Content uploaded by Aina Casaponsa
Author content
All content in this area was uploaded by Aina Casaponsa on Jul 02, 2020
Content may be subject to copyright.
1 | Page
Incidental changes in orthographic processing in the native
1
language as a function of learning a new language late in life
2
Maria Borragan1; Aina Casaponsa2; Eneko Antón3; Jon Andoni Duñabeitia4,5*
3
!
!
BCBL, Basque Center on Cognition, Brain and Language; San Sebastian, Spain
4
!
"
Deparment of Linguisitcs and English Language, Lancaster University; Lancaster, England
5
!
#
Humanitate eta Hezkuntza Zientzien fakultatea, Mondragon Unibertsitatea; Mondragon, Spain
6
!
$
Centro de Ciencia Cognitiva (C3), Universidad Nebrija; Madrid, Spain
7
!
%!!
Department of Languages and Culture, Faculty of Humanities, Social Sciences, and Education, The Arctic University of Norway;
8
Tromsø, Norway.
9
10
11
Abstract
12
Acquiring a second alphabetic language also entails learning a new set of orthographic rules and
13
specific patterns of grapheme combinations (namely, the orthotactics). The present longitudinal
14
study aims to investigate whether orthotactic sensitivity changes over the course of a second
15
language learning program. To this end, a group of Spanish monolingual old adults completed a
16
Basque language learning course. They were tested in different moments with a language
17
decision task that included pseudowords that could be Basque-marked, Spanish-marked or
18
neutral. Results showed that the markedness effect varied as a function of second language
19
acquisition, showing that learning a second language changes the sensitivity not only to the
20
orthographic patterns of the newly acquired language, but to those of the native language too.
21
These results demonstrate that the orthographic representations of the native language are not
22
static and that experience with a second language boosts markedness perception in the first
23
language.
24
25
26
27
*Contact information: jdunabeitia@nebrija.es
28
29
30
31
Keywords
32
Orthotactics; orthographic regularities; markedness; second language learning.
33
34
35
2 | Page
Introduction
36
Learning a new language not only involves acquiring new vocabulary, grammar,
37
phonology and syntactic rules, but also acquiring the implicit statistical probabilities regarding
38
the new language’s orthographic structure, such as orthotactics. Orthotactics are the patterns
39
of grapheme combinations in written words (see Conway, Bauernschmidt, Huang, & Pisoni,
40
2010; Krogh, Vlach, & Johnson, 2013), and they are learned implicitly by extracting the sub-
41
lexical regularities of words. People become sensitive to these regularities even after little
42
exposure to printed words (Chetail & Content, 2017), developing a high sensitivity to letter
43
sequences belonging to one’s language (Miller, Bruner, & Postman, 1954; Owsowitz, 1963).
44
When readers are exposed to one or several languages, they pick up statistical
45
orthotactic regularities in an unconscious manner, and these seemingly automatically extracted
46
patterns guide ulterior language processing. For instance, a Spanish-English bilingual can easily
47
detect that the word txerri (the Basque word for pig) is neither an English nor a Spanish word
48
solely on the basis of the statistical orthotactic regularities of its constituents, since the bigram
49
tx is not present in the English or Spanish vocabulary. Hence, native speakers of English or
50
Spanish do not need to know the meaning of txerri, or have any knowledge of Basque, in order
51
to decide that this word does not belong to their native language. When we learn a second
52
language (L2) with an alphabet that maps onto our native one, a similar process of extracting
53
statistical orthotactic regularities takes place (Bordag, Kirschenbaum, Rogahn, & Tschirner,
54
2017; Comesaña, Soares, Sánchez-Casas, & Lima, 2012). Thus, it seems plausible that as we
55
become more proficient in a second language, the new statistical regularities would be better
56
integrated within the preexisting set, leading to a change in our sensitivity to them. In other
57
words, it seems reasonable to predict that the general sensitivity to the orthotactics of both first
58
and second language would change once the new regularities have entered into the system.
59
With this in mind, this study aims to investigate how learning a second language could change
60
the sensitivity to statistical orthotactic regularities from both languages.
61
3 | Page
Previous research exploring language detection mechanisms by manipulating
62
orthographic markedness (namely, the use of language-specific letter combinations) has
63
demonstrated that young bilingual adults, as well as young monolingual adults, are highly
64
sensitive to violations of the statistical orthotactic regularities of the native language
65
(Casaponsa, Carreiras, & Duñabeitia, 2014; Vaid & Frenck-Mestre, 2002). In the study by
66
Casaponsa et al. (2014), Spanish monolinguals and Spanish-Basque bilinguals performed a
67
language decision task on Spanish and Basque words. Critically, some of the Basque words
68
included highly distinctive marked bigrams, while others did not. All groups were faster at
69
detecting letter strings that violated Spanish orthotactics as compared to other strings, showing
70
a recognition advantage for Basque words with marked bigrams (e.g., etxe, the Basque word for
71
house, which contains the bigram ‘tx’ that does not exist in Spanish). These results showed that
72
even monolinguals can easily detect letter patterns that do not align with their previous implicit
73
orthographic knowledge. Importantly, this suggests that people develop a certain degree of
74
sensitivity to letter sequences that do not conform to their native orthotactic rules, regardless
75
of whether they know the language of the words or not.
76
Results from monolingual and bilingual samples thus suggest that orthotactic processing
77
occurs at an early, semantics-free stage of visual word recognition. Consequently, language
78
attribution mechanisms triggered by orthotactic patterns appear to take place at a sub-lexical
79
level, before access to lexical and semantic representations (see BIA+ extended model, Van
80
Kesteren, Dijkstra, & de Smedt, 2012; see also BIA+ S model, Casaponsa, Thierry, & Duñabeitia,
81
2020). Studies exploring the influence of sub-lexical orthographic cues on event-related
82
potential (ERP) patterns related to automatic and unconscious processing of language switches
83
corroborate this idea (e.g., Casaponsa, Carreiras, & Duñabeitia, 2015; Casaponsa, Thierry, &
84
Duñabeitia, 2020; Hoversten, Brothers, Swaab, & Traxler, 2017). Therefore, it seems plausible
85
that sub-lexical factors such as orthotactic distinctiveness play a key role in determining the
86
4 | Page
language of words during visual word recognition (see also Oganian, Conrad, Aryani, Heekeren,
87
& Spalek, 2016; Vaid & Frenck-Mestre, 2002). And, in the absence of additional contextual cues,
88
multilingual single word recognition is a process that initially requires a fast-acting language
89
detection mechanism. Consequently, orthotactics should have a direct impact in second
90
language learning through correct and efficient language categorization.
