and response accuracy. They also examined ERP components, namely EPN and LPC components reflecting early (automatic) and late (higher-order) lexical processes respectively, to investigate whether those components are affected comparably or differently by word valence in L1 and in L2.

 

While the study is principally well design and has the potential to answer those research questions, I am highly concerned by the generalisability of the results given the small sample size and the unbalanced groups. While the study includes 3 within-subject factors (i.e., valence: positive, neutral, negative; language of target stimulus: English, Spanish; and concreteness: abstract, concrete) and 3 between-subject factors (i.e., language status; language dominance; age of acquisition), the total sample size is 26 participants. This results in a very limited number of participants per language group (i.e., 6, 17 and 3), which is even more concerning since language groups are further subdivided for dominance and age of acquisition. For instance, I wonder whether this study (in its current state) can investigate language dominance as a factor since only some L1 Spanish participants became dominant in EN, while no L1 English participant became dominant in Spanish. I am conscious that EEG studies are complex to conduct with hundreds of participants, but I believe more participants are necessary to investigate that many factors, especially if you compute 4-way interactions.

 

Moreover, the paper discusses a huge number of results. In my humble opinion, it might benefit from either focussing on some results while leaving some others out, or form reorganising the result section in a less redundant way with a clearer overview for the reader (or maybe even from a combination of both strategies). Regarding the clarity of the manuscript, some sentences are repeated twice  (e.g. line 493-494 and line 497-498) and sometimes are even restated in a different way later on (e.g. line  505-510).

 

I strongly encourage the authors to collect more data and resubmit their paper once the manuscript has been revised.  I also include some additional, more specific comments attached.

 

Comments for author File: Comments.pdf

Reviewer 3 Report

Thank you for the opportunity to review the manuscript titled “Early and late ERP responses to L1 and L2 emotion words” in Languages (ID: 1667419). This paper examines whether factors related to language experience (i.e., AoA, dominance, and language status) impact emotion processing in Spanish-English. Using a lexical decision task, the authors found that language status and age of acquisition influenced early and late time windows, respectively. Language dominance had minimal effects on emotionality, except among balanced bilinguals. While I find the research question to be fascinating and addresses a gap in the field of bilingual emotion processing, I cannot recommend this paper for publication. For a between subjects design, the sample sizes for each factor within each group is far too low (also unbalanced sample sizes), calling the reliability and validity of the findings into question. Furthermore, the authors did not plot the ERP waveforms by group, making it difficult to verify the time windows, how noisy the EEG signal is within each group, and whether the waveforms were baseline corrected. The paper would also greatly benefit from additional proofreading.

Introduction

• Paragraph 3 (lines 73-100) could be edited down. 
• The authors did a good job reviewing the literature and providing clear objectives, research questions, and hypotheses.

Methods

• The authors mentioned that data from one participant were missing due to an excessive number of artifacts in the EEG. What type of artifacts (ocular, motor, noise, etc.)? What percentage of trials?
• Why was a cut-off of 5 used to differentiate early from late bilinguals? Was this established using a median split or based on previous literature?
• Table 1: Please also report the mean self-ratings for the L2 (i.e., both languages) for each group (L1 English, L1 Spanish, and Both). Why were the mean self-ratings for the balanced group (Both) not reported? This will help the reader understand the group categorization.
• A larger sample size in each group is required. The groups are also unbalanced. Unbalanced sample sizes can lead to unequal variances between samples, which impacts statistical power and Type I error.
• Dominance: 15 English-dominant, 5 Spanish-dominant, 6 balanced
• L1 status: 6 L1 English vs. 17 L1 Spanish vs. 3 Both 
• English AoA: 12 early bilinguals vs. 14 late bilinguals
• Spanish AoA: 22 early bilinguals vs. 4 late bilinguals
• Appendix B is missing.
• Provide a bit more information about Wuggy. 
• Line 303-304: Stimuli were matched according to grammatical category. I don’t see the means reported in Table 2. I recommend adding the non-significant p-values at the end of the sentence (..., all ps > .xx).

