RESEARCH ARTICLE

Clustering and Switching in Verbal Fluency Across
Varying Degrees of Cognitive Control Demands:
Evidence From Healthy Bilinguals and
Bilingual Patients With Aphasia

Erin Carpenter

, Claudia Peñaloza

, Leela Rao, and Swathi Kiran

Aphasia Research Laboratory, Department of Speech, Language and Hearing Sciences, Sargent College of Health &
Rehabilitation Sciences, 波士顿大学, 波士顿, 嘛, 美国

关键词: bilingual aphasia, semantic executive control, language control, lexical access, semantic
fluency, phonemic fluency

抽象的

Different linguistic contexts place varying amounts of cognitive control on lexical retrieval in
bilingual speakers, an issue that is complicated in bilingual patients with aphasia (BPWA) 到期的
to subsequent language and cognitive deficits. Verbal fluency tasks may offer insight into the
interaction between executive and language control in healthy bilinguals and BPWA, 经过
examining conditions with varying cognitive control demands. The present study examined
switching and clustering in verbal fluency tasks in BPWA and healthy bilinguals across single-
and dual-language conditions. We also examined the influence of language processing and
language proficiency on switching and clustering performance across the dual-language
状况. Thirty-five Spanish-English BPWA and twenty-two Spanish-English healthy
bilinguals completed a language use questionnaire, tests of language processing, and two
verbal fluency tasks. The semantic category generation task included four conditions: 二
single-language conditions (No-Switch L1 and No-Switch L2) that required word production in
each language separately; one dual-language condition that allowed switching between
languages as desired (Self-Switch); and one dual-language condition that required switching
between languages after each response (Forced-Switch). The letter fluency task required word
production in single-language contexts. 全面的, healthy bilinguals outperformed BPWA
across all measures. Results indicate that switching is more sensitive to increased control
demands than clustering, with this effect being more pronounced in BPWA, underscoring the
interaction between semantic executive processes and language control in this group.
此外, for BPWA switching performance relies on a combination of language abilities
and language experience metrics.

介绍

Understanding the relationship between different levels of control during language production
in bilingual speakers remains a central issue in the field of bilingualism. Lexical retrieval in
healthy bilingual speakers may require a combination of language control processes as well
as semantic executive control processes to manage competition from the unintended language

开放访问

杂志

引文: Carpenter, E., Peñaloza, C。,
饶, L。, & Kiran, S. (2021). Clustering
and switching in verbal fluency across
varying degrees of cognitive control
需要: Evidence from healthy
bilinguals and bilingual patients with
aphasia. Neurobiology of Language,
2(4), 532–557. https://doi.org/10.1162
/nol_a_00053

DOI:
https://doi.org/10.1162/nol_a_00053

已收到: 10 二月 2021
公认: 31 八月 2021

利益争夺: 作者有
declared that competing interests exist.

通讯作者:
Erin Carpenter
evc5102@bu.edu

处理编辑器:
Karen Emmorey

版权: © 2021
麻省理工学院
在知识共享下发布
归因 4.0 国际的
(抄送 4.0) 执照

麻省理工学院出版社

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

Language control:
The management of competition
from language task schemas via
inhibitory control mechanisms for
appropriate production of the
intended language.

Semantic executive control:
The use of control mechanisms that
direct activation and inhibition of
lexical and phonological
representations at the lexical level.

Aphasia:
An acquired language disorder
affecting language comprehension
and production following acquired
brain injury.

as well as competing lexical representations. The relationship between these different levels of
control is further complicated in bilingual patients with aphasia (BPWA) as impairments to the
language system make it difficult to differentiate the two levels of control in this group. Verbal
fluency tasks provide an interesting opportunity to differentiate these two levels of control in
both healthy bilinguals and BPWA by examining quantitative and qualitative performance
across tasks with different language demands.

When a speaker wants to produce a word in one language (例如, 狗), semantically related
representations are also activated (例如, 猫) via spreading activation. Once the lexical nodes
are activated, activation spreads downward to the corresponding phonological units, which in
turn send activation upward to phonologically related lexical representations (例如, the lexical
node dog sends activation to its corresponding phonological units /d-o-g/, which in turn acti-
vate the phonologically related node dot) (柯林斯 & Loftus, 1975; 戴尔, 1986; 戴尔 &
O’Seaghdha, 1992). This mechanism requires semantic executive control processes to manage
activation of concepts and inhibit unintended nodes. These semantic executive control pro-
cesses direct and control lexical activation in a contextually appropriate manner (Jefferies
等人。, 2008). For bilinguals, semantic representations activate lexical nodes in both languages
in parallel even when only one language is in use (例如, dog activates cat in English, 也
perro and gato in Spanish) (Bialystok et al., 2009; Colomé, 2001; Costa et al., 2000; Kroll
等人。, 2006), a phenomenon which is modulated by relative proficiency in each language
(Dijkstra & van Heuven, 1998, 2002; Dijkstra et al., 2019; Kroll & 斯图尔特, 1994). Since both
languages are simultaneously activated, inhibitory control mechanisms are required to man-
age competition from the non-target language for successful language production (绿色的,
1998; 绿色的 & Abutalebi, 2013). 所以, in bilingual speakers, control mechanisms are
required at two levels, first at the level of language task schemas, which designate the lan-
guage of intended use, 第二个, at the level of lexical competitors, which aid in selection
of the appropriate lexical node.

More recent theories propose that different linguistic contexts require varying amounts of
认知过程, including inhibition, goal maintenance, conflict monitoring, and interfer-
ence suppression for successful language use (绿色的 & Abutalebi, 2013). The Adaptive
Control Hypothesis (ACH; 绿色的 & Abutalebi, 2013) outlines three interactional contexts
for bilingual speakers, which consist of single-language, dual-language, and dense code-
switching environments. In single-language contexts, speakers use one language exclusively,
which requires continuous inhibition of the non-target language in order to avoid cross-
language intrusions. In dual-language contexts, bilinguals switch between their two languages
in a constrained manner (IE。, in responses to external environmental language cues), typically
when communicating with different speakers in the same environment. This context, 所以,
requires increased control processes to inhibit competitors from the non-target language.
反过来, in dense code-switching environments, bilinguals switch between their two lan-
guages as desired (IE。, in the absence of external environmental language constraints), requir-
ing more opportunistic planning rather than direct inhibition of the non-intended language. 在
这个上下文, speakers opportunistically use joint language activation, which allows them to
make use of alternative forms of expression in whichever language is most readily available.
One main assumption of the ACH is that in the single-language and dual-language contexts,
the two languages are in a competitive relationship, while in the dense code-switching context
the two languages are in a cooperative one (绿色的 & Abutalebi, 2013). Because of this, 这
dual-language context is thought to place the highest cognitive control demands on language
加工, whereas the dense code-switching context is thought to place the least.
此外, for unbalanced bilinguals, higher levels of activation in the dominant language,

Neurobiology of Language

533

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

leads to increased inhibitory demands when speaking in the less-dominant single-language
语境 (绿色的, 1998).

Within the context of verbal fluency tasks, which has been extensively studied in bilingual
个人 (Friesen et al., 2015; Kiran et al., 2014; Luo et al., 2010; Patra et al., 2020A; 罗伯茨
& Le Dorze, 1998) and is the focus of the current study, the ACH provides a framework to
predict the influence of interactional contexts on the degree of control mechanisms required
to successfully complete the task (Carpenter et al., 2020; Jevtovic et al., 2020). Within this
框架, we can examine two levels of control in bilingual speakers. The first is language
控制, which manages competition from language task schemas via inhibitory control mech-
anisms for appropriate production of the intended language within the parameters of different
interactional contexts. This may be examined by imposing different language demands across
状况 (例如, requiring responses only in one language vs. requiring switching between
languages for each response). The second is semantic executive control, which directs activa-
tion and inhibition of lexical and phonological representations at the lexical level and may be
examined by analyzing qualitative aspects of verbal fluency performance (例如, switching and
clustering performance).

数字 1 aims to differentiate the control mechanisms required at the language control and
semantic executive control levels in the three interactional contexts outlined by the ACH
(绿色的 & Abutalebi, 2013) as they apply to verbal fluency tasks described in this article.
第一的, when healthy bilinguals produce items in the single-language context (represented by
the No-Switch context in Figure 1A), inhibition originates from the level of the language task
schemas, where activated concepts in the unintended language (in this case L2) are inhibited
at the language control level rather than at the semantic executive control level. 在这种情况下,
semantic executive control is required to inhibit the co-activated L1 lexical representations
仅有的. When subjects are allowed to freely switch between languages for consecutive items,
as in the dense code-switching context, speakers make use of whatever lexical nodes are most
readily available to facilitate lexical access through spreading activation (例如, dog and possi-
bly gato). In this instance, there are no language constraints, 因此, no inhibitory processes
are imposed at the language level, 反而, inhibitory mechanisms at the semantic executive
control level are required to inhibit non-intended lexical nodes from both languages (看到
Self-Switch context in Figure 1B). 最后, in the dual-language context (represented by the
Forced-Switch context in Figure 1C), which requires controlled switching between target lan-
guages (例如, dog and then gato), inhibitory mechanisms are required from both the language
control and semantic executive control levels. 在这种情况下, language control is required to
inhibit the previously activated L2 language task schema, and since representations at the lex-
ical level were previously activated to produce an item in the L2, semantic executive control is
required to inhibit these activated representations. This results in high cognitive demands on
language control and semantic executive control mechanisms. 最后, semantic pro-
cessing may become more effortful as cognitive processes are recruited to successfully inhibit
the non-target language while simultaneously managing competition from lexical
陈述.

The relationship between these different control levels in bilingual speakers remains a cen-
tral issue in the field of bilingualism; 然而, this issue is further complicated in BPWA, 在
whom language and cognitive control deficits are common following focal brain damage
(Gray & Kiran, 2016; Helm-Estabrooks, 2002). To better understand the complex relationship
between language control and semantic executive control, especially in BPWA, we assert that
analyzing switching and clustering performance during verbal fluency tasks provides a means
for teasing apart these two levels of control. Clustering is measured as the number of lexical

Neurobiology of Language

534

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

数字 1.
(A) Language control and semantic executive control schema for bilingual speakers in the No-Switch (NS) 语境, where speakers
use one language exclusively. Inhibition is represented with subtraction signs. Automatic activation is represented with arrows; forward ac-
tivation is shown using solid lines, and backward activation is shown using dashed lines. Activation for the intended word, 狗, is represented
with addition signs. (乙) Language control and semantic executive control schema in the Self-Switch (SS) context where bilingual speakers
switch between languages in an unconstrained manner (IE。, in the absence of external environmental language cues), which allows speakers
to switch between languages as desired. Inhibition is represented with subtraction signs. Automatic activation is represented with arrows;
forward activation is shown using solid lines, and backward activation is shown using dashed lines. Activation for the intended word, 狗,
is represented with addition signs. (C) Language control and semantic executive control schema in the Forced-Switch (FS) 语境, 在哪里
bilingual speakers switch between languages in a constrained manner (IE。, in response to external environmental cues). Inhibition is repre-
sented with subtraction signs. Automatic activation is represented with arrows; forward activation is shown using solid lines, and backward
activation is shown using dashed lines. Activation for the intended word, 狗, is represented with addition signs.