91
Preceding research on language categorization suggests differential development of
92
sequential bilinguals’ linguistic systems (Segalowitz, 1991; Van Kesteren et al., 2012); it assumes
93
that the native language is stable through time while the second language is the one that
94
changes the most throughout acquisition and consolidation. It is thus expected that the native
95
language influences the second language, and not the other way around. Evidence in support of
96
this assumption comes from studies showing that second language learners normally exhibit
97
difficulties with L2 accent and prosody, with spillover or transfer effects from their L1. This
98
evident L2 malleability has led some authors to characterize the native language as stable and
99
resistant and the L2 weak as impressionable (Frenck-Mestre & Pynte, 1997; Hernandez, Bates,
100
& Avila, 1994). However, and not surprisingly, recent evidence shows that not the L2 but also
101
the L1 changes during learning (see, among many others, Baus, Costa, & Carreiras, 2013; Kroll,
102
Dussias, Bice, & Perrotti, 2015).
103
While L2 language learning abilities can extend beyond young adulthood, the malleability of
104
the native language as a function of the acquisition of a new language seems to diminish with
105
increasing age (Macwhinney, 2007; Schmid & Köpke, 2017). In spite of the cognitive decline
106
associated with ageing (Harris et al., 2009), language learning can effectively take place late in
107
life ( see Antoniou & Wright, 2017; Ramos, Fernández García, Antón, Casaponsa, & Duñabeitia,
108
2017; Ware et al., 2017). The question of interest here is whether L2 acquisition late in life
109
impacts L1 orthotactic structure. Thus, the present study focuses on older adults as a critical test
110
group. It is worth noting that the sensitivity to violations of the orthotactic rules of the first and
111
5 | Page
second language has already been shown in younger bilingual adults to certain extent (Oganian,
112
Conrad, Aryani, Spalek, & Heekeren, 2015), suggesting that L2 learning might have an impact in
113
L1 orthotactics. However, it is unclear whether similar L1 changes can be observed in older
114
populations, when presumably the resistance to change and stability of L1 is at its peak, and the
115
malleability and plasticity of the language system is at its lowest.
116
Hence, the present longitudinal study aims to investigate whether older adults are sensitive
117
to markedness before learning a second language, and how this learning process changes their
118
sensitivity to orthographic regularities. Specifically, we tested whether language learning late in
119
life and the progressive improvement in L2 skills modulated learners’ sensitivity not only to L2
120
orthotactics, but also to the orthotactic structure of the L1. To this end, older native Spanish
121
speaker adults immersed in a Basque language-learning course for two consecutive academic
122
years were tested in three critical moments (before, during, and after language learning) on their
123
sensitivity to orthotactics via a language discrimination task. We decided to use a two-
124
alternative forced-choice language decision task on pseudowords to minimize the influence of
125
pre-existing L1 lexical and semantic knowledge (see Oganian et al., 2016, for a similar
126
procedure).
127
128
Methods
129
Participants
130
Thirty retired Spanish monolingual adults took part in this longitudinal experiment.
131
However, only twenty participants remained through the two year sessions (8 females; mean
132
age = 66.57; SD = 5.56). All participants were living in the Basque Country, a Spanish region with
133
two coexisting co-official languages, Spanish and Basque. None of the participants had prior
134
knowledge of Basque, neither could they understand or produce linguistic structures in any
135
6 | Page
other language than Spanish (see below). All participants reported having normal or corrected-
136
to-normal vision, and none of them had any history of chronic neuropsychological disorders.
137
Participants were recruited by advertisement at a Center of Continuing Education for
138
Adults where free Basque lessons were offered to retired Spanish monolingual adults with no
139
prior knowledge of Basque. This experiment was part of a larger project suported by the Basque
140
Government to study the impact of second language acquisition in the elderly on other cognitive
141
capacities, such as inhibitory control (Antón, Fernández García, Carreiras, & Duñabeitia, 2016)
142
and switching ability (Ramos et al., 2017). Participants signed a written consent form approved
143
by the Ethics and Research Committees of the Basque Center on Cognition, Brain, and Language
144
(BCBL) before the start of the research and educational actions.
145
Participants undertook Basque lessons for two whole academic years at the Center of
146
Continuing Education for Adults. They attended Basque lessons for a period of eight months
147
each year. Small groups of a maximum of 10 participants per class were arranged. In total, five
148
hours and a half of training were set per week, distributed in three sessions held during working
149
days. Participants were tested at the beginning of the academic year (T1), at the end of that
150
same academic year (T2), and at the end of the second year of taking Basque lessons (T3). The
151
linguistic project was coordinated by the Department of Education, Linguistic Policy and Culture
152
of the Basque Government, and managed by native Basque-Spanish bilingual professional
153
language trainers specialized in adult teaching.
154
155
At the beginning of the first academic year, all participants completed a general
156
assessment consisting of a series of cognitive and language proficiency tasks. Age-related
157
cognitive functioning was assessed using the Spanish version of the Mini-Mental State
158
Examination (MMSE; see Lobo, Ezquerra, Gómez, Sala, & Seva, 1979). Participants’ linguistic
159
profile before learning Basque was characterized via self-report measures of proficiency, and all
160
participants were asked to rate their knowledge of Spanish and Basque on a scale from 1 to 10
161
7 | Page
(see Table 1). Also, teachers evaluated participants’ Basque proficiency based on their own
162
perception before the lessons started, ensuring that they did not have previous knowledge of
163
Basque . Self- and teacher-perceived Basque proficiency levels were also assessed at the end of
164
the learning process, together with additional objective measures of Basque knowledge. As
165
objective measures of language learning, participants completed a picture naming test (de Bruin,
166
Carreiras, & Duñabeitia, 2017) in which participants had to name sixty-five common names in
167
Basque (see Table 1), and the beginner language test (A1 level) of the Common European
168
Framework for Reference (CEFR, Council of Europe, 2011), with a maximum score of 20.
