Motor cortex pre-activation by standing facilitates word retrieval in aphasia

Abstract

Background and Objectives

A tight link between linguistic functions and activation of motor areas has been consistently reported, indicating that the two systems share functional neural resources. However, little efforts have been made to explore whether this knowledge could aid the rehabilitation of neurological conditions affecting language functions (e.g., aphasia). Moreover, the few previous studies that assessed the impact of motor-language interactions in aphasia used manual gestures, which are an integral component of language comprehension and production.

Methods

In the present study we assessed whether pre-activation of the leg-motor cortex, during standing vs. sitting, can facilitate language production in chronic aphasia. In a cross-over within-subject design, we assessed performance on a picture naming task and controlled for effects on processing speed and simple verbal reaction time.

Results

We found that standing compared to sitting had a beneficial effect on the number of semantic self-corrections that resulted in correct naming. In the absence of effects on motor or general processing speed, this points to a specific effect on lexical retrieval and selection. This was further corroborated by an error pattern analysis. Successful semantic self-corrections during standing were only found when there was already partial activation of the target semantic network, i.e., when self-corrections were preceded by an incorrect, but semantically associated naming response.

Discussion

Our findings show that pre-activation of the motor system, that extends beyond the intrinsic link between manual gestures and language, can facilitate lexical access in chronic aphasia and may open new directions in aphasia rehabilitation.

Introduction

Approximately 30% of stroke patients suffer from aphasia which affects language production and comprehension.¹ The most frequent aphasia symptom is impaired word-retrieval (anomia).² Despite the availability of several standardized intervention strategies, chronic anomia is relatively resistant to intervention.³ Thus, novel training-adjuvant therapies need to be devised.

One exciting new direction in aphasia rehabilitation research is exploitation of the tight link between the language and the action (motor) systems that has been discovered in recent years.^{4, 5} It has been suggested that close interactions between these two systems are a relic of the evolution of spoken language from facial and manual hand gestures.^{6, 7} Recently, studies in healthy individuals have assessed whether activity in one domain (e.g., the language system) can facilitate activity in the other domain (e.g., the motor system), or vica versa. Indeed, the excitability of the hand and mouth motor cortex is enhanced during speech comprehension and production tasks, but not during non-linguistic control tasks. This effect has been shown for action and non-action related language materials.^8–14 Moreover, Liuzzi et al.¹⁵ recently provided evidence for a pre-excitation of the leg motor cortex during non-action specific language processing, indicating that language-motor system interactions are more extensive than previously thought (i.e., extending beyond the hand and the mouth) and might not be limited to action specific language materials. It has also been shown that motor activity can affect language processing. Executing or observing manual gestures facilitates lexical retrieval and semantic processing^16–18 and preventing manual gestures can slow down speech production.^{19, 20}

This language-motor interaction system could represent an intriguing new `backdoor' approach to accessing the language system for rehabilitation.²¹ Unlike direct language therapies, this system, in principle, is independent of remaining linguistic functions but is unspecifically connected to the language system. Activating the motor system in aphasia may support intensive speech therapy by providing better “working conditions” to activate the remaining parts of the language network. In particular, it has been shown that even though 2/3 of the neural activity during word-processing is category specific (e.g., hand related word elicit activity in the somatotopic representation of the hand area of the motor cortex), the remaining activity pattern is shared across semantic categories.²² Pre-activation of these common areas of the network (e.g., the inferior frontal cortex) or those which are tightly connected (e.g., the ventral premotor cortex)^{23, 24} might have a beneficial effect on language functions across semantic categories.

Several previous studies demonstrated that naming impairments in aphasia can be reduced by speech accompanying symbolic or iconic gestures²⁵ and even simple pointing gestures that are unrelated to the target words can have beneficial effects on naming performance in non-fluent aphasia.^26–28 Still, pre-activation of the motor system as such was confounded by possible pre-activation of the gestural system by the pointing movement and the precise mechanisms by which meaningful and non-meaningful hand gestures improve language functions in aphasia remain unclear. Some explanations include facilitation of lexical retrieval through increased net activity in a common semantic network,²⁵ a spill-over of activity from the hand motor cortex to the face area,²⁶ or facilitation of compensatory right frontal activity.^{27, 29}

