Journal of the Neurological Sciences 338 (2014) 77–86
Contents lists available at ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Training in rapid auditory processing ameliorates auditory comprehension in aphasic patients: A randomized controlled pilot study Elzbieta Szelag a,b,⁎,1, Monika Lewandowska a, Tomasz Wolak a,c, Joanna Seniow d, Renata Poniatowska d, Ernst Pöppel e,f, Aneta Szymaszek a,b,1 a
Laboratory of Neuropsychology, Nencki Institute of Experimental Biology, 3 Pasteur Str., Warsaw, Poland University of Social Sciences and Humanities, 19/31 Chodakowska Str., Warsaw, Poland Institute of Physiology and Pathology of Hearing, 1 Zgrupowania AK “Kampinos” Str., Warsaw, Poland d Institute of Psychiatry and Neurology, 9 Sobieski Str., Warsaw, Poland e Human Science Center, Ludwig-Maximilian University, 11 Prof.-Max-Lange-Platz, Bad Tölz, Germany f Institute of Medical Psychology, Ludwig-Maximilian University, Munich, Germany b c
a r t i c l e
i n f o
Article history: Received 11 June 2013 Received in revised form 10 December 2013 Accepted 10 December 2013 Available online 18 December 2013 Keywords: Rapid auditory processing Aphasia Auditory comprehension Acoustic temporal training Stroke Neurorehabilitation Restoration of function
a b s t r a c t Experimental studies have often reported close associations between rapid auditory processing and language competency. The present study was aimed at improving auditory comprehension in aphasic patients following specific training in the perception of temporal order (TO) of events. We tested 18 aphasic patients showing both comprehension and TO perception deficits. Auditory comprehension was assessed by the Token Test, phonemic awareness and Voice-Onset-Time Test. The TO perception was assessed using auditory Temporal-Order-Threshold, defined as the shortest interval between two consecutive stimuli, necessary to report correctly their before–after relation. Aphasic patients participated in eight 45–minute sessions of either specific temporal training (TT, n = 11) aimed to improve sequencing abilities, or control nontemporal training (NT, n = 7) focussed on volume discrimination. The TT yielded improved TO perception; moreover, a transfer of improvement was observed from the time domain to the language domain, which was untrained during the training. The NT did not improve either the TO perception or comprehension in any language test. These results are in agreement with previous literature studies which proved ameliorated language competency following the TT in language-learning-impaired or dyslexic children. Our results indicated for the first time such benefits also in aphasic patients. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The relationship between rapid auditory processing (RAP) and language has been discussed for many years. Several studies indicated deficits in the processing of single units of language–phonemes (e.g., stopconsonants limited in time up to about 40 ms) due to RAP deficits. They resulted in deficient phonological awareness underlying receptive language deficits. These deficits have been often accompanied by disordered RAP in the time domain of some tens of milliseconds and reflected, for example, in auditory perception of temporal order (TO) of two stimuli presented in rapid succession. Using such a paradigm, auditory temporal-order-threshold (TOT) can be measured and defined as the shortest time gap between two consecutive stimuli which is necessary for a subject to report correctly their relation ‘before–after’. Some ⁎ Corresponding author at: Laboratory of Neuropsychology, Nencki Institute of Experimental Biology, 02-093 Warsaw, 3 Pasteur Street, Poland. Tel.: +48 22 5892286; fax: +48 22 8225342. E-mail address: [email protected] (E. Szelag). 1 These authors contributed equally to this work. 0022-510X/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jns.2013.12.020
evidence indicated that for normal healthy volunteers this gap usually ranges from ca. 30 ms up to 80 ms [1–4]. Its duration is usually longer in elderly subjects than in younger ones, resulting in higher TOT values because of overlapping age-related declines in both timing and mental activity [5–8]. Elevated TOTs were also reported in patients with aphasia following left hemisphere brain lesions [9,2,10–12], children with language-learning-impairment [13,14] and children or adults with dyslexia [15,16]. Based on these data, some authors argued that proper processing of rapid temporal changes in the incoming speech signal is required for correct phonological processing [17]. Early studies by Efron [17] and Swisher & Hirsh [2] emphasized that timing deficits in aphasics were observed in ordering both auditory and visual stimuli (e.g. two diodes of different colours). This would suggest the existence of a central neural mechanism underlying the perception of TO, independent of the sensory modality. More recent studies confirmed significantly higher auditory TOTs in Wernicke's aphasics (about 120 ms) than in matched controls [9,10,12]. The close association between language and timing disorders was supported by neuroanatomical overlapping of structures involved in RAP and receptive language functions [9,10,12,18,19]. These structures comprised left
78
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
hemispheric gyrus temporalis superior, gyrus temporalis medius and surrounding white matter. Although many authors noted the overlapping between timing and language deficits [20–23], others doubt the existence of a cause–effect between these two functions [24–27]. In light of this evidence we should mention two important publications by Merzenich et al. [28] and Tallal et al. [29]. They reported amelioration of language functions following TIP training in languagelearning-impaired children. These authors developed several training tasks in which three variables were adaptively adjusted to each subject's performance i.e., duration, frequency and inter-stimulus-interval in the presented stimuli. Four weeks of such intensive training resulted in significant improvement of RAP (i.e., TOTs lowered from 400 ms pretraining to 100 ms post-training) which was accompanied by improvements in phonological processing and auditory comprehension. These language functions remained untrained during the applied intervention [28,29]. Based on these reports, a training programme Fast ForWord (FFW) was developed by Scientific Learning (http://www.scilearn.com) and broadly implemented into clinical practice [30]. FFW was also applied in children and adults with dyslexia, bringing positive therapeutic effects. Nevertheless, some authors did not confirm any positive effects of FFW on writing, reading or spelling [30, for a recent review]. Despite existing evidence on clinical benefits of temporal training in language-disordered children, data indicating positive effects of temporal training in aphasic sample are very limited. The only early report by v. Steinbüchel et al. [31] demonstrated slight improvement in both auditory perception of TO and phoneme discrimination in aphasic patients following training in temporal discrimination. Their training procedure comprised 4 or 8 h of training in TO perception and a verbal feedback was provided on the correctness achieved. However, detailed information on the number of tested aphasics, description of the patient sample, temporal and control training parameters were missing from this preliminary report. To the best of our knowledge, there are no more reports on the application of temporal training in aphasic patients. It should be mentioned that aphasia constitutes a major medical and social problem in modern society. As estimated by the demographic data, up to 270,000 patients with stroke-related aphasia are diagnosed worldwide each year [32]. Only a minority of them recover completely from their language deficits. Examinations of typical recovery curves revealed that only 25% of patients have a chance for the full restoration of disturbed language. It indicates the importance of implementation of new techniques of rehabilitation to change the burden of languagedisordered individuals with the adoption of new intervention based on neuronal mechanisms underlying language. Considering the importance of RAP in language, in the present study we tested whether timing deficits in aphasics could be reduced by the specific acoustic training in RAP. Furthermore, we investigated whether a transfer of improvement could be observed from the trained time domain to the language domain which was untrained at that time. Taking into account the number of aphasic patients, such a new therapeutic approach could be applied in neurorehabilitation programmes targeting aphasic patients. This study offers some important scientific indications for new components of therapy programmes. 2. Methods 2.1. Participants Eighteen patients (9 male and 9 female) suffering from aphasia after first-ever stroke (haemorrhage or infarction, lesion age x ± SD = 17 ± 11 weeks) participated in the study. They varied in age from 34 to 76 years (x ± SD = 55.4 ± 11.2 years), were right-handed [33], Polish native speakers and had normal hearing level verified by screening audiometry (audiometer AS 208), using frequencies of 250, 500, 750, 1000, 1500, 2000 and 3000 Hz [34]. Apart from stroke they had neither