Neurobiology of Language

535

Clustering and switching in bilingual aphasia

items produced within a given semantic subcategory in semantic category generation and suc-
cessively producing words with overlapping phonemic properties in letter fluency. Switching
is a change from one semantic subcategory or phonemic cluster to another during a verbal
fluency task (数字 2). Clustering reflects the relatively automatic processes of spreading ac-
tivation to related concepts from an activated concept within a given subcategory (休斯 &
Bryan, 2002; Troyer et al., 1997; Unsworth et al., 2011), and is therefore associated more with
the lexical component of verbal fluency tasks as it relies upon verbal memory and accessing
and using the word store (Troyer et al., 1997). 反过来, switching is the ability to generate
new clusters. This measure reflects strategic and controlled shifts between subcategories, 和
因此, is thought to measure, to a greater extent, semantic executive control abilities
(Gruenewald & Lockhead, 1980; 休斯 & Bryan, 2002; Raboutet et al., 2010; 罗森
等人。, 2005; Unsworth et al., 2011). Studies have shown that effective performance on verbal
fluency tasks requires a balance between clustering and switching processes (Troyer et al.,
1997).

Previous studies investigating switching and clustering performance in patients with apha-
是 (PWA) have shown that monolingual PWA produce fewer correct words, smaller cluster
sizes, and fewer switches than healthy monolinguals (Bose et al., 2017), reflecting difficulties
in both lexical (fewer words and smaller clusters) and executive (fewer switches) 成分
of this task. Comparably, Kiran and colleagues (2014) found that BPWA produce reduced
number of words, smaller mean semantic cluster sizes, and switch fewer times across both
languages compared to healthy bilinguals. 然而, BPWA showed a dissociation between
switching and clustering performance (例如, switching was more impaired than clustering),
indicating that BPWA may show different degrees of impairment in automatic versus con-
trolled search mechanisms (Kiran et al., 2014). These studies highlight the need for research
to investigate the interaction between these two variables and how they may be differentially
impacted with changing language demands of the task.

Recent studies have used a fluency difference score (FDS) as a measure of cognitive control
ability in verbal fluency tasks (Friesen et al., 2015; Patra et al., 2020A). FDS is calculated by
taking the difference in performance between the category generation and letter fluency tasks

数字 2.
Schema of clustering and switching processes during verbal fluency tasks. This figure represents an example of clustering and
switching processes during a semantic category generation task using the category “animals.” The first cluster consists of the subcategory
“felines” where the speaker first produces the subcategory exemplar lion; in turn related concepts are then activated via automatic spreading
activation, which leads the speaker to produce the exemplars tiger and panther. Once this subcategory is exhausted, the speaker must dis-
engage from that cluster, implement controlled search processes to find another item to produce, and reengage with a new cluster, 哪里的
process begins again.

Neurobiology of Language

536

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

as a proportion of correct responses on the category generation task. Given that letter fluency
is thought to have higher cognitive control demands while semantic category generation is
thought to place greater emphasis on lexico-semantic abilities (Luo et al., 2010; Patra et al.,
2020乙; Shao et al., 2014), a smaller FDS is thought to be indicative of better cognitive control
能力 (Friesen et al., 2015; Patra et al., 2020乙). A recent study by Patra et al. (2020乙) 成立
that BPWA produced fewer correct words, had larger FDSs, and switched fewer times as com-
pared to healthy bilinguals; 然而, both groups demonstrated similar cluster scores.
此外, they found that BPWA with better inhibitory control produced more correct
words and a greater number of semantic switches, further supporting the role of cognitive con-
trol in switching performance.

To expand these early and interesting results, the present study examines switching and
clustering strategies in healthy bilinguals and BPWA with the goal of disentangling the poten-
tial interactions among language proficiency, bilingual language and semantic executive con-
控制, and focal brain damage. Previous work has examined the influence of varying cognitive
control demands in semantic category generation performance in both healthy bilinguals
(Jevtovic et al., 2020) and BPWA (Carpenter et al., 2020). These studies implemented four lan-
guage conditions (two single-language conditions and two dual-language conditions), 作为一个
way of varying the cognitive control demands of the task. The single-language conditions con-
sisted of two No-Switch (NS-L1 and NS-L2) 状况, where participants produced lexical
items within a given semantic category using one language only (Figure 1A). The first dual-
language condition was the Self-Switch (SS) 状况, where participants switched between
languages as desired (Figure 1B). 最后, the second dual-language condition was the Forced-
Switch (FS) 状况, in which participants were required to switch languages after each re-
响应 (Figure 1C). Both studies found superior performance in the SS and NS-L1 conditions
compared to the NS-L2 and FS conditions for the healthy bilinguals, and Carpenter and col-
联赛 (2020) additionally found that BPWA demonstrated comparable performance to
healthy bilinguals only in the SS condition. These results highlight that BPWA were more sen-
sitive to the impact of cognitive control on lexical retrieval in the conditions where the two
languages are in a competitive relationship (IE。, NS-L1, NS-L2, and FS conditions) 但不在
the condition where the two languages are in a cooperative relationship (IE。, SS condition),
suggesting that increased control demands arising from the level of language control may
hinder communicative effectiveness in BPWA.

所以, in the current study, we examined verbal fluency switching and clustering per-
formance within four different language contexts (NS-L1, NS-L2, SS, and FS) as a method for
teasing apart language and semantic executive control processes in BPWA and healthy bilin-
guals. The following research questions were posed.

(我) How do cognitive control abilities differ between L1 and L2? 第一的, we hypothesized that
healthy bilinguals would demonstrate smaller FDSs than BPWA, reflective of control
impairments following acquired brain injury (ABI). 此外, it was expected that
FDSs would be larger in L2 compared to L1, indicating greater control abilities for both
groups in L1.

(二) How does switching and clustering performance differ for BPWA and healthy bilinguals
across the different conditions of the semantic category generation task? Given the find-
ings of previous research (Kiran et al., 2014; Patra et al., 2020A), it was hypothesized
that healthy bilinguals would produce larger mean semantic cluster sizes and a greater
number of switches compared to BPWA, reflective of varying degrees of language im-
pairment in the BPWA group. When looking at performance across conditions, 我们

Neurobiology of Language

537

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

anticipated that clustering performance would be comparable across the four condi-
tions for both groups as it is a measure of automatic spreading activation and should
therefore be minimally impacted by increasing language control demands. 反过来,
we expected that in the conditions where both languages are in a competitive relation-
船 (IE。, in the NS-L1, NS-L2, and FS conditions), switching performance would be
compromised for BPWA, reflective of competing control processes between language
task schemas and semantic executive control processes.

(三、) How does switching and clustering performance differ for BPWA and healthy bilinguals
across the different conditions of the letter fluency task? Like the semantic category gen-
eration task, it was hypothesized that healthy bilinguals would outperform BPWA on
both switching and clustering measures. 此外, it was expected that, 尽管
healthy bilinguals would perform better in LF-L1, BPWA would demonstrate reduced
performance in both languages, reflective of the increased semantic executive control
demands imposed by this task, as semantic competition during phonological searching
requires more controlled search processes rather than reliance on automatic activation
of related concepts.

(四号) What factors predict clustering and switching performance in the SS and FS conditions? 它
was hypothesized that there would be a dissociation between measures that contributed to
switching and clustering performance for both groups. 具体来说, standardized assess-
ment scores including scores on the Pyramids and Palm Trees test (PAPT; 霍华德 &
帕特森, 1992) and selected subtests of the Bilingual Aphasia Test (BAT; Paradis,
1989) would better predict clustering performance in both conditions, as higher activation
of lexical representations would lend itself to better ability to extend semantic clusters.
此外, it was hypothesized that FS switching performance would be predicted by
language experience measures (language use questionnaire, or LUQ, 指标) 如-
creased proficiency would lead to greater efficiency at implementing language control,
resulting in less competition between the language control and semantic executive control
级别. 最后, for BPWA, it was hypothesized that Raven’s Coloured Progressive Matrices
(RCPM; Kertesz, 2006) scores, a measure of matrix reasoning (Fong et al., 2020), 会
be predictive of performance on FS switching performance as better nonverbal cognitive
function would lead to more controlled search processes as task demands increase.

材料和方法

参加者

Thirty-five Spanish-English BPWA (19 女性; mean age = 52.9, 标准差= 16.4; mean education =
14.9, 标准差= 2.7; mean age of acquisition (AoA) of the second language (L2) = 10.5, 标准差= 7.9) 和
22 age-matched Spanish–English healthy bilinguals (19 女性; mean age = 47.2; 标准差= 15.4)
participated in this study [t(1, 55) = −1.3268, p = 0.19]. While the groups were matched for
年龄, there were between group differences in education (healthy bilinguals mean = 17.6, 标准差=
5.2) [t(1, 51) = 2.5041, p = 0.016]) and L2 AoA (healthy bilinguals mean = 17.0, 标准差= 11.9)
[t(1, 52) = 2.4968, p = 0.016]). 尽管如此, 25 BPWA and 21 healthy bilinguals reported
Spanish as their first-acquired language (L1), so both groups consisted of mainly L1 Spanish
speakers. All BPWA were at least 6 months post-onset (mean MPO = 61.0, 标准差= 86.2) 和
presented with aphasia secondary to stroke (n = 33), traumatic brain injury (n = 1), or tumor
(n = 1). All participants gave their written informed consent.

