SPECIAL ISSUE:
The Role of the Cerebellum in Language Comprehension and Production

No Evidence for Semantic Prediction Deficits in
Individuals With Cerebellar Degeneration

开放访问

杂志

Maedbh King1*

, Sienna Bruinsma1*

, and Richard B. Ivry1,2

1心理学系, 加州大学, 伯克利, CA, 美国
2Helen Wills Neuroscience Institute, 加州大学, 伯克利, CA, 美国
* 这些作者的贡献同等.

关键词: cerebellar degeneration, cerebellum, internal model, prediction, semantic prediction

抽象的

Cerebellar involvement in language processing has received considerable attention in the
neuroimaging and neuropsychology literatures. Building off the motor control literature, 一
account of this involvement centers on the idea of internal models. In the context of language,
this hypothesis suggests that the cerebellum is essential for building semantic models that, 在
concert with the cerebral cortex, help anticipate or predict linguistic input. 迄今为止, supportive
evidence has primarily come from neuroimaging studies showing that cerebellar activation
increases in contexts in which semantic predictions are generated and violated. Taking a
neuropsychological approach, we put the internal model hypothesis to the test, asking if
individuals with cerebellar degeneration (n = 14) show reduced sensitivity to semantic
prediction. Using a sentence verification task, we compare reaction time to sentences that vary
in terms of cloze probability. We also evaluated a more constrained variant of the prediction
假设, asking if the cerebellum facilitates the generation of semantic predictions when the
content of a sentence refers to a dynamic rather than static mental transformation. The results
failed to support either hypothesis: Compared to matched control participants (n = 17),
individuals with cerebellar degeneration showed a similar reduction in reaction time for
sentences with high cloze probability and no selective impairment in predictions involving
dynamic transformations. These results challenge current theorizing about the role of the
cerebellum in language processing, pointing to a misalignment between neuroimaging and
neuropsychology research on this topic.

引文: 国王, M。, Bruinsma, S。, & Ivry,
右. 乙. (2022). No evidence for semantic
prediction deficits in individuals with
cerebellar degeneration. Neurobiology
of Language. Advance publication.
https://doi.org/10.1162/nol_a_00083

DOI:
https://doi.org/10.1162/nol_a_00083

支持信息:
https://doi.org/10.1162/nol_a_00083

已收到: 31 可能 2022
公认: 21 九月 2022

利益争夺: 作者有
included a competing interest section
at back of article.

介绍

通讯作者:
Maedbh King
maedbhking@gmail.com

处理编辑器:
Julie Fiez

版权: © 2022
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

Anatomical, neuropsychological, and neuroimaging work over the last 35 years has implica-
ted the cerebellum in functions extending well beyond the motor domain. Throwing down the
gauntlet for a “cognitive cerebellar revolution,” Leiner et al. (1986) highlighted the expansion
of the cerebellum across vertebrate evolution and, 尤其, a parallel expansion of the
neocerebellum and prefrontal cortex. They hypothesized that this pattern likely reflected
extensive communication between the cerebellum and cortical association regions, connec-
tions that could support mental coordination in a manner analogous to how cerebellocortical
connectivity supported motor coordination. Over the following decades, the neuropsycholog-
伊卡尔, neuroimaging, and brain stimulation literatures have provided ample evidence in line
with this general perspective. Behavioral impairments on a variety of cognitive and affective
tasks have been described in individuals with cerebellar dysfunction (Ivry & Fiez, 2000; Kansal
等人。, 2017; Schmahmann & Sherman, 1998; Sokolov et al., 2017), and activation is

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

/

/

.

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

consistently observed in the cerebellum during a broad range of tasks that is not readily related
to motor preparation or production (金等人。, 2019; Stoodley & Schmahmann, 2009乙).

One prominent domain of interest is language. In their seminal PET (positron emission
tomography) 学习, Petersen et al. (1989) used a subtractive logic to identify brain regions
involved with different linguistic processes. Their results revealed prominent activation of
the right cerebellum in a contrast designed to show regions involved in semantic retrieval
while controlling for articulatory preparation. This observation has been confirmed in many
subsequent neuroimaging studies investigating semantic retrieval and prediction, 与
activation centered in right medial-to-lateral Crus I/II (Fiez et al., 1996; Lesage et al., 2017;
Moberget et al., 2016). The neuropsychology literature presents a more problematic picture.
While speech dysarthria is a prominent feature in some forms of cerebellar ataxia (Ackermann
& Hertrich, 1994; Richter et al., 2007), the evidence is mixed in terms of whether cerebellar
dysfunction, at least when emerging in adulthood, disrupts semantic processing. In a widely
cited case study by Fiez et al. (1992), a patient with a right inferior cerebellar lesion had
marked difficulty on a semantic generation task, either coming up with an inappropriate
semantic associate to a target noun or failing to follow instructions and simply repeating the
target word. 然而, subsequent studies failed to find group-level differences on similar tasks
(Gasparini et al., 1999; Helmuth et al., 1997; Richter et al., 2004; Riva, 1998). While it is
possible that these null results reflect the heterogeneity in the patient samples, it may be that
semantic generation tasks are not sensitive to cerebellar-related language impairments or opti-
mal for assessing computations performed by the cerebellum in semantic processing.

Theoretically, researchers have turned to the sensorimotor control literature in considering
how the cerebellum may contribute to semantic processing. One key idea has centered on the
notion of prediction and, 具体来说, how the cerebellum may generate internal models to
anticipate future input based on the current context (Ito, 2008; Wolpert et al., 1998). 在里面
motor context, an internal model may represent the dynamic properties of an object (例如,
body part, tool), allowing the agent to anticipate the sensory consequences of an action within
a particular environment (Wolpert et al., 1995). Extending this idea to language, an internal
model would operate on abstract semantic knowledge and input from the current discourse to
generate semantic expectancies; 例如, to anticipate the speaker’s next utterance or
anticipate when the speaker would pause to facilitate fluid turn taking (Lesage et al., 2012,
2017; Moberget et al., 2014).

Semantic verification tasks have provided one test bed of the internal model hypothesis. 在
an functional magnetic resonance imaging (功能磁共振成像) 学习, Moberget et al. (2014) compared the
hemodynamic response to three types of sentences: (1) congruent, in which the last word was
highly predictable given the preceding, base phase; (2) incongruent, in which the last word
violated a prediction established by the base phase; 和 (3) scrambled, in which the words in
the base phrase were randomly shuffled, thus precluding the generation of a prediction. 一个
analysis time-locked to the appearance of the last word revealed two key cerebellar results.
第一的, violations of predictions produced bilateral activation encompassing posterolateral Crus
I/II compared to either congruent or scrambled sentences. 第二, compared to scrambled
句子, congruent sentences produced activation in a more circumscribed portion of pos-
terolateral cerebellum, now restricted to the right cerebellar hemisphere only. 因此, these data
are in accord with the hypothesis that the cerebellum not only processes violations of predic-
系统蒸发散 (which might suggest a role in error-detection/error-correction), but also is engaged in the
generation of predictions. Lesage et al. (2017) reported similar effects in an event-related fMRI
design that manipulated sentence predictability. Not only did they find that activity in the right
posterolateral cerebellum correlated with the predictability of the target word, but they also

Internal model:
Model that represents the dynamic
properties of an object (例如, 身体
部分, tool), allowing the agent to
anticipate the sensory consequences
of an action within a particular
环境.

Neurobiology of Language

2

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

/

/

.

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

demonstrated that this same region was engaged in phonological, but not semantic or ortho-
graphic processing.