neurological nor psychiatric disorders and reported no history of head injuries. Description of the patient sample is given in Table 1. All patients displayed auditory comprehension deficits evidenced by the Token Test [35]; Phoneme Discrimination Test (PDT) [36] and Voice-Onset-Time (VOT) Test [37]. Their language deficits were accompanied by disordered RAP (for detailed description see below). We adopted the following exclusion criteria: recurrent stroke, global aphasia with poor verbal contact, poor general health, participation in other rehabilitation programmes during our study, and older age to minimize undiagnosed cognitive deficits which might have negative effects on the training results. These inclusion/exclusion criteria allowed to diminish the risk uncontrolled side-effects which could influence the experimental verification of therapy effectiveness. But such therapy is addressed in future broader patients population suffering from also other language disorders. It was a blinded randomized controlled study. The patients were randomly assigned into two groups according to age, gender, lesion age, level of comprehension deficits and RAP deficits (Tables 1–3). The first group was assigned to temporal training (TT, n = 11) and the second one to nontemporal control training (NT, n = 7). For detailed description of the training protocols see below. Evidence has suggested that the above mentioned subject-related factors generate vulnerability with respect to restoration of language and aphasia rehabilitation. Synopsis of pre- (white column) and post-training (grey column) performance profile in each individual subject is given in Tables 2 and 3. Using U Mann–Whitney test, all pre-training between-group differences for all these variables were nonsignificant, i.e., for age: z = −1,09, p b .28; lesion age: z = 0.18, p b .86; RAP assessed by TOT: z = 0.45, p b .66, comprehension level assessed by the Token test: z = − 0.58, p b .57; PDT z = 0.73, p b .45 (Table 2), VOT z = −0.14, p b .90. It indicated a balance between TT and NT groups (Tables 2 and 3), allowing the comparison of the effectiveness of two different training procedures in groups with relatively similar subject-related characteristics, as well as pre-training performance, thus, similar prognosis for language recovery. The place of the lesion was evidenced by CT or MRI. Fig. 1 presents the summarized damaged regions in TT and NT groups. Neuroanatomical analyses confirmed that in both groups lesions were localized mainly in the left hemispheric temporal lobe and covered the classical areas engaged in both auditory comprehension and time perception [10,12,38,39]. In particular, in the TT group lesioned areas comprised superior temporal gyrus, Heschl's gyrus, Rolandic operculum and insula. Almost exactly the same areas were damaged in the NT group: superior and middle temporal gyrus, Heschl's gyrus, Rolandic operculum, insula, putamen and frontal inferior orbital lobe. Furthermore, the synopsis of brain damaged regions in individual patients assigned to TT and NT is documented in Fig. 2. Considering individual patients' data evidenced in Tables 1–3 and Fig. 2 it may be assumed that the TT and NT groups were as matched as possible. 2.2. Ethical approval The research was approved by the Bioethics Commission at the Institute of Psychiatry and Neurology in Warsaw (permission no 5/2005 from February 2nd 2005) as well as by the Bioethics Commission at the Warsaw Medical University (permission no 5/2010 from January 26th 2010). The study was conducted according to the principles expressed in the Helsinki Declaration; the written informed consent from each participant was obtained prior to testing. 2.3. Procedures The study comprised both assessment and training procedures. The assessment procedures included TOT measurement, language and attention tests which were performed before (pre-training assessment) and after the training (post-training assessment). Moreover, the follow-up
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
79
Table 1 Description of the patient sample included in temporal training and non-temporal control training (abbreviations: M = male, F = female; I = infarction, H = haemorrhage stroke). No
Gender
Patient age (years)
Type of stroke
Lesion age (weeks)
Training Group
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
M M M M M M F F F F F M M M F F F F
70.4 61.2 57.9 51.2 43.4 34.7 70.4 58.5 52.5 45.1 37.2 63.6 61.4 48.6 76 57.4 54.8 52.5
I I I I I I I I H I H I I I I I H H
13 10 10 9 36 5 8 14 14 40 26 5 7 36 17 8 15 24
Temporal training (TT)
assessment was repeated about 6 months after the training completion in TT group. Only 6 patients from the TT group took part in follow-up assessment, as the remaining 5 patients during the relatively long period of 6 months between post-training and follow-up assessment suffered from recurrent stroke or other systemic diseases. As these acquired additional disorders may affect long-term language recovery, these 5 patients could be not considered in the follow-up assessment (compare our patient inclusion criteria). The training procedures comprised either TT or NT. 2.3.1. Assessment procedures 2.3.1.1. TOT measurement. Stimuli were pairs of 1 ms rectangular clicks presented in rapid succession with varied inter-stimulus-intervals (ISIs). The paired-clicks were presented monaurally, i.e., one click was presented to the one ear, followed by a second click to the other ear. The stimuli were generated by a 16-bit Sound Blaster Extigy Sound Card and delivered at a comfortable listening level through the headphones SONY MDR-CD 480. The task was to report the TO of clicks within each pair by pointing to one of two response cards. Each card displayed the possible order of presented clicks, i.e., right–left or left–
Nontemporal control training (NT)
right. Each stimulus pair was preceded by a warning signal of 400 ms, delivered 1500 ms before the first click. The ISIs varied from 1 to 600 ms according to the adaptive maximum-likelihood-based algorithm [40]. In each trial ISI was calculated on the basis of correctness of previous responses. This tracking procedure estimated individual TOTs as the minimum ISI between two clicks at which a subject reported their order at 75% correctness. TOT values were assessed using ‘Yet Another Adaptive Procedure’ [41] on the basis of the maximum likelihood parameter estimation. The measurement was continued until the TOT value was located with a probability of 95% inside a + 5-ms interval around the currently estimated threshold [42]. The proper data collection followed after an introductory session in which the patient reported the order of clicks presented with relatively long ISI of 600 ms. If the pre-defined criterion of 11 correct responses in last 12 presentations was reached, the proper measurement started. The task was completed in a soundproof room and repeated twice during two sessions separated by a few days. For each patient the mean value of TOT from the two sessions was calculated (for individual data see Table 2) and compared to TOT of healthy volunteers matched for age. At these comparisons we based on our normative data were collected