Participants completed an LUQ (Kastenbaum et al., 2019), which captured information
regarding each participant’s language use, exposure, and self-rated proficiency for each

Neurobiology of Language

538

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

language separately over their lifetime. 具体来说, the LUQ obtained information regarding
L2 AoA; language usage on weekdays and weekends measured on an hourly basis; 寿命
exposure for hearing, 请讲, and reading in each language; lifetime confidence for hear-
英, 请讲, and reading in each language; family proficiency for parents and siblings;
educational history regarding language use and preferences of the participant and peers dur-
ing elementary school, high school, and college; and language ability rating of each partic-
ipant’s self-rated abilities in speaking, listening, 阅读, and writing for each language.
Language usage and language ability ratings in each language were collected both pre-
and post-ABI for the BPWA to reflect changes in these metrics after aphasia onset. 为了
BPWA, caregivers were present to corroborate LUQ metrics including patients’ language
use patterns pre- and post-ABI, language abilities ratings, and exposure in each language.
桌子 1 summarizes the LUQ information of the BPWA and the healthy bilinguals.
Independent samples t tests revealed no differences in their levels of relative language ex-
经历. 桌子 2 summarizes lesion information and aphasia subtypes of the BPWA.

Standardized Assessments

Participants also completed additional language testing in English and Spanish that included
(A) a 60-item naming screener consisting of high frequency and concrete items (Peñaloza,
Grasemann, 等人。, 2019) 和 (乙) the Boston Naming Test (BNT; 卡普兰等人。, 2001;
Kohnert et al., 1998) to assess picture naming abilities; (C) selected subtests of the
Psycholinguistic Assessment of Language Processing in Aphasia (PALPA; Kay et al., 1992)
and its Spanish translation (EPLA; Kay et al., 1995) including PALPA 47/ EPLA 45, PALPA
48/ EPLA 46, PALPA 49/ EPLA 47 and PALPA 50/EPLA 48 to assess lexico-semantic process-
英; (d) four subtests of the BAT (Paradis, 1989) including semantic categories, synonyms, 一个-
tonyms I, and antonyms II to assess semantic processing; 和 (e) the PAPT test (霍华德 &
帕特森, 1992) to assess nonverbal semantic knowledge. 此外, matrix reasoning
was evaluated only in the BPWA using the RCPM subtest from the Western Aphasia

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

桌子 1.

Average percentage of the Language Use Questionnaire metrics for bilingual patients with aphasia and healthy bilinguals.

Pre-ABI
LAR

Post-ABI
LAR

Pre-ABI
language use

Post-ABI
language use

团体
BPWA n

L1
L2
30 30 34

L1
32

L2
32

Lifetime
exposure
L2
L1

Lifetime
confidence
L2
L1

Family
proficiency
L2
L1

Educational
历史

意思是 (%) 93.5

81.6

59.5 49.7 51.6

48.4

65.0

35.1 57.6

42.4

91.7

61.7

94.5

58.6

72.2

27.8

标清

11.3

17.2

21.3 20.2 30.0

30.0

27.0

27.0 20.4

20.4

13.6

23.4

12.3

32.6

26.9

N/A N/A 22

意思是 (%) 98.0

83.1 N/A N/A 45.5

54.5

N/A N/A 66.6

33.4

97.6

49.5

96.9

43.7

81.5

18.5

标清

5.9

16.4 N/A N/A 33.4

33.4

N/A N/A 18.8

18.8

5.5

27.5

12.7

25.8

25.2

0.091 0.742 N/A N/A 0.292 0.291 N/A N/A

0.104 0.104 0.059 0.086 0.509 0.085 0.207 0.085

p
价值

笔记. LAR = language ability rating; BPWA = bilingual patients with aphasia; HB = healthy bilingual; ABI = acquired brain injury; N/A = no data available.

Neurobiology of Language

539

L2 Aphasia
subtype

全球的

Broca’s

Clustering and switching in bilingual aphasia

桌子 2.

Lesion and clinical information for bilingual patients with aphasia.

Code
BPWA1

Sex MPO
年龄
82 中号 411

Lesion information

Span. Right posterior parieto-occipital infarct

L1 WAB
AQ
55.7

L1 Aphasia
subtype

Broca’s

L2 WAB
AQ
29.6

BPWA2

BPWA3

BPWA4

44 中号

BPWA5

BPWA6

BPWA7

BPWA8

BPWA9

48 中号

Span.

Left border zone infarcts between the

74.1

Anomic

68.5

ACA and MCA

Eng. Multiple CVAs secondary to

93.4

Anomic

81.4

Anomic

Moyamoya disease

Eng.

Left MCA CVA

Span.

Left CVA

Span.

Left CVA

130

Span.

Left tumor in posterior centrum semiovale

Span.

Left CVA

Span. TBI

89.8

Anomic

N/A

27.3

77.5

N/A

Broca’s

Anomic

N/A

84.5

20.7

37.3

Anomic

全球的

Broca’s

67.6 Wernicke’s

N/A

BPWA10

66 中号 339

Span.

Left MCA CVA

BPWA11

BPWA12

53 中号

Span.

Left MCA temporoparietal infarct

79.1

Conduction

54.4

Broca’s

Span.

Left CVA

51.3 Wernicke’s

47.5 Wernicke’s

BPWA13

104

Eng.

Left frontoparietal hemorrhage

BPWA14

77 中号

BPWA15

BPWA16

70 中号

BPWA17

BPWA18

BPWA19

Span.

Left MCA CVA involving precentral gyrus

Span.

Left MCA hemorrhage

Span.

Left frontal lobe CVA

Span.

Left CVA

Eng.

Left MCA and PCA CVAs

Span.

Left CVA

BPWA20

54 中号

Span.

Subacute left MCA CVA

BPWA21

69 中号

BPWA22

BPWA23

BPWA24

46 中号

BPWA25

56 中号

BPWA26

39 中号

BPWA27

42 中号

BPWA28

62 中号

Eng.

Left CVA

Eng.

Left CVA

Span.

Left anterior MCA infarct

Span. TIA

Span.

Left CVA

Span.

Left CVA

Eng.

Left CVA

Eng.

Left CVA

98.2

67.4

78.9

57.3

72.3

N/A

Anomic

Broca’s

Anomic

Conduction

Anomic

N/A

35.9 Wernicke’s

96.5

82.4

Anomic

N/A

64.7

76.8

39.8

66.4

68.8

69.8

Broca’s

Conduction

Broca’s

Conduction

N/A

46.5

60.8

71.2

Conduction

Broca’s

Conduction

N/A

81.2

Anomic

94.6

89.2

Broca’s

Anomic

39.5

57.8

78.6

Anomic

Broca’s

Anomic

BPWA29

164

Span.

Left CVA, right occipital and right

N/A

cerebellar infarcts

Neurobiology of Language

540

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

桌子 2.

(continued )

Code
BPWA30

年龄
21

Sex MPO
23
F

L1
Span.

Left CVA

Lesion information

BPWA31

64 中号

Eng.

Left CVA, Hx of right CVA

BPWA32

BPWA33

56 中号

BPWA34

BPWA35

Span.

Left MCA CVA

Span.

Left CVA

Span.

Left MCA CVA

Eng.

Left MCA CVA

L1 WAB
AQ
34.4

47.7

L1 Aphasia
subtype

Broca’s

Conduction

92.6

Anomic

N/A

L2 WAB
L2 Aphasia
AQ
subtype
53.3 Wernicke’s

16.4

70.4

97.2

N/A

Broca’s

Conduction

Anomic

N/A

笔记. Aphasia subtypes are based on WAB-R (Kertesz, 2006) classifications; for more detailed information regarding classification of aphasia subtypes refer
to the WAB-R scoring manual. MPO = months post-onset; L1 = first-acquired language; Span. = Spanish; Eng. = English; ACA = anterior cerebral artery;
MCA = middle cerebral artery; CVA = cerebrovascular accident; TBI = traumatic brain injury; Hx = prior medical history; PCA = posterior cerebral artery;
TIA = transient ischemic attack; WAB = Western Aphasia Battery; AQ = Aphasia Quotient; N/A = no data available; L2 = second-acquired language.

Battery–Revised ( WAB-R; Kertesz, 2006). A summary of performance on standardized mea-
sures for BPWA and healthy bilinguals is included in Table 3.

Experimental Tasks

Two verbal fluency tasks were administered: a modified semantic category generation task
(IE。, producing items in a given semantic category in 60 s) and a traditional letter fluency task
(IE。, producing items beginning with a given letter in 60 s in both languages). In the category
generation task (Carpenter et al., 2020), participants were to produce words in four semantic
类别: 动物, 衣服, 食物, and modes of transportation. The conditions for the cate-
gory generation task consisted of (A) two No-Switch conditions (NS-L1 and NS-L2) where par-
ticipants produced responses in one language exclusively, (乙) one Self-Switch condition (SS)

桌子 3.

Average percentage of correct responses on standardized measures for bilingual patients with aphasia and healthy bilinguals.

Naming screener

BNT

PALPA/EPLA composite

BAT composite

Nonverbal assessments

Verbal assessments

团体
BPWA n

意思是 (%)

56.5

标清

29.1

意思是 (%)