169
170
Table 1. Descriptive statistics of the assessment
171
Before Basque lessons
Age
65.2 (3.81)
Cognitive function (MMSE)
28.8 (1.24)
Self-perceived Spanish competence
8.1 (0.55)
Self-perceived Basque competence
0
Teacher-perceived Basque competence
0
After Basque lessons
Self-perceived Basque competence
5.75 (1.45)
Teacher-perceived Basque competence
6.15 (1.09)
A1 level score
19.7 (4.28)
Picture naming
27.85 (10.26)
Note. Values reported correspond to the means (and standard deviations in parenthesis) of the age in years, result of the MMSE
172
test, self-perceived Spanish and Basque skills (0-10 scale), teacher-perceived Basque competence (0-10 scale), score in the A1 level
173
test (with a maximum score of 20), and number of correctly named pictures in a picture naming test.
174
175
176
Materials and procedure
177
First, a corpus of bigrams was constructed with the Spanish words from the B-PAL (Davis
178
& Perea, 2005) and Basque words from the E-HITZ (Perea et al., 2006) databases, and filtered
179
with the items contained in the SYLLABARIUM database (Duñabeitia, Cholin, Corral, Perea, &
180
Carreiras, 2010). Words that contained letters that do not exist in the other language were
181
removed (e.g., c, ñ, v). Bigrams that did not appear in any form in the other language were
182
considered illegal and were selected for the construction of the marked pseudowords. Bigrams
183
8 | Page
were considered neutral in both languages if they had a frequency of appearance of at least 10
184
times in different words of each database. One hundred and thirty-five pseudowords were
185
generated with the help of Wuggy (Keuleers & Brysbaert, 2010), manipulating the presence or
186
absence of distinctive bigrams of each language. Forty-five of these pseudowords were Spanish-
187
marked items, forty-five were Basque-marked items, and forty-five were language-neutral
188
pseudowords. Marked pseudowords were created making sure that at least one of the
189
constituent bigrams violated the orthotactics of the other language. For instance, ‘txamur’ is
190
considered a Basque-marked pseudoword because the bigram ‘tx’ does not exist in Spanish
191
(namely, the ‘tx’ bigram has a frequency of use of 0 in Spanish). On the other hand, neutral
192
pseudowords were created using bigrams that were plausible in both languages, such as the
193
bigram ‘rd’ that exists in words such as cerdo (the Spanish word for pig), or ardi (the Basque
194
word for sheep). Those neutral bigrams were controlled to have equal mean frequency of use in
195
Spanish and Basque, t(44)=0.03, p =.976, Cohen´s!d=.033 (see Table 2). The position- and length-
196
dependent mean bigram frequency of each pseudoword as provided by B-PAL and E-Hitz
197
databases was calculated the sets of pseudowords were matched based on this measure. This
198
way, neutral pseudowords had an overall mean bigram frequency similar to that of Spanish-
199
marked pseudowords when measured according to the Spanish statistics, and similar to that of
200
Basque-marked pseudowords when measured according to the Basque statistics. This ensured
201
that neutral pseudowords were equally legal in both languages when position-dependent and
202
length-dependent measures were taken into account. Furthermore, the number of orthographic
203
neighbours in Spanish and Basque were controlled to be similar for neutral pseudowords and
204
for marked pseudowords (see Table 2).
205
206
207
208
9 | Page
209
Table 2. Descriptive statistics of characteristics of the materials
210
Neutral
pseudowords
Spanish-marked
pseudowords
Basque-marked
pseudowords
Length
5.88 (1,46)
6.11 (1,35)
6.11 (1,54)
Neighbors in Spanish
1.75 (2,49)
0.35 (1,19)
1.91 (2,19)
Neighbors in Basque
1.6 (2,15)
1.64 (2,67)
0.2 (0,69)
Mean bigram frequency in Basque
2.06 (0,35)
0.68 (0,29)
2.10 (0,29)
Mean bigram frequency in Spanish
2.24 (0,54)
2.25 (0,52)
0.91 (0,39)
Illegal bigram frequency in Basque
0 (0)
1.31 (0,51)
0 (0)
Illegal bigrams frequency in Spanish
0 (0)
0 (0)
1.37 (0,61)
Note. Values reported are means and standard deviation in parenthesis on word length (number of letters), orthographic
211
neighbours (number of words that share all letters but one), mean bigram frequency (position- and length-dependent mean
212
bigram frequency as extracted by B-PAL and E-HITZ), and illegal bigrams (extracted from total counts of the LEXESP and
213
SILLABARIUM databases).
214
215
Participants were tested individually in a quiet room within the educational institution
216
by a trained research assistant who accompanied them during the course of the whole language
217
learning process. The same computer was used at all test moments in order to avoid any
218
hardware-related differences across sessions. The experiment was programmed in Experiment
219
Builder (SR Research, Ontario). The start of each trial was marked by a fixation cross appearing
220
in the middle of the screen for 500 ms, immediately followed by the target word for 3000 ms or
221
until participants’ response. At the beginning of the task, participants performed some trials as
222
practice. They were instructed to decide whether the string of letters appearing on the screen
223
could belong to Spanish or Basque (i.e., forced-choice), and to do so as fast as possible.
224
Participants were asked to press one out of two buttons in a handheld controller to indicate
225
whether each string could belong to Spanish or Basque. Participants were informed that none
226
of the strings appearing on the screen were real words. Participants were asked to perform this
227
task before (T1) and after (T2) the first academic year, and one year later (T3).
228
229
Data analysis
230
10 | Page
Accuracy and reaction times were collected, and all statistical analyses were carried out
231
in the statistical environment R (R core team, 2013). Before data analysis, responses below 200
232
ms (0.01 %) and timeouts (0.04%) were excluded. Also, responses that deviated 3.5 standard
233
deviations above and below the mean from all within-subjects (1.05% of outliers) or within-
234
items (0.43% of outliers) factors were excluded from the analyses, leading to a final rejection of
235
1.33% of the data.