Table 2 Synopsis of pre- and post-training individual performance profiles on TOT, Token Test and PDT in particular patients assigned to TT and NT groups. Numbers of individual patients correspond with those in Table 1 and Fig. 2. Asterisks indicate situations where patients were not able to perform the task. No
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Training group
Temporal training (TT)
Nontemporal control training (NT)
Token test (percentage of errors)
PDT (percentage of errors)
Pre
TOT (mean; ms) Post
Pre/post difference
Pre
Post
Pre/post difference
Pre
Post
Pre/post difference
386 294 141 289 168 102 377 180 522 460 395 112 265 229 444 250 391 169
159 158 63 96 67 48 277 99 497 186 99 86 178 114 510 182 285 206
227 136 78 193 101 54 100 81 25 274 296 26 87 115 −66 68 106 −37
58 * 44 58 * 18 64 30 * * 48 22 70 64 58 34 58 46
56 * 12 48 * 8 36 4 * * 38 12 78 64 58 18 52 34
2 * 32 10 * 10 28 26 * * 10 10 −8 0 0 16 6 12
6.25 34.38 15.63 34.38 31.25 0 43.75 0 46.88 15.63 3.13 9.38 31.25 6.25 25 0 15.63 9.38
12.5 21.88 0 12.5 21.88 0 15.63 0 50 125 3.13 0 37.5 6.25 9.38 0 12.5 9.38
−6.25 12.5 15.63 21.88 9.37 0 28.12 0 −3.12 3.13 0 9.38 −6.25 0 15.62 0 3.13 0
80
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
Table 3 Synopsis of pre- and post-training individual performance profiles on VOT in particular patients assigned to TT and NT groups. The typical voicing zone comprises VOT values of −100, −90, −80, −70, −60 and −50 ms whereas, the typical unvoicing zone, the VOT values are +5, +10, +20, +30, +40, +50, +60, +70, +80 and +90 ms. For each of these two zones the mean percentage of correct responses is provided. The typical transition zone comprising VOT values of −40, −30, −20, −10 and 0 ms was dropped, as words characterized by these VOT values are usually reported randomly either as /DOMEK/ or as /TOMEK/ (compare Fig. 5). Numbers of individual patients correspond with those in Tables 1, 2 and Fig. 2. No
Training group
The mean percentage of reported words Voiced words
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Temporal training (TT)
Nontemporal control training (NT)
Unvoiced words
Pre
Post
Post/pre difference
Pre
Post
Post/pre difference
86 47 100 56 100 100 39 100 75 39 89 100 58 92 86 94 70 100
100 75 100 78 97 100 92 100 70 56 89 100 78 95 94 100 72 97
14 28 0 22 −3 0 53 0 −5 17 0 0 20 3 8 6 2 −3
98 50 100 78 95 97 75 100 62 65 93 97 68 50 72 95 62 98
100 78 100 70 98 100 92 98 72 80 85 100 85 83 67 100 55 93
2 28 0 −8 3 3 17 −2 10 15 −8 3 17 33 −5 5 −7 −5
from 86 volunteers [4] or 168 volunteers [3] varied in age from 20 to 69 years. Accordingly, in the present study the TOT cut-off value for patient inclusion was 1.5 SD above the mean value evidenced in healthy volunteers at a given age. Such inclusion criterion met about of 80% pretested patients. Patients included to the study were, next, assigned to our training if their temporal ordering deficits were accompanied by comprehension deficits, evidenced at least in the Token Test. 2.3.1.2. Language tests. Auditory comprehension was tested using 3 tests: (1) Token Test; (2) Phoneme Discrimination Test and (3) VoiceOnset-Time Test. Token Test constitutes a part of the Aachener Aphasie Test [35] and it is remarkably sensitive to disrupted linguistic processes that are central to aphasic disorder. Plastic tokens (coloured squares and circles of two sizes: big and little) were presented and a patient carried out oral commands, e.g., “Touch the little red circle and the big yellow square”. The test consists of 50 commands given in 5 sections of increasing complexity. Percentage of total number of committed errors was analysed. Phoneme Discrimination Test (PDT) comprises 64 paired-words presented in 8 series of 8 pairs each (75% were different and 25% the same). The words within each pair differed in consonants, contrasted for place of articulation, plosive, fricative, voicing and nasality. The task was to judge whether paired words were the same or different. Responses were given by pointing to one of two response cards. Alternative versions of 4 series of this test were used in consecutive assessments (pre-, post-training and follow-up). The outcome measure was the percentage of committed errors. Voice-Onset-Time (VOT) Test comprises differentiation between voiced and unvoiced contrast in two Polish words /TOMEK/ and / DOMEK/ (in English: Tom/house). On the basis of two words naturally spoken by the Polish native speaker, 21 pseudo-word stimuli varying in the VOT of the initial stop consonant were synthesized. They were created by a manipulation in the duration of an interval between the burst and the onset of laryngeal pulsing in the initial consonant in naturally uttered word /TOMEK/ (unvoiced). According to the Polish phonology, confirmed by our normative data collected from 67 healthy volunteers [37], pseudo-words of the VOT from − 100 to − 50 ms (voiced zone) are typically recognized as voiced /DOMEK/, whereas, from + 5 to + 90 ms (voiceless zone) as unvoiced /TOMEK/. There is
also a transition zone for VOT from −40 to +5 ms, where the discrimination is at the chance level (50% voiced and 50% voiceless consonants). The test comprised 6 series; each of them contained 21 randomly ordered pseudo-words of VOT from −100 to +90 ms in 10 ms steps. The patient's task was to decide whether the presented pseudo-word was / TOMEK/ or /DOMEK/. Responses were given by pointing to one of two response cards which displayed pictures corresponding to these two words. The percentage of reported /TOMEK/ or /DOMEK/ at particular VOTs was analysed. 2.3.1.3. Attention tests. Two aspects of attention, i.e. alertness and vigilance were assessed in order to control attentional deficits which may influence patient performance [43]. Accordingly, alertness is defined as the ability to raise and maintain a high level of attention in anticipation of a test stimulus, whereas, vigilance is the ability to maintain attention over a longer period of time. Alertness was assessed by measuring a simple reaction time in response to a visual stimulus presented as a cross in the monitor centre and cued by an auditory warning signal. In the vigilance task, a low and a high tone were presented sequentially. The task was to press the button as soon as possible when two identical tones were presented in a row. Simple reaction time was measured in both attention tests. 2.3.2. Training procedures Two computer auditory training procedures were applied: training in temporal processing (TT) and nontemporal control training (NT). 2.3.2.1. Training in temporal information processing. TT procedure was similar to that of TOT assessment, i.e., two clicks were presented via headphones with various ISIs and the task was to reproduce the TO of two clicks presented in consecutive pairs. The task difficulty was adaptively adjusted for each patient on the basis of correctness achieved (see above for the description of adaptive procedure described in TIP assessment). The paired clicks were presented in 10-trial blocks. For each patient the starting ISI was calculated individually on the basis of the mean TOT value from two sessions obtained in pre-training assessment. This mean TOT was increased by a constant 20 ms interval, to create a relatively comfortable difficulty level. Within each block paired clicks were presented at constant ISI, which varied between blocks according
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
81
Fig. 1. Schematic display of the common lesioned areas for TT and NT subjects. Localization of lesions comprised only left hemispheric regions.
to an algorithm related to the patient's actual TOT. When at least 8 correct responses were given within one block (classified as ‘correct block’), ISI in the next block decreased, making the task more difficult. Thus, the adaptive algorithm used in TT depended on the subject's performance. If the actual TOT was longer than 100 ms, ISI decreased by 5 ms; in case of TOT between 50 and 100 ms by 2 ms; whereas for TOT below 50 ms by 1 ms. If seven or fewer correct responses were obtained within
one block (‘incorrect block’), in the next block the ISI increased by 1 ms, regardless of patient's actual TOT. 2.3.2.2. Nontemporal control training. In case of NT, paired 1-second tones with constant, relatively long ISI of 3 s were presented via headphones at different volumes. Within each pair one tone was louder than the other. The task was to report which tone was louder: the first or the second.