84.6

标清

9.6

42.7

29.0

78.6

15.5

44.1

26.2

76.8

15.2

32.8

24.0

62.8

18.0

82.6

10.0

93.2

6.9

74.8

13.2

87.5

8.1

69.1

26.7

92.7

11.0

64.6

20.3

83.5

17.0

89.2

10.7

94.4

5.7

PAPT
29

RCPM
25

p value

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.044 74.5 16.9 N/A N/A N/A N/A Note. BPWA = Bilingual patients with aphasia; HB = healthy bilingual; BNT = Boston Naming Test; PALPA/EPLA = Psycholinguistic Assessment of Language Processing in Aphasia in English and Spanish respectively; BAT = Bilingual Aphasia Test; PAPT = Pyramids and Palm Trees; RCPM = Raven’s Coloured Progressive Matrices; L1 = first-acquired language; L2 = second-acquired language; N/A = no data available. Neurobiology of Language 541 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Clustering and switching in bilingual aphasia where participants could switch between languages as desired, and (c) one Forced-Switch condition (FS) where participants were required to switch between languages after each re- sponse. Each condition was administered twice across two separate days. On day one, instruc- tions were administered only in L1 and tasks NS-L1, NS-L1, SS, and FS were administered, and on day two all instructions were provided in L2, and tasks NS-L2, NS-L2, SS, and FS were ad- ministered. The category-to-condition assignments were counterbalanced across participants to account for the potential impact of semantic category knowledge on condition performance. In the letter fluency task, participants were to produce words beginning with the letters “F,” “A,” and “S” in English (Controlled Oral Word Association Test; Benton & Hamsher, 1976) and “P,” “M,” and “R” in Spanish (Peña-Casanova et al., 2009), as these letter combinations are fre- quently administered and show similar norms across English and Spanish respectively. The order of administration for the letter fluency task was counterbalanced across participants. Both tasks have been described in detail elsewhere (Carpenter et al., 2020). Data Coding and Scoring Procedures All data including bilingual language history metrics, performance on standardized language assessments, and performance on the semantic category generation task were coded as being produced in each participant’s first-acquired L1 or second-acquired L2 as self-reported in their LUQ. Responses in the verbal fluency tasks were recorded in audio and in written form during the testing session and verified for accuracy afterwards via the audio recording. In the semantic category generation task, since each condition was administered twice, scores were created for each condition by averaging the total number of correct words across the two administra- tions. Responses were correct if they were unique words in the target category, produced in the target language, were not a repetition of a previously produced response, and contained no more than one phonemic substitution, omission, or addition. Fluency difference scores were computed as the difference in the number of correct re- sponses between category generation and letter fluency tasks as a proportion of correct re- sponses in the category generation task, to reflect the role of cognitive control in verbal fluency tasks (Friesen et al., 2015; Patra et al., 2020b). FDSs were computed in each lan- guage by using the number of correct responses in the NS and LF conditions in L1 and L2 separately. Switching and clustering analyses closely followed the methods outlined by Troyer et al. (1997), where semantic clustering was defined as successively producing words that shared a semantic subcategory (e.g., dog, cat (pets) or horse, pig, cow (farm animals)), and phonemic clustering was defined as successively generating words that fulfilled any one of the following criteria: (i) words that began with the same first two sounds (e.g., art, arm), (ii) words that dif- fered only by a vowel sound regardless of the actual spelling (e.g., fat, fit, foot), (iii) words that rhymed (e.g., sand, stand ), or (iv) words that were homonyms (e.g., sum, some). Cluster size was calculated starting with the second word in each cluster, such that a single word (e.g., dog) was given a cluster size of zero, two-word clusters (e.g., dog, cat) were given a cluster size of one, and so on. Mean cluster size was calculated by adding the size of each cluster and dividing the total score by the number of clusters, such that dog, cat, horse, pig, cow would receive a mean cluster size of 1.5 (a cluster size of 1 for dog, cat plus a cluster size of 2 for horse, pig, cow averaged to get 1.5). Switches were calculated by tallying the number of tran- sitions between clusters, including single words (e.g., dog, cat, horse, pig, cow, dolphin would contain two switches, after cat and after cow). Errors and repetitions were included in both the clustering and switching analyses. Neurobiology of Language 542 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Clustering and switching in bilingual aphasia Statistical Analyses First, to assess cognitive control abilities in L1 and L2, we conducted a two-way repeated mea- sures ANOVA with Group (BPWA and HB) as the between-subject factor and Language (L1 and L2) as the within-subject factor for FDSs (research question 1). To compare the mean semantic cluster size and number of switches produced by the two groups across the four experimental conditions of the semantic category generation task (research question 2), two 2-way repeated measures ANOVAs were conducted with Group (BPWA and HB) as the between-subject factor and Condition (SS, NS-L1, NS-L2, and FS) as the within-subject factor for mean semantic cluster size and number of semantic switches separately. Similarly, for research question 3, similar analyses were conducted with Group as the between-subject and Condition (LF-L1 and LF-L2) as the within-subject factor for both mean phonemic cluster size and number of phonemic switches. Additionally, since different between-group effects were expected for different conditions, planned paired-samples t tests were conducted across conditions for each group separately, even if no significant interaction effect was observed. Furthermore, to examine a main effect of Group, planned independent samples t tests were conducted to investigate specific differences between groups in each condition. All p values were corrected for multiple comparisons using the FDR approach. To determine which pre- and post-ABI LUQ factors loaded onto different components (re- search question 4), principal component analyses (PCAs) were first conducted for BPWA and healthy bilinguals separately as done in previous research (Kastenbaum et al., 2019; Peñaloza, Barrett, & Kiran, 2019). These PCAs were performed for L1 and L2 separately for a larger sam- ple of BPWA (n = 59) and healthy bilinguals (n = 31) that overlapped with this project (Marte et al., 2021). (Results of the PCA for the entire sample of BPWA can be found at https://www .bu.edu/aphasiaresearch/resources/principal-component-analysis-of-bilinguals-with-aphasia/ and for healthy bilinguals at https://www.bu.edu/aphasiaresearch/resources/principal -component-analysis-of-luq-metrics-in-bilingual-healthy-controls/.) Once these PCAs were performed, individual factor scores were extracted for the present cohort of BPWA (n = 32) and healthy bilinguals (n = 22) for whom data were available. A varimax normalized factor rotation was used to examine the factor loadings of components with factor loadings >0.6, 和
the scores were extracted using the default regression function from the psych package in R
(https://cran.r-project.org/web/packages/psych/index.html). Individual factor scores were then
included in four linear regression models along with language scores (L1 BAT, L2 BAT, 和
PAPT) for both groups and RCPM scores for BPWA only, to determine which factors predicted
performance on SS average cluster size, SS number of switches, FS average cluster size, and FS
number of switches. The regression analyses were only conducted in the dual-language con-
版本, as we were interested in teasing apart the impact of language factors that contributed
to performance in the condition with the least demands of cognitive control (例如, SS) com-
pared to the condition with the highest demands of cognitive control (例如, FS). Backward step-
wise regressions were conducted to determine which factors best contributed to the models.
These four linear regressions were conducted for BPWA and healthy bilinguals separately.

结果

While the number of correct responses was not the focus of the current study and has
previously been reported elsewhere for a subset of patients included in this study
(Carpenter et al., 2020), we have summarized performance on the category generation
and letter fluency tasks below as they provide the basis for results reported in this article
(桌子 4).

Neurobiology of Language

543

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

桌子 4. Number of correct responses produced across the two verbal fluency tasks.

BPWA (n = 35)

HB (n = 22)

Condition
SS

NS-L1

NS-L2

LF-L1

LF-L2

意思是
7.34

7.63

5.71

5.31

4.01

3.02

标清
5.10

4.52

4.87

4.25

3.38

3.52

意思是
18.05

16.68

15.50

12.64

13.11

10.51

标清
6.43

6.51

5.32

5.95

3.97

3.23

p value
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 Cohen’s d 1.85 1.61 1.92 1.42 2.47 2.22 Note. BPWA = bilingual patients with aphasia; HB = healthy bilinguals; SS = Self-Switch; NS-L1 = No Switch (L1); NS-L2 = No Switch (L2); FS = Forced-Switch; LF-L1 = Letter Fluency, L1; LF-L2 = Letter Fluency, L2. BPWA produced significantly fewer correct responses than healthy bilinguals across all conditions. Pairwise comparisons revealed superior performance in the SS condition compared to the NS-L2 ( p = 0.004) and FS ( p < 0.001) conditions, superior perfor- mance in the NS-L1 condition compared to the NS-L2 ( p = 0.012) and FS conditions ( p < 0.001), and superior performance in the NS-L2 condition compared to the FS condition ( p = 0.019). No differences were observed between the SS and NS-L1 condition ( p = 0.268). Additionally, participants showed superior performance in LF-L1 compared to LF-L2 ( p < 0.001). Fluency Difference Score 2 = 0.127] and Language [F(1, 54) = 5.113, p = 0.028, (cid:1) p Results from the repeated-measures ANOVA revealed a significant main effect of Group [F(1, 54) = 7.847, p = 0.007, (cid:1) 2 = 0.087]; p however, the Group × Language interaction was not significant [F(1, 54) = 0.356, p = 0.553, (cid:1)p 2 = 0.007], indicating that the BPWA had significantly larger FDSs than healthy bilinguals, and that overall FDSs in L1 were smaller compared to L2 ( p = 0.002) for both groups (Figure 3). Independent samples t tests revealed that BPWA produced significantly larger FDSs than l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d / . l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Figure 3. Fluency difference scores (FDSs) of the bilingual patients with aphasia (BPWA) and healthy bilinguals (HB). Boxplots are shown separately for each group across the two languages (L1 and L2). BPWA showed significantly larger FDSs relative to the HB in both languages (statistical significance at the 0.05 level). Neurobiology of Language 544 Clustering and switching in bilingual aphasia healthy bilinguals in L1 [t(54) = 2.209, p = 0.031, p-adj = 0.031, d = 0.618] and L2 [t(54) = −2.644, p = 0.011, p-adj = 0.022, d = 0.484]. Semantic Clustering and Switching Analysis Semantic clustering Results from the repeated-measures ANOVA revealed a significant main effect of Group [F(1, 55) = 6.721, p = 0.012, (cid:1) 2 = 0.095], although the main effect of Condition [F(3, 53) = 0.957, p p = 0.420, (cid:1) 2 = 0.045] and the Group × Condition interaction were not significant [F(3, 53) = p 0.616, p = 0.608, (cid:1) 2 = 0.076], indicating that BPWA produced smaller mean semantic cluster p sizes than healthy bilinguals, although groups did not demonstrate differences in performance across conditions (Figure 4). Independent samples t tests revealed that BPWA produced signif- icantly smaller average semantic cluster sizes than healthy bilinguals in the SS [t(55) = −2.291, p = 0.026, p-adj = 0.037, d = 0.648], NS-L1 [t(55) = −2.256, p = 0.028, p-adj = 0.037, d = 0.591], and NS-L2 [t(55) = −2.491, p = 0.016, p-adj = 0.037, d = 0.676] conditions; however no group differences were observed in the FS condition [t(55) = −0.514, p = 0.609, p-adj = 0.609, d = 0.146], indicating comparable mean semantic cluster sizes across groups in this condition. Semantic switching 2 = 0.490] and Condition [F(3, 53) = 6.140, p = 0.001, (cid:1) p Results from the repeated-measures ANOVA revealed a significant main effect of Group [F(1, 55) = 53.870, p < 0.001, (cid:1) 2 = 0.258], p although the Group × Condition interaction was not significant [F(3, 53) = 0.966, p = 0.416, (cid:1) 2 = 0.052], indicating that BPWA produced fewer semantic switches than healthy bilinguals p and performance differed across conditions for both groups (Figure 5). Specifically, pairwise comparisons revealed significantly more switches produced in the SS compared to the NS-L2 ( p = 0.015) and FS ( p = 0.002) conditions, and significantly more switches produced in the l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d / . l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Figure 4. Mean semantic cluster sizes of the bilingual patients with aphasia (BPWA) and healthy bilinguals (HB) on the semantic category generation task. Boxplots are shown separately for each group across the four conditions (SS, NS-L1, NS-L2, and FS). BPWA showed signif- icantly smaller mean semantic cluster sizes relative to the HB in all conditions except for the FS condition. Neurobiology of Language 545 Clustering and switching in bilingual aphasia Figure 5. Number of semantic switches produced by the bilingual patients with aphasia (BPWA) and healthy bilinguals (HB) on the semantic category generation task. Boxplots are shown separately for each group across the four conditions (SS, NS-L1, NS-L2, and FS). BPWA produced significantly fewer semantic switches relative to the HB in all conditions. l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 NS-L1 condition compared to the NS-L2 ( p = 0.001) and FS ( p = 0.004) conditions. No signif- icant differences in number of switches were observed between the SS and NS-L1 conditions and the NS-L2 and FS conditions ( p > 0.05 in all cases). Independent samples t tests revealed
that BPWA produced significantly fewer switches than healthy bilinguals in all conditions
( p-adj < 0.001 in all cases). To further investigate performance across conditions within each group, paired-samples t tests were performed for BPWA and healthy bilinguals separately. Results revealed that BPWA produced significantly more switches in SS compared to FS [t(34) = 2.517, p = 0.017, p-adj = 0.034] and NS-L1 compared to both NS-L2 [t(34) = 3.276, p = 0.002, p-adj = 0.006] and FS [t(34) = 3.688, p = 0.001, p-adj.= 0.006]; however no differences arose between SS and NS-L1 [t(34) = 1.586, p = 0.112, p-adj = 0.168], SS and NS-L2 [t(34) = 1.322, p = 0.195, p-adj = 0.234], or NS-L2 and FS [t(34) = 0.993, p = 0.328, p-adj = 0.328]. Healthy bilinguals did not significantly differ across any conditions after correcting for multiple com- parisons ( p-adj > 0.192 in all cases).

