Perception of grammatical tone in Akan patients with left and right hemisphere brain damage

ABSTRACT

It remains a matter of debate what roles the left and right hemispheres play in processing speech prosody. Brain lesion studies have demonstrated that lexical tone perception among native speakers of tonal languages is more disrupted in left hemisphere damaged (LHD) individuals than right hemisphere damaged (RHD) individuals. This has been taken to suggest that linguistically-relevant prosodic cues are predominantly left-lateralised, whereas non-linguistic stimuli are predominantly right-lateralised. However, this phenomenon has only been examined in lexical tone, leaving grammatical tone perception unexplored. The aim of this study was twofold: Firstly, to examine how individuals with LHD and RHD perceive grammatical tone, and secondly to compare grammatical tone to non-linguistic tone perception. Therefore, native Akan speakers with LHD, RHD and no-brain damage (NBD) controls were tested in two discrimination tasks that examined linguistic and non-linguistic tone perception. The results showed that while both the individuals with LHD and RHD show impairment in grammatical tone perception, there was a trend of a better performance for the RHD group. Nonetheless, for non-linguistic tone perception, individuals with LHD outperformed the RHD individuals, although both had reduced performance compared to the NBD individuals. A further analysis revealed that the reduced perceptual abilities of both the LHD and RHD groups in grammatical tone perception can be attributed to grammatical problems rather than tone per se. We conclude that there is potentially a bilateral involvement of the two hemispheres in grammatical tone processing, with the left being the dominant hemisphere.

KEYWORDS:

• Grammatical tone
• non-linguistic tone
• brain damage
• hemispheric lateralisation
• Akan

Introduction

Research on the neuronal underpinnings of speech prosody processing dates back to Monrad-Krohn’s (1947) study, that investigated a woman who had difficulties in producing phonemic tone contrasts in her native Norwegian dialect after left hemisphere brain damage, even though her ability to sing was unaffected. Over the last several decades, the issue of the distinct roles that the left and the right hemispheres play in speech perception has received considerable attention in the literature. Most studies have shown that the left hemisphere (LH) is more involved in the processing of phonemic units such as phonemes, syllables, and words, while the right hemisphere (RH) is specialised for melodic, prosodic and affective (e.g. music, pitch contours) processing (Kimura, 1961, 1964; Van Lanker, 1980; Zatorre & Belin, 2001). The use of lexical tone (i.e. when tone is used to distinguish lexical meanings) in tonal languages has provided significant insights into understanding the interplay between the functions of the two hemispheres in speech perception (Gandour, 2006). This is because pitch variations in tonal languages are relevant to signal both linguistic (differences in lexical meaning) and acoustic (e.g. pitch contour) information (Gandour, 2006). However, while a large number of studies have examined the hemispheric roles in lexical tone perception (Gandour et al., 1992, 2002; Gu et al., 2013), the neural underpinnings of grammatical tone perception have not been uncovered to date. This study aims to address this gap in the literature by examining tone perception in left and right brain damaged speakers of Akan, a tonal language that uses tone for grammatical functions.

Hemispheric lateralisation in lexical tone processing

The precise mechanisms underlying the hemispheric asymmetry for speech prosody remains controversial. The two dominant accounts proposed for the hemispheric asymmetry of human pitch perception are the function-dependent brain asymmetry and the acoustic-dependent brain asymmetry models. The function-dependent brain asymmetry model, on the one hand, predicts LH lateralisation when pitch information such as tone or intonation are used for linguistic purposes (Gandour et al., 2000; Van Lanker, 1980; for a review, see Zatorre & Gandour, 2008). For instance, processing of lexical tone will be left lateralised since in this case tones are used to provide semantic information at the word level (Gu et al., 2013; Xi et al., 2010). The acoustic-dependent brain asymmetry model, however, predicts RH lateralisation of lexical tone because pitch information is processed on the basis of the acoustic signals it provides, irrespective of its functions (Ge et al., 2015; Poeppel, 2003; Ren et al., 2009; Sidtis & Van Lancker-Sidtis, 2003; Van Lancker & Sidtis, 1992; Zatorre & Belin, 2001).