82
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
The following four frequencies were used: 400, 600, 800, and 1000 Hz. Only one frequency was presented within each pair or block. According to our pretesting data, the difficulty level of TT and NT was comparable as much as possible and demanded comparable mental loud. Furthermore, both the training duration, protocol and motivation system were identical. Although the NT has to involve, as does every cognitive task, some temporal cues related to the stimulus duration or ISI, it concerned multisecond timing, and thus a totally different time range compared to the topic of the TT which was focussed on millisecond timing. During the TT and NT each pair of sounds was preceded by a visual warning signal. Subjects answered by pointing to the response cards placed in front of them. Next, the experimenter entered their responses into the computer programmes. After each trial, a visual feedback on correctness was provided on the monitor, i.e. a smile or sad face after each correct or incorrect response, respectively. In TT and NT the same motivation system was used. For each correct response patients obtained one point. After completion of each block (10 trials), the total number of points was displayed on the monitor. Finally, after each ‘correct block’ the patient was rewarded with a puzzle which was collected on the monitor during each session. Each training type comprised eight 45-minute sessions, held three times a week. The whole therapy was completed within 3 weeks and comprised about14 individual sessions with each patient performed in the Laboratory of Neuropsychology, Nencki Institute. During each session rest-pauses were provided on patient's request. The duration of these pauses was excluded from the effective training time.
2.4. Statistical analyses To verify the effect of TT or NT on temporal processing (TOT), language and attention functions (pre- vs. post-training), as well as the stability of these effects (follow-up vs. post-training and follow-up vs. pretraining) Wilcoxon Signed-Rank Test for two dependent samples was performed. 3. Results 3.1. The effect of TT and NT on timing and language functions: within-group comparisons for pre-training versus post-training assessment 3.1.1. TOT measurement In TT group, TOT values were significantly lower (improved performance) post- in comparison to pre-training (n = 11; z = 2.93; p b .003). The corresponding value after NT was nonsignificant (n = 7; z = 1.52; p b .13, Fig. 3). These results are illustrated by synopsis of pre- and post-training individual data (Table 2). Despite relatively similar TOT values pre-training, the differences between pre- and posttraining performance in NT seem smaller than these in TT group. 3.1.2. Language processing 3.1.2.1. Token test. After TT the percentage of errors decreased significantly in comparison to pre-training (n = 7; z = 2.37; p b .018). After NT the performance remained at a stable level, i.e., the difference between pre- and post-training was nonsignificant (n = 7; z = 1.48; p b .14; Fig. 4A). These relationships are reflected in individual data, indicating post-training the reduced number of errors in patients following TT, in comparison to NT (Table 2).
Fig. 2. The brain damaged regions in individual patients assigned to temporal or nontemporal control training. Numbers of individual patients correspond with those in Table 1. For patient nos. 3, 11, 12, 15, and 16, the neuroanatomical data were not available. Abbreviations: LH—left hemisphere, RH—right hemisphere.
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
83
Fig. 3. The mean Temporal-Order-Thresholds in ms (with SEM) for pre- and post-training assessment in TT and NT groups. The significant difference is indicated by an asterisk (p b .003). The number of subjects is given in brackets.
3.1.2.2. PDT. After TT the mean percentage of errors tended to be lower than in pre-training assessment (n = 11; z = 1.89; p b .059). In the NT group, the difference between pre- and post-training assessment was nonsignificant (n = 7; z = 1.10; p b .28, Fig. 4B). Looking at the performance profiles in individual patients (Table 2), despite a huge improvement after TT in some patients (e.g., patients no 2, 3, 4, 7), in the other ones the progress was less clear, or even missing (e.g., patients no 1, 9). Similarly, unclear relations on PDT were evidenced after NT. 3.1.2.3. Voice-Onset-Time test. After TT improved performance was found in both the voiced and unvoiced zone. In the voiced zone (at VOT = − 90 ms and VOT = − 60 ms) the increased percentage of voiced words /DOMEK/ was reported (z = 1.89, p b .06; z = 2.20, p b .028, respectively). It was accompanied by the improved performance in the unvoiced zone (at VOT = 40 ms), where significantly more unvoiced words /TOMEK/ were reported (z = 2.11, p b .035). Additionally, the mean percentage of reported unvoiced words /TOMEK/ increased significantly in the transition zone (at VOT = 0; z = 2.02, p b .044; Fig. 5). In contrast, after NT the differences between the numbers of reported voiced words /DOMEK/ and unvoiced /TOMEK/ were nonsignificant for any VOT value, i.e. in the voiced, unvoiced or transition zone. The individual performance profiles in particular patients are provided on Table 3.
Individual performance profiles in particular patients (Table 3) indicated after TT for the typical voicing zone higher improvements in the mean of percent of reported voicing words /DOMEK/ than after the NT. On the other hand, these differences for the typical unvoicing zone seem more similar. 3.1.3. Attentional processing In both TT and NT groups, attentional performance remained at a similar level in post-training assessment. The difference between postand pre-training in the mean reaction times was nonsignificant for both alertness (TT group: z = 0.09; p b .93; NT group: z = 0.94; p b .35) and vigilance (TT group: z = 1.15, p b .25; NT group: z = 0.54, p b .60). 3.2. The stability of improvements following TT As explained in the Participants Section, the follow-up assessment was completed only for participants undergoing TT. Due to reduced patient sample (objective reasons, see more explanations in the Procedure Section), statistical analyses were performed only on 6 patients. In Token Test only 4 patients were tested, therefore, no statistical analyses were performed. Table 4 presents results of comparisons between particular assessments in the reduced patient sample.
Fig. 4. The mean percentage of errors (with SEM) in TT and NT groups committed in: (A) Token Test and (B) Phoneme Discrimination Test in pre- and post-training assessment. The number of subjects is given in brackets; some subjects were not able to perform the task (compare Table 2). (*)—significant difference (p b .018); (#)—tendency towards significance (p b .059).
84
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
Fig. 5. The mean percentage of reported voiced and unvoiced words at particular VOT values in pre- and post-training assessment after TT and NT. (*)—significant differences (see text for detailed level); (#)—tendency towards significance (p b .06).