From a theoretical perspective, studies in healthy subjects demonstrated that language-motor facilitation is not restricted to the upper extremities or the face area, but also comprises other motor circuits, including the leg motor system.^{4, 9, 13–15, 30, 31} However, none of the available studies that examined facilitation of language via the motor system in aphasia had tried to activate the motor system without making use of hand gestures. This might also be of clinical relevance, as a number of aphasia patients that have a comorbid upper extremity paresis which may prevent uni- or bimanual gestures. In addition, there might be impaired ability to produce co-speech gestures in aphasia patients per se.²⁵

Thus, based on previous work that demonstrated increased excitability of the leg-motor cortex during language tasks,¹⁵ we assessed whether picture naming in aphasia can be improved by pre-activation of the leg motor cortex. For the purpose of the present exploratory proof-of-principle study, we chose to utilize a simple task manipulation: In a cross-over within-subject design, twenty patients with chronic non-fluent aphasia and anomia named visually presented pictures while either sitting or standing. It has been shown that standing (vs. sitting) results in relatively increased activity in the leg-motor cortex.³² To explore effects on performance, we assessed naming accuracy and latency during the two conditions. In addition, we assessed latency and accuracy during two tasks that aimed at controlling for effects of standing on motor or general processing speed.

Methods

Patients were recruited through a stroke database (Department of Neurology of the University of Münster) and referrals by local neurologists and speech-language pathologists. Inclusion criteria comprised chronic non-fluent aphasia due to cerebrovascular stroke (> 12 months post-stroke), and right-handedness prior to stroke (Edinburgh Inventory Laterality Index > +70).³³ Patients with other aphasia types were excluded as previous studies have only found effects of motor facilitation in patients with non-fluent aphasia.^{26, 27, 29} Patients were screened for paresis using the NIH stroke scale (www.ninds.nih.gov/doctors/NIH_Stroke_Scale.pdf) and patients with severe paresis of the leg that prevented unassisted standing for more than 20 minutes were excluded from the study. Other exclusion criteria comprised severe psychiatric comorbidities (e.g., depression or psychosis as determined by the patients' medical records and the Beck's Depression Inventory, BDI),³⁴ dementia (Mini Mental State Examiniation, MMSE)³⁵ and severe apraxia of speech.

A total of twenty patients with aphasia and mild to moderate anomia were recruited (7 women, 13 men; age mean 55.3 years, range 39–67; time since stroke mean 98.5 months, range 41–258; years of education mean 12.8, range 9–19; BDI mean 6.6, range 1–15; MMSE mean 27.1, range 22–29; for details see Table 1). Aphasia was due to a lesion within the territory of the left middle cerebral artery in 19 patients. One patient had aphasia due to occlusion of the right middle cerebral artery (crossed aphasia; patient #10). Data were analyzed with and without this participant, and removal of this patient did not affect the outcome in the variables of interest. Twelve of the patients suffered from Broca's aphasia and eight had amnestic aphasia as determined by the Aachen Aphasia Test (AAT).³⁶ See Table 1 for details of the AAT-naming subtest and the profile score, a measure of overall aphasia severity. None of the patients were above the threshold for clinical depression (BDI: <17) or dementia (MMSE < 25; in one patient (# 04) the lower score was explained by the primary language disorder; Table 1).

Table 1.