This long-standing debate regarding hemispheric lateralisation of lexical tones has resulted in a large body of studies using a wide range of methods such as dichotic listening, neuroimaging, and behavioural testing in brain lesioned individuals, to observe the functions of the two hemispheres in tonal processing. In dichotic listening studies, in which two verbal stimuli are presented to both the left and the right ears simultaneously, presentation of linguistic (tonal) stimuli often yields a right ear advantage among speakers of tonal languages, thus indicating processing in the language-dominant left hemisphere (Gandour, 2006; Kimura, 1967; Wang et al., 2004). However, cross-linguistics studies have shown that there is no right ear advantage for native speakers of non-tonal languages listening to lexical tone stimuli (Van Lancker & Fromkin, 1973; Wang et al., 2001, 2004). That is, the left hemisphere is dominant for the processing of pitch in lexical tone among native speakers of tonal languages, but not for speakers of non-tonal languages. This modulatory effect of language experience in functional brain asymmetry has been replicated in studies (Gandour et al., 2002; Wong et al., 2004) that have used functional Magnetic Resonance Imaging (fMRI), positron emission tomography (PET) and Event-Related Potentials (ERPs).

In an fMRI study, Gandour et al. (2002) presented Thai speech (varying pitch and vowel duration) and non-speech stimuli (hums with varying pitch contour and duration) to both Thai and Chinese speakers and found activation in the left inferior prefrontal cortex in the speech condition for the Thai group only. This suggests that the linguistics aspects of pitch and duration were unique to Thai listeners because they perceived them as having linguistic properties. However, regardless of language experience, both Thai and Chinese groups exhibited similar right fronto-parietal activation patterns when judging nonspeech hums for both spectral and temporal cues, indicating that language-specific effects disappear when stimuli are purely acoustic in nature.

Consistent with the neuroimaging data, electrophysiological studies have also demonstrated a right-lateralised pre-attentive processing of lexical tone (Chandrasekaran et al., 2007; Gu et al., 2012; Luo et al., 2006; Ren et al., 2009), suggesting that the auditory processing of lexical tones is shaped mainly by acoustic properties at a pre-attentive processing stage (Chandrasekaran et al., 2009). Consequently, Luo et al. (2006) proposed a two-stage model for processing linguistic pitch contours: a pre-attentive stage of processing involved in a lower-level pitch analysis in the right hemisphere, and an attentive stage of processing driven by higher-level linguistic representations with activation of neural circuits lateralised to the left hemisphere. However, recent neurophysiological data on lexical tone perception in native speakers of Mandarin Chinese have shown parallel processing of acoustic and phonological information (Xi et al., 2010; Yu et al., 2014, 2017), contrary to the serial processing mechanism proposed by Luo et al. (2006). Further, a recent meta-analysis found support for convergent activation in bilateral inferior prefrontal and superior temporal regions as well as the right caudate during lexical tone perception in tonal languages (Kwok et al., 2017). In sum, the literature shows mixed results regarding hemispheric lateralisation of lexical tone processing using brain imaging techniques.

Turning now to brain lesion studies, it has been cross-linguistically attested that both individuals with brain damage to the left hemisphere (LHD) and right hemisphere (RHD) show impairments in lexical tone processing, with the former performing worse than the latter (Baum & Pell, 1997; Gandour & Dardarananda, 1983; Gandour et al., 1992; Kadyamusuma et al., 2011a, 2011b; Packard, 1986). Gandour and Dardarananda (1983) observed that Thai speaking individuals with LHD demonstrated more difficulties in distinguishing Thai words that differed only in tone than individuals with RHD, who were in turn less accurate than the non-brain damaged speakers. Similarly, Moen and Sundet (1996) showed that Norwegian speakers with LHD were impaired in their identification of two Norwegian tones (pitch accents), whereas such impairment was not apparent in the RHD group. Yiu and Fok (1995) also demonstrated that Cantonese tone processing was disrupted in LHD individuals with aphasias, while it was intact in speakers with dysarthria. Thus, tone processing deficits are associated with lesion sites (especially in the left hemisphere) that result in aphasia. However, Yui and Fok did not include individuals with RHD. Note that much of the data on tone processing in clinical population (as discussed above) come from tonal languages spoken in Asia. There are only two studies that have examined lexical tone processing in individuals with unilateral brain damage in an African tonal language, Shona, a Bantu language spoken in Zimbabwe (Kadyamusuma et al., 2011a, 2011b). Kadyamusuma et al. (2011a) showed that for lexical tone perception, Shona speaking individuals with left hemisphere damage were poorer than individuals with right hemisphere damage, who were in turn lower than the non-brain-damaged individuals who performed at ceiling. In a follow-up experiment, Kadyamusuma et al. (2011b) examined the perceptual abilities of Shona speaking LHDs and RHDs to discriminate pitch in Shona words (with lexical tone) and in their low-pass filtered analogues. Their results showed that in both tasks, the LHDs showed more impairment than the RHDs. This suggests that for the low-pass filtered speech task, although there was no segmental information present in the Shona words, the low-pass filtered stimuli still remained linguistic in nature, and hence, were processed like Shona words (Kadyamusuma et al., 2011b). Taken together, there is evidence for a bilateral processing of lexical tone in tonal languages, with the left hemisphere being dominant (Kadyamusuma et al., 2011b).