To summarize, in the follow-up assessment a relatively stable level of performance was observed compared to post-training. It is important to note that for all tasks in follow-up assessment the performance was not lower than in pre-training for all analysed outcome measures, i.e., milliseconds (TOT measurement), percentage of committed errors (Phoneme Discrimination and Token Test), percentage of reported voiced/ unvoiced words in the voiced and unvoiced zones (VOT test). 4. Discussion 4.1. Summary of results The TT in aphasic patients significantly improved both temporal ordering and language competency. In contrast, the control NT did not Table 4 Stability of improvements for particular tasks in consecutive assessments following TT. The statistical significance of differences between particular assessments in reduced patient sample was indicated. The other difference between consecutive assessments were nonsignificant. No statistical comparisons were performed for the Token Test (n = 4). Results obtained in particular assessments Task
N pre-training 296
p
Contents lists available at ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Training in rapid auditory processing ameliorates auditory comprehension in aphasic patients: A randomized controlled pilot study Elzbieta Szelag a,b,⁎,1, Monika Lewandowska a, Tomasz Wolak a,c, Joanna Seniow d, Renata Poniatowska d, Ernst Pöppel e,f, Aneta Szymaszek a,b,1 a
Laboratory of Neuropsychology, Nencki Institute of Experimental Biology, 3 Pasteur Str., Warsaw, Poland University of Social Sciences and Humanities, 19/31 Chodakowska Str., Warsaw, Poland Institute of Physiology and Pathology of Hearing, 1 Zgrupowania AK “Kampinos” Str., Warsaw, Poland d Institute of Psychiatry and Neurology, 9 Sobieski Str., Warsaw, Poland e Human Science Center, Ludwig-Maximilian University, 11 Prof.-Max-Lange-Platz, Bad Tölz, Germany f Institute of Medical Psychology, Ludwig-Maximilian University, Munich, Germany b c
a r t i c l e
i n f o
Article history: Received 11 June 2013 Received in revised form 10 December 2013 Accepted 10 December 2013 Available online 18 December 2013 Keywords: Rapid auditory processing Aphasia Auditory comprehension Acoustic temporal training Stroke Neurorehabilitation Restoration of function
a b s t r a c t Experimental studies have often reported close associations between rapid auditory processing and language competency. The present study was aimed at improving auditory comprehension in aphasic patients following specific training in the perception of temporal order (TO) of events. We tested 18 aphasic patients showing both comprehension and TO perception deficits. Auditory comprehension was assessed by the Token Test, phonemic awareness and Voice-Onset-Time Test. The TO perception was assessed using auditory Temporal-Order-Threshold, defined as the shortest interval between two consecutive stimuli, necessary to report correctly their before–after relation. Aphasic patients participated in eight 45–minute sessions of either specific temporal training (TT, n = 11) aimed to improve sequencing abilities, or control nontemporal training (NT, n = 7) focussed on volume discrimination. The TT yielded improved TO perception; moreover, a transfer of improvement was observed from the time domain to the language domain, which was untrained during the training. The NT did not improve either the TO perception or comprehension in any language test. These results are in agreement with previous literature studies which proved ameliorated language competency following the TT in language-learning-impaired or dyslexic children. Our results indicated for the first time such benefits also in aphasic patients. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The relationship between rapid auditory processing (RAP) and language has been discussed for many years. Several studies indicated deficits in the processing of single units of language–phonemes (e.g., stopconsonants limited in time up to about 40 ms) due to RAP deficits. They resulted in deficient phonological awareness underlying receptive language deficits. These deficits have been often accompanied by disordered RAP in the time domain of some tens of milliseconds and reflected, for example, in auditory perception of temporal order (TO) of two stimuli presented in rapid succession. Using such a paradigm, auditory temporal-order-threshold (TOT) can be measured and defined as the shortest time gap between two consecutive stimuli which is necessary for a subject to report correctly their relation ‘before–after’. Some ⁎ Corresponding author at: Laboratory of Neuropsychology, Nencki Institute of Experimental Biology, 02-093 Warsaw, 3 Pasteur Street, Poland. Tel.: +48 22 5892286; fax: +48 22 8225342. E-mail address: [email protected] (E. Szelag). 1 These authors contributed equally to this work. 0022-510X/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jns.2013.12.020
evidence indicated that for normal healthy volunteers this gap usually ranges from ca. 30 ms up to 80 ms [1–4]. Its duration is usually longer in elderly subjects than in younger ones, resulting in higher TOT values because of overlapping age-related declines in both timing and mental activity [5–8]. Elevated TOTs were also reported in patients with aphasia following left hemisphere brain lesions [9,2,10–12], children with language-learning-impairment [13,14] and children or adults with dyslexia [15,16]. Based on these data, some authors argued that proper processing of rapid temporal changes in the incoming speech signal is required for correct phonological processing [17]. Early studies by Efron [17] and Swisher & Hirsh [2] emphasized that timing deficits in aphasics were observed in ordering both auditory and visual stimuli (e.g. two diodes of different colours). This would suggest the existence of a central neural mechanism underlying the perception of TO, independent of the sensory modality. More recent studies confirmed significantly higher auditory TOTs in Wernicke's aphasics (about 120 ms) than in matched controls [9,10,12]. The close association between language and timing disorders was supported by neuroanatomical overlapping of structures involved in RAP and receptive language functions [9,10,12,18,19]. These structures comprised left
78
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
hemispheric gyrus temporalis superior, gyrus temporalis medius and surrounding white matter. Although many authors noted the overlapping between timing and language deficits [20–23], others doubt the existence of a cause–effect between these two functions [24–27]. In light of this evidence we should mention two important publications by Merzenich et al. [28] and Tallal et al. [29]. They reported amelioration of language functions following TIP training in languagelearning-impaired children. These authors developed several training tasks in which three variables were adaptively adjusted to each subject's performance i.e., duration, frequency and inter-stimulus-interval in the presented stimuli. Four weeks of such intensive training resulted in significant improvement of RAP (i.e., TOTs lowered from 400 ms pretraining to 100 ms post-training) which was accompanied by improvements in phonological processing and auditory comprehension. These language functions remained untrained during the applied intervention [28,29]. Based on these reports, a training programme Fast ForWord (FFW) was developed by Scientific Learning (http://www.scilearn.com) and broadly implemented into clinical practice [30]. FFW was also applied in children and adults with dyslexia, bringing positive therapeutic effects. Nevertheless, some authors did not confirm any positive effects of FFW on writing, reading or spelling [30, for a recent review]. Despite existing evidence on clinical benefits of temporal training in language-disordered children, data indicating positive effects of temporal training in aphasic sample are very limited. The only early report by v. Steinbüchel et al. [31] demonstrated slight improvement in both auditory perception of TO and phoneme discrimination in aphasic patients following training in temporal discrimination. Their training procedure comprised 4 or 8 h of training in TO perception and a verbal feedback was provided on the correctness achieved. However, detailed information on the number of tested aphasics, description of the patient sample, temporal and control training parameters were missing from this preliminary report. To the best of our knowledge, there are no more reports on the application of temporal training in aphasic patients. It should be mentioned that aphasia constitutes a major medical and social problem in modern society. As estimated by the demographic data, up to 270,000 patients with stroke-related aphasia are diagnosed worldwide each year [32]. Only a minority of them recover completely from their language deficits. Examinations of typical recovery curves revealed that only 25% of patients have a chance for the full restoration of disturbed language. It indicates the importance of implementation of new techniques of rehabilitation to change the burden of languagedisordered individuals with the adoption of new intervention based on neuronal mechanisms underlying language. Considering the importance of RAP in language, in the present study we tested whether timing deficits in aphasics could be reduced by the specific acoustic training in RAP. Furthermore, we investigated whether a transfer of improvement could be observed from the trained time domain to the language domain which was untrained at that time. Taking into account the number of aphasic patients, such a new therapeutic approach could be applied in neurorehabilitation programmes targeting aphasic patients. This study offers some important scientific indications for new components of therapy programmes. 