ID	sex	age	handedness *	education (years)	duration of aphasia (months)	AAT syndrome classification **	aphasia severity (AAT-profile) ***	AAT naming (max. 120)	BDI	MMSE	NIH stroke scale (arm/leg)
01	m	47	100% right	12	100	Broca	49.6	89	15	27	0/0
02	m	55	100% right	14	258	Broca	60.4	106	5	28	0/1
03	m	53	100% right	18	109	Broca	50.5	91	9	25	1/0
04	m	67	90% right	11	50	Broca	59.9	109	5	22	1/1
05	m	50	100% right	11	95	Broca	51.9	72	1	27	0/0
06	f	57	100% right	11	115	Broca	49.4	74	3	26	2/2
07	m	65	100% right	11	47	Broca	64.0	98	9	29	0/0
08	m	39	100% right	18	132	Broca	58.5	90	4	29	1/1
09	f	41	100% right	11	117	Broca	58.3	111	1	27	0/0
10	m	59	80% right	11	72	Broca	51.4	89	5	29	0/0
11	m	66	100% right	13.5	42	Broca	49.4	73	7	26	1/1
12	m	63	100% right	10	121	Broca	59.1	88	4	26	1/1
	mean	55.1		12.6	104.8		55.2	90.8	5.6	26.7	0.6/0.6
	SD	9.5		2.7	57.3		5.2	13.4	3.9	2.0	0.7/0.7
13	m	61	100% right	19	61	amnestic	65.0	103	3	30	1/1
14	f	50	100% right	9	41	amnestic	59.9	94	11	28	1/2
15	f	57	100% right	12.5	52	amnestic	69.7	117	6	29	0/0
16	f	57	100% right	11	45	amnestic	63.7	107	1	28	0/0
17	m	58	100% right	13	158	amnestic	53.9	73	12	26	2/1
18	f	42	100% right	13	75	amnestic	61.2	113	14	25	1/1
19	f	61	100% right	11	231	amnestic	62.9	107	11	27	1/1
20	m	58	100% right	15.5	49	amnestic	72.6	118	6	29	0/0
	mean	55.5		13.0	89.0		63.6	104.0	8.0	27.7	0.7/0.7
	SD	6.4		3.0	68.7		5.7	14.7	4.6	1.6	0.7/0.7

M = male, F = female; SD = standard deviation

^*assessed with the Edinburgh Handedness Inventory (Oldfield, 1971)

^**aphasic syndromes according to the criteria of the Aachen Aphasia Test (AAT; Huber et al., 1983)

^***the profile score of the AAT represents a weighted average and t-transformed score of the 5 AAT subtests (Token Test, Repeating, Written Language, Naming, Comprehension). Higher scores indicate better recovery

BDI: Beck Depression Inventory (Hautzinger et al., 1994); MMSE: Mini Mental State examination (Folstein et al., 1975); critical cut-off BDI: > 17, MMSE: < 25

[Note: the lower score of patient 04 was explained by the primary language disorder]

NIH: National Institute of Health Stroke Scale and scoring guidelines can be found at http://www.ninds.nih.gov/doctors/NIH_Stroke_Scale.pdf

Written informed consent was obtained from all participants. The study was approved by the institutional review board of the University of Münster and was conducted in accordance with the Helsinki Declaration.

Experimental design and tasks

Data were collected for three tasks: The primary outcome measure was a picture naming task. Control tasks were a simple verbal reaction time task (saying the word `yes' in response to a visual cue) and a visual discrimination task (deciding whether two geometrical objects where identical or different by a button press). The two control tasks were included to assess whether potential latency differences between the two conditions during picture naming could simply be explained by a more general motor facilitation (verbal reaction time) or increased processing speed (visual discrimination). We employed a cross-over, within-subject design in which each patient had to execute each task while sitting (condition 1) and while standing (condition 2). In condition 1, the participants were instructed to sit in a comfortable and relaxed position in a medical armchair with back support. Upper and lower extremities were positioned on arm and leg rests. During condition 2 the patients were instructed to stand on the floor with their weight on both feet in an upright position (Figure 1A+B illustrates the experimental set-up during the naming task for the two conditions).

Figure 1.

illustrates the experimental set-up for the picture naming task during (A) standing and (B) sitting. Below, the time line of an exemplary session is shown (C)

Presentation of the experimental stimuli was accomplished by a head-mounted pair of Liquid Crystal Display goggles (Sony Glasstron, PLM-S700E) to ensure that visual presentation (e.g., visual angle, viewing distance, brightness) was comparable between conditions and tasks. Stimuli were presented using presentation software (Neurobehavioral Systems, Inc., Albany, CA).

All three experiments and both conditions were administered on the same day with short breaks in between the tasks. All subjects started with the naming tasks, afterwards the two control tasks were administered. For half of the subjects the experimental session began with condition 1 and for the other half with condition 2. All three tasks for each condition were administered consecutively within one session. For the naming task, data from two sessions were obtained, with the order randomized across subjects (Figure 1C shows an exemplary session).