Behavioral assays of cerebellar involvement in language have come from two sources, 非-
invasive brain stimulation targeted at the cerebellum in healthy individuals (Argyropoulos, 2016;
D’Mello et al., 2017; Gatti et al., 2021; Lesage et al., 2012) or experiments involving individuals
with cerebellar pathology (Alexander et al., 2012; Kansal et al., 2017; Stoodley & Schmahmann,
2009A). Lesage et al. (2012) instructed participants to shift their gaze to a peripheral object that
was predicted by the semantic content of a spoken sentence. After repetitive transcranial
magnetic stimulation (rTMS) was applied targeting right posterolateral cerebellum, 参与者
took longer to fixate on the predicted object. 相比之下, rTMS had no effect when the semantic
content did not predict the target object, and the selective effect on predictive looking was not
observed when the rTMS was directed at the vertex, a control stimulation location.

Along the same lines, D’Mello et al. (2017) 表明, after transcranial direct current
stimulation (tDCS) over the right lateral cerebellum, the blood oxygen level dependent (大胆的)
response in the cerebellum was selectively elevated in response to predictive sentences. 国际米兰-
estingly, the change in activation was not accompanied by a change in behavior. 参加者
were faster to verify predictable sentences compared to non-predictive, and this advantage was
similar in the tDCS and control conditions. 相似地, Moberget et al. (2016) failed to observe
any difference in behavior between individuals with cerebellar pathology and matched control
参与者. 因此, there is a mismatch between the imaging/neurostimulation and neuropsy-
chology results. Consistent with the internal model hypothesis, increased activation is
observed in the right posterolateral cerebellum when the semantic content of a sentence sup-
ports the formation of a prediction. 然而, the evidence is mixed in terms of whether lesions
encompassing this region disrupts the ability to make semantic-based predictions. 更多的
一般来说, the neuropsychological literature fails to provide strong evidence of a role of the
cerebellum in semantic processing. While individuals with cerebellar degeneration (CD)
consistently show impairment in tests of fluency or word completion (Alexander et al.,
2012; Kansal et al., 2017; Stoodley & Schmahmann, 2009A, 2009乙), these tasks heavily tax
processes associated with cognitive control.

In the present study, we set out to conduct a new neuropsychological test of the internal
model hypothesis, comparing the performance of individuals with cerebellar degeneration and
matched controls on a sentence verification task. Similar to our earlier studies (Moberget et al.,
2014, 2016), participants made a two-alternative forced choice response on a semantic pre-
diction task, indicating if each sentence was meaningful or meaningless. We emphasized reac-
tion time in the current study, under the assumption that this would provide a more sensitive
measure of performance than accuracy. We manipulated the strength of the prediction of the
final word in the sentence given its base phrase. 为此, we manipulated cloze probability,
a common linguistic metric that quantifies the context-dependent predictability of a word. 我们的
main analysis focused on the meaningful sentences. We expected that control participants
would have faster reaction times in response to target words with high cloze probability com-
pared to target words with low cloze probability (Staub et al., 2015), reflecting the benefits of
predictive processing when the base phrase allows the participant to anticipate the target
word. If the integrity of the cerebellum is important for generating semantic predictions, 我们
expect that the CD group would show an attenuated effect of cloze probability: 具体来说,
the CD group should show a reduced benefit to target words with high cloze probability.

As a secondary test of the prediction hypothesis, we analyzed reaction times for the mean-
ingless sentences to examine how the two groups responded to violations of semantic

Cloze probability:
Linguistic metric that quantifies the
context-dependent predictability of
a word.

Neurobiology of Language

3

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

.

/

/

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

Continuous operations of
representational transformation
(CoRT):
A theory proposing that the
cerebellum facilitates mental
processing when an internal
representation undergoes a
continuous and dynamic
transformation but is not essential
when the mental operations entail
more discrete or transformations
maintenance of static
陈述.

预测. This analysis is problematic for two reasons. 第一的, these sentences entail two types of
prediction violations. There is the general semantic violation given that the final word results in a
nonsensical sentence. 和, at least for the high cloze sentences, there is a prediction violation in
that the final word does not match the expected word. 第二, we did not have a strong a priori
hypothesis here, even for the controls. Reaction times may be faster when the base phrase
suggests a strong prediction since the violation would be easier to detect. 或者, reaction
times might be slower when the base phrase suggests a strong prediction since anticipation of the
expected final word might cause interference in processing the target word. 尽管如此, 这
meaningless sentences provide a test bed to see if the CD and control groups showed different
response time patterns to semantic prediction violations.

We also designed the experiment to test a second hypothesis concerning predictive func-
tions of the cerebellum. Even in the motor domain, prediction is not the exclusive domain of
the cerebellum; 的确, one could argue that most neural activity is fundamentally about pre-
措辞 (弗里斯顿, 2009). This leads to the question of what might be constraints on cerebellar-
dependent predictions. Various candidates have been proposed over the years, some that
specify the task (例如, sensory consequences of movements generated by the skeletal muscu-
lature; Day et al., 1998) or computational (例如, 定时, error-based learning; Keele & Ivry,
1990; Wolpert et al., 1998) domain. McDougle et al. (2022) have recently proposed that
the cerebellum is essential for predictions that require a continuous representational transfor-
运动 (CoRT). The key idea is that the cerebellum facilitates mental processing when an inter-
nal representation undergoes a continuous dynamic transformation but is not essential when
the mental operations entail more discrete or transformations or maintenance of static repre-
句子. In our initial work on this problem, the CoRT constraint provided a parsimonious
account of the pattern of impairments (and spared performance) in the domains of visual cog-
nition and arithmetic. 例如, individuals with CD were impaired in the rate at which
they mentally manipulated a visual image (Shepard & Metzler, 1971), whereas they showed a
normal processing rate in the iterative evaluation of a series of discrete spatial representations.

In the context of language, we hypothesize that a CoRT-like operation would be relevant
when the semantic content of a sentence conveys a dynamic representational transformation.
例如, the sentence “The man cut the rope with a pair of scissors” evokes a strong,
dynamic mental simulation, with a conceptualization of the context (an agent, desiring to
cut a rope), helping anticipate the final word. 相比之下, the sentence “Spring was her favorite
season of the year” is relatively static. One could say it requires an internal model, one that
captures the idea that a person has their idiosyncratic preferences for a season. But this model
does not evoke a dynamic simulation. Based on this reasoning, we also designed the exper-
iment such that we could compare sentences that described dynamic transformations (CoRT-
dependent) with those that described static situations (non-CoRT). Although we do not have an
a priori prediction for the control participants on this dimension, we predicted that the CD
group would be relatively impaired on the sentence verification task for the dynamic
句子.

材料和方法

We tested individuals with CD and matched controls on a semantic verification task. Partici-
pants were asked to determine whether a sentence was meaningful or meaningless based on
the final word of the sentence. We manipulated two variables, prediction strength and
dynamic context in a 2 × 2 设计 (数字 1). All testing, including the neurological evaluation
of the CD participants was conducted online.

Neurobiology of Language

4

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

.

/

/

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

数字 1. Task design. (A) Representative sentences for evaluating the internal and CoRT models.
Sentence predictability is manipulated via high and low cloze probabilities, and representational
transformations are manipulated via dynamic and static simulations. (乙) Example trial. Each word of
the stem (“The man loosened the tie around his”) is presented for 500 多发性硬化症, followed by a fixation
叉 (500 多发性硬化症). Participants then have 2,000 ms to make a key press indicating “true” or “false” on
the target word (“neck”). True indicates a meaningful sentence, while false indicates a meaningless
句子.

We first describe the procedures involved in creating the stimulus materials for the study
and then turn to the procedure for the main experiment. The study was approved by the insti-
tutional review board at the University of California, 伯克利.

Construction and Evaluation of Stimuli

Sentences were drawn from two sentence databases, both of which included cloze probability
ratings (Block & Baldwin, 2010; Peelle et al., 2020). 一起, 有 3,583 sentences in the
two databases. 这 498 sentences in the Block and Baldwin database have a minimum cloze
probability of 0.3. A cloze probability is calculated as the proportion of participants who give
the same response to an incomplete sentence or passage (Kutas & Hillyard, 1984). We applied
this same minimum criterion to the Peelle database, which excluded 923 的 3,085 sen-
tences in that database, yielding a total of 2,660 sentences in the initial inclusion stimulus set.