So far, all the studies on tonal processing in individuals with damaged and healthy brains have focused on lexical tone, leaving grammatical tone an untapped area of research. Thus, it is not known yet how grammatical tone is processed. Typologically, tonal languages in Africa are different from those spoken in Asia, in terms of tone inventories and rule systems (Gandour, 2006; J. T. Gandour, 1983). This has made it insufficient to formulate theories which could be generalised to African tonal languages. For example, Akan has a grammatical function of tone which is non-existent in Asian languages (such as Chinese and Thai). Additionally, while lexical tone may require only a word level processing (i.e. differences in lexical meaning), grammatical tone may require processing beyond the word level (e.g. a temporal relation between an adverb and a verb, expressed through grammatical tone in Akan). Therefore, examining grammatical tone processing in individuals with brain damage will give more insights into the already existing theories about tone processing in tonal languages. Therefore, in this study we investigate this issue using Akan, a Kwa language spoken in Ghana, which has grammatical tone.

Features of Akan

Akan is a tonal language with two basic tones: High tone and Low tone, which are pronounced on a relatively level pitch (Abakah, 2000, 2005; Dolphyne, 1988). The meaning of a sentence or a word in Akan is not only dependent on the vowels and the consonants that make up the words, but also on the pitch with which each syllable is produced (Dolphyne, 1988; Osam, 2003, 2008). Pitch in Akan can be used for both lexical (eg. pàpá – ‘father’, pápá – ‘good’, pàpà – ‘fan’) and grammatical functions (see Examples 1a-b). That is, unlike in other tonal languages (such as Chinese) where pitch is used only for contrasts of lexical meaning, in Akan, certain grammatical categories such as verb inflection (tense/aspect) can be distinguished by tone (Dolphyne, 1988). Below are examples of grammatical tone in Akan.

Table

(1a)	Papa	no	twèrέ	lέtὲ
	Man	the	write-HAB	letter.
	‘The man writes letter(s)’

Table

(1b)	Papa	no	twèrὲὲ	lέtὲ
	Man	the	write:PAST	letter
	‘The man wrote letter(s)’

In Example 1a – b, the same verb ‘-twerε-’ (meaning ‘to write’) is perceived as either present habitual as in (1a) or past tense as in (1b), depending on the tonal marking of the syllabic units of the verb. Whereas the tonal pattern on the disyllabic verb for the habitual aspect in (1a) is Low – High, the past has a Low-Low tonal pattern, with a prolonged tone (written as two vowels) on the last syllable (1b). For Akan mono-syllabic verbs, the tone of the present habitual and the past are always High and Low, respectively. Thus, the difference between the habitual and the past is indicated solely by tone and duration. In the section below, we discuss the morphological implications of the past and the present habitual verbs.