2. Methods 2.1. Participants Eighteen patients (9 male and 9 female) suffering from aphasia after first-ever stroke (haemorrhage or infarction, lesion age x ± SD = 17 ± 11 weeks) participated in the study. They varied in age from 34 to 76 years (x ± SD = 55.4 ± 11.2 years), were right-handed [33], Polish native speakers and had normal hearing level verified by screening audiometry (audiometer AS 208), using frequencies of 250, 500, 750, 1000, 1500, 2000 and 3000 Hz [34]. Apart from stroke they had neither
neurological nor psychiatric disorders and reported no history of head injuries. Description of the patient sample is given in Table 1. All patients displayed auditory comprehension deficits evidenced by the Token Test [35]; Phoneme Discrimination Test (PDT) [36] and Voice-Onset-Time (VOT) Test [37]. Their language deficits were accompanied by disordered RAP (for detailed description see below). We adopted the following exclusion criteria: recurrent stroke, global aphasia with poor verbal contact, poor general health, participation in other rehabilitation programmes during our study, and older age to minimize undiagnosed cognitive deficits which might have negative effects on the training results. These inclusion/exclusion criteria allowed to diminish the risk uncontrolled side-effects which could influence the experimental verification of therapy effectiveness. But such therapy is addressed in future broader patients population suffering from also other language disorders. It was a blinded randomized controlled study. The patients were randomly assigned into two groups according to age, gender, lesion age, level of comprehension deficits and RAP deficits (Tables 1–3). The first group was assigned to temporal training (TT, n = 11) and the second one to nontemporal control training (NT, n = 7). For detailed description of the training protocols see below. Evidence has suggested that the above mentioned subject-related factors generate vulnerability with respect to restoration of language and aphasia rehabilitation. Synopsis of pre- (white column) and post-training (grey column) performance profile in each individual subject is given in Tables 2 and 3. Using U Mann–Whitney test, all pre-training between-group differences for all these variables were nonsignificant, i.e., for age: z = −1,09, p b .28; lesion age: z = 0.18, p b .86; RAP assessed by TOT: z = 0.45, p b .66, comprehension level assessed by the Token test: z = − 0.58, p b .57; PDT z = 0.73, p b .45 (Table 2), VOT z = −0.14, p b .90. It indicated a balance between TT and NT groups (Tables 2 and 3), allowing the comparison of the effectiveness of two different training procedures in groups with relatively similar subject-related characteristics, as well as pre-training performance, thus, similar prognosis for language recovery. The place of the lesion was evidenced by CT or MRI. Fig. 1 presents the summarized damaged regions in TT and NT groups. Neuroanatomical analyses confirmed that in both groups lesions were localized mainly in the left hemispheric temporal lobe and covered the classical areas engaged in both auditory comprehension and time perception [10,12,38,39]. In particular, in the TT group lesioned areas comprised superior temporal gyrus, Heschl's gyrus, Rolandic operculum and insula. Almost exactly the same areas were damaged in the NT group: superior and middle temporal gyrus, Heschl's gyrus, Rolandic operculum, insula, putamen and frontal inferior orbital lobe. Furthermore, the synopsis of brain damaged regions in individual patients assigned to TT and NT is documented in Fig. 2. Considering individual patients' data evidenced in Tables 1–3 and Fig. 2 it may be assumed that the TT and NT groups were as matched as possible. 2.2. Ethical approval The research was approved by the Bioethics Commission at the Institute of Psychiatry and Neurology in Warsaw (permission no 5/2005 from February 2nd 2005) as well as by the Bioethics Commission at the Warsaw Medical University (permission no 5/2010 from January 26th 2010). The study was conducted according to the principles expressed in the Helsinki Declaration; the written informed consent from each participant was obtained prior to testing. 2.3. Procedures The study comprised both assessment and training procedures. The assessment procedures included TOT measurement, language and attention tests which were performed before (pre-training assessment) and after the training (post-training assessment). Moreover, the follow-up
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
79
Table 1 Description of the patient sample included in temporal training and non-temporal control training (abbreviations: M = male, F = female; I = infarction, H = haemorrhage stroke). No
Gender
Patient age (years)
Type of stroke
Lesion age (weeks)
Training Group
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
M M M M M M F F F F F M M M F F F F
70.4 61.2 57.9 51.2 43.4 34.7 70.4 58.5 52.5 45.1 37.2 63.6 61.4 48.6 76 57.4 54.8 52.5
I I I I I I I I H I H I I I I I H H
13 10 10 9 36 5 8 14 14 40 26 5 7 36 17 8 15 24
Temporal training (TT)
assessment was repeated about 6 months after the training completion in TT group. Only 6 patients from the TT group took part in follow-up assessment, as the remaining 5 patients during the relatively long period of 6 months between post-training and follow-up assessment suffered from recurrent stroke or other systemic diseases. As these acquired additional disorders may affect long-term language recovery, these 5 patients could be not considered in the follow-up assessment (compare our patient inclusion criteria). The training procedures comprised either TT or NT. 2.3.1. Assessment procedures 2.3.1.1. TOT measurement. Stimuli were pairs of 1 ms rectangular clicks presented in rapid succession with varied inter-stimulus-intervals (ISIs). The paired-clicks were presented monaurally, i.e., one click was presented to the one ear, followed by a second click to the other ear. The stimuli were generated by a 16-bit Sound Blaster Extigy Sound Card and delivered at a comfortable listening level through the headphones SONY MDR-CD 480. The task was to report the TO of clicks within each pair by pointing to one of two response cards. Each card displayed the possible order of presented clicks, i.e., right–left or left–
Nontemporal control training (NT)
right. Each stimulus pair was preceded by a warning signal of 400 ms, delivered 1500 ms before the first click. The ISIs varied from 1 to 600 ms according to the adaptive maximum-likelihood-based algorithm [40]. In each trial ISI was calculated on the basis of correctness of previous responses. This tracking procedure estimated individual TOTs as the minimum ISI between two clicks at which a subject reported their order at 75% correctness. TOT values were assessed using ‘Yet Another Adaptive Procedure’ [41] on the basis of the maximum likelihood parameter estimation. The measurement was continued until the TOT value was located with a probability of 95% inside a + 5-ms interval around the currently estimated threshold [42]. The proper data collection followed after an introductory session in which the patient reported the order of clicks presented with relatively long ISI of 600 ms. If the pre-defined criterion of 11 correct responses in last 12 presentations was reached, the proper measurement started. The task was completed in a soundproof room and repeated twice during two sessions separated by a few days. For each patient the mean value of TOT from the two sessions was calculated (for individual data see Table 2) and compared to TOT of healthy volunteers matched for age. At these comparisons we based on our normative data were collected