We enlisted raters to evaluate the degree to which these 2,660 sentences conveyed a
dynamic event (CoRT-ness rating). We initially recruited undergraduate participants (n = 8,
4 female/4 male) through a research participant pool at the University of California, 伯克利.
These individuals completed the rating procedure in the lab. 然而, with the onset of insti-
tutional restrictions related to the COVID-19 pandemic, we transitioned to online methods.
For these ratings, we recruited participants (n = 61) through Prolific (https://www.prolific.co/),
an online participant recruitment platform. We verified via self-report that the first language of
all participants was English. Participants received monetary compensation for their time.

Participants were asked to rate each sentence in terms of how strongly the content evoked a
dynamic event, using a 5-point Likert scale where 1 indicated strongly disagree (IE。, strongly

Neurobiology of Language

5

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

/

.

/

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

/

.

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

non-CoRT) 和 5 indicated strongly agree (IE。, strongly CoRT). Participants were provided
examples of sentences conveying a strong dynamic event (例如, “The man loosened the tie
around his neck.”) and a weak dynamic event (例如, “Her job was easy most of the time.”).
As a primer, participants were given a list of five sentences with the correct ratings to ensure
that they understood the goal of the task.

Each trial began with the presentation of a fixation cross. 后 500 多发性硬化症, a target sentence was
displayed at the center of the screen, along with the instruction prompt at the bottom of the
屏幕, “Does this sentence evoke strong, changing imagery?” Participants had 10 s to indicate
their response using the number keys “1” to “5.” The fixation cross appeared as soon as the
rating was registered. The in-person participants rated all 500 of the sentences in the Block
& Baldwin data set. The stimulus order was randomized across individuals and breaks were
provided every 180 试验. For the online participants, we used the Gorilla platform (https://
gorilla.sc/) to run the experiment. 这 2,160 sentences retained from the Peelle database were
randomly divided into 12 sets of 180 sentences each. Each participant rated one set, 与一个
break provided after 90 试验. The stimuli within a set were randomized across individuals.

For the final stimulus set, we used the cloze probability information and CoRT ratings to
选择 125 sentences for each of the four cells created by the factorial crossing of prediction
and dynamics (Figure 1A). On the prediction dimension, sentences had to have a cloze prob-
ability between 0.3 和 0.5 to be selected as weak exemplars, or between 0.8 和 1.0 成为
selected as strong exemplars. 这样, we created a similar cloze range for each level. 为了
the dynamics dimension, sentences had to have a mean rating of less than 2 (静止的) or mean
rating greater than 4 (dynamic). For high cloze sentences, dynamic conditions had an average
CoRT rating of 4.214 (标准差= 0.24), and static conditions had a rating of 1.616 (标准差= 0.306). 这
average values of high cloze probabilities were roughly equivalent across dynamic (米=
0.885, 标准差= 0.053) and static (米= 0.878, 标准差= 0.049) ratings. For low cloze sentences,
dynamic conditions had an average CoRT rating of 4.357 (标准差= 0.298), and static conditions
had an average rating of 1.655 (标准差= 0.305). The average values of low cloze probabilities
were roughly equivalent across dynamic (米= 0.407, 标准差= 0.058) and static (米= 0.398, 标准差=
0.057) ratings.

As a final step, we conducted a pilot of the semantic verification task, with participants
making speeded responses to indicate if the sentence was meaningful or non-meaningful
(detailed below). 参加者 (n = 77) were recruited via Prolific and tested across six blocks
with the full set of 500 句子, with a break provided approximately every 80 试验. Based
on the results of the pilot study, we designed the main experiment to include five blocks each
comprising 64 试验. We opted to reduce the number of blocks from six (in the pilot) to five
(main experiment) to reduce the experiment time for our patient group. We divided the 500
pilot sentences into four conditions: (1) dynamic, high cloze; (2) dynamic, low cloze; (3) 静止的,
high cloze; (4) 静止的, low cloze. Within each of those groups, we chose 80 sentences that had
the lowest standard deviation across the pilot participants, resulting in a total of 320 句子
for the main experiment.

Main Experiment

参加者

Seventeen individuals with CD and 19 年龄- and education-matched control participants were
recruited, drawing on a patient database maintained by our lab. The database includes indi-
viduals recruited from ataxia support groups around the country and through advertisements
posted on the National Ataxia Foundation webpage. Researchers from our group sent

Neurobiology of Language

6

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

/

/

.

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

recruitment letters to support group leaders and these leaders shared our letters with their
members. Interested individuals contacted the lab and after completing a series of screening
程序, were entered into the CD database if they met inclusion criteria (Saban & Ivry,
2021). We then invited these individuals to participate in our experiment. Control participants
were recruited via advertisements circulated on internet forums (例如, craigslist). Although we
have limited demographic data on the control group, we assume they are selected from a
similar geographic distribution as the CD group given that the only restriction was that they
reside in the United States. Participants were compensated with either $20 或一个 $20 亚马逊
gift card for their participation.

桌子 1 provides demographic, diagnostic, neurological, and neuropsychological informa-
tion for each individual in the CD group, along with summary information for the control

桌子 1. Demographic and clinical assessment for cerebellar degeneration patients (p01–p17) and healthy controls.

ID
p01

p02

p03

p04

p05

p06

p07

p08

p09

p10

p11

p12

p13

p14

p15

p16

p17

控制 (中号 )

控制 (标清)

Patient (中号 )

Patient (标清)

Hand
右

L

N/A

右

右

右

右

右

右

右

右

右

右

右

右

L

右

–

–

–

–

年龄
48

性别
F

55

48

51

39

89

80

79

57

42

67

42

48

81

61

53

59

52.44

11.86

60.5

15.31

F

F

F

F

F

F

中号

中号

F

中号

F

F

中号

F

F

F

–

–

–

–

MoCA
29

27

N/A

26

27

19

26

22

29

29

26

28

27

24

22

27

27

SARA
4

10

N/A

2

13

15

18.5

8

6

10

15

20

10

10

7

17

9

Years of education

25

18

22

12

16

12

12

16

20

14

20

19

20

16

12

18

12

26.39

2.72

25.93

2.83

–

–

10.64

5.36

16.25

2.54

16.63

3.86

Etiology
SCA-1

Included
是的

SAOA

SCA-3

SCA-6

SAOA

SCA-3

SAOA

SCA-6

SCA-3

SAOA

SCA-5

SAOA

SAOA

SAOA

SCA-6

SAOA

SCA-6

–

–

–

–

是的

是的

是的

是的

不

是的

不

是的

是的

不

是的

是的

是的

是的

是的

是的

–

–

–

–

笔记. MoCA scores are out of a total of 29 点 (perfect score). SARA scores are out of a total of 40 点 (most severe ataxia). MoCA = Montreal Cognitive
Assessment; SARA = Scale for the Assessment and Rating of Ataxia; SCA = spinocerebellar ataxia; SAOA = spontaneous adult-onset ataxia. Participants p06, p08
and p11 were excluded from the final sample due to their poor performance on the semantic evaluation task.

Neurobiology of Language

7

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

.

/

/

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

Spinocerebellar ataxia (SCA):
Progressive and degenerative
hereditary disorder that results in
atrophy of the cerebellum. Symptoms
typically include uncoordinated
gait, loss of balance, and cognitive
deficits.

团体. For the purposes of this study, we included individuals who had a diagnosis of cere-
bellar ataxia based on clinical exam (and MRI evidence as indicated by self-report) or diag-
nosis of spinocerebellar ataxia based on genetic testing and/or family history. 我们排除了
individuals with Friedrich’s ataxia. For the neurological exam of ataxia, we used the Scale
for the Assessment and Rating of Ataxia (SARA; Schmitz-Hübsch et al., 2006), 其中有一个
total possible score of 40 (maximum impairment). The test was modified for online adminis-
翻译. 具体来说, items assessing gait, 姿态, and sitting were scored using a flow chart of
questions rather than through direct observation, given concerns about remote testing. 到
establish general neuropsychological status, we used the Montreal Cognitive Assessment
(MoCA; Nasreddine et al., 2005). For online testing, we eliminated the trail making item.
Although the maximum score on the modified MoCA is 29, the scores reported below were
adjusted so that they fall on a 30-point scale. Participants in the control group were also
administered the MoCA.