236
Accuracy was analyzed with logistic mixed-effects models and reaction times with linear
237
mixed-effects models (Baayen, Davidson, & Bates, 2008; Barr, 2013; Jaeger, 2008), using lme4
238
package for R (Bates, Mächler, Bolker, & Walker, 2015). We first fitted maximal random
239
structure models. When the data did not support the execution of the maximal model random
240
structure, we then reduced the model complexity in order to arrive at a parsimonious model. To
241
do so, we computed principal component analyses (PCA) of the random structure (see Bates et
242
al., 2015), and then kept the number of principal components that cumulatively accounted for
243
100% of the variance. Type-III Anova Wald-tests was obtained to assess the significance of fixed
244
effects for binary data, and Type-III Anova F-tests with Satterwhite approximations to degrees
245
of freedom were obtained for response latency analysis. Averaged reaction times and accuracy
246
rates per condition are presented in Table 3. Considering that decisions made on neutral
247
pseudowords cannot be characterized as correct or incorrect responses in the absence of
248
language cues, response latencies for these items were modelled by the type of response. The
249
response tendency was based on the given response of the participants, being dummy coded as
250
‘1’ if participants responded Spanish and ‘0’ if they responded Basque (see Table 3). In contrast,
251
in the case of marked pseudowords, the percentage of correct responses was analyzed based
252
on the presence of language cues, and reaction times were analysed using only correct answers
253
(see Table 3).
254
First, we investigated whether Type of Markedness (Neutral, Spanish-marked, Basque-
255
marked) and Test Moment (T1, T2, T3) had an overall impact on participants’ language choice.
256
11 | Page
Then, we analysed marked and neutral pseudowords separately, given the low proportion of
257
“other" language choices for marked conditions (i.e., Spanish-marked pseudowords and Basque-
258
marked pseudowords were correctly categorised as Spanish and Basque, respectively, more
259
than 90% of the cases; see Table 3). Note also that responses for neutral pseudowords cannot
260
be categorized as correct or incorrect responses for obvious reasons. Thus, response latencies
261
for neutral psuedowords were analysed including Test Moment (T1, T2, T3) and Response Type
262
(Spanish, Basque) as predictors. Reaction times and accuracy data of marked pseudowords was
263
analysed including Test Moment (T1, T2, T3) and Type of Markedness (Basque-marked, Spanish-
264
marked) as predictors.
265
266
Table 3. Descriptive statistics for the language decision task in the three different test moments (T1, T2
267
and T3). The values reported correspond to the means and standard deviations (in parenthesis) of the
268
accuracy rates (% of hits) and of the reaction times (in milliseconds).
269
Marked
Neutral
ACCURACY
Basque
Spanish
Basque tendency
Spanish tendency
T1
92.44 (26.45)
91.12 (28.46)
32.36 (20.56)
67.64 (46.81)
T2
94.33 (23.14)
94.72 (22.38)
28.03 (18.79)
71.97 (44.94)
T3
93.2 (25.18)
91.74 (27.54)
28.04 (21.38)
71.96 (44.95)
REACTION TIMES
Basque
Spanish
Basque tendency
Spanish tendency
T1
873 (296)
1003 (446)
1295 (559)
1050 (454)
T2
897 (306)
953 (380)
1269 (526)
1029 (467)
T3
883 (283)
898 (297)
1276(519)
1014 (441)
270
Results
271
Overall response choices
272
First, we analyzed the tendency of Spanish responses based on Type of Markedness
273
(Spanish-marked, Basque-marked, and Neutral) across Test Moments (T1, T2, and T3 ). Analyses
274
revealed a main effect of language markedness [χ2(2)= 199.72, p<.001], such that the tendency
275
of Spanish responses was higher for Spanish-marked pseudowords as compared to neutral
276
pseudowords [b=2.32, SE=0.44, z=5.23, p<.001], and for neutral pseudowords as compared to
277
12 | Page
Basque-marked pseudowords [b=5.25, SE=0.42, z=12.55, p<.001]. We did not find a significant
278
main effect of the Test Moment [χ2(2)=2.63, p=.275]. The interaction between Type of
279
Markedness and Test Moment was significant [χ2(4)=13.03, p=.01]. However, post-hoc analyses
280
revealed no significant differences across Test Moment for neutral (all ps>.26), Basque-marked
281
(all ps>.40), or Spanish-marked pseudowords (all ps>.21).
282
283
Figure 1. Percentage of Spanish responses to Basque-marked, neutral, and Spanish-marked
284
pseudowords before language learning (T1), after one year of language learning (T2), and after
285
two years (T3). Error bars represent ±1 standard error (SE) of the mean.
286
287
Neutral pseudowords
288
Analysis of the reaction times on neutral pseudowords based on the Response
289
(Spanish, Basque) and Test Moment (T1, T2, T3) showed that participants were faster at
290
classifying neutral pseudowords as Spanish (see Figure 2) than Basque [F(1,22.02)=10.14,
291
p=.004; b=125.84, SE=39.53]. The main effect of Test Moment [F(2,19.99)=0.05, p=.95] and the
292
interaction between Test Moment and Response [F(2,2202.14)=2.07, p=.13] were not
293
significant.
294
0
20
40
60
80
100
Bas qu e- mar ked Neu tr al Spanish-marked
Spanish responses (%)
T1 T2 T3
13 | Page
295
Marked pseudowords
296
The analysis of congruent language selection responses based on the presence of Type
297
of Markedness (Spanish-marked, Basque-marked) across Test Moments did not reveal any
298
significant main effect or interaction (all ps > .12). Overall accuracy ratings were already close to
299
ceiling at T1 for both the Spanish-marked and Basque-marked items (see Table 3).
300
Analyses of reaction times on marked pseudowords revealed a main effect of Type of
301
Markedness [F(1,36.3)=6.46, p=0.02], indicating that participants were overall slower at
302
detecting Spanish-marked than Basque-marked pseudowords [b=90.25, SE=35.51; see Figure 2].