Picture naming task

Prior to the experiment, four sets of 30 line drawings from an online database³⁷ depicting common objects and animals were matched for word frequency (CELEX),³⁸ length (number of letters), name agreement and visual complexity [frequency: mean 1.9–2.0, F(3,116)= .1, p= .9; length: mean 6.2–6.6 letters, F= .3, p= .9; name agreement (max. 1): mean .89–.91, F= .1, p= .9; visual complexity: mean 18898–20062 pixels, F= .08, p= .97]. Two of these matched sets were used during each of the experimental conditions (sitting_1,2 and standing_1,2, i.e., all four sets were used for each patients), and the respective sets were counterbalanced between subjects.

The patients were told that they would see object pictures on a screen for 5 seconds. Their task was to name the pictures as quickly and accurately as possible during this time interval. Each picture presentation was preceded by a blank screen (2.8 seconds). The entire session was digitally recorded via a microphone attached to the goggle system and transferred to a PC for offline analysis.

The number of correct responses during each experimental condition was assessed separately for:

• (1)Correctly produced responses: these responses included immediately correct responses (including minor phonemic errors, i.e., >2/3 of the letters were correctly produced), hesitant correct responses (e.g., mhm, ehm, … `correct response') or corrections for minor phonemic errors (mirrot -> mirror); see Tables 2+3 for details of the two response types
• (2)Semantic paraphasias that were successfully corrected (e.g., dog -> cat]

Table 2.

	syndrome	Correctly named objects (N, max. 120) *	Latency correct naming (msec) **	Semantic self-Corrections (N) **	Semantic paraphasias	other errors (omissions, neologisms)	Correct visual Discrimination (N, max. 88)	Latency visual discrimination (msec)	Latency verbal reaction time (msec)
01	Broca	78 (1)	3154	3	28	11	86	900	1495
02	Broca	104 (2)	2299	3	11	2	85	964	1360
03	Broca	87 (3)	2759	1	14	18	84	1146	1296
04	Broca	91 (3)	2498	3	21	5	84	1749	1567
05	Broca	86 (12)	2579	4	15	15	86	1102	1291
06	Broca	49 (21)	2359	11	38	22	70	2555	1455
07	Broca	82 (5)	2551	1	29	8	82	1548	1460
08	Broca	92 (10)	2317	9	13	6	81	765	1348
09	Broca	91 (24)	2459	5	11	13	88	1358	1414
10	Broca	88(15)	2575	2	17	13	84	789	1431
11	Broca	81 (29)	2408	3	7	29	77	2036	1414
12	Broca	84 (20)	2395	6	20	10	86	1411	1327
	mean	84.4 (12.0)	2529.9	4.2	18.7	12.7	82.7	1360.7	1405.1
	SD	13.0 (9.6)	236.3	3.0	9.0	7.6	4.9	542.4	83.7
13	amnestic	99 (12)	2286	7	12	2	81	1965	1285
14	amnestic	101 (2)	2391	1	13	5	88	1548	1502
15	amnestic	117 (2)	2319	0	2	1	81	937	1455
16	amnestic	97 (1)	2203	2	19	2	85	1075	1505
17	amnestic	51 (5)	3218	0	28	41	87	1775	1413
18	amnestic	96 (3)	2318	1	16	7	81	1618	1458
19	amnestic	82 (0)	2449	2	13	23	88	1689	1339
20	amnestic	95 (1)	2651	1	20	4	82	970	1399
	mean	92.2 (3.2)	2480.0	1.8	15.3	10.6	84.1	1447.6	1420.0
	SD	19.2 (3.8)	327.1	2.2	7.5	14.1	3.2	396.4	77.1
	between group differences	p= .28 (p= .02)	p= .69	p= .06	p= .40	p= .67	p= .49	p= .70	p= .69

^*Numbers represent the total number of correct responses (excluding semantic self corrections); in brackets the number of hesitant responses or minor phonemic corrections.

^**Only immediately correct responses were included in the latency analysis; SD = standard deviation

Table 3.