Morphological processing in right and left hemisphere

In the extant literature, nonfluent aphasia, which is often characterised by agrammatism, is typically associated with damage in the left hemisphere. Agrammatism is often manifested by the omission or substitution of grammatical morphemes and function words. In languages with word-based morphology, such as English and Japanese, omissions are limited to free-standing morphemes. In contrast, in morphologically complex languages like Hebrew, Swahili, and Italian, substitution of morphemes is prevalent. Generally, morphologically complex structures are problematic for this population (Menn & Obler, 1990). However, verbal morphology is selectively impaired, with subject-verb agreement typically preserved better than tense. There is cross-linguistic attestation to tense and/or impairment among individuals with this type of brain damage (see Friedmann & Grodzinsky, 1997 for Hebrew and Palestinian Arabic, Martínez-Ferreiro & Bastiaanse, 2013 for Spanish, Miceli et al., 1989 for Italian, Dickey et al., 2008 for English, Stavrakaki & Kouvava, 2003 for Greek, and Wenzlaff & Clahsen, 2004 for German, T. O. Abuom & Bastiaanse, 2013 for Swahili). Interestingly, Tsiwah et al. (2020) showed that native Akan speakers with agrammatic aphasia have difficulties with Akan grammatical tone, particularly verbs referring to the past. This selective impairment was accounted for by the Past Discourse Linking Hypothesis (PADILIH: Bastiaanse, 2013; Bastiaanse et al., 2011). PADILIH suggests that past verbs require discourse linking (for a comprehensive discussion on time reference processing in individuals with LHD, particularly agrammatic aphasia, see Tsiwah et al., 2020), and thus are problematic for individuals with agrammatic aphasia. Non-past verbs such as the present, on the other hand, do not require reference to the discourse since the action is a here-and-now event.

Although affixation problems are mostly reported among LHD individuals, some studies (Marangolo & Piras, 2008; Marangolo et al., 2003) have reported a distinct dissociation between impaired derivational morphology (e.g. fall-ire (to fail) → fall-imento (failure) in Italian) and intact inflectional processing among RHD individuals, despite the participants having no other linguistic impairments. Marangolo and Piras (2008) reported that five out of nine RHD individuals they tested, selectively failed in deriving nouns from verbs (i.e. osservare (to observe) → osservazione (observation) in Italian), by mostly substituting the derived noun with a frequent inflectional suffix of the verb paradigm (i.e. osservato (observed) instead of osservazione (observation)). Note that these studies reported derivational morphological problems and not tense, as in the current study.

The current study

The goal of this study was to address the mechanisms of grammatical tone perception by answering two primary research questions. The first question was: do individuals with LHD and those with RHD differ in their ability to perceive grammatical tone? Our prediction was that Akan speakers with LHD will show more impairment in grammatical tone perception than RHD speakers. This is because in grammatical tone, pitch signals grammatical processes, which in the previous literature have been found to be problematic for individuals with LHD (Tsiwah et al., 2020), whereas individuals with RHD generally are not expected to present with such problems with grammatical processing. Further, we also examine if present habitual and past verbs selectively influence grammatical tone perception.

The second research question was: do individuals with LHD and RHD show different impairment patterns when pitch is used for linguistic functions (such as grammatical contrast) compared to when it is used for non-linguistic purposes (pure tone perception)? For the non-linguistic task, we predicted that the individuals with LHD should have fewer problems than the individuals with RHD. Further, individuals with LHD were predicted to perform better in non-linguistic tone perception than in grammatical tone perception, since the former does not involve linguistic operations. In contrast, the individuals with RHD were expected to show equal performance in grammatical tone and non-linguistic tone perception.

These predictions were based on the findings that the processing of pitch variations is right lateralised when it contains purely acoustic units (non-linguistic), but when tone is used for linguistic functions, processing is left lateralised (Liberman & Whalen, 2000; Whalen & Liberman, 1987). We addressed these questions by examining the perceptual discrimination abilities of native Akan speakers with LHD, RHD and without brain damage, in both a grammatical tone (linguistic) and a pure tone (non-linguistic) task. Here we focus on the Akan present habitual and the past which are solely distinguished by tonal height and duration (see Example 1a-b earlier).

Materials and methods

Participants

The present study included two clinical groups and ten non-brain damaged (NBD) participants. The clinical groups consisted of six individuals with LHD and agrammatic aphasia (see Tsiwah et al., 2020 for the diagnosis criteria) and six individuals with RHD and no aphasia. The individuals with LHD were four males and two females, with a mean age of 52 (range: 19–76, SD = 16.9) years, while the RHD group included 5 males and 1 female with a mean age of 53 (range: 34–67, SD = 13.3) years. The NBDs comprised five males and five females with a mean age of 51 (range: 35–71, SD = 11.5) years. The minimum number of years of formal education among all three groups was 9 years (range: 9–18). All participants spoke Akan as their native language, and had been using Akan as their primary language since birth. The data of five of the six individuals with LDH as well as the ten NBDs have been reported in a previous study by Tsiwah et al. (2020). The selection of the five LDH participants in the present study was predicated on their prior participation in a non-linguistic tonal task, as described in the aforementioned study.