Three individuals in the CD group (p06, p08, p11) and one control (c05) were not included
in the final analysis because of poor performance on the sentence verification task (全面的
accuracy below 70% on meaningful and meaningless sentences). Although this criterion
may eliminate individuals with the most severe language problems, we were concerned it
was more reflective of disengagement with the task or failure to understand the task instruc-
系统蒸发散. 的确, the mean accuracy score for these three was well below that of any of the other
参与者, with two performing poorly on the meaningful sentences and one responding
“meaningful” to all of the meaningless sentences. For the remaining 14 individuals in the
CD group, the mean score on the SARA test was 10.64 (标准差= 5.36), a value that corresponds
to mild-to-moderate impairment. At the time of testing, an average of 6.8 yr (标准差= 7.6) 有
elapsed since diagnosis of cerebellar ataxia, and the participants reported symptom onset
preceded diagnosis by about 4 yr.

We selected individuals for the control group by referencing our database to produce
matches in terms of age and years of education (yoe). As can be seen in Table 1, the group
means were comparable for the CD (M age = 60.5; M yoe = 16.63) and control (M age =
52.44; M yoe = 16.25) 团体, and there was no statistically significant difference between
the groups for either years of education (t 30 = 0.34, p = 0.74) or age (t30 = 1.733, p = 0.09).

Although not part of our recruitment criteria, there was no significant difference between
the CD (米= 25.93, 标准差= 2.83) and control (米= 26.39, 标准差= 2.72) groups on the MoCA test
(t 30 = 0.49, p = 0.63). Note that a MoCA score below 26 is indicative of possible cognitive
impairment. Three of the CD participants in the final sample scored below this criterion as did
six of the control participants. Given our focus on one important aspect of cognition, 语言,
we did not exclude participants based on their MoCA score. 而且, there was no relation-
ship between the overall MoCA score and two performance metrics: task accuracy (R = 0.28
p = 0.12) and reaction time (R = 0.01, p = 0.59).

程序

Participants were recruited via an email sent to individuals in our participant database. 这
email provided a brief overview of the experimental goal and requirements (例如, English as
first language, completion of the 2 min (62 审判) experiment in a single session). A link was
provided to the experiment and participants were informed that they could initiate testing
whenever they were ready since the testing was done in an automated manner. Each partic-
ipant was provided with a unique link, the code that allowed us to match the participant with
their experimental data.

Neurobiology of Language

8

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

.

/

/

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

The link took the participant to the Gorilla platform where the participant worked through a
series of introductory screens that culminated in a page where they could provide electronic
informed consent. Once this was obtained, the CD and control participants completed a 2-min
task designed to provide a brief assessment of American English fluency. We modified the
English LexTALE task (Lemhöfer & Broersma, 2012) for our online platform. On each trial, A
fixation screen was presented for 250 多发性硬化症, followed by a letter string. The participant indicated
if the letter string formed an English word (press “J” on keyboard) or did not form an English
word (press “F”). If a response was not entered within 2,000 多发性硬化症, the trial was recorded as
“incorrect” and the program moved to the next trial. 有 62 试验, each with a unique
target word (32 字, 32 non-words). All participants performed at a high level (overall M =
96%, range = 70%–100%).

Following the fluency assessment, the program moved on to the semantic verification task.
The general procedure was similar to that employed in Moberget et al. (2014). The start of each
trial was marked by the presentation of a fixation screen for 500 多发性硬化症 (Figure 1B). 这是
followed by the serial presentation of the sentence stem (例如, “The man loosened the tie around
his”), with each word displayed on the screen for 500 多发性硬化症. We opted to use a serial presentation
mode to minimize demands on eye movements. Following the presentation of the stem, 这
fixation screen re-appeared for 500 多发性硬化症, cueing the participant that the next screen would dis-
play the target word. The target word was then presented and the participant made a speeded
response, indicating if the target word resulted in a meaningful sentence (例如, “neck”) 或非-
meaningful sentence (例如, “banana”). Responses were made by pressing the “J” key with the
right index finger or “F” key with the left index finger for meaningful and non-meaningful sen-
时态, 分别. The target word remained visible until a response was detected or 2 s
elapsed. The screen then went blank for 500 ms before the onset of the next trial. 参加者
were instructed that the primary dependent variable was response time; 像这样, while they
should seek to achieve a high level of accuracy, they should respond as quickly as possible.

Participants first completed a practice block of 24 trials in which 16 of the sentences were
meaningful and eight were non-meaningful. This was followed by five test blocks of 64 试验
each, or a total of 320 test trials. Across the 320 试验, there were 80 sentences for each of the
four conditions (Figure 1A), providing our assessment of prediction (high vs. low cloze prob-
能力) and CoRT (dynamic vs. static inference). Subject to this constraint, the stimulus set for
each participant was based on random selection (without replacement) from the full set of 344
句子, with the unselected items used in the practice trials. Within each of the four con-
版本, the target word was the final word for that sentence in the stimulus set on 70% 的
试验 (和, 因此, formed a meaningful sentence). For the other 30% of the trials, the target word
was determined by randomly generating a target word from all other target words. The exper-
imenter then manually checked that the target word that was randomly chosen was subject to
the constraint that it formed a non-meaningful sentence.

Each block took approximately 8 min and summary feedback in the form of percent correct
for the previous block was provided at the end of each block, along with a reminder of the
keyboard mappings. The participant moved on to the next block by pressing the spacebar.
Upon completing the sentence verification task, the participant was asked to fill out a brief
feedback form concerning their experience with the online testing platform. The entire exper-
iment took approximately 50 min (longer if participants took longer breaks).

We did not include reaction times for the non-meaningful trials or for trials in which the
response was incorrect. The mean values were then entered in a 2 (团体) × 2 (预言) × 2
(CoRT) within-subject analysis of variance (ANOVA).

Neurobiology of Language

9

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

p

:
/
/

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
/

d
哦

我
/

我

/

/

.

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

.