ID	syndrome	Correct sitting (N) *	Correct standing (N) *	Latency sitting (msec) **	Latency standing (msec) **	Semantic self-corrections sit (N)	Semantic self-corrections stand (N)
01	Broca	37 (0)	41 (1)	3398.4	2911.1	0	3
02	Broca	54 (1)	50 (1)	2277.3	2321.1	0	3
03	Broca	43 (1)	44 (2)	2800.0	2718.5	0	1
04	Broca	44 (2)	47 (1)	2501.2	2496.2	1	2
05	Broca	41 (3)	45 (9)	2620.4	2539.2	1	3
06	Broca	26 (12)	23 (9)	2263.6	2455.3	2	9
07	Broca	38 (5)	44 (0)	2553.6	2549.4	0	1
08	Broca	46 (4)	46 (6)	2318.9	2315.5	3	6
09	Broca	47 (13)	44 (11)	2368.9	2550.4	0	5
10	Broca	43 (7)	45 (8)	2557.9	2593.1	1	1
11	Broca	40 (14)	41 (15)	2427.7	2388.9	0	3
12	Broca	42 (13)	42 (7)	2407.5	2383.7	2	4
	mean	41.7 (6.2)	42.6 (5.8)	2541.3	2518.56	0.8	3.4
	SD	6.6 (5.3)	6.6 (4.8)	311.4	171.9	1.0	2.3
	within group comparison	t(11) =1.0, p= .33		t(11)= −.46, p= .65		t(11)= 4.6, p=. 0007
13	amnestic	52 (8)	47 (4)	2203.6	2368.9	3	4
14	amnestic	51 (1)	50 (1)	2481.7	2301.4	0	1
15	amnestic	58 (1)	59 (1)	2324.4	2315.4	0	0
16	amnestic	48 (0)	49 (1)	2254.1	2153.2	1	1
17	amnestic	24 (2)	27 (3)	3306.0	3130.8	0	0
18	amnestic	46 (3)	50 (0)	2331.7	2305.3	0	1
19	amnestic	39 (0)	43 (0)	2523.6	2375.9	1	1
20	amnestic	46 (1)	49 (0)	2733.6	2569.8	1	0
	mean	45.5 (2.0)	46.7 (1.2)	2519.8	2440.1	0.7	1.0
	SD	10.2 (2.6)	9.1 (1.4)	360.4	301.9	1.0	1.3
	within group comparison	t(7)= 1.2, p= .29		t(7)= −1.9, p= .10		t(7)= 1.0, p= .35

^*Numbers represent the sum of correct responses (excluding semantic self-corrections), in brackets the number of hesitant responses or minor phonemic corrections

^**Only immediately correct responses were included in the latency analysis; SD = standard deviation

Errors were grouped in two categories and analyzed separately: (1) Omissions or other types of responses (e.g., neologisms, where 2/3 or more of the word was produced incorrectly or other unrelated utterances) and (2) semantic paraphasias that were not corrected.

For the precise assessment of verbal response times (response latency) we used a digital stereo recorder with two channels. On the first channel we recorded a computer triggered signal that was sent one second prior to the picture onset and that was not audible for the patients. On the second channel the verbal response of the subject was recorded. The two channels were then combined to a stereo signal with an analog audio mixer. For the analysis of the recorded data, we used a self-written software (http://www.staff.unimarburg.de/~sommerje/) that allowed us to manually determine the onset of the verbal response (i.e., onset of correct name) relative to the trigger signal. For this purpose the software played a short sequence of the second channel in a continuous loop. (Note: For the latency analysis only immediately correct responses were included; i.e., self-corrections with longer and more variable latencies were excluded).

Correctness and latency of the respective responses were assessed by one of the authors (UW) who was blinded to the respective experimental condition (sit/stand).

Control Conditions: Verbal reaction time and visual discrimination tasks

For the verbal reaction time task the subjects were told that they would see a black dot in the middle of a white screen. Their task was to respond to the dot by saying the word `yes' aloud as quickly as possible. A total of 32 dots were displayed at randomized intervals of 2–5 seconds during each of the conditions (sit/stand). To assess the latency of the verbal responses, we used the same dual channel methodology described above. All patients performed at ceiling level during this task, thus, all responses were included in the analysis.

For the visual discrimination task we created a set of simple geometric shapes (supplementary Figure 1). Two of these shapes were presented simultaneously on the screen and the patients had to indicate whether they were identical (e.g., the same stimulus, with the same orientation and size) or different. Subjects responded manually with right index and middle fingers on a response box device. For each condition 44 pairs of shapes were presented (50% identical and non-identical shapes). The order was randomized across and within subjects. The response box was either placed comfortably in the lap of the patient (sitting) or on a height adjustable table so that the subjects could assume a relaxed arm position (standing). Response latencies were calculated for correct responses only.