The two clinical groups were recruited from the Speech and Language Therapy Centre, and the Physiotherapy Center at Korle Bu Teaching Hospital, Ghana. CT scans showed that all participants in these groups had suffered from a single stroke either in the left or right hemisphere. All participants were right- handed (with exception of one individual with RHD, who was ambidextrous). The time post-onset of stroke ranged from 3 to 36 months (see Appendix A for full demographic data). None had vision or hearing problems. The study received approval from the boards of the Research Ethical Review Committee (CETO) of the Faculties of Arts, Philosophy, and Theology and Religious Studies, University of Groningen, and the Korle Bu Teaching Hospital-Institutional Review Board (KBTH-IRB), Accra, Ghana. All participants gave their written informed consent.

Materials and procedure

The materials and procedure used in Tsiwah et al. (2020) study were also used for the current study. For a more detailed description, see Tsiwah et al. (2020).

Non-linguistic tone perception task

The non-linguistic task comprised a Tonal Screening Test (TST: Bruder et al., 2011; Kayser, 2011; Stevens et al., 2000; Wexler et al., 1998), in which the members of a pair of non-linguistic tones must be judged to be either the same or different. The presentation of the tonal pairs was done via headphones. The tones were made up of 300-ms sine waves with frequencies between 325 Hz and 1994 Hz. For example, tonal pairs with frequencies of 1328 Hz − 1129 Hz and 680 Hz − 680 Hz made up different and same tones, respectively. There was a total of 70 tonal pairs: 10 practice items, and 60 main trials (30 same, 30 different tones pairs).

Linguistic tone perception task

The comprehension subtest of the African version of the Test for Assessing Reference of Time (TART: T. Abuom & Bastiaanse, 2010) was adapted to Akan for the linguistic tone perception task.

A total of 16 transitive verbs (e.g. to drink; to write) were used in a spoken-sentence-to-picture-matching paradigm. Each of the 16 experimental verbs appeared twice; once in each time frame (past and present), making a total of 32 experimental trials. The order of the 32 trials was randomised. Each action was depicted by two pictures, one placed above the other (see Figure 1).

Figure 1. Example of the Linguistic task, with the target sentence ‘the man ate the orange’. © University of Groningen.

Target pictures for the past time frames were contrasted by a picture with the same action in the present time and vice versa. An example of the items is given below.

Table

(2)	Experimenter	Pàpá	nó	hwànèè	ànkáá	nó
		Man	the	peel:PAST	orange	the
		The man peeled the orange.

A pair of pictures was presented to the participants and the sentence was read aloud by the experimenter. The participants had to point to the picture that matched the spoken sentence. A response was correct when the participants pointed to the picture that matched the time frame encoded in the sentence that was read out loud.

Data analysis

To test for differences between the three groups across tasks, generalised linear modelling (GLM) was performed using the glm function of the lme4 package (Bates et al., 2013) and the glht function of the multcomp package (Hothorn et al., 2013) in R (R Core Team, 2013). The dependent variable (Accuracy) in the model was accuracy (1 = correct, 0 = incorrect). The model included the fixed effects Group (LHD, RHD and NBDs) and Task (LT and Non-LT), with and without the interactions between these variables. The best model, which included an interaction between Group and Task, was chosen based on the AIC and the anova test (using chi-square) of the full model with the effect (interaction) in question against the model without the effect (interaction) in question (with significance at p < 0.001). The purpose of this analysis was to examine the overall perceptual abilities for linguistic versus non-linguistic tone. Since the linguistic tone task used past and present habitual verbs, a follow-up analysis was performed on only the linguistic task to explore potential differences in perceiving these two verb types. The model for this analysis included the fixed effects Group (LHD, RHD and NBDs) and Condition (past vs habitual), with and without the interactions between these variables, with the model with interactions between Condition and Group proving to be the best performing model (the same model selection procedure for the first analysis was used).