Phonemic Switching and Clustering Analysis

Phonemic clustering

Results from the repeated-measures ANOVA revealed a significant main effect of Group [F(1,
54) = 11.257, p < 0.001, (cid:1) 2 = 0.164], p and Group × Condition interaction [F(1, 54) = 5.900, p = 0.019, (cid:1) 2 = 0.098], indicating that, p overall, the BPWA produced smaller mean phonemic cluster sizes than healthy bilinguals and only healthy bilinguals showed superior performance in LF-L1 compared to LF-L2 (Figure 6). 2 = 0.173], Condition [F(1, 54) = 10.560, p = 0.002, (cid:1) p Phonemic switching Results from the repeated-measures ANOVA revealed a significant main effect of Group [F(1, 54) = 76.290, p < 0.001, (cid:1) 2 = 0.568], although the main effect of Condition [F(1, 54) = p 0.349, p = 0.557, (cid:1) 2 = 0.006] and the Group × Condition interaction were not significant p Neurobiology of Language 546 Clustering and switching in bilingual aphasia Figure 6. Mean phonemic cluster sizes of the bilingual patients with aphasia (BPWA) and healthy bilinguals (HB) on the letter fluency task. Boxplots are shown separately for each group across the two conditions (LF-L1 and LF-L2). BPWA produced significantly smaller mean pho- nemic cluster sizes relative to the HB in the LF-L1 condition but not the LF-L2 condition. [F(1, 54) = 1.140, p = 0.290, (cid:1) 2 = 0.021], indicating that BPWA produced fewer phonemic p switches than healthy bilinguals and groups demonstrated similar performance across condi- tions (Figure 7). Independent samples t tests revealed that BPWA produced significantly fewer switches in both conditions ( p-adj < 0.001 in both cases). l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Figure 7. The number of phonemic switches produced by the bilingual patients with aphasia (BPWA) and healthy bilinguals (HB) on the letter fluency task. Boxplots are shown separately for each group across the two conditions (LF-L1 and LF-L2). BPWA showed significantly fewer phonemic switches relative to the HB in both conditions. Neurobiology of Language 547 Clustering and switching in bilingual aphasia Factors That Predict Performance on FS and SS For BPWA, the PCA conducted for L1 revealed three components with eigenvalues greater than 1 that explained 72% of the variance. Using a factor loading threshold of 0.6, it was de- termined that Component 1 represents L1 Background, Component 2 represents L1 Use and Exposure, and Component 3 represents L1 Environment. A varimax normalized factor rotation was used to examine the factor loadings for the first three components. The PCA conducted for L2 revealed two components with eigenvalues greater than 1 that explained 63% of the var- iance. The factor loadings for L2 indicate that Component 1 represents L2 Background and Environment, and Component 2 represents L2 Use and Exposure. Again, a varimax normalized factor rotation was used to examine the factor loadings for the first two components. (Table 5). For healthy bilinguals, the PCA conducted for L1 revealed one component with an eigen- value greater than 1 that explained 50% of the variance. Again, using a threshold of 0.6, the factor loadings indicate that Component 1 represents L1 Background and Exposure. A varimax normalized factor rotation was used to examine which factors loaded ono this component. The PCA conducted for L2 revealed two components with eigenvalues greater than 1 that ex- plained 72% of the variance. The factor loadings for L2 indicate that Component 1 represents L2 Background and Environment, and Component 2 represents L2 Use and Exposure. Again, a varimax normalized factor rotation was used to examine which factors loaded onto these components (Table 6). As Table 7 shows, backward stepwise regressions were conducted for each of the measures for BPWA. Results of the first multiple linear regression including PAPT, L1 BAT, L1 Use and Exposure, and L2 Background and Environment explained 64% of the variance of SS cluster size in BPWA [F(4, 8) = 6.251, p = 0.014], with L1 BAT as the only significant predictor in the model. Results of the second multiple linear regression including L2 BAT, RCPM, L1 Use and Exposure, L2 Background and Environment, and L2 Use and Exposure explained 71% of the variance of SS switches [F(5, 7) = 6.738, p = 0.013], with L2 BAT, L1 Use and Exposure, L2 Background and Environment, and L2 Use and Exposure significantly predicting SS switches; l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Table 5. Factor loadings for L1 and L2 Language Use Questionnaire totals for BPWA. RC 1 (Background) – L1 RC 2 (Use/Exposure) – RC 3 (Environment) – RC 1 (Background/Environment) −0.84 RC 2 (Use/Exposure) −0.17 L2 0.85 0.46 −0.02 0.29 −0.08 0.74 0.50 0.72 29% 0.23 −0.08 0.90 0.83 0.18 0.41 0.68 0.03 28% −0.01 0.58 −0.05 0.14 0.89 0.08 0.21 0.08 15% 0.46 0.01 0.15 0.15 0.86 0.50 0.52 0.85 33% 0.63 0.58 0.75 0.77 0.07 0.58 0.69 0.19 31% L2 AoA Pre-ABI LAR Post-ABI LAR Pre-ABI use Post-ABI use Family proficiency Education Exposure Confidence Variance Note. Factor loadings greater than 0.6 are in bold. L2 AoA = L2 age of acquisition; ABI = acquired brain injury; LAR = language ability rating; RC = rotated component. Neurobiology of Language 548 Clustering and switching in bilingual aphasia Table 6. Factor loadings for L1 and L2 Language Use Questionnaire totals for healthy bilinguals. L1 RC 1 (Background/Exposure) – L2 RC 1 (Background/Environment) −0.81 RC 2 (Use/Exposure) −0.17 0.84 0.50 0.39 0.94 0.81 0.81 50% 0.21 −0.05 0.80 0.64 0.58 0.69 37% 0.83 0.77 −0.04 0.58 0.72 0.58 36% L2 AoA LAR Use Family proficiency Education Exposure Confidence Variance Note. Factor loadings greater than 0.6 are in bold. AoA = age of acquisition; LAR = language ability ratings; RC = rotated component. however, L2 Background and Environment did not survive corrections for multiple compari- sons. Results of the third multiple linear regression including PAPT, L1 BAT, L2 BAT, RCPM and L1 Environment explained 50% of the variance of FS cluster size [F(5, 7) = 3.355, p = 0.073]; however, no individual predictors were significant ( p values > 0.05 in all cases).
Results of the fourth multiple linear regression including PAPT, L1 BAT, L2 BAT, L1
Background, L2 Background and Environment, and L2 Use and Exposure explained 33% 的
the variance of FS switches [F(6, 6) = 1.964, p = 0.216], with PAPT, L1 Background, L2
Background and Environment, and L2 Use and Exposure significantly predicting FS switches.

For the healthy bilinguals, 桌子 8 shows results of the first multiple linear regression in-
cluding L2 Background and Environment that explained 7% of the variance of SS cluster size
[F(1, 20) = 2.558, p = 0.125], and no individual predictors were significant. Results of the
second multiple linear regression including PAPT, L1 Background and Exposure, and L2
Background and Environment explained 50% of the variance of SS switches [F(3, 18) =
7.983, p = 0.001], with PAPT significantly predicting SS switches. Results of the third multiple
linear regression including PAPT and L1 BAT explained 7% of the variance of FS cluster size
[F(2, 19) = 1.782, p = 0.195], and no individual predictors were significant. Results of the
fourth multiple linear regression including L1 BAT scores explained 17% of the variance of
FS switches [F(1, 20) = 7.299, p = 0.033], with L1 BAT significantly predicting FS switches;
然而, this did not survive corrections for multiple comparisons.

讨论

The purpose of this study was to examine how varying degrees of cognitive control demands
on lexical retrieval impacted performance in two verbal fluency tasks in healthy bilinguals and
BPWA using four language contexts. The conditions implemented in the semantic category
generation task included two No-Switch conditions (NS-L1 and NS-L2), where participants
responded only in one language, one Self-Switch condition (SS), where participants switched
between languages as desired, and one Forced-Switch condition (FS), where participants were
required to switch languages after each response. 此外, participants completed a tra-
ditional letter fluency task in each language separately (LF-L1 and LF-L2). 全面的, we found
that healthy bilinguals, 一般来说, outperformed BPWA across all measures. This is consistent

Neurobiology of Language

549

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

桌子 7. Multiple linear regression results for BPWA.