Statistical analyses

Statistical analyses were conducted with JMP 4.0 (SAS Institute, Inc.). Initially, the effects of the two experimental conditions were assessed for the entire patient sample and all dependent variables by two-tailed paired t-tests. To qualify whether there were differential effects of the task manipulation (sit/stand) for the two groups of patients (amnestic/Broca's aphasia), we first compared patients with Broca's aphasia and amnestic aphasia with regard to (a) the clinical and demographic variables and (b) the dependent variables (i.e., latencies and correct responses for the three experimental tasks). For the latter analyses, the two experimental conditions were pooled (i.e., average number of responses or latencies for sitting and standing). Then, the effect of the posture manipulation for the two groups of patients was further qualified using repeated measures ANOVAs and post-hoc t-tests.

Associations between naming ability and aphasia severity (AAT naming subtests and profile score) were assessed with Pearson correlation coefficients.

Results

All patients tolerated the experimental tasks well and were able to complete the study.

A. Results for the entire patient sample

Effects of sitting and standing on picture naming tasks

For the total number of correct naming responses (including hesitant responses) and response latencies, no differences between sitting and standing were observed (number of correct responses: t(19)=1.55, p=.14; latency: t(19)=−1.34, p=.19). However, a significantly larger number of semantic self-corrections were observed during standing as compared to sitting (t(19)=3.89, p=.001).

Effects of sitting and standing on control tasks

No differences were observed between sitting and standing during the control tasks (latency verbal reaction time: t(19))=.63, p=.53; number of correct responses visual discrimination task: (t(19)=1.12, p=.27; latency visual discrimination: t(19)=−.79, p=.43).

B. Post-hoc analysis for the two groups of aphasia patients (amnestic/Broca's aphasia)

Differences between aphasic syndromes

The two patients groups were comparable with respect to age (F(1,18)=.00, p=.93), education (F(1,18)=.08, p=.77), duration of aphasia (F(1,18)=.31, p=.58), and BDI and MMSE scores (BDI F(1,18)=1.4, p=.24; MMSE F(1,18)=1.3, p=.25). As expected, aphasia was more severe in patients with Broca's aphasia as indicated by a lower AAT profile score (F(1,18)=11.3, p=.003) and there was a trend for lower naming scores in the AAT (F(1,18)=4.2, p=.053; see Table 1).

Similarly, the total number of correct responses (pooled data for sitting and standing) during the experimental naming task (including minor phonemic errors or delayed responses) was higher in amnestic patients compared to patients with Broca's aphasia (mean/SD 92.2±19.2 vs. 84.4±13.0); however, this difference was not statistically significant (F(1,18)=1.2, p=.28). In line with the more severe aphasia in patients with Broca's aphasia, these patients produced more delayed or hesitant correct responses ((F(1,18)=5.9, p=.02) and there was a trend for more self-corrections of semantic paraphasias (F(1,18)=3.8, p=.06; please note that this latter comparison was also significant when we removed patient 13, who produced a disproportionably large number of semantic self-corrections relative to the other patients with amnestic aphasia; Table 2). Even though patients with Broca's aphasia on average produced more semantic paraphasias, omissions, and neologisms, the distributions of error types was statistically not significantly different between the groups (all p>.40, Table 2).

Response latencies between the two groups were comparable for the three tasks (naming: F(1,18)=.15, p=.69; verbal reaction time: F(1,18)=.16, p=.69; visual discrimination: F(1,18)=.15, p=.70). No differences were found for the number of correct responses during the visual discrimination task for the two groups (F(1,18)=.47, p=.49; Table 2).

Post-hoc comparison between aphasia subtypes

A subsequent repeated measures ANOVA revealed that the overall significant beneficial effect of standing (vs. sitting) on the number of semantic self corrections was driven solely by the subgroup of patients with Broca's aphasia who had more severe aphasia. These patients produced more semantic self-corrections during standing vs. sitting compared to the amnestic patients (Interaction Aphasia Type x Posture: F(1,18)= 10.6, p= .004; amnestic aphasia: t(7)=1.0, p=.35; Broca's aphasia: t(11)=4.63, p=.0007; Table 3; Figure 2). This is also corroborated by the fact that aphasia severity (AAT profile score) was negatively correlated with the relatively larger number of self-corrections during standing compared to sitting (r=−.52, p=.01); i.e., patients with more severe aphasia produced more semantic self-corrections during standing compared to sitting.