Results

In Figure 2 and Table 1, the mean percentage of correct responses on the two tasks are given for the LHD, RHD, and the NBD groups. Since the performance of the NBD group for both tasks were significantly different from both the RHD (LT: β = −2.20, SE = 0.33, z = −6.74, p < 0.001; Non-LT: β = −2.31, SE = 0.28, z = −8.15, p < 0.001) and the LHD (LT: β = −2.20, SE = 0.33, z = −6.74, p < 0.001; Non-LT: β = −1.43, SE = 0.31, z = −4.66, p < 0.001) groups, we excluded the NBD group from further analyses. To test for the effect of time post onset of stroke, we included this variable in the model. However, this was not found to be significant (β = 0.002, SE = 0.04, z = 0.41, p > 0.05). The model with interaction between Group and Task was significantly better than the model without (model with interaction: AIC = 175.48; model without interaction: AIC = 194.92; X² (1) = 21.43, p < 0.001).

Figure 2. Percent mean accuracy scores per task (LT = Linguistic task; NON_LT = Non-linguistic task) per group of participants, including standard error bars.

Table 1. Percentage of accuracy scores on non-/linguistic tone discrimination task.

Groups	Grammatical tone	Pure Tone
LHD1	56*	98
LHD2	71*	90*
LHD3	59*	92*
LHD4	53*	72*
LHD5	65*	90*
LHD6	59*	96
Mean (SD)	61 (6.77)	90 (9.51)
RHD1	78*	75*
RHD2	69*	80*
RHD3	66*	71*
RHD4	69*	53*
RHD5	75*	96
RHD6	78*	93*
Mean (SD)	72 (5.38)	78 (15.77)
NBD Mean (SD)	96 (2.9)	97 (1.96)
(range)	88–100	95–100

LHD = left hemisphere damage; RHD = right hemisphere damage; NBD = non-brain damage; SD = standard deviation; *outside the range of NBD.

The post hoc analysis showed that on the group level, there was only a marginal significant difference between the individuals with RHD and LHD on their performance on the linguistic task (β = 0.520, SE = 0.22, z = 2.37, p = 0.081). This trend indicated that the RHD group was better than the LHD group. However, the non-linguistic task showed a significant difference in performance between the RHD and LHD groups (β = −0.881, SE = 0.22, z = −4.09, p < 0.001), with the former performing worse than the latter. Across task comparison showed that the individuals with LHD performed significantly better on the non-linguistic task than on the linguistic task (β = 1.72, SE = 0.23, z = 7.55, p < 0.001). The individuals with RHD showed no difference in performance on the linguistic and the non-linguistic tasks (β = 0.32, SE = 0.21, z = 1.56, p = 0.401).

For the follow-up analyses, Figure 3 shows the mean percentage accuracy responses for present habitual and past verbs for the linguistic task across the three groups. Similar to the initial analyses, the NBD group’s overall performance significantly differed from both the RHD (β = −2.45, SE = 0.34, z = −7.15, p < 0.001) and LHD (β = −3.14, SE = 0.34, z = −9.15, p < 0.001) groups. However, a pairwise comparison revealed that for the present habitual verbs, both the LHD (β = −0.73, SE = 0.62, z = 1.17, p > 0.05) and the RHD (β = −1.17, SE = 0.57, z = −2.03, p > 0.05) groups’ perceptual abilities were not significantly different from that of the NBD group. For past verbs, the NBD group were significantly better than the LHDs (β = 3.88, SE = 0.43, z = 9.07, p < 0.001) and the RHDs (β = −2.77, SE = 0.42, z = −6.67, p < 0.001). Interestingly, the LHD (β = −3.65, SE = 0.48, z = −7.62, p < 0.001) and the RHD (β = −2.10, SE = 0.41, z = −5.18, p < 0.001) groups were significant worse on past verbs than the habitual verbs, although the latter group was relatively better on past verbs (β = 1.10, SE = 0.31, z = 3.61, p = 0.015) than the former group. There was no significant difference between the two groups on their perception of habitual verbs (β = −0.44, SE = 0.55, z = −0.80, p > 0.05).

Figure 3. Percent mean accuracy scores on present habitual and past verbs per group of participants, including error bars.