/

我

F

乙
y
G
你
e
s
t

t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

No semantic prediction deficits in cerebellar degeneration

结果

正如预期的那样, mean accuracy was very high for both groups (CD (n = 14): 米= 0.97, 标准差= 0.05;
控制 (n = 18): 米= 0.96, 标准差= 0.07) across all four conditions (图2A). While recognizing the
limited value of an error analysis given that performance was near ceiling, we employed a three-
way within-subject ANOVA with the factors (团体) × 2 (预言) × 2 (CoRT). None of the main
effects was significant (团体: F1,120 = 0.04, p = 0.85, ηp2 < 0.001; Prediction: F1,120 = 0.46, p = 0.51, ηp2 < 0.01; CoRT: F1,120 = 0.48, p = 0.49, ηp2 < 0.01). The Group factor did not interact with either Prediction (F1,1200 = 0.29, p = 0.59, ηp2 < 0.001) or CoRT (F1,120 = 0.19, p = 0.66, ηp2 < 0.001), and the three-way interaction was not significant (F1,120 = 0.0001, p = 0.99, ηp2 < 0.001). We analyzed the meaningful and meaningless trials separately in the analyses of the reac- tion time data, restricting these analyses to the trials in which the participants were accurate. For the meaningful sentences, the CD group was slower than the control group by 206 ms (main effect of group: F1,30 = 23.34, p < 0.001, ηp2 = 0.44). As can be seen in Figure 2B, participants became slightly faster over the course of the five test blocks (F4,120 = 5.54, p < 0.001, ηp2 = 0.16), but the difference between the two groups was maintained. Given the absence of a Group × Block interaction (F4,120 = 0.19, p = 0.94, ηp2 < 0.001), we collapsed the data across the five blocks in the subsequent analyses. Our main interest centers on how the semantic judgments were influenced by the semantic content of the base phrases. Specifically, we asked how the degree of prediction and dynamic simulation influenced participants’ response times on the verification task. We first focused on the meaningful sentences given that the time to process the final word should be faster when that word is predicted by the context. There was a main effect of Prediction (F1,120 = 12.55, p < 0.001, ηp2 = 0.09), with faster response times on trials in which the target word had high cloze probability compared to trials in which the target word had low cloze probability (Figure 3A and 3B; Figure S1). Importantly, the Group × Prediction interaction was not significant (F1,120 = 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 / d o i / l / . / 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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 2. Behavioral performance summary for the cerebellar degeneration (CD) and control (CO) groups. (A) Mean accuracy averaged across blocks is close to ceiling performance for both groups. (B) Mean reaction time (RT) on correct trials for meaningful sentences as a function of test block (64 trials/block). The CD group responded slower than the CO group, with both groups showing a reduction in RT over the experimental session. Neurobiology of Language 10 No semantic prediction deficits in cerebellar degeneration 0.008, p = 0.93, ηp2 < 0.001): Participants in the CD group showed a similar benefit to that of the control participants when the base phrase enabled the anticipation of the target word. Thus, this analysis fails to support the prediction derived from the internal model hypothesis, namely, that the difference between high and low cloze conditions would be attenuated in the CD group. Turning to the CoRT hypothesis, we next turned to the comparison of verification times when the base phrase described a dynamic or static context. As noted in the Introduction, we had no a priori reason to suppose that verification times for the control group would be sensitive to the relative dynamics; our focus was on whether the CD group would be selectively affected on the dynamic sentences. For both groups, the results indicate that response times did not differ for sentences that described a dynamic event compared to sentences that described a static event (F1,120 = 0.02, p = 0.89, ηp2 < 0.001, Figure 3C and 3D; Figure S1 in the Supporting Information, available at https://doi.org/10.1162/nol_a_00083). Importantly, this factor did not interact with Group (F1,120 = 0.07, p = 0.79, ηp2 < 0.001). Thus, the results fail to support the hypothesis that the integrity of the cerebellum may be disproportionately important when the semantic content engages a dynamic mental transformation. The three-way Group × Prediction × CoRT interaction was also not significant (F1,120 = 0.04, p = 0.85, ηp2 < 0.001). We also examined the meaningless sentences to ask how the two groups responded to vio- lations of semantic predictions. To address this question, we used a three-way ANOVA, with the factors Group, Prediction, and Trial Type (meaningful vs. meaningless). Response times were longer to target words that rendered the sentence meaningless compared to target words that resulted in a meaningful sentence (Figure 4, F1,120 = 12.85, p < 0.01, ηp2 = 0.097). There 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 / d o i / l / / . 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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. No differences were observed between the cerebellar degeneration (CD) and control (CO) groups in their ability to use predictive semantic information. (A) Mean response times (RTs) to target words as a function of cloze probability. (B) Benefit on verification task to target words that are highly predictable (low cloze–high cloze). Both the CD and CO groups were faster when the target word had high predictability, and the benefit was similar for the two groups. (C) Mean RTs to target words as a function of the dynamics conveyed by the base phrase. (D) There was no differ- ence in RT for the dynamic and static sentences, a null effect that was similar for both groups. RTs in all graphs are limited to trials in which the participants correctly verified that the sentence was meaningful. CoRT = continuous representational transformation. Neurobiology of Language 11 No semantic prediction deficits in cerebellar degeneration Figure 4. No differences observed between the cerebellar degeneration (CD) and control (CO) groups in evaluating sentences that entail prediction violations (i.e., meaningless trials). (A) Response time (RT) difference between meaningless and meaningful sentences. Participants were slower when the target word did not fit with the context, and this effect was similar for the CO and CD groups. (B) When the target word did not fit with the context, mean RT was numerically (but not statistically) faster when the context allowed for strong expectation (target word would have been high cloze). The magnitude of this effect was similar for the two groups. was no main effect of Prediction for meaningless trials (F1,60 = 0.39, p = 0.54, ηp2 < 0.01) and Trial Type did not interact with Group (F1,120 = 0.015, p = 0.91, ηp2 < 0.001) or Prediction (F1,120 = 1.95, p = 0.17, ηp2 < 0.001). Thus, while prediction violations slowed response times, the results failed to show that this effect was sensitive to the degree of prediction conferred by the base phrase and, importantly, failed to identify any difference in how the CD and control groups respond to violations of semantic expectancies. ( We note that although the target words were selected to yield semantic predictions, a few of the violations were syntactic rather than semantic. However, there were not enough syntactic violations to do a separate analysis of these sentences.) DISCUSSION To determine whether the cerebellum plays an important role in semantic processing, we tested patients with CD and age- and education-matched healthy controls on a sentence ver- ification task. Specifically, we were interested in evaluating two hypotheses concerning how the cerebellum might contribute to semantic processing. One hypothesis was based on the idea that the cerebellum forms internal models. Extending this idea in the language domain, we hypothesized that the integrity of the cerebellum would be required to generate predictions based on semantic content. The second hypothesis is based on the idea that the cerebellum is essential for mental operations that entail a CoRT. When applying this constraint to language, we hypothesized that the integrity of the cerebellum would be essential for linguistic predic- tions that require a dynamic, rather than static, simulation of semantic content. The results failed to provide support for either hypothesis. We parametrically manipulated semantic predictability (high vs. low cloze probability), and the degree to which the semantic content suggested simulation (dynamic vs. static). We Neurobiology of Language 12 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 / d o i / l / . / 