Table 2 Synopsis of pre- and post-training individual performance profiles on TOT, Token Test and PDT in particular patients assigned to TT and NT groups. Numbers of individual patients correspond with those in Table 1 and Fig. 2. Asterisks indicate situations where patients were not able to perform the task. No
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Training group
Temporal training (TT)
Nontemporal control training (NT)
Token test (percentage of errors)
PDT (percentage of errors)
Pre
TOT (mean; ms) Post
Pre/post difference
Pre
Post
Pre/post difference
Pre
Post
Pre/post difference
386 294 141 289 168 102 377 180 522 460 395 112 265 229 444 250 391 169
159 158 63 96 67 48 277 99 497 186 99 86 178 114 510 182 285 206
227 136 78 193 101 54 100 81 25 274 296 26 87 115 −66 68 106 −37
58 * 44 58 * 18 64 30 * * 48 22 70 64 58 34 58 46
56 * 12 48 * 8 36 4 * * 38 12 78 64 58 18 52 34
2 * 32 10 * 10 28 26 * * 10 10 −8 0 0 16 6 12
6.25 34.38 15.63 34.38 31.25 0 43.75 0 46.88 15.63 3.13 9.38 31.25 6.25 25 0 15.63 9.38
12.5 21.88 0 12.5 21.88 0 15.63 0 50 125 3.13 0 37.5 6.25 9.38 0 12.5 9.38
−6.25 12.5 15.63 21.88 9.37 0 28.12 0 −3.12 3.13 0 9.38 −6.25 0 15.62 0 3.13 0
80
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
Table 3 Synopsis of pre- and post-training individual performance profiles on VOT in particular patients assigned to TT and NT groups. The typical voicing zone comprises VOT values of −100, −90, −80, −70, −60 and −50 ms whereas, the typical unvoicing zone, the VOT values are +5, +10, +20, +30, +40, +50, +60, +70, +80 and +90 ms. For each of these two zones the mean percentage of correct responses is provided. The typical transition zone comprising VOT values of −40, −30, −20, −10 and 0 ms was dropped, as words characterized by these VOT values are usually reported randomly either as /DOMEK/ or as /TOMEK/ (compare Fig. 5). Numbers of individual patients correspond with those in Tables 1, 2 and Fig. 2. No
Training group
The mean percentage of reported words Voiced words
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Temporal training (TT)
Nontemporal control training (NT)
Unvoiced words
Pre
Post
Post/pre difference
Pre
Post
Post/pre difference
86 47 100 56 100 100 39 100 75 39 89 100 58 92 86 94 70 100
100 75 100 78 97 100 92 100 70 56 89 100 78 95 94 100 72 97
14 28 0 22 −3 0 53 0 −5 17 0 0 20 3 8 6 2 −3
98 50 100 78 95 97 75 100 62 65 93 97 68 50 72 95 62 98
100 78 100 70 98 100 92 98 72 80 85 100 85 83 67 100 55 93
2 28 0 −8 3 3 17 −2 10 15 −8 3 17 33 −5 5 −7 −5
from 86 volunteers [4] or 168 volunteers [3] varied in age from 20 to 69 years. Accordingly, in the present study the TOT cut-off value for patient inclusion was 1.5 SD above the mean value evidenced in healthy volunteers at a given age. Such inclusion criterion met about of 80% pretested patients. Patients included to the study were, next, assigned to our training if their temporal ordering deficits were accompanied by comprehension deficits, evidenced at least in the Token Test. 2.3.1.2. Language tests. Auditory comprehension was tested using 3 tests: (1) Token Test; (2) Phoneme Discrimination Test and (3) VoiceOnset-Time Test. Token Test constitutes a part of the Aachener Aphasie Test [35] and it is remarkably sensitive to disrupted linguistic processes that are central to aphasic disorder. Plastic tokens (coloured squares and circles of two sizes: big and little) were presented and a patient carried out oral commands, e.g., “Touch the little red circle and the big yellow square”. The test consists of 50 commands given in 5 sections of increasing complexity. Percentage of total number of committed errors was analysed. Phoneme Discrimination Test (PDT) comprises 64 paired-words presented in 8 series of 8 pairs each (75% were different and 25% the same). The words within each pair differed in consonants, contrasted for place of articulation, plosive, fricative, voicing and nasality. The task was to judge whether paired words were the same or different. Responses were given by pointing to one of two response cards. Alternative versions of 4 series of this test were used in consecutive assessments (pre-, post-training and follow-up). The outcome measure was the percentage of committed errors. Voice-Onset-Time (VOT) Test comprises differentiation between voiced and unvoiced contrast in two Polish words /TOMEK/ and / DOMEK/ (in English: Tom/house). On the basis of two words naturally spoken by the Polish native speaker, 21 pseudo-word stimuli varying in the VOT of the initial stop consonant were synthesized. They were created by a manipulation in the duration of an interval between the burst and the onset of laryngeal pulsing in the initial consonant in naturally uttered word /TOMEK/ (unvoiced). According to the Polish phonology, confirmed by our normative data collected from 67 healthy volunteers [37], pseudo-words of the VOT from − 100 to − 50 ms (voiced zone) are typically recognized as voiced /DOMEK/, whereas, from + 5 to + 90 ms (voiceless zone) as unvoiced /TOMEK/. There is
also a transition zone for VOT from −40 to +5 ms, where the discrimination is at the chance level (50% voiced and 50% voiceless consonants). The test comprised 6 series; each of them contained 21 randomly ordered pseudo-words of VOT from −100 to +90 ms in 10 ms steps. The patient's task was to decide whether the presented pseudo-word was / TOMEK/ or /DOMEK/. Responses were given by pointing to one of two response cards which displayed pictures corresponding to these two words. The percentage of reported /TOMEK/ or /DOMEK/ at particular VOTs was analysed. 2.3.1.3. Attention tests. Two aspects of attention, i.e. alertness and vigilance were assessed in order to control attentional deficits which may influence patient performance [43]. Accordingly, alertness is defined as the ability to raise and maintain a high level of attention in anticipation of a test stimulus, whereas, vigilance is the ability to maintain attention over a longer period of time. Alertness was assessed by measuring a simple reaction time in response to a visual stimulus presented as a cross in the monitor centre and cued by an auditory warning signal. In the vigilance task, a low and a high tone were presented sequentially. The task was to press the button as soon as possible when two identical tones were presented in a row. Simple reaction time was measured in both attention tests. 2.3.2. Training procedures Two computer auditory training procedures were applied: training in temporal processing (TT) and nontemporal control training (NT). 2.3.2.1. Training in temporal information processing. TT procedure was similar to that of TOT assessment, i.e., two clicks were presented via headphones with various ISIs and the task was to reproduce the TO of two clicks presented in consecutive pairs. The task difficulty was adaptively adjusted for each patient on the basis of correctness achieved (see above for the description of adaptive procedure described in TIP assessment). The paired clicks were presented in 10-trial blocks. For each patient the starting ISI was calculated individually on the basis of the mean TOT value from two sessions obtained in pre-training assessment. This mean TOT was increased by a constant 20 ms interval, to create a relatively comfortable difficulty level. Within each block paired clicks were presented at constant ISI, which varied between blocks according
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
81
Fig. 1. Schematic display of the common lesioned areas for TT and NT subjects. Localization of lesions comprised only left hemispheric regions.
to an algorithm related to the patient's actual TOT. When at least 8 correct responses were given within one block (classified as ‘correct block’), ISI in the next block decreased, making the task more difficult. Thus, the adaptive algorithm used in TT depended on the subject's performance. If the actual TOT was longer than 100 ms, ISI decreased by 5 ms; in case of TOT between 50 and 100 ms by 2 ms; whereas for TOT below 50 ms by 1 ms. If seven or fewer correct responses were obtained within
one block (‘incorrect block’), in the next block the ISI increased by 1 ms, regardless of patient's actual TOT. 2.3.2.2. Nontemporal control training. In case of NT, paired 1-second tones with constant, relatively long ISI of 3 s were presented via headphones at different volumes. Within each pair one tone was louder than the other. The task was to report which tone was louder: the first or the second.