303
The main effect of Test Moment was not significant [F(2,19)=.83, p=.45]. Importantly, a
304
significant interaction was found between Type of Markedness and Test Moment
305
[F(2,4778)=20.89, p<.001]. Planned comparisons revealed that whilst in T1 participants were
306
significantly slower at responding to Spanish-marked pseudowords as compared to Basque-
307
marked pseudowords [i.e., markedness effect; b=155.73, SE=36.97, t(42.6)=4.21, p<.001], this
308
difference diminished after language learning [T2: b=67.81, SE= 436.90, t(42.3)=1.84, p=.07; T3:
309
47.21, SE=36.995, t(42.7)=1.28, p=.21]. This reduction of the markedness effect over the
310
different test moments was due to an overall reduction in response latencies to Spanish-marked
311
pseudowords [b=96.31, SE=41.19, t(20.9)=2.34, p=.03], whilst Basque-marked pseudoword
312
response latencies remained constant [b=-12.21, SE=41.13, t(20.7)=-.30, p=.77].
313
314
14 | Page
315
Figure 2. Bar plots depicting participants’ response latencies in the language decision task for
316
neutral (left) and marked (right) pseudowords in T1, T2, and T3. For neutral pseudowords, all
317
Basque (dark grey) and Spanish (light grey) responses are included. For marked pseudowords,
318
only responses congruent with the marked type are included. Error bars represent ±1 standard
319
error (SE) of the mean.
320
321
Discussion
322
The present longitudinal study investigated changes in orthotactic sensitivity in a group
323
of older Spanish monolingual adults before and after they learned Basque. Our results confirmed
324
previous findings showing that older adults were highly sensitive to orthotactic markedness in
325
Basque, as shown by their faster reaction times when responding to Basque-marked words
326
(Casaponsa et al., 2014; Duñabeitia, Ivaz, & Casaponsa, 2016; Oganian, Conrad, Aryani,
327
Heekeren, & Spalek, 2016). This sensitivity to L2 markedness was shown even before
328
participants learned Basque, and it persisted during the learning process. However, and more
329
importantly, we also found that participants demonstrated increased sensitivity to orthotactic
330
markedness in their native language after learning a second language, evidenced by faster
331
reaction times. This strongly suggests that sensitivity to native orthotactics changes due to the
332
accommodation of newly acquired regularities from a second language.
333
As shown in the current study, before and after learning a second language, the
334
presence of language-specific orthotactic cues guides and aids language classification. Our
335
600
800
1000
1200
1400
T1 T2 T3
Reaction Times (ms)
Marked
Basque
Spanish
600
800
1000
1200
1400
T1 T2 T3
Reaction Times (ms)
Neutral
15 | Page
participants were able to easily classify Basque-marked and Spanish-marked pseudowords as
336
Basque and Spanish, respectively, despite their complete lack of Basque knowledge. However,
337
when participants classified seemingly neutral pseudowords without language-specific
338
orthotactic cues, they showed a strong preference to classify them as belonging to their native
339
language, Spanish. This effect was also accompanied by faster reaction times for the neutral
340
pseudowords which they deemed to be Spanish. One possible explanation for this finding is that
341
readers consider familiar orthotatic patterns to be part of their previous knowledge. In line with
342
this assumption, previous research (Ellis & Beaton, 1993) has shown that people prefer to learn
343
letter sequences that follow sequences found in their native language, suggesting they have a
344
preference for patterns that follow or align with the L1 orthotactic rules.
345
In general terms, participants showed high sensitivity levels to orthographic
346
markedness, responding significantly faster to marked than to neutral pseudowords in both
347
languages, both before and during second language learning. Even though older adults were
348
equally accurate at detecting Spanish-marked as they were at detecting Basque-marked
349
pseudowords, they responsed more quickly to Basque-marked pseudowords. However, this was
350
only true in the T1, when they had not yet learned Basque. This suggests that before learning
351
the second language, participants could easily realize that Basque-marked pseudowords did not
352
conform to the L1 orthotactic regularities. These results are in line with previous findings
353
showing that even monolinguals are very sensitive to letter sequences that violate the
354
orthotactic rules of their native language (Casaponsa et al., 2014; Casaponsa & Duñabeitia, 2016;
355
Oganian, Conrad, Aryani, Heekeren, & Spalek, 2016).
356
While the finding of an inherent sensitivity of monolinguals to detect strings that deviate
357
from the orthotactic standards set by the orthographic distributional properties of the native
358
language is not a trivial one, other findings provide additional insights regarding the dynamic
359
nature of the orthographic system. Interestingly, results from the two other test moments (T2
360
16 | Page
and T3) suggest that the probabilistic distribution of regularities in the native language changes.
361
While accuracy in detecting the language of marked pseudowords remained very high and
362
relatively constant across the three test moments, the analysis of reaction times showed
363
significant variations depending on the type of pseudowords. Basque-marked pseudowords
364
were detected equally fast across the three test moments, but reaction times to Spanish-marked
365
pseudowords decreased significantly as a function of increased exposure to the new language.
366
It could be tentatively argued that this reduction in reaction times associated with Spanish-
367
marked pseudowords could be associated with a change in the response strategy. In a first test
368
moment, participants could have had carefully evaluated if the pseudowords belonged to
369
Spanish by assessing their degree of similarity with known Spanish words, and then stop using
370
this strategy once they became familiar with the task, resulting in faster reaction times across
371
sessions. However, this may not seem to be a valid explanation that fits all the data, since if this
372
were the case, participants would have shown faster responses over time for all types of
373
pseudowords. We believe that similar automatic sub-lexical and lexical competition and
374
selection mechanisms guided participants’ responses in the three test sessions, as predicted by
375
current bilingual interactive activation models.