Figure 2.

shows the effects of the two experimental conditions, expressed as percentage of semantic self-corrections relative to the number of the individual patient's number of incorrect responses; i.e., the number of possible self-corrections. Formula: [(semantic self-corrections_{stand or sit}/60-correct_{stand or sit}) × 100], where 60 is the maximum number of correct responses possible). Combined refers to the combined group of patients

The same analysis was conducted for all the other dependent variables, even though no significant effects of the posture manipulation were found when we included the entire patient sample. No significant interactions were found for these variables (all F=.05–.67, all p>.42).

However, in the less severely affected amnestic subgroup, naming latency was shorter during standing compared to sitting in seven of the eight patients (see Table 3). However, the direct comparison between the two conditions was not significant (t(7)= –1.9, p= .1).

Error patterns and semantic self-corrections

We further analyzed the patterns of semantic errors and classified them in related (i.e., whether or not they belonged to the same broad semantic category or shared features, e.g., cat vs. dog) or unrelated errors (e.g., dog vs. moon). Additionally, the semantic self-corrections were classified by whether or not they belonged to the same broad category as the initial (incorrect) response. As illustrated in supplementary Table 1A both patient groups produced more semantically related than unrelated errors during both conditions (total number of related/unrelated errors: 288/57; related errors; Broca: related > unrelated errors during sitting/standing F(1,22)=27.2/28.7, both p<.0001; amnestic: F(1,14)=19.2/8.0, both p<.02; Please note: the larger number of semantic paraphasias during sitting in patients with Broca's aphasia is explained by the fact that a larger number of self-corrections occurred during standing, see supplementary Table 1B). Thus, in about 80% of the trials in which the patients produced a semantic paraphasia, the response shared semantic features of the target picture. Moreover, when we analyzed the types of semantic self-corrections, in 58 out of 65 instances (89%) the semantic self-correction was preceded by a semantically related response (supplementary Table 1B), i.e., the patients activated a common semantic network, but initially failed to activate the target lexical form.

Discussion

The main finding of the present study was that a simple task manipulation, aimed at increasing activity in the leg motor cortex, can improve word-retrieval in patients with chronic aphasia. Specifically we found a greater number of semantic self-corrections that resulted in correct naming of objects during standing compared to sitting. On the other hand, the total number of immediately correct responses (excluding semantic self corrections) and response latencies were comparable between the two conditions, the number of phonemic error corrections was comparable and, no facilitatory effects were found on the control tasks. Taken together, these findings point to a specific impact of the task manipulation on lexical retrieval and/or selection, rather than effects on processing speed or a simple spill-over of activity from the motor cortex. This is also supported by the fact that, successful semantic self-corrections were only found when there was partial activation of the target semantic network, i.e., when self-corrections were preceded by an incorrect, but semantically associated naming response.

A tight link between linguistic functions and activation of motor areas has been consistently reported by many researchers, indicating that motor and the language systems share functional neural activation patterns.⁵ Owing to the action recognition system, language simultaneously activates the motor system to increase access to lexical information and to facilitate meaning construction.⁵ Moreover, the involvement of the motor system in language functions expands far beyond its role in speech production. Studies have demonstrated that language processing has effects on both the mouth and facial motor systems,¹⁰ the hand motor system,^{11, 12} as well as the leg motor system.¹⁵ Thus, action-perception networks constitute a core mechanism of language production and comprehension and activation of the language system leads to automatic and significant activation of the motor system.

Conversely, it should be possible to enhance activity in language areas by pre-activation of different parts of the motor action system. In other words, motor stimulation could be used to activate the language network and to increase cortical activation in regions within the network, such as Broca's area.³⁹ Rapid functional interaction and information exchange between these two brain systems during language comprehension was found to contribute to semantic processing and to facilitate responses to linguistic stimuli.^{14, 16, 40, 41} In fact, this mechanism and implications for aphasia rehabilitation have already been suggested by findings from a few previous studies: Hanlon et al. demonstrated that a pointing gesture with the (affected) right arm has beneficial effects on naming performance in non-fluent aphasic patients.²⁶ Crosson et al. found that a two week anomia treatment, where naming trials were preceded by a complex left hand/arm movement, leads improved lexical retrieval.²⁸ However, the potentially beneficial impact of motor system pre-activation on language processing has not been studied extensively in aphasia patients. Moreover, none of the previous studies has assessed the impact of other parts than the hand motor cortex on language functions in aphasia and such studies are confounded by the fact that manual gestures are critically involved in language comprehension and production per se²⁵ and do not allow to assess the effects of more general motor cortex activation. Given this theoretical background, our findings now provide for the first time experimental evidence that there is indeed a general pre-activation of the language system by the motor system that extends beyond the hand motor cortex, and can be used to facilitate language functions in aphasic patients.