Statistical Analyses

• Why were two different time windows chosen for each ERP component? Is 300-400ms still considered “early” for an EPN? Could this component be an N400 instead? The N400 has been previously found in a lexical decision task (e.g., Xu, Kang, Sword, & Guo, 2017).

Xu, X., Kang, C., Sword, K., & Guo, T. (2017). Are Emotions Abstract or Concrete? Experimental Psychology, 64(5), 315-324.

Results

• It is very important to plot the ERP waveforms to check whether the waveforms were baseline corrected and if the correct time windows were selected.
• Please include effect sizes.
• Table 3: I would recommend reporting the means and standard deviations for each factor/type of analysis rather than just for L1 status. Furthermore, in this table, the data for balanced bilinguals are missing.
• Lines 530-535: In order to break down the 4-way interaction, two separate 3-way interactions, while keeping one factor constant, would need to be conducted. Double check all significant 4-way interactions.
• Figures 1-4:
• y-axis should be labelled
• Error bars are missing
• Line 881: “...and this advantage held true…” Why is a larger LPC in response to concrete words than abstract words an advantage? Please clarify.
• The authors could potentially look at Age of English and Spanish Acquisition along a continuum rather than dichotomically.
• The Figures at the moment are unclear and difficult to interpret. 

Discussion

• I found the Results and Discussion sections difficult to follow.
• I recommend explicitly stating the research question again in the examples below.
• Line 1132: “Overall, with regard to the research question formulated earlier,” 
• Lines 1209/1237/1385
• One possible explanation for the reduced LPC provided by the authors is the difference in word grammatical class. Is this the same as grammatical category? According to the information in the Materials and Methods section, the stimuli matched in grammatical category.
• Considering the participants were living in an immersive environment, is the demarcation between groups large enough to justify looking at L1 status and dominance? Were any of these participants heritage speakers? If not, including a few sentences about how the current findings may shed light on how heritage speakers process emotions could be a potential avenue for future research.
• The authors should highlight the implications and significance of their work in the Discussion.

Supplementary Materials:

• The link to the supporting materials is missing.

Minor Comments/Edits:

A careful read through the paper will improve the manuscript immensely.

• “Language History Questionnaire (LQH)” → LHQ
• Line 283 “...and 14 late English bilinguals.” The second half of the sentence is missing.
• Data is plural.
• Line 340: “The EEG data was filtered.” → “The EEG data were filtered.”
• Line 385: “The data obtained for those electrodes was averaged” → “were”
• Line 327: fixation signal → fixation cross
• All figures: x-axis → y-axis
• For all t-tests, a comma should proceed the t value and not a semi-colon
• For the AoA analyses (RTs, Accuracy rates, and ERPs), it would be easier to process if the authors reported the English and Spanish AoA findings separately under different subheadings.
• Try to be consistent with regards to significant digits. In some cases, the authors rounded to two significant digits, whereas in other places, the authors rounded to four significant digits.
• Lines 497-499: “there was a significant two-way interaction between language of the target stimulus and valence, F(2, 2345.0) = 25.9187; p < .001” is repeated already in Lines 493-494. 
• Additional spaces inserted throughout the manuscript. Please delete the extra spaces between words/sentences.
• Lines 617/1088: “targes” → targets
• Line 643: a space after < is missing
• Line 672: that → than
• Headings on lines 788 and 937 are missing “ms” 
• Line 837: reposes → responses
• Lines 873-877: Please remove the additional colons 
• Lines 879-888: Different font size? The font looks smaller.
• For the Figures illustrating L1 status, I would recommend changing the group names to “L1 English, L1 Spanish, Both” instead of “English, Spanish, Both” in order to be consistent with the factor labels used in the text.
• Figure 3 Dominance: Title should be “LPC 550-650 ms”
• In a number of locations throughout the results section, the word "targets" or "words" is sometimes missing for the second comparison. For example, in line 953: "..where concrete Spanish targets elicited a larger LPC than abstract targets"