376
Hence, the current pattern could be readily accounted for by bilingual word
377
identification models that predict different processing mechanisms as a function of the sub-
378
lexical characteristics of the items (i.e., see BIA+ extended, Van Kesteren et al., 2015; see also
379
BIA+ S, Casaponsa et al., 2020). In the case of neutral words, responses were mainly influenced
380
by the formal similarity with existing lexical entries from the native language lexicon,
381
consequently leading to faster Spanish choices compared to Basque choices (see Figure 2). Not
382
surprisingly, responses to neutral pseudowords were heavily influenced by the native language
383
even after learning Basque, leading to similar choices and response latencies across sessions. It
384
should be noted in this regard that the general level of L2 achieved was admittedly low (namely,
385
17 | Page
A1 level of CEFR), and accordingly the degree of L2 lexical consolidation was low too. In this line,
386
Casaponsa, Antón, Pérez and Duñabeitia (2015) showed that at A1 levels, the speed of response
387
to L2 words is indeed heavily influenced by L1 knowledge, coinciding with the findings of the
388
current study. In the case of marked pseudowords, the mechanisms that underlie language
389
identification differ for Spanish-like and Basque-like strings. On the one hand, responses to
390
Basque-marked pseudowords were mainly driven by the earliest stages of orthographic
391
processing, leading to faster reaction times as compared to Spanish-marked pseudowords.
392
Importantly, these decisions were not affected by L2 proficiency, leading to similar reaction
393
times across sessions (see Casaponsa et al., 2014, for similar results; see also BIA+S, Casaponsa
394
et al., 2020). On the other hand, responses to Spanish-marked pseudowords appeared to be less
395
mediated by sub-lexical stages of processing and more mediated by lexical search routines at
396
initial stages of language learning, resulting in significantly slower reaction times at T1. We
397
suggest that the reliance on specific L1 and L2 orthotactic information progressively increased
398
as participants learned Basque, and that the response criteria for Spanish-marked pseudowords
399
shifted from a lexical search at T1 to a sub-lexical strategy at T2 and T3, allowing participants to
400
speed up their language decision process for strings that violated L2 orthotactics.
401
This account fits well with current bilingual interactive activations models that include
402
sub-lexical language nodes (see BIA+ extended, Van Kesteren et al., 2015; see also BIA+S,
403
Casaponsa et al., 2019). These accounts predict that the activation of the sub-lexical language
404
nodes due to the presence of language-specific sub-lexical cues will speed up language decision
405
processes once the orthotactic rules of the first and the second language are integrated in the
406
system. In the absence of sub-lexical language cues, the language decision process would be
407
guided by lexical language nodes, and hence influenced by lexical competition and selection
408
mechanisms. Thus, the current results fit well with these accounts, and they suggest that
409
participants developed increased sensitivity to orthotactic regularities specific to their native
410
18 | Page
language as a function of second language learning. This finding is particulary interesting as it
411
suggests that learning a second language changes the sensitivity to the orthotactics of the native
412
language (see also Casaponsa et al., 2014, suggesting that bilinguals’ sensitivity to markedness
413
changes as a function of proficiency).
414
Learning a language implies, among many other things, integrating new words within
415
the set of existent representations of the native language. Therefore, while learning a second
416
language, people also learn the similarities and differences between the to-be-incorporated
417
words and the already known ones. The construction of the orthotactic repertoire is thus an
418
automatic and spontaneous parallel process that takes place as a result of visual word
419
processing. Learners need to implicitly acquire new orthotactic regularities and compare these
420
with already known (native) patterns in order to make links between the new and the existing
421
pieces of information. Thus, it seems plausible that as readers compare the new patterns with
422
the old ones, they become more sensitive to the specificities of the old ones, consequently
423
perceiving native orthotatic regularities differently. In other words, we propose that after
424
learning a second language, readers may be better able to detect strings with native language-
425
specific cues due to increased saliency as pieces of orthotactic information that contrast with
426
the newly acquired language.
427
The idea that the native language may be permeable challenges the assumption that the
428
L1 remains static over time. Whilst the second language can be influenced by native language
429
processing (Frenck-Mestre & Pynte, 1997; Hernandez et al., 1994; Segalowitz, 1991), the native
430
language itself has been typically considered as relatively impermeable and immutable.
431
However, results in this study suggest that L1 orthotactic sensitivity changes while learning a
432
second language. The idea that bilinguals’ whole linguistic system displays adaptive changes was
433
already proposed by Kroll, Dussias, Bice, and Perrotti (2015; see also Dussias and Sagarra, 2007).
434
They hypothesized that the linguistic system is permeable in both languages, especially when
435
19 | Page
high proficiency in L1 is achieved. The idea of native language changes pursuant to language
436
learning fits well with preceding studies suggesting that learning new words and grammar
437
interacts with the existing language in a dynamic way, changing the linguistic system as a whole
438
(Baus et al., 2013; Chang, 2013; Kartushina, Frauenfelder, & Golestani, 2016; Linck, Kroll, &
439
Sunderman, 2009). Following these premises, our results demonstrate that changes in the
440
linguistic system due to L2 learning can emerge even when the malleability of the native
441
language is presumably at its lowest. By means of testing older samples over a period of two
442
years of language learning, we were able to show that lifelong exposure to a unique language
443
system (namely, the native language), does not eliminate permeability to the properties of a
444
new language. Furthermore, our results suggest that the sub-lexical mechanisms underpinning
445
second language learning across the lifespan are relatively stable and qualitatively similar for old
446
and young learners. Similar to young adults (see Oganian et al., 2016), older learners successfully
447
rely on the acquisition of implicit knowledge when learning a second language, focusing on the
448
statistical regularities of the sub-lexical units of their languages.
449
Taken together, the present results support the view that the native language is
450
permeable and changes during second language learning. Specifically, learning a new language
451
that does not share native language orthotactics can change the perception of orthotactics in
452
the native language already at early stages of L2 acquisition. Future research should explore
453
what other aspects of the native language may change as consequence of second language
454
learning, and correctly characterize the stages and rythms at which these changes take place.
455
This research will lead to a better understanding of the relationship between the native
456
language and the multiligual linguistic system.
457
458
20 | Page
459
Acknowledgements
460
This study was partially supported by grants RED2018-102615-T and PGC2018-097145-
461
B-I00 from the Spanish Government, H2019/HUM-5705 from the Comunidad de Madrid, and
462
by a fellowship from La Caixa foundation (ID 100010434, code LCF/BQ/ES16/11570003).