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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 No semantic prediction deficits in cerebellar degeneration observed a main effect of prediction such that participants were faster to respond to sentences in which the target word was strongly predictable from the base phrase compared to sentences in which the target word was weakly predictable. This result corroborates previous literature that has reported an influence of cloze probability on reaction time (Staub et al., 2015). Despite this evidence that the task was sensitive to the operation of predictive mechanisms, the individuals with cerebellar degeneration showed response time advantage for the high cloze words. We also did not find support for the more constrained CoRT hypothesis, as we did not observe a selective cost for the CD group in response times to the sentences that might be expected to generate dynamic simulations. Moreover, the results provide indirect evidence against another task-independent hypothesis of cerebellar function, error-based processing (Fiez et al., 1992). Participants were slower to respond to the meaningless sentences than to meaningful sentences (D’Mello et al., 2017). Taking this as a measure of sensitivity to predic- tion violations, the CD and control groups showed a similar increase in response time in judg- ing a sentence as meaningless. Using similar tasks, neuroimaging (D’Mello et al., 2017; Lesage et al., 2017; Moberget et al., 2014) and neurostimulation studies (D’Mello et al., 2017; Lesage et al., 2012) have pro- duced evidence to suggest a cerebellar role in semantic prediction, with the emphasis in those studies on the internal model hypothesis. For example, compared to scrambled word strings, predictive sentences produce an increase in the BOLD response in a circumscribed region of posterolateral right cerebellum, and violations of these predictions produce a strong response that is roughly symmetric for right and left posterolateral cerebellum (Moberget et al., 2014). These results stand in contrast to the null results of the present study, one in which we set out to provide a more direct test of a general and constrained variant of the internal model hypoth- esis through our parametric manipulations. Taken together, we find an unsatisfying misalignment of neuroimaging, neurostimulation, and neuropsychology findings in terms of how the cerebellum contributes to semantic processing. It is no doubt foolhardy to wish for a perfect marriage; different methods have their strengths and weaknesses. With respect to the cerebellum and cognition, much of the impetus has come from the neuroimaging literature, a method that is quite powerful in identifying areas that show activity correlated with tasks and/or processes. Neuropsychology offers a powerful tool to test functional hypotheses, although neural plasticity places some limitations in interpretation, especially for null results. It is also important to keep in mind one important limitation with cerebellar neuro- imaging: The BOLD signal appears to be a function of activity at the mossy fiber terminals (Mathiesen et al., 2000) rather than activity reflective of intra-cerebellar processing (e.g., Purkinje cell firing). Thus, cerebellar neuroimaging is providing a window concerning the inputs to the cerebellum, but mute in terms of how that information is transformed by the cerebellum. Caution is, of course, warranted when interpreting null results. We can consider a few pos- sible explanations for why we failed to observe support for either hypothesis. First, we do note that there were performance differences between the CD and control groups, namely, that the CD group was slower overall. This is to be expected given their ataxia. In previous studies involving manual responses for choice RT tasks, we have observed an increase in reaction time of between 50 and 150 ms (Breska & Ivry, 2018; Helmuth et al., 1997). In the present study, the increase was 220 ms. It is possible that the larger difference in the present study is motoric in nature, related to the fact the testing was performed online with a non-optimal response device (computer keyboard) for individuals with ataxia. On the other hand, it is pos- sible that the increase in response time does reflect an impairment in sentence comprehension or some more generic aspect of cognitive processing, and that our experimental manipulations are insensitive to this impairment. Neurobiology of Language 13 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 / d o i / l . / / 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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 No semantic prediction deficits in cerebellar degeneration Second, our patient sample is quite heterogeneous, composed of individuals with different etiologies. It is possible that, when treated as a group, the pathology does not consistently impact critical regions within the cerebellum that are essential for semantic predictions (e.g., right posterolateral cerebellum). However, it is unlikely that the null results are attribut- able to divergent patterns of degeneration as the literature indicates that pathology in postero- lateral regions of the cerebellum are generally associated with the etiologies of our sample (Hernandez-Castillo et al., 2018; Lukas et al., 2006). Nonetheless, it would be useful to recruit a larger sample size in future work to allow for voxelwise morphometry to relate patterns of pathology and behavior (Hernandez-Castillo et al., 2018; Kansal et al., 2017). We were not able to do any analysis along these lines in the present study: Not only is our sample small, but also we do not have scans for most of the participants. Another approach would be to recruit individuals with unilateral lesions encompassing posterolateral regions and compare the effect of right and left cerebellar lesions. Another approach might be to focus on individuals with cerebellar degeneration who show language impairments on standard neuropsychological tests of language (e.g., range of seman- tic fluency tasks, word completion tasks, Boston Naming Test (Kaplan et al., 1983)). This infor- mation would allow us to ask if performance on our experimental task of semantic prediction relates to profiles of language competence as assessed by these instruments. While our screen- ing test for cognitive impairment (MoCA) included a semantic fluency task, it is likely neither comprehensive nor sensitive enough to detect subtle language deficits. We note that the patient group in the current study presented with relatively mild cognitive symptoms (as reported by MoCA). It is possible that individuals with milder impairment are more likely to participate in online studies. Third, our experimental design may not have been optimal for testing the prediction and/or CoRT hypotheses. For example, we imposed a 500 ms delay prior to the presentation of the target word, with this interval serving as a cue that the next word would be the target. How- ever, the delay period may have masked the cost of an impaired prediction process, one that operates more slowly in the CD group compared to the control group. That is, the delay period could have provided a “slack” window that provided sufficient time for the CD group to generate a robust semantic prediction (Pashler, 1994). This hypothesis can be addressed by repeating the experiment without the delay period, perhaps using a change in color or font to designate the target word. In terms of the CoRT hypothesis, sentence processing in general may entail the dynamic manipulation of the constituent words and phrases, independent of the semantic content of the sentences. By this view, a CoRT impairment might impact both our dynamic and static sentences. Tasks involving word pair judgments would present another means to compare semantic comprehension without imposing the more temporally extended demands of sentence processing. For example, would the CD group show weaker priming for dynamic word pairs (ball–throw) compared to static word pairs (sunset–red). Fourth, drawing on the motor control literature, we have conceptualized prediction as a dynamic process, one in which an input is fed into an internal model to generate an expec- tancy of the expected sensory consequences (i.e., the target word). It is possible that semantic expectancies do not entail simulation of this form, but rather emerge from memory retrieval (Firth, 1957; Mikolov et al., 2013; Smith & Levy, 2013). That is, after hearing the phrase, “The boss refused to give him a —,” anticipating the word “raise” may not require running a sim- ulation; rather, “raise” may be primed because it tends to co-occur (sadly) with the words “boss” and “refused.” Indeed, a