82
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
The following four frequencies were used: 400, 600, 800, and 1000 Hz. Only one frequency was presented within each pair or block. According to our pretesting data, the difficulty level of TT and NT was comparable as much as possible and demanded comparable mental loud. Furthermore, both the training duration, protocol and motivation system were identical. Although the NT has to involve, as does every cognitive task, some temporal cues related to the stimulus duration or ISI, it concerned multisecond timing, and thus a totally different time range compared to the topic of the TT which was focussed on millisecond timing. During the TT and NT each pair of sounds was preceded by a visual warning signal. Subjects answered by pointing to the response cards placed in front of them. Next, the experimenter entered their responses into the computer programmes. After each trial, a visual feedback on correctness was provided on the monitor, i.e. a smile or sad face after each correct or incorrect response, respectively. In TT and NT the same motivation system was used. For each correct response patients obtained one point. After completion of each block (10 trials), the total number of points was displayed on the monitor. Finally, after each ‘correct block’ the patient was rewarded with a puzzle which was collected on the monitor during each session. Each training type comprised eight 45-minute sessions, held three times a week. The whole therapy was completed within 3 weeks and comprised about14 individual sessions with each patient performed in the Laboratory of Neuropsychology, Nencki Institute. During each session rest-pauses were provided on patient's request. The duration of these pauses was excluded from the effective training time.
2.4. Statistical analyses To verify the effect of TT or NT on temporal processing (TOT), language and attention functions (pre- vs. post-training), as well as the stability of these effects (follow-up vs. post-training and follow-up vs. pretraining) Wilcoxon Signed-Rank Test for two dependent samples was performed. 3. Results 3.1. The effect of TT and NT on timing and language functions: within-group comparisons for pre-training versus post-training assessment 3.1.1. TOT measurement In TT group, TOT values were significantly lower (improved performance) post- in comparison to pre-training (n = 11; z = 2.93; p b .003). The corresponding value after NT was nonsignificant (n = 7; z = 1.52; p b .13, Fig. 3). These results are illustrated by synopsis of pre- and post-training individual data (Table 2). Despite relatively similar TOT values pre-training, the differences between pre- and posttraining performance in NT seem smaller than these in TT group. 3.1.2. Language processing 3.1.2.1. Token test. After TT the percentage of errors decreased significantly in comparison to pre-training (n = 7; z = 2.37; p b .018). After NT the performance remained at a stable level, i.e., the difference between pre- and post-training was nonsignificant (n = 7; z = 1.48; p b .14; Fig. 4A). These relationships are reflected in individual data, indicating post-training the reduced number of errors in patients following TT, in comparison to NT (Table 2).
Fig. 2. The brain damaged regions in individual patients assigned to temporal or nontemporal control training. Numbers of individual patients correspond with those in Table 1. For patient nos. 3, 11, 12, 15, and 16, the neuroanatomical data were not available. Abbreviations: LH—left hemisphere, RH—right hemisphere.
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
83
Fig. 3. The mean Temporal-Order-Thresholds in ms (with SEM) for pre- and post-training assessment in TT and NT groups. The significant difference is indicated by an asterisk (p b .003). The number of subjects is given in brackets.
3.1.2.2. PDT. After TT the mean percentage of errors tended to be lower than in pre-training assessment (n = 11; z = 1.89; p b .059). In the NT group, the difference between pre- and post-training assessment was nonsignificant (n = 7; z = 1.10; p b .28, Fig. 4B). Looking at the performance profiles in individual patients (Table 2), despite a huge improvement after TT in some patients (e.g., patients no 2, 3, 4, 7), in the other ones the progress was less clear, or even missing (e.g., patients no 1, 9). Similarly, unclear relations on PDT were evidenced after NT. 3.1.2.3. Voice-Onset-Time test. After TT improved performance was found in both the voiced and unvoiced zone. In the voiced zone (at VOT = − 90 ms and VOT = − 60 ms) the increased percentage of voiced words /DOMEK/ was reported (z = 1.89, p b .06; z = 2.20, p b .028, respectively). It was accompanied by the improved performance in the unvoiced zone (at VOT = 40 ms), where significantly more unvoiced words /TOMEK/ were reported (z = 2.11, p b .035). Additionally, the mean percentage of reported unvoiced words /TOMEK/ increased significantly in the transition zone (at VOT = 0; z = 2.02, p b .044; Fig. 5). In contrast, after NT the differences between the numbers of reported voiced words /DOMEK/ and unvoiced /TOMEK/ were nonsignificant for any VOT value, i.e. in the voiced, unvoiced or transition zone. The individual performance profiles in particular patients are provided on Table 3.
Individual performance profiles in particular patients (Table 3) indicated after TT for the typical voicing zone higher improvements in the mean of percent of reported voicing words /DOMEK/ than after the NT. On the other hand, these differences for the typical unvoicing zone seem more similar. 3.1.3. Attentional processing In both TT and NT groups, attentional performance remained at a similar level in post-training assessment. The difference between postand pre-training in the mean reaction times was nonsignificant for both alertness (TT group: z = 0.09; p b .93; NT group: z = 0.94; p b .35) and vigilance (TT group: z = 1.15, p b .25; NT group: z = 0.54, p b .60). 3.2. The stability of improvements following TT As explained in the Participants Section, the follow-up assessment was completed only for participants undergoing TT. Due to reduced patient sample (objective reasons, see more explanations in the Procedure Section), statistical analyses were performed only on 6 patients. In Token Test only 4 patients were tested, therefore, no statistical analyses were performed. Table 4 presents results of comparisons between particular assessments in the reduced patient sample.
Fig. 4. The mean percentage of errors (with SEM) in TT and NT groups committed in: (A) Token Test and (B) Phoneme Discrimination Test in pre- and post-training assessment. The number of subjects is given in brackets; some subjects were not able to perform the task (compare Table 2). (*)—significant difference (p b .018); (#)—tendency towards significance (p b .059).
84
E. Szelag et al. / Journal of the Neurological Sciences 338 (2014) 77–86
Fig. 5. The mean percentage of reported voiced and unvoiced words at particular VOT values in pre- and post-training assessment after TT and NT. (*)—significant differences (see text for detailed level); (#)—tendency towards significance (p b .06).
To summarize, in the follow-up assessment a relatively stable level of performance was observed compared to post-training. It is important to note that for all tasks in follow-up assessment the performance was not lower than in pre-training for all analysed outcome measures, i.e., milliseconds (TOT measurement), percentage of committed errors (Phoneme Discrimination and Token Test), percentage of reported voiced/ unvoiced words in the voiced and unvoiced zones (VOT test). 4. Discussion 4.1. Summary of results The TT in aphasic patients significantly improved both temporal ordering and language competency. In contrast, the control NT did not Table 4 Stability of improvements for particular tasks in consecutive assessments following TT. The statistical significance of differences between particular assessments in reduced patient sample was indicated. The other difference between consecutive assessments were nonsignificant. No statistical comparisons were performed for the Token Test (n = 4). Results obtained in particular assessments Task
N pre-training 296
p