Independent variable

Dependent variable
SS cluster size

(Intercept)

PAPT

L1 BAT

乙
1.278

−2.684

SE
1.564

1.969

t
0.817

−1.363

Pr (> |t|)
0.438

0.210

p-adj
0.438

0.263

3.478

0.941

3.697

0.006**

0.030*

L1 Component 2 (Use/Exposure)

−0.544

0.255

−2.130

L2 Component 1 (Background/Environment)

0.595

0.300

SS switches

(Intercept)

L2 BAT

RCPM

L1 Component 2 (Use/Exposure)

4.724

2.240

6.042

2.098

−5.465

−3.673

3.173

0.773

1.987

2.109

2.880

−1.722

−4.753

0.066

0.082

0.073

0.137

0.087

0.024*

0.047*

0.129

0.002**

0.013*

L2 Component 1 (Background/Environment)

1.280

0.497

2.577

0.037*

0.055

L2 Component 2 (Use/Exposure)

FS cluster size

(Intercept)

PAPT

L1 BAT

L2 BAT

RCPM

L1 Component 3 (环境)

FS switches

(Intercept)

PAPT

L1 BAT

L2 BAT

−3.459

−2.063

0.905

2.782

8.248

5.069

4.680

2.846

3.715

2.887

−3.823

−0.742

1.627

1.645

1.287

−11.672

5.352

−2.181

0.377

0.330

16.089

4.672

1.143

3.443

−17.338

6.018

−2.881

8.273

4.191

1.974

−7.574

4.146

−1.827

L1 Component 1 (Background)

2.611

0.968

L2 Component 1 (Background/Environment)

2.690

0.858

L2 Component 2 (Use/Exposure)

2.451

0.863

2.697

3.134

2.839

0.007**

0.020*

0.482

0.148

0.144

0.239

0.066

0.291

0.012*

0.028*

0.096

0.118

0.036*

0.020*

0.030*

0.482

0.296

0.349

0.296

0.349

0.050*

0.112

0.117

0.050*

笔记. All significant results are marked in bold. SS = Self-Switch; FS = Forced-Switch; L1 BAT = Bilingual Aphasia Test, L1; L2 BAT = Bilingual Aphasia Test, L2;
PAPT = Pyramids and Palm Trees; RCPM = Raven’s Coloured Progressive Matrices. p-adj = adjusted p value. *p < 0.05; **p < 0.01. Adjusted p values were calculated via the Benjamini Hochberg procedure using the Multcomp package in R (https://cran.r-project.org/web/packages/multcomp/index.html). with previous research (Carpenter et al., 2020; Kiran et al., 2014; Patra et al., 2020a) and re- flects varying degrees of language impairment in the BPWA group. First, we examined FDSs in L1 and L2 as a measure of cognitive control abilities and found that both groups demonstrated smaller FDSs (indicative of better control) in L1 compared to L2. Our second aim examined semantic clustering and switching performance across the four conditions of the semantic cat- egory generation task. No group differences were found in mean semantic cluster size across the four conditions. Additionally, while healthy bilinguals did not differ in the number of switches produced across conditions, BPWA produced more switches in SS compared to FS Neurobiology of Language 550 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Clustering and switching in bilingual aphasia Table 8. Multiple linear regression results for healthy bilinguals. Dependent variable SS cluster size (Intercept) Independent variable B 1.995 SE 0.139 t 14.334 Pr (> |t|)
<0.001*** p-adj <0.001*** L2 Component 1 (Background/Environment) 0.250 0.156 1.599 0.125 0.125 SS switches (Intercept) −18.343 6.170 −2.973 0.008** 0.0163* PAPT 27.133 6.497 4.176 <0.001*** 0.002** L1 Component 1 (Background/Exposure) 0.838 0.438 1.911 L2 Component 1 (Background/Environment) FS cluster size (Intercept) FS switches PAPT L1 BAT (Intercept) L1 BAT −0.790 −2.929 0.436 3.238 −1.811 −0.905 9.486 5.111 1.856 −4.478 −1.804 2.660 3.730 −1.683 −0.484 0.072 0.087 0.377 0.079 0.109 0.634 9.176 3.995 2.297 0.033* 0.087 0.087 0.377 0.163 0.163 0.634 0.065 Note. All significant results are marked in bold. SS = Self-Switch; FS = Forced-Switch; L1 BAT = Bilingual Aphasia Test, L1; L2 BAT = Bilingual Aphasia Test, L2; PAPT = Pyramids and Palm Trees (Howard & Patterson, 1992). *p < 0.05; **p < 0.01; ***p < 0.001. Adjusted p values were calculated via the Benjamini Hochberg procedure using the Multcomp package in R (https://cran.r-project.org/web/packages/multcomp/index.html). and NS-L1 compared to NS-L2 and FS. Our third aim examined phonemic clustering and switching performance in the two conditions of the letter fluency task. The results of this anal- ysis showed group differences in mean phonemic cluster sizes produced in the two conditions, indicating that healthy bilinguals performed better in LF-L1 compared to LF-L2 while BPWA showed similar performance across conditions. Additionally, no differences were observed in the number of switches between the two conditions for either group. Finally, our fourth aim examined which language experience measures (LUQ metrics) and standardized assessment scores predicted switching and clustering performance in the two dual-language conditions (SS and FS) for both groups separately. For both groups, switching performance was more de- pendent on language factors than clustering, with BPWA switching performance relying on both language experience and standardized assessment scores, while healthy bilingual perfor- mance relied solely on standardized assessment scores. The main findings of this study are reported in Table 9. First, in regard to cognitive control abilities across L1 and L2, we found that healthy bilin- guals demonstrated better control than BPWA overall, as evidenced by smaller FDSs. Additionally, both groups demonstrated superior performance in L1 compared to L2. This was anticipated, as better control demands in L1 are reflective of less inhibitory processes re- quired to suppress the less-dominant L2. For BPWA, these results are consistent with reduced switching performance observed in NS-L2 compared to NS-L1, as in this condition BPWA are less able to manage increased control demands arising from the level of language control in order to inhibit their more dominant L1, leading to reduced ability to systematically search within the lexicon for new lexical candidates. Second, in line with previous research (Kiran et al., 2014), BPWA showed smaller mean semantic cluster sizes and number of switches compared to the healthy bilinguals. Of note, both groups did not differ in terms of mean semantic cluster size produced across the four Neurobiology of Language 551 l D o w n o a d e d f r o m h t t p : / / d i r e c t . m i t . e d u n o / l / l a r t i c e - p d f / / / / 2 4 5 3 2 1 9 7 9 6 9 8 n o _ a _ 0 0 0 5 3 p d . / l f b y g u e s t t o n 0 7 S e p e m b e r 2 0 2 3 Clustering and switching in bilingual aphasia Table 9. Summary of main findings. Measure Fluency difference scores Between group effects Across condition effects HB > BPWA

L1 > L2

Mean semantic cluster size

HB > BPWA (except FS)

No differences

Number of semantic switches

HB > BPWA in all conditions

SS > FS; NS-L1 > NS-L2 and FS for BPWA only

Mean phonemic cluster size

HB > BPWA in LF-L1

LF-L1 > LF-L2 for HB only

Number of phonemic switches

HB > BPWA in both conditions

LF-L1 = LF-L2 for HB and BPWA

SS clusters

SS switches

FS clusters

FS switches

Predicted by standardized assessment scores for BPWA.

Predicted by standardized assessment scores for HB and BPWA.
Predicted by language experience measures for BPWA.

Not predicted by any measures for either group.

Predicted by standardized assessment scores for HB and BPWA.
Predicted by language experience measures for BPWA.

笔记. BPWA = bilingual patients with aphasia; HB = healthy bilinguals; SS = Self-Switch; NS-L1 = No Switch (L1); NS-L2 = No Switch (L2); FS = Forced-Switch.

conditions of the semantic category generation task. This comparable clustering performance
across the four conditions was expected, as clustering reflects the automatic process of
spreading activation and therefore is less likely to be impacted by increased language con-
trol demands. 第二, when examining switching performance, differences were observed
in number of switches produced across the four conditions. 尤其, while healthy bi-
linguals demonstrated comparable switching performance across the four conditions, BPWA
produced more switches in the SS condition compared to FS condition, as well as in the NS-
L1 compared to the NS-L2 and FS conditions. These results highlight the interaction between
language control and semantic executive control underlying lexical retrieval across the dif-
ferent conditions. 具体来说, in the FS condition (Figure 1C), participants are tasked with
switching between languages for each new item produced, which requires speakers to in-
hibit competition across two levels of control, language control and semantic executive con-
控制. Results indicate that the increased top-down inhibitory control demands arising from the
level of language control in order to inhibit the previously activated language, impedes
BPWA’s abilities to implement semantic executive control processes needed for successful
semantic switching performance. 此外, BPWA showed reduced switching perfor-
mance in the NS-L2 condition also, highlighting that BPWA may have more difficulty resolv-
ing conflict at the language control level in their weaker language (Figure 1A). 具体来说,
compared to the NS-L1 condition, in the NS-L2 condition, speakers are required to use
larger amounts of inhibitory processes at the language control level in order to suppress
the dominant L1 in favor of the weaker L2. For BPWA, damage to the language system
reduces the amount of cognitive resources they are able to dedicate to each level of control;
therefore in the NS-L2 condition, the increased resources being allocated to the language
control level reduces the ability of BPWA to implement the controlled search mechanisms
required to initiate a new cluster. 全面的, these results are consistent with findings reported
by Troyer and colleagues (1997), where a divided attention condition led to a reduced num-
ber of switches but not clustering, reinforcing that switching and clustering are dissociable
components of verbal fluency tasks and increased cognitive demands may impact controlled
流程 (switching) more than automatic processes (clustering). 此外, 为了

Neurobiology of Language

552

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

individuals with focal brain damage, reduced cognitive resources leads to worse perfor-
mance not only in the condition with the highest cognitive control demands (IE。, FS), 但
also in the condition with moderate cognitive control demands (IE。, NS-L2).

第三, in the letter fluency task, we found again that healthy bilinguals overall produced
larger mean phonemic cluster sizes and switched more times compared to BPWA.
有趣的是, the interaction effect observed for these mean phonemic cluster sizes indicated
that only healthy bilinguals produced larger phonemic cluster sizes in L1 compared to L2,
although neither group differed in number of switches produced in either language.
Performance of the healthy bilinguals directly contrasts the results of the semantic category
generation task, perhaps reflecting the distinctions in search mechanisms on letter fluency
tasks compared to semantic category generation tasks. As noted in the Introduction, 期间
phonological retrieval in letter fluency tasks, suppression of semantic competition in favor
of phonologically related words requires more controlled processes rather than reliance on
automatic activation of related concepts (数字 2). These increased processing demands
would be further impacted by proficiency as the L2 phonological system is generally less well
developed than the L1 phonological system (Dijkstra & van Heuven, 1998, 2002; Dijkstra
等人。, 2019; Kroll & 斯图尔特, 1994), leading to less readily available lexical representations.
Since the healthy bilinguals demonstrated larger mean phonemic cluster sizes in the LF-L1
condition compared to the LF-L2, but BPWA did not, it appears that the BPWA were more
sensitive to the increased demands imposed by this task. While healthy bilinguals were able
to make greater use of semantic executive control processes to systematically search within
the lexicon for phonemically related concepts in L1, where there is less competition arising
from the level of language control, BPWA demonstrated reduced clustering performance in
both L1 and L2.