memory-based account might be especially appropriate for high cloze sentences given that they tend to be more formulaic than low cloze sentences (Conklin & Schmitt, 2012). In general, cerebellar degeneration does not impact memory Neurobiology of Language 14 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 / d o i / l / . / 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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 No semantic prediction deficits in cerebellar degeneration retrieval (Appollonio et al., 1993; McDougle et al., 2022). The similar sensitivity of the CD and control groups to cloze probability in the current study may be another manifestation of spared memory and memory retrieval in cerebellar degeneration. CONCLUSIONS Dating back to the earliest neuroimaging studies, language studies have figured prominently in the quest to understand the role of the cerebellum in cognition. Many of the studies have involved tasks that tax semantic retrieval, with the results showing consistent activation in right posterolateral cerebellum when people generate semantic associates (Petersen et al., 1989), manipulate information in verbal working memory (Marvel & Desmond, 2010), or verify meaningfulness (D’Mello et al., 2017; Moberget et al., 2014). Given that cerebellar pathology is not associated with profound aphasia, functional hypotheses have focused on how the cer- ebellum facilitates linguistic processing. Prediction has been central to most of these hypoth- eses given that fluent communication requires feedforward processing, defined within the domain of language comprehension as the ability to anticipate linguistic intent that arises during a conversation with another individual or reading a text. The internal model and CoRT hypotheses provide two ways in which the cerebellum might support prediction in the domain of language, and indeed, support cognition more generally (Diedrichsen et al., 2019; McDougle et al., 2022). However, we failed to obtain support for either hypothesis when put to what we see as rigorous tests. We have outlined limitations with our study, and certainly see a need for future studies that provide further tests of the internal model and CoRT hypotheses, as well as other dimensions of semantic processing. We also recognize that pre- diction, at least as conceptualized here, may not be the appropriate kernel for understanding how the cerebellum supports language. ACKNOWLEDGMENTS Thanks to Jonathan Tsay and Amanda LeBel for providing helpful feedback on the manuscript. FUNDING INFORMATION Richard B. Ivry, National Institutes of Health (https://dx.doi.org/10.13039/100000002), Award ID: NS092079. Richard B. Ivry, National Institutes of Health (https://dx.doi.org/10.13039 /100000002), Award ID: NS105839. AUTHOR CONTRIBUTIONS Maedbh King: Conceptualization: Equal; Data curation: Equal; Formal analysis: Equal; Investi- gation: Equal; Methodology: Equal; Project administration: Equal; Software: Lead; Supervision: Equal; Validation: Equal; Visualization: Equal; Writing—original draft: Equal; Writing—review & editing: Equal. Sienna Bruinsma: Data curation: Equal; Formal analysis: Equal; Methodology: Equal; Visualization: Equal; Writing—original draft: Equal; Writing—review & editing: Supporting. Richard B. Ivry: Conceptualization: Equal; Funding acquisition: Lead; Project administration: Equal; Supervision: Equal; Writing—original draft: Equal; Writing—review & editing: Equal. COMPETING INTERESTS Richard B. Ivry is a co-founder with equity in Magnetic Tides, Inc. The other authors declare no competing interests exist. Neurobiology of Language 15 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 / d o i / l . / / 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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 No semantic prediction deficits in cerebellar degeneration DATA AVAILABILITY STATEMENT All of the sentences used across experimental task blocks (including cloze and CoRT ratings) are available to download on Open Science Framework at https://osf.io/aegkm/?view_only =None. The code used to analyze these data is available to download on Github at https:// github.com/maedbhk/NOL_cerebellum. REFERENCES Ackermann, H., & Hertrich, I. (1994). Speech rate and rhythm in cerebellar dysarthria: An acoustic analysis of syllabic timing. Folia Phoniatrica et Logopaedica, 46(2), 70–78. https://doi.org /10.1159/000266295, PubMed: 8173615 Alexander, M. P., Gillingham, S., Schweizer, T., & Stuss, D. T. (2012). Cognitive impairments due to focal cerebellar injuries in adults. Cortex, 48(8), 980–990. https://doi.org/10.1016/j .cortex.2011.03.012, PubMed: 21549360 Appollonio, I. M., Grafman, J., Schwartz, V., Massaquoi, S., & Hallett, M. (1993). Memory in patients with cerebellar degener- ation. Neurology, 43(8), 1536–1544. https://doi.org/10.1212 / WNL.43.8.1536, PubMed: 8351008 Argyropoulos, G. P. (2016). Experimental use of transcranial direct current stimulation (tDCS) in relation to the cerebellum and language. In The Linguistic Cerebellum (pp. 377–407). Academic Press. Block, C. K., & Baldwin, C. L. (2010). Cloze probability and com- pletion norms for 498 sentences: Behavioral and neural valida- tion using event-related potentials. Behavior Research Methods, 42(3), 665–670. https://doi.org/10.3758/ BRM.42.3.665, PubMed: 20805588 Breska, A., & Ivry, R. B. (2018). Double dissociation of single- interval and rhythmic temporal prediction in cerebellar degener- ation and Parkinson’s disease. Proceedings of the National Academy of Sciences, 115(48), 12283–12288. https://doi.org/10 .1073/pnas.1810596115, PubMed: 30425170 Conklin, K., & Schmitt, N. (2012). The processing of formulaic lan- guage. Annual Review of Applied Linguistics, 32, 45–61. https:// doi.org/10.1017/S0267190512000074 Day, B. L., Thompson, P. D., Harding, A. E., & Marsden, C. D. (1998). Influence of vision on upper limb reaching movements in patients with cerebellar ataxia. Brain, 121(2), 357–372. https://doi.org/10.1093/brain/121.2.357, PubMed: 9549511 Diedrichsen, J., King, M., Hernandez-Castillo, C., Sereno, M., & Ivry, R. B. (2019). Universal transform or multiple functionality?: Understanding the contribution of the human cerebellum across task domains. Neuron, 102(5), 918–928. https://doi.org/10.1016 /j.neuron.2019.04.021, PubMed: 31170400 D’Mello, A. M., Turkeltaub, P. E., & Stoodley, C. J. (2017). Cerebel- lar tDCS modulates neural circuits during semantic prediction: A combined tDCS-fMRI study. Journal of Neuroscience, 37(6), 1604–1613. https://doi.org/10.1523/JNEUROSCI.2818-16.2017, PubMed: 28069925 Fiez, J. A., Petersen, S. E., Cheney, M. K., & Raichle, M. E. (1992). Impaired non-motor learning and error detection associated with cer- ebellar damage: A single case study. Brain, 115(Part 1), 155–178. https://doi.org/10.1093/brain/115.1.155, PubMed: 1559151 Fiez, J. A., Raichle, M. E., Balota, D. A., Tallal, P., & Petersen, S. E. (1996). PET activation of posterior temporal regions during auditory word presentation and verb generation. Cerebral Cortex, 6(1), 1–10. https://doi.org/10.1093/cercor/6.1.1, PubMed: 8670633 Firth, J. R. (1957). A synopsis of linguistic theory, 1930–1955. In Studies in linguistic analysis. Basil Blackwell. Friston, K. (2009). The free-energy principle: A rough guide to the brain? Trends in Cognitive Sciences, 13(7), 293–301. https://doi .org/10.1016/j.tics.2009.04.005, PubMed: 19559644 Gasparini, M., Di Piero, V., Ciccarelli, O., Cacioppo, M. M., Pantano, P., & Lenzi, G. L. (1999). Linguistic impairment after right cerebellar stroke: A case report. European Journal of Neurology, 6(3), 353–356. https://doi.org/10.1046/j.1468-1331 .1999.630353.x, PubMed: 10210918 Gatti, D., Vecchi, T., & Mazzoni, G. (2021). Cerebellum and semantic memory: A TMS study using the DRM paradigm. Cortex, 135, 78–91. Helmuth, L. L., Ivry, R. B., & Shimizu, N. (1997). Preserved perfor- mance by cerebellar patients on tests of word generation, dis- crimination learning, and attention. Learning & Memory, 3(6), 456–474. https://doi.org/10.1101/ lm.3.6.456, PubMed: 10456111 Hernandez-Castillo, C. R., King, M., Diedrichsen, J., & Fernandez- Ruiz, J. (2018). Unique degeneration signatures in the cerebellar cortex for spinocerebellar ataxias 2, 3, and 7. NeuroImage: Clinical, 20, 931–938. https://doi.org/10.1016/j.nicl.2018.09 .026, PubMed: 30308379 Kansal, K., Yang, Z., Fishman, A. M., Sair, H. I., Ying, S. H., Jedynak, B. M., Prince, J. L., & Onyike, C. U. (2017). Structural cerebellar correlates of cognitive and motor dysfunctions in cerebellar degeneration. Brain, 140(3), 707–720. https://doi.org/10.1093 /brain/aww327, PubMed: 28043955 Kaplan, E. F., Goodglass, H., & Weintraub, S. (1983). The Boston naming test (2nd ed.). Lea & Febiger. Keele, S. W., & Ivry, R. (1990). Does