463
464
21 | Page
References
465
Antón, E., Fernández García, Y., Carreiras, M., & Duñabeitia, J. A. (2016). Does bilingualism
466
shape inhibitory control in the elderly? Journal of Memory and Language, 90, 147–160.
467
https://doi.org/10.1016/j.jml.2016.04.007
468
Antoniou, M., & Wright, S. M. (2017). Uncovering the mechanisms responsible for why
469
language learning may promote healthy cognitive aging. Frontiers in Psychology, 8(DEC),
470
1–12. https://doi.org/10.3389/fpsyg.2017.02217
471
Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed
472
random effects for subjects and items. Journal of Memory and Language, 59(4), 390–412.
473
https://doi.org/10.1016/j.jml.2007.12.005
474
Barr, D. J. (2013). Random effects structure for testing interactions in linear mixed-effects
475
models. Frontiers in Psychology, 4(June), 3–4. https://doi.org/10.3389/fpsyg.2013.00328
476
Bates, D., Mächler, M., Bolker, B. M., & Walker, S. C. (2015). Fitting linear mixed-effects models
477
using lme4. Journal of Statistical Software, 67(1). https://doi.org/10.18637/jss.v067.i01
478
Baus, C., Costa, A., & Carreiras, M. (2013). On the effects of second language immersion on
479
first language production. Acta Psychologica, 142(3), 402–409.
480
https://doi.org/10.1016/j.actpsy.2013.01.010
481
Bordag, D., Kirschenbaum, A., Rogahn, M., & Tschirner, E. (2017). The role of orthotactic
482
probability in incidental and intentional vocabulary acquisition L1 and L2. Second
483
Language Research, 33(2), 147–178. https://doi.org/10.1177/0267658316665879
484
Casaponsa, A., Antón, E., Pérez, A., & Duñabeitia, J.A. (2015). Foreign language comprehension
485
achievement: insights from the cognate facilitation effect. Frontiers in Psychology, 6:588.
486
Casaponsa, A., Carreiras, M., & Duñabeitia, J. A. (2014). Discriminating languages in bilingual
487
contexts: The impact of orthographic markedness. Frontiers in Psychology, 5(MAY), 1–10.
488
https://doi.org/10.3389/fpsyg.2014.00424
489
Casaponsa, A., Carreiras, M., & Duñabeitia, J. A. (2015). How do bilinguals identify the language
490
of the words they read? Brain Research, 1624(August), 153–166.
491
https://doi.org/10.1016/j.brainres.2015.07.035
492
Casaponsa, A., & Duñabeitia, J. A. (2016). Lexical organization of language-ambiguous and
493
language-specific words in bilinguals. Quarterly Journal of Experimental Psychology,
494
69(3), 589–604. https://doi.org/10.1080/17470218.2015.1064977
495
Casaponsa, A., Thierry, G., & Duñabeitia, J. A. (2020). The role of orthotactics in language
496
switching: An ERP investigation using masked language priming. Brain Sciences, 10(1), 22.
497
https://doi.org/10.3390/brainsci10010022
498
Chang, C. B. (2013). A novelty effect in phonetic drift of the native language. Journal of
499
Phonetics, 41(6), 520–533. https://doi.org/10.1016/j.wocn.2013.09.006
500
Chetail, F., & Content, A. (2017). The perceptual structure of printed words: The case of silent E
501
words in French. Journal of Memory and Language, 97, 121–134.
502
https://doi.org/10.1016/j.jml.2017.07.007
503
22 | Page
Comesaña, M., Soares, A. P., Sánchez-Casas, R., & Lima, C. (2012). Lexical and semantic
504
representations in the acquisition of L2 cognate and non-cognate words: Evidence from
505
two learning methods in children. British Journal of Psychology, 103(3), 378–392.
506
https://doi.org/10.1111/j.2044-8295.2011.02080.x
507
Conway, C. M., Bauernschmidt, A., Huang, S. S., & Pisoni, D. B. (2010). Implicit statistical
508
learning in language processing: Word predictability is the key. Cognition, 114(3), 356–
509
371. https://doi.org/10.1016/j.cognition.2009.10.009
510
Council of Europe, C. (2011). Common European framework of reference for languages:
511
Learning, teaching, assessment. Cambridge, UK: Press Syndicate of the University of
512
Cambridge.
513
Davis, C. J., & Perea, M. (2005). BuscaPalabras: A program for deriving orthographic and
514
phonological neighborhood statistics and other psycholinguistic indices in Spanish.
515
Behavior Research Methods, 37(4), 665–671. https://doi.org/10.3758/BF03192738
516
de Bruin, A., Carreiras, M., & Duñabeitia, J. A. (2017). The BEST dataset of language proficiency.
517
Frontiers in Psychology, 8(MAR). https://doi.org/10.3389/fpsyg.2017.00522
518
Duñabeitia, J. A., Cholin, J., Corral, J., Perea, M., & Carreiras, M. (2010). SYLLABARIUM: An
519
online application for deriving complete statistics for Basque and Spanish orthographic
520
syllables. Behavior Research Methods, 42(1), 118–125.
521
https://doi.org/10.3758/BRM.42.1.118
522
Duñabeitia, J. A., Ivaz, L., & Casaponsa, A. (2016). Journal of Cognitive Psychology
523
Developmental changes associated with cross-language similarity in bilingual children.
524
https://doi.org/10.1080/20445911.2015.1086773doi.org/10.1080/20445911.2015.10867
525
73
526
Dussias, P. E., & Sagarra, N. (2007). The effect of exposure on syntactic parsing in Spanish -
527
English bilinguals. Bilingualism, 10(1), 101–116.
528
https://doi.org/10.1017/S1366728906002847
529
Ellis, N. C., & Beaton, A. (1993). Psycholinguistic Determinants of Foreign Language Vocabulary
530
Learning. Language Learning, 43(4), 559–617. https://doi.org/10.1111/j.1467-
531
1770.1993.tb00627.x
532
Frenck-Mestre, C., & Pynte, J. (1997). Syntactic ambiguity resolution while reading in second
533
and native languages. The Quarterly Journal of Experimental Psychology, 50A(1), 119–
534
148.