最后, the regression analyses revealed differences in factors that contributed to switching
and clustering performance across conditions and groups. 第一的, for the BPWA, L1 BAT scores
significantly contributed to SS clustering performance indicating that when language is not a
约束, clustering performance solely depends on language abilities (as measured by stan-
dardized assessments), rather than language experience. 第二, L2 BAT scores, L1 Use and
Exposure, and L2 Use and Exposure significantly contributed to SS number of switches pro-
推导出来的. These results indicate that the higher an individual’s L2 BAT scores the more they
switched between semantic subcategories, suggesting that greater semantic ability aids in sys-
tematically searching the mental lexicon for new lexical candidates once a semantic subcat-
egory has been exhausted. 此外, the negative slopes for Use and Exposure in L1 and L2
suggest that as language experience increases, the number of semantic switches produced by
BPWA declines. This indeed may indicate that individuals with greater language experience
are able to further extend semantic subcategories, therefore reducing the need to switch.

更远, FS clustering performance was not predicted by any standardized assessment
scores or language experience measures. This finding was surprising, as it was expected that
better performance on standardized assessments would lend itself to clustering performance
when cognitive demands increased. 然而, it may indeed be the case that the more errors
and overall fewer responses elicited by this condition may contribute to a floor effect in the
数据. 最后, FS switching performance was significantly predicted by PAPT scores, L1
Background, L2 Background and Environment, and L2 Use and Exposure. These results indi-
cate that greater nonverbal semantic knowledge (PAPT scores) and language experience in
both languages benefits switching performance when the language demands of the task in-
折痕. 进一步来说, semantic switching in the FS condition places the highest control
demands on both language control and semantic executive control (see Figure 1C and

Neurobiology of Language

553

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

数字 2). BPWA, who present with reduced cognitive resources to allocate to different levels
of control, are less able to manage competition between language control and semantic ex-
ecutive control. 所以, 在这个组内, it may be possible that their L1 and L2 language
experience aids in compensating for these deficits by dedicating fewer cognitive resources to
language control, allowing them to better implement semantic executive control processes
required when attempting to switch between clusters.

For the healthy bilinguals, SS and FS clustering was not predicted by standardized assessment
scores or language experience measures. This finding may relate to the fact that when there are
no language constraints, automatic spreading activation and not language experience may drive
the outcome. 第二, PAPT scores significantly predicted switching performance on the SS
condition underscoring the notion that nonverbal semantic knowledge (as measured by PAPT)
lends itself to more readily available items when switching subcategories. 最后, FS switching
was not predicted by any measures after correcting for multiple comparisons.

Of note, some interesting patterns arose when looking at the regression results across the
two groups. 第一的, switching performance, regardless of condition was dependent on standard-
ized assessment scores for both healthy bilinguals and BPWA, whereas clustering performance
was not predicted by standardized assessment scores or language experience measures. 这些
results suggest that the controlled search processes required of switching are highly dependent
on the strength of semantic and lexical representations and access to them. 尤其, BPWA
switching performance was dependent on language experience measures in the FS condition.
This indicates that for BPWA, switching performance when language demands of the task in-
crease relies on both standardized assessment scores and language experience in L1 and L2.
This may underscore the relationship between impairment and language proficiency in BPWA
and the importance of language control for successful production, as reduced language con-
trol abilities lead to greater competition with semantic executive control processes, 所以
disrupting lexical production in BPWA. BPWA with greater language experience in L1 and L2
may be better at resolving conflict at the level of language task schemas in order to implement
controlled semantic executive control processes for better semantic switching performance.
This is consistent with the findings of Carpenter et al. (2020), which highlights that increased
language control demands may, at least in part, lead to reduced communicative success in this
人口. This highlights the importance of understanding language control in BPWA and
how it may contribute to more functional communication outcomes post-ABI.

One limitation of this study is that for BPWA, RCPM performance did not make a significant
contribution to any of the predictive models, suggesting that RCPM, a measure of matrix rea-
soning (Fong et al., 2020), may not sufficiently tap the executive functions intrinsically related
to the mechanisms used in switching and clustering performance. Future studies should care-
fully select measures of executive functions as to better underscore the mechanisms used across
the different conditions. Of note, this study consisted of primarily Spanish dominant Spanish-
English bilinguals. 然而, follow-up analyses to examine whether there was an effect of par-
ticipants’ reported L1 on their verbal fluency performance found no differences in switching
and clustering scores when separately grouping individuals by their reported L1 (Spanish or
英语). This was consistent when examining just the BPWA group. While there were fewer
L1 English individuals in this study, the lack of differences suggests that both groups were per-
forming comparably with the L1 Spanish individuals across the four conditions, suggesting
these results may be generalizable to English dominant Spanish–English bilinguals. 此外,
healthy bilinguals had higher education overall compared to BPWA, which may influence their
performance on verbal fluency tasks; future studies may consider using education as a covariate
when investigating verbal fluency performance in similar populations. 最后, lower verbal

Neurobiology of Language

554

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

fluency scores for the BPWA in general may have contributed to a potential floor effect in the
数据. 然而, reanalysis of the patient data including only individuals with higher verbal flu-
ency scores did not change the overall results.

结论

The present findings suggest that switching and clustering are dissociable components of ver-
bal fluency performance, with switching being more sensitive to the impact of cognitive con-
trol demands. 具体来说, as language demands of the task increased, BPWA’s abilities to
implement semantic executive control was hindered. 此外, switching and clustering
performance may differ on semantic category generation tasks as compared to letter fluency
tasks due the underlying mechanisms required of each task. 此外, language abilities
may play a greater role in the ability to implement controlled processes (switching) 而不是
automatic ones (clustering) for both healthy bilinguals and BPWA. 重要的, for BPWA stan-
dardized assessment scores as well as language experience in L1 and L2 greatly influence se-
mantic switching abilities when language demands of the task increase, highlighting the
importance of language control for successful lexical production post-ABI.

致谢

This work was supported by the National Institute on Deafness and Other Communication
Disorders of the National Institutes of Health (grant U01DC014922) awarded to Swathi
Kiran. The content is solely the responsibility of the authors and does not necessarily represent
the official views of the National Institutes of Health.

资金信息

Swathi Kiran, National Institute on Deafness and Other Communication Disorders (https://dx
.doi.org/10.13039/100000055), 奖项ID: U01DC014922.

作者贡献
Erin Carpenter: 数据管理: 平等的; 形式分析: 带领; 方法: 平等的; Writing –
original draft: 带领; 写作——复习 & 编辑: 平等的. Claudia Peñaloza: 概念化:
平等的; 数据管理: 平等的; 调查: 平等的; 方法: 平等的; 项目管理:
平等的; 写作——复习 & 编辑: 配套. Leela Rao: 概念化: 平等的; 数据
curation: 平等的; 方法: 平等的. Swathi Kiran: 概念化: 带领; Funding acqui-
位置: 带领; 调查: 带领; 方法: 带领; 项目管理: 带领; Writing –
review & 编辑: 带领.

COMPETING INTERESTS

Swathi Kiran serves as a consultant for Constant Therapy Health with no scientific overlap with
the present study. All other authors have no financial or nonfinancial conflicts of interest.

参考

Benton, A. L。, & Hamsher, K. (1976). Multilingual aphasia examina-

的 (2nd 版。). AJA Associates.

Bialystok, E., Craik, F. 我. M。, 绿色的, D. W., & Gollan, 时间. H. (2009).
Bilingual minds. Psychological Science in the Public Interest, 10(3),
89–129. https://doi.org/10.1177/1529100610387084, 考研:
26168404

Bose, A。, Wood, R。, & Kiran, S. (2017). Semantic fluency in aphasia:
Clustering and switching in the course of 1 minute. 国际的
Journal of Language and Communication Disorders, 52(3), 334–345.
https://doi.org/10.1111/1460-6984.12276, 考研: 27767243
Carpenter, E., 饶, L。, Peñaloza, C。, & Kiran, S. (2020). Verbal flu-
ency as a measure of lexical access and cognitive control in

Neurobiology of Language

555

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

bilingual persons with aphasia. Aphasiology, 34(11), 1341–1362.
https://doi.org/10.1080/02687038.2020.1759774, 考研:
34366537

柯林斯, A. M。, & Loftus, 乙. F. (1975). A spreading-activation theory
of semantic processing. 心理评论, 82(6), 407–428.
https://doi.org/10.1037/0033-295X.82.6.407

Colomé, À. (2001). Lexical activation in bilinguals’ speech produc-
的: Language-specific or language-independent? 杂志
记忆与语言, 45(4), 721–736. https://doi.org/10.1006
/jmla.2001.2793

Costa, A。, Caramazza, A。, & Sebastian-Galles, 氮. (2000). The cog-
nate facilitation effect: Implications for models of lexical access.
实验心理学杂志: 学习, 记忆, 和
认识, 26(5), 1283–1296. https://doi.org/10.1037/0278
-7393.26.5.1283, 考研: 11009258

戴尔, G. S. (1986). A spreading-activation theory of retrieval in sen-
tence production. 心理评论, 93(3), 283–321. https://
doi.org/10.1037/0033-295X.93.3.283, 考研: 3749399

戴尔, G. S。, & O’Seaghdha, 磷. G. (1992). Stages of lexical access
in language production. 认识, 42(1–3), 287–314. https://
doi.org/10.1016/0010-0277(92)90046-K, 考研: 1582160
Dijkstra, T。, & van Heuven, 瓦. J. 乙. (1998). The BIA model and
bilingual word recognition. 在J. Grainger, A. 中号. Jacobs, & A.
Jacobs (编辑。), Localist connectionist approaches to human cog-
尼尼申 (PP. 189–225). Psychology Press.