the cerebellum provide a com- mon computation for diverse tasks? A timing hypothesis. Annals of the New York Academy of Sciences, 608(1), 179–211. https://doi .org/10.1111/j.1749-6632.1990.tb48897.x, PubMed: 2075953 King, M., Hernandez-Castillo, C. R., Poldrack, R. A., Ivry, R. B., & Diedrichsen, J. (2019). Functional boundaries in the human cerebellum revealed by a multi-domain task battery. Nature Neuroscience, 22(8), 1371–1378. https://doi.org/10.1038 /s41593-019-0436-x, PubMed: 31285616 Kutas, M., & Hillyard, S. A. (1984). Brain potentials during reading reflect word expectancy and semantic association. Nature, 307(5947), 161–163. https://doi.org/10.1038/307161a0, PubMed: 6690995 Ito, M. (2008). Controls of mental activities by internal models in the cerebellum. Nature Reviews Neuroscience, 9(4), 304–313. https://doi.org/10.1038/nrn2332, PubMed: 18319727 Ivry, R. B., & Fiez, J. A. (2000). Cerebellar contributions to cognition and imagery. In M. S. Gazzaniga (Ed.), The new cognitive neuro- sciences (2nd ed., pp. 999–1011). MIT Press. Leiner, H. C., Leiner, A. L., & Dow, R. S. (1986). Does the cerebel- lum contribute to mental skills? Behavioral Neuroscience, 100(4), 443–454. https://doi.org/10.1037/0735-7044.100.4.443, PubMed: 3741598 Lemhöfer, K., & Broersma, M. (2012). Introducing LexTALE: A quick and valid lexical test for advanced learners of English. Behavior Neurobiology of Language 16 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 / d o i / l / . / 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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 No semantic prediction deficits in cerebellar degeneration Research Methods, 44(2), 325–343. https://doi.org/10.3758 /s13428-011-0146-0, PubMed: 21898159 Lesage, E., Hansen, P. C., & Miall, R. C. (2017). Right lateral cere- bellum represents linguistic predictability. Journal of Neurosci- ence, 37(26), 6231–6241. https://doi.org/10.1523/JNEUROSCI .3203-16.2017, PubMed: 28546307 Lesage, E., Morgan, B. E., Olson, A. C., Meyer, A. S., & Miall, R. C. (2012). Cerebellar rTMS disrupts predictive language processing. Current Biology, 22(18), R794–R795. https://doi.org/10.1016/j .cub.2012.07.006, PubMed: 23017990 Lukas, C., Schöls, L., Bellenberg, B., Rüb, U., Przuntek, H., Schmid, G., Köster, O., & Suchan, B. (2006). Dissociation of grey and white matter reduction in spinocerebellar ataxia type 3 and 6: A voxel-based morphometry study. Neuroscience Letters, 408(3), 230–235. https://doi.org/10.1016/j.neulet.2006.09.007, PubMed: 17005321 Marvel, C. L., & Desmond, J. E. (2010). Functional topography of the cerebellum in verbal working memory. Neuropsychology Review, 20(3), 271–279. https://doi.org/10.1007/s11065-010 -9137-7, PubMed: 20563894 Mathiesen, C., Caesar, K., & Lauritzen, M. (2000). Temporal coupling between neuronal activity and blood flow in rat cere- bellar cortex as indicated by field potential analysis. Journal of Physiology, 523(Part 1), 235–246. https://doi.org/10.1111/j .1469-7793.2000.t01-1-00235.x, PubMed: 10673558 McDougle, D. S., Tsay, J. S., Pitt, B., King, M., Saban, W., Taylor, J. A., & Ivry, R. B. (2022). Continuous manipulation of mental representations is compromised in cerebellar degeneration. Brain, Article awac072. Advance online publication. https://doi .org/10.1093/brain/awac072, PubMed: 35202465 Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013). Distributed representations of words and phrases and their compositionality. In Proceedings of the 26th International Conference on Neural Information Processing Systems (Vol. 2, pp. 3111–3119). Curran Associates Inc. Moberget, T., Gullesen, E. H., Andersson, S., Ivry, R. B., & Endestad, T. (2014). Generalized role for the cerebellum in encoding inter- nal models: Evidence from semantic processing. Journal of Neuroscience, 34(8), 2871–2878. https://doi.org/10.1523 /JNEUROSCI.2264-13.2014, PubMed: 24553928 Moberget, T., Hilland, E., Andersson, S., Lundar, T., Due- Tønnessen, B. J., Heldal, A., Ivry, R. B., & Endestad, T. (2016). Patients with focal cerebellar lesions show reduced auditory cor- tex activation during silent reading. Brain and Language, 161, 18–27. https://doi.org/10.1016/j.bandl.2015.08.004, PubMed: 26341544 Nasreddine, Z. S., Phillips, N. A., Bédirian, V., Charbonneau, S., Whitehead, V., Collin, I., Cummings, J. L., & Chertkow, H. (2005). The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. Journal of the American Geriatrics Society, 53(4), 695–699. https://doi.org/10 .1111/j.1532-5415.2005.53221.x, PubMed: 15817019 Pashler, H. (1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin, 116(2), 220–244. https://doi.org /10.1037/0033-2909.116.2.220, PubMed: 7972591 Peelle, J. E., Miller, R. L., Rogers, C. S., Spehar, B., Sommers, M. S., Van Engen, K. J. (2020). Completion norms for 3085 English sen- tence contexts. Behavior Research Methods, 52(4), 1795–1799. https://doi.org/10.3758/s13428-020-01351-1, PubMed: 31993960 Petersen, S. E., Fox, P. T., Posner, M. I., Mintun, M., & Raichle, M. E. (1989). Positron emission tomographic studies of the processing of singe words. Journal of Cognitive Neuroscience, 1(2), 153–170. https://doi.org/10.1162/jocn.1989.1.2.153, PubMed: 23968463 Richter, S., Gerwig, M., Aslan, B., Wilhelm, H., Schoch, B., Dimitrova, A., Gizewski, E. R., Zielger, W., Karnath, H.-O., & Timmann, D. (2007). Cognitive functions in patients with MR-defined chronic focal cerebellar lesions. Journal of Neurology, 254(9), 1193–1203. https://doi.org/10.1007/s00415-006-0500-9, PubMed: 17380238 Richter, S., Kaiser, O., Hein-Kropp, C., Dimitrova, A., Gizewski, E., Beck, A., Aurich, V., Ziegler, W., & Timmann, D. (2004). Preserved verb generation in patients with cerebellar atrophy. Neuropsychologia, 42(9), 1235–1246. https://doi.org/10.1016/j .neuropsychologia.2004.01.006, PubMed: 15178175 Riva, D. (1998). Language deficits in a child with omolateral (left) temporo-basal and cerebellar lesions. Neuropsychologia, 36(1), 71–75. https://doi.org/10.1016/S0028-3932(97 )00095-X, PubMed: 9533389 Saban, W., & Ivry, R. B. (2021). PONT: A protocol for online neu- ropsychological testing. Journal of Cognitive Neuroscience, 33(11), 2413–2425. https://doi.org/10.1162/jocn_a_01767, PubMed: 34347867 Schmahmann, J. D., & Sherman, J. C. (1998). The cerebellar cogni- tive affective syndrome. Brain, 121(Part 4), 561–579. https://doi .org/10.1093/brain/121.4.561, PubMed: 9577385 Schmitz-Hübsch, T., Tezenas du Montcel, S., Baliko, L., Berciano, J., Boesch, S., Depondt, C., Giunti, P., Globas, C., Infante, J., Kang, J.-S., Kremer, B., Mariotti, C., Melegh, B., Pandolfo, M., Rakowicz, M., Ribai, P., Rola, R., Schöls, L., Szymanski, S., … Klockgether, T. (2006). Scale for the assessment and rating of ataxia: Development of a new clinical scale. Neurology, 66(11), 1717–1720. https://doi.org/10.1212/01.wnl.0000219042.60538 .92, PubMed: 16769946 Shepard, R. N., & Metzler, J. (1971). Mental rotation of three- dimensional objects. Science, 171(3972), 701–703. https://doi .org/10.1126/science.171.3972.701, PubMed: 5540314 Smith, N. J., & Levy, R. (2013). The effect of word predictability on reading time is logarithmic. Cognition, 128(3), 302–319. https:// doi.org/10.1016/j.cognition.2013.02.013, PubMed: 23747651 Sokolov, A. A., Miall, R. C., & Ivry, R. B. (2017). The cerebellum: Adaptive prediction for movement and cognition. Trends in Cognitive Sciences, 21(5), 313–332. https://doi.org/10.1016/j .tics.2017.02.005, PubMed: 28385461 Staub, A., Grant, M., Astheimer, L., & Cohen, A. (2015). The influ- ence of cloze probability and item constraint on cloze task response time. Journal of Memory and Language, 82, 1–17. https://doi.org/10.1016/j.jml.2015.02.004 Stoodley, C. J., & Schmahmann, J. D. (2009a). The cerebellum and language: Evidence from patients with cerebellar degeneration. Brain and Language, 110(3), 149–153. https://doi.org/10.1016/j .bandl.2009.07.006, PubMed: 19664816 Stoodley, C. J., & Schmahmann, J. D. (2009b). Functional topogra- phy in the human cerebellum: A meta-analysis of neuroimaging studies. NeuroImage, 44(2), 489–501. https://doi.org/10.1016/j .neuroimage.2008.08.039, PubMed: 18835452 Wolpert, D. M., Ghahramani, Z., & Jordan, M. I. (1995). An internal model for sensorimotor integration. Science, 269(5232), 1880–1882. https://doi.org/10.1126/science.7569931, PubMed: 7569931 Wolpert, D. M., Miall, R. C., & Kawato, M. (1998). Internal models in the cerebellum. Trends in Cognitive Sciences, 2(9), 338–347. https://doi.org/10.1016/S1364-6613(98)01221-2, PubMed: 21227230 Neurobiology of Language 17 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 / d o i / l / / . 1 0 1 1 6 2 n o _ a _ 0 0 0 8 3 2 0 6 2 7 8 7 n o _ a _ 0 0 0 8 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

下载pdf