535
Harris, S. E., Houlihan, L. M., Corley, J., Gow, A. J., Starr, J. M., Penke, L., … Rafnsson, S. B.
536
(2009). Age-associated cognitive decline. British Medical Bulletin, 92(1), 135–152.
537
https://doi.org/10.1093/bmb/ldp033
538
Hernandez, A. E., Bates, E. A., & Avila, L. X. (1994). On-line sentence interpretation in Spanish–
539
English bilinguals: What does it mean to be “in between”? Applied Psycholinguistics,
540
15(4), 417–446. https://doi.org/10.1017/s014271640000686x
541
Hoversten, L. J., Brothers, T., Swaab, T. Y., & Traxler, M. J. (2017). Early processing of
542
orthographic language membership information in bilingual visual word recognition:
543
Evidence from ERPs. Neuropsychologia, 103, 183–190.
544
https://doi.org/10.1002/cncr.27633.Percutaneous
545
23 | Page
Jaeger, T. F. (2008). Categorical data analysis: Away fron ANOVAs. Journal of Memory and
546
Language, 59(4), 434–446. https://doi.org/10.1016/j.jml.2007.11.007.Categorical
547
Kartushina, N., Frauenfelder, U. H., & Golestani, N. (2016). How and When Does the Second
548
Language Influence the Production of Native Speech Sounds: A Literature Review.
549
Language Learning, 66, 155–186. https://doi.org/10.1111/lang.12187
550
Keuleers, E., & Brysbaert, M. (2010). Wuggy: A multilingual pseudoword generator. Behavior
551
Research Methods, 42(3), 627–633. https://doi.org/10.3758/BRM.42.3.627
552
Krogh, L., Vlach, H. A., & Johnson, S. P. (2013). Statistical learning across development: Flexible
553
yet constrained. Frontiers in Psychology, 3(JAN), 1–11.
554
https://doi.org/10.3389/fpsyg.2012.00598
555
Kroll, J. F., Dussias, P. E., Bice, K., & Perrotti, L. (2015). Bilingualism , Mind , and Brain. Annual
556
Review Linguistics, 1(1), 377–394. https://doi.org/10.1146/annurev-linguist-030514-
557
124937
558
Linck, J. A., Kroll, J. F., & Sunderman, G. (2009). Linck, J. A., Kroll, J. F., & Sunderman, G. (2009).
559
Losing access to the native language while immersed in a second language: Evidence for
560
the role of inhibition in second-language learning. Psychological Science, 20(12), 1507–
561
1515. https://doi.org/10.1038/mp.2011.182.doi
562
Lobo, A., Ezquerra, J., Gómez, F. B., Sala, J. M., & Seva, A. D. (1979). ognocitive mini-test (a
563
simple practical test to detect intellectual changes in medical patients). Actas Luso-
564
Espanolas de Neurologia, Psiquiatria y Ciencias Afines, 7(3), 189–202.
565
Macwhinney, B. (2007). A Unified Model A Unified Model. Social Sciences.
566
Miller, G. A., Bruner, J. S., & Postman, L. (1954). Familiarityofletter
567
sequencesandtachistoscopicidentification. The Journal of General Psychology, 50(1), 129–
568
139.
569
Oganian, Y., Conrad, M., Aryani, A., Spalek, K., & Heekeren, H. R. (2015). Activation patterns
570
throughout the word processing network of L1-dominant bilinguals reflect language
571
similarity and language decisions. Journal of Cognitive Neuroscience, 27(11), 2197–2214.
572
Oganian, Yulia, Conrad, M., Aryani, A., Heekeren, H. R., & Spalek, K. (2016). Interplay of bigram
573
frequency and orthographic neighborhood statistics in language membership decision.
574
Bilingualism, 19(3), 578–596. https://doi.org/10.1017/S1366728915000292
575
Owsowitz, S. E. (1963). TheEffectsofWordFamiliarityandLetterStructure
576
FamiliarityonthePerceptionofWords. SantaMonica,CA: RandCorporation Publications.
577
Perea, M., Urkia, M., Davis, C. J., Agirre, A., Laseka, E., & Carreiras, M. (2006). E-Hitz: A word
578
frequency list and a program for deriving psycholinguistic statistics in an agglutinative
579
language (Basque). Behavior Research Methods, 38(4), 610–615.
580
https://doi.org/10.3758/BF03193893
581
Ramos, S., Fernández García, Y., Antón, E., Casaponsa, A., & Duñabeitia, J. A. (2017). Does
582
learning a language in the elderly enhance switching ability? Journal of Neurolinguistics,
583
43, 39–48. https://doi.org/10.1016/j.jneuroling.2016.09.001
584
Schmid, M. S., & Köpke, B. (2017). The relevance of first language attrition to theories of
585
bilingual development. Linguistic Approaches to Bilingualism, 7(6), 637–667.
586
24 | Page
Segalowitz, N. (1991). Does Advanced Skill in a Second Language Reduce Automaticity in the
587
First Language? Language Learning, 41(1), 59–83. https://doi.org/10.1111/j.1467-
588
1770.1991.tb00676.x
589
Vaid, J., & Frenck-Mestre, C. (2002). Do orthographic cues aid language recognition? A
590
laterality study with French-English bilinguals. Brain and Language, 82(1), 47–53.
591
https://doi.org/10.1016/S0093-934X(02)00008-1
592
Van Kesteren, R., Dijkstra, T., & de Smedt, K. (2012). Markedness effects in Norwegian–English
593
bilinguals: Task-dependent use of language-specific letters and bigrams. The Quarterly
594
Journal of Experimental Psychology, 65(11), 2129–2154.
595
Ware, C., Damnee, S., Djabelkhir, L., Cristancho, V., Wu, Y. H., Benovici, J., … Rigaud, A. S.
596
(2017). Maintaining cognitive functioning in healthy seniors with a technology-based
597
foreign language program: A pilot feasibility study. Frontiers in Aging Neuroscience,
598
9(MAR), 1–10. https://doi.org/10.3389/fnagi.2017.00042
599
600