Dijkstra, T。, & van Heuven, 瓦. J. 乙. (2002). The architecture of the
bilingual word recognition system: From identification to deci-
锡安. Bilingualism: Language and Cognition, 5(3), 175–197.
https://doi.org/10.1017/S1366728902003012

Dijkstra, T。, Wahl, A。, Buytenhuijs, F。, Van Halem, N。, Al-Jibouri, Z。,
De Korte, M。, & Rekké, S. (2019). Multilink: A computational
model for bilingual word recognition and word translation.
Bilingualism: Language and Cognition, 22(4), 657–679. https://
doi.org/10.1017/S1366728918000287

Fong, 中号. C. M。, Hui, 氮. Y。, Fung, 乙. S. W., Ma, 中号. K. H。, Law,
时间. S. T。, 王, X。, & 王, 瓦. S. (2020). Which cognitive func-
tions subserve clustering and switching in category fluency?
Generalisations from an extended set of semantic categories
using linear mixed-effects modelling. Quarterly Journal of
实验心理学, 73(12), 2132–2147. https://doi.org/10
.1177/1747021820957135, 考研: 32972306

Friesen, D. C。, Luo, L。, Luk, G。, & Bialystok, 乙. (2015). Proficiency
and control in verbal fluency performance across the lifespan for
monolinguals and bilinguals. 语言, 认识, 和
神经科学, 30(3), 238–250. https://doi.org/10.1080
/23273798.2014.918630, 考研: 25642427

Gray, T。, & Kiran, S. (2016). The relationship between language
control and cognitive control in bilingual aphasia. Bilingualism:
Language and Cognition, 19(3), 433–452. https://doi.org/10.1017
/S1366728915000061

绿色的, D. [W.] (1998). Mental control of the bilingual lexico-
semantic system. Bilingualism: Language and Cognition, 1(2),
67–81. https://doi.org/10.1017/S1366728998000133

绿色的, D. W., & Abutalebi, J. (2013). Language control in bilin-
guals: The adaptive control hypothesis. 认知杂志
心理学, 25(5), 515–530. https://doi.org/10.1080/20445911
.2013.796377, 考研: 25077013

Gruenewald, 磷. J。, & Lockhead, G. 右. (1980). The free recall of cat-
egory examples. 实验心理学杂志: 人类
Learning and Memory, 6(3), 225–240. https://doi.org/10.1037
/0278-7393.6.3.225

Helm-Estabrooks, 氮. (2002). Cognition and aphasia: A discussion
Journal of Communication Disorders, 35(2),

and a study.

171–186. https://doi.org/10.1016/S0021-9924(02)00063-1,
考研: 12036150

霍华德, D ., & 帕特森, K. (1992). Pyramids and palm trees test.

皮尔逊.

休斯, D. L。, & Bryan, J. (2002). Adult age differences in strategy
use during verbal fluency performance. Journal of Clinical and
Experimental Neuropsychology, 24(5), 642–654. https://doi.org
/10.1076/jcen.24.5.642.1002, 考研: 12187447

Jefferies, E., 帕特森, K., & Ralph, 中号. A. L. (2008). Deficits of
knowledge versus executive control in semantic cognition:
Insights from cued naming. Neuropsychologia, 46(2), 649–658.
https://doi.org/10.1016/j.neuropsychologia.2007.09.007,
考研: 17961610

Jevtovic, M。, Duñabeitia, J. A。, & de Bruin, A. (2020). 怎么办
bilinguals switch between languages in different interactional
上下文? A comparison between voluntary and mandatory lan-
guage switching. Bilingualism: Language and Cognition, 23(2),
401–413. https://doi.org/10.1017/S1366728919000191

卡普兰, E., 古德格拉斯, H。, & Weintraub, S. (2001). Boston naming

测试 (2nd 版。). Lippincott, 威廉姆斯, & Wilkins.

Kastenbaum, J。, Bedore, L。, Pena, E., 盛, L。, Mavis, 我。, Sebastian,
R。, & Kiran, S. (2019). The influence of proficiency and language
combination on bilingual lexical access. Bilingualism: 语言
a n d C o g n i t i o n 2 2 ( 2 ) , 1 – 3 1 . h t t p s : / / d o i . o r g / 1 0 . 1 0 1 7
/S1366728918000366, 考研: 30983875

Kay, J。, Coltheart, M。, & Lesser, 右. (1992). Psycholinguistic assess-
ments of language processing in aphasia. Psychology Press.
Kay, J。, Lesser, R。, & Coltheart, 中号. (1995). EPLA: Evaluación del pro-
cesamiento lingüístico en la afasia. 埃尔鲍姆 (英国) 泰勒 &
Francis.

Kertesz, A. (2006). Western aphasia battery—revised ( WAB-R).

PsychCorp. https://doi.org/10.1037/t15168-000

Kiran, S。, Blachandran, 我。, & Lucas, J. (2014). The nature of lexical-
semantic access in bilingual aphasia. Behavioural Neurology,
文章 389565. https://doi.org/10.1155/2014/389565, 考研:
24825956

Kohnert, K. J。, Hernandez, A. E., & Bates, 乙. (1998). Bilingual per-
formance on the Boston naming test: Preliminary norms in
Spanish and English. Brain and Language, 65(3), 422–440.
https://doi.org/10.1006/brln.1998.2001, 考研: 9843612

Kroll, J. F。, Bobb, S. C。, & Wodniecka, Z. (2006). Language selec-
tivity is the exception, 不是规则: Arguments against a fixed lo-
cus of language selection in bilingual speech. Bilingualism:
Language and Cognition, 9(2), 199–135. https://doi.org/10.1017
/S1366728906002483

Kroll, J. F。, & 斯图尔特, 乙. (1994). Category interference in translation
and picture naming: Evidence for asymmetric connections
between bilingual memory representations. Journal of Memory
and Language, 33(2), 149–174. https://doi.org/10.1006/jmla
.1994.1008

Luo, L。, Luk, G。, & Bialystok, 乙. (2010). Effect of language proficiency
and executive control on verbal fluency performance in bilinguals.
认识, 114(1), 29–41. https://doi.org/10.1016/j.cognition.2009
.08.014, 考研: 19793584

Marte, M。, Carpenter, E., Scimeca, M。, Falconer, 我。, Abdollahi, F。,
Peñaloza, C。, & Kiran, S. (2021). LEX-BADAT: Language experi-
ence in bilinguals with and without aphasia dataset [Manuscript
in preparation]. Department of Speech, Language and Hearing
科学, Sargent College of Health & Rehabilitation Sciences,
波士顿大学.

Paradis, 中号. (1989). The bilingual aphasia test. 埃尔鲍姆.
Patra, A。, Bose, A。, & Marinis, 时间. (2020A). Lexical and cognitive
underpinnings of verbal fluency: Evidence from Bengali-English

Neurobiology of Language

556

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
n
哦

我
/

我

A
r
t
我
C
e
–
p
d

F
/

2
4
5
3
2
1
9
7
9
6
9
8
n
哦
_
A
_
0
0
0
5
3
p
d

我

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Clustering and switching in bilingual aphasia

bilingual aphasia. Behavioral Sciences, 10(155), 1–28. https://土井
.org/10.3390/bs10100155, 考研: 33050055

Patra, A。, Bose, A。, & Marinis, 时间. (2020乙). Performance difference in
verbal fluency in bilingual and monolingual speakers.
Bilingualism: Language and Cognition, 23(1), 204–218. https://
doi.org/10.1017/S1366728918001098

Peña-Casanova, J。, Quiñones-Úbeda, S。, Gramunt-Fombuena, N。,
Quintana-Aparicio, M。, Aguilar, M。, Badenes, D ., Cerulla, N。,
Molinuevo, J. L。, Ruiz, E., Robles, A。, Sagrario Barquero, M。,
Antúnez, C。, Martínez-Parra, C。, Frank-García, A。, Fernández,
M。, Alfonso, 五、, 索尔, J. M。, & Blesa, 右. for the NEURONORMA
Study Team. (2009). Spanish multicenter normative studies
(NEURONORMA Project): Norms for verbal fluency tests.
Archives of Clinical Neuropsychology, 24(4), 395–411. https://
doi.org/10.1093/arclin/acp042, 考研: 19648583

Peñaloza, C。, Barrett, K., & Kiran, S. (2019). The influence of pre-
stroke proficiency on poststroke lexical-semantic performance in
bilingual aphasia. Aphasiology, 34(10), 1223–1240. https://土井
.org/10.1080/02687038.2019.1666082, 考研: 33281268
Peñaloza, C。, Grasemann, U。, Dekhtyar, M。, 米库莱宁, R。, &
Kiran S. (2019). BiLex: A computational approach to the effects
of age of acquisition and language exposure on bilingual lexical
使用权. Brain and Language, 195, 文章 104643. https://doi.org
/10.1016/j.bandl.2019.104643, 考研: 31247403

Raboutet, C。, Sauzéon, H。, Corsini, M。, Rodrigues, J。, Langevin, S。,
& N’Kaoua, 乙. (2010). Performance on a semantic verbal fluency
task across time: Dissociation between clustering, switching, 和

categorical exploitation processes. Journal of Clinical and
Experimental Neuropsychology, 32(3), 268–280. https://doi.org
/10.1080/13803390902984464, 考研: 19657912

罗伯茨, 磷. M。, & Le Dorze, G. (1998). Bilingual aphasia: 语义学
组织, strategy use, and productivity in semantic verbal
fluency. Brain and Language, 65(2), 287–312. https://doi.org/10
.1006/brln.1998.1992, 考研: 9784272

罗森, V. M。, Sunderland, T。, 征收, J。, Harwell, H。, McGee, L。,
Hammond, C。, Bhupali, D ., Putman, K., Bergeson, J。, &
Lefkowitz, C. (2005). Apolipoprotein E and category fluency:
Evidence for reduced semantic access in healthy normal controls
at risk for developing Alzheimer’s disease. Neuropsychologia,
43(4), 647–658. https://doi.org/10.1016/j.neuropsychologia
.2004.06.022, 考研: 15716154

Shao, Z。, Janse, E., Visser, K., & 迈耶, A. S. (2014). What do verbal
fluency tasks measure? Predictors of verbal fluency performance
in older adults. 心理学前沿, 5, 文章 772. https://土井
.org/10.3389/fpsyg.2014.00772, 考研: 25101034

Troyer, A. K., Moscovitch, M。, & Winocur, G. (1997). Clustering
and switching as two components of verbal fluency from youn-
ger and older healthy adults. Neuropsychology, 11(1), 138–146.
https://doi.org/10.1037/0894-4105.11.1.138, 考研: 9055277
Unsworth, N。, Spillers, G. J。, & Brewer, G. A. (2011). Variation in
verbal fluency: A latent variable analysis of clustering, switching,
and overall performance. Quarterly Journal of Experimental
心理学, 64(3), 447–466. https://doi.org/10.1080/17470218
.2010.505292, 考研: 20839136