In view of the neuronal language model based on Hebbian-learning mechanisms,^{5, 42} language acquisition at the neuronal level leads to the formation of semantic networks. These networks are distributed across cortical areas, mainly in left fronto-temporal language regions, and prefrontal motor regions.⁴³ Moreover, within these distributed networks common lexical and semantic processes which overlap with the inferior frontal action recognition system can be distinguished from processes that code for distinct physical and psychological features of the stimuli (e.g., category specific somatotopic activations related to arm vs. mouth related words).²² Two critical features of these distributed representations are that (i) even partial stimulation can activate these networks, and (ii) feedforward and feedbackward connections between different cortical areas amplify cortical activation within the networks. Thus, inducing unspecific activity in the motor system may increase net activity in areas responsible for lexical retrieval across different categories (the above mentioned common parts of the network) which may have facilitated lexical access in our patient sample. Indeed, it has been demonstrated that unspecific pre-activation of the language network with transcranial magnetic stimulation facilitated picture-word verification in healthy subjects.⁴⁴ In the present study, this was expressed as a larger number of semantic self-corrections from semantically related associates, suggesting that motor system pre-activation was only sufficient to enhance naming performance under the condition of partial access to an object name (i.e., category specific pre-activation). Indeed, as shown in supplementary Table 1, approximately 90% of semantic self-corrections that resulted in correct naming of the pictures were preceded by an incorrect but related response. This might also explain why the patients with amnestic aphasia, who in general had more correct responses and less semantic errors, on average did not benefit from the standing intervention.

Also, Pulvermüller et al.²² demonstrated that about 2/3 of the neurometabolic activity during word processing was accounted for by category specific activation, indicating that unspecific input in more general nodes of the network might enhance net activity and facilitate selection processes, on the other hand, this may not be sufficient to improve word retrieval without category specific pre-activation. This remains to be explored in future studies, using a larger number of stimuli belonging to different semantic categories (e.g., leg vs. hand related pictures).

Based on the present data and the fact that lesion information was not available for all of the patients, we can only speculate about the neural mechanisms responsible for improved naming performance during leg-motor cortex pre-activation in our study. It is known that the left inferior frontal gyrus (Broca's area) and the ventral premotor cortex (vPMC), areas which are tightly linked anatomically, show a common activation when healthy subjects observe gestures⁴⁵ and during gait control (e.g., standing).⁴⁶ Moreover, anatomical studies show that part of Broca's area (BA44) and the ventral premotor cortex (BA6) share specific cytoarchitectonic properties.^{23, 24} Thus, due to shared cytoarchitectonic features, tight anatomical links, functional connectivity, and topographic proximity, the ventral premotor cortex (vPMC) may be a good candidate for the remapping of linguistic functions in aphasia patients. This hypothesis needs to be scrutinized in future studies which should combine structural and functional imaging data with brain-stimulation targeting specifically at the vPMC.

Supplementary Material

Suppl Fig 1

Supplementary Figure 1 shows examples of stimuli used for the visual discrimination task (A) examples of the stimuli (B) example for “same object” (C) example for “different object”

Clinical implications.

In summary, we demonstrated that inducing activity in the leg-motor cortex selectively facilitates naming in aphasia patients. This facilitation was limited to patients with more severe (Broca's) aphasia, however, for patients with amnestic aphasia, there was a trend towards reduced naming latency. Thus, the positive findings in this proof-of-principle study provide a rationale for assessing the effectiveness of motor cortex stimulation by more sophisticated and clinically applicable techniques such as stimulation with anodal transcanial direct current stimulation (aTDCS)^47–49 or high frequency repetitive transcranial magnetic stimulation (rTMS).⁵⁰ Future studies should explore whether the combination of speech and language therapy and motor cortex stimulation can enhance treatment outcome.