Discussion

The first goal of the current study was to investigate whether Akan individuals with LHD and those with RHD show impairments in their abilities to perceive grammatical tone. The second goal was to examine whether linguistic (grammatical) and non-linguistic (pure pitch) tone perception differed for individuals with LHD and RHD. Our findings will be discussed in the light of the two dominant models that have been proposed to account for the hemispheric asymmetry of human pitch perception, namely, the function-dependent brain asymmetry and the acoustic-dependent brain asymmetry models.

Grammatical tone processing

With respect to the first research question, we found that both the LHD and the RHD speakers of Akan had difficulties in perceiving grammatical tone as compared to the non-brain damaged individuals. That is, an impairment in either the left or the right hemisphere resulted in a reduced ability to perceive grammatical tone. Comparatively, this is consistent with brain lesion data for lexical tone perception in other tonal languages: both individuals with left and right hemisphere brain damage show impairments in lexical tone perception (Norwegian: Moen & Sundet, 1996; Thai: Gandour & Dardarananda, 1983; Gandour et al., 1992; Shona: Kadyamusuma et al., 2011a). Surprisingly, the right hemisphere damaged individuals did not perform significantly better (or worse) than the individuals with left hemisphere damage in our study. Thus, problems with grammatical tone perception were, to a large degree, equally pronounced in individuals with left and right hemisphere damage. However, with a small sample size, caution must be applied, as our results might not entirely be representative of how these two groups perceive grammatical tone. It is important to note that the mean accuracy of the individuals with LHD on the linguistic task was considerably lower than that of the RHD group, although the statistical model did not show a clear significant difference. Additionally, the individual scores showed that all individuals with RHD were better than the LHD individuals. Therefore, the absence of statistical significance might be due to the small sample size of our data. Since we did not investigate lexical tone in the current study, it is essential to exercise caution when interpreting the grammatical tone findings by drawing a direct comparison with the results of earlier studies on lexical tone.

On the question of whether the reduced performance by both groups in the linguistic task could be attributed to the differences in the verb morphology (present habitual and past verbs), the findings from our follow-up analyses demonstrated this to be the case. Both the LHD and the RHD groups showed a significantly lower performance for past verbs whereas present habitual verbs remained relatively intact, although the former group performed relatively worse on past verbs. Our data did not show any significant perceptual differences in habitual verbs across the groups. That is, both the LHD, RHD and NBD performed equally good on the habitual verbs.

Given that the LHD group’s performance on a non-linguistic task was close to ceiling, we argue that their reduced performance on the linguistic tone task cannot be solely reduced to perceptual problems, but a result of morphosyntactic problems which are commonly reported for this population (Bastiaanse, 2008). It should be noted that all participants in the LHD group had been diagnosed with language problems. Further, the worse performance of this group on the past verbs can be attributed to past time reference deficit, as captured by the Past Discourse Linking Hypothesis (PADILIH: Bastiaanse, 2013; Bastiaanse et al., 2011). PADILIH suggests that past verbs require discourse linking, which is pragmatic in nature (for a comprehensive discussion on time reference processing in individuals with LHD, particularly agrammatic aphasia, see Tsiwah et al., 2020), and thus, are problematic for individuals with agrammatic aphasia. Interestingly, the RHDs who were diagnosed as not having language problems, also showed a selective impairment on Akan past verb perception. One potential explanation could be the fact that for this linguistics task, the picture depicting past time was always contrasted with a present picture in which an action is taking place. Therefore, it is possible that the patient groups (LHDs and RHDs) showed a preference for present verbs by pointing to pictures depicting present because it contained an action which was ongoing whereas the action in the past pictures had already taken place. Another potential explanation for this past and habitual asymmetry for the RHD group could be attributed to affixation. The affixation for the Akan past verbs are seemingly complex since they require a prolongation of the final tone, which is not the case for the present habitual. Although affixation problems are mostly reported among LHD individuals, some studies (Marangolo & Piras, 2008; Marangolo et al., 2003) have reported a distinct dissociation between impaired derivational morphology (e.g. fall-ire (to fail) → fall-imento (failure) in Italian) and intact inflectional processing among RHD individuals, despite having no other linguistic impairments. Note that these studies reported derivational morphological problems, and not tense, as in the current study.

Taken together, this study has demonstrated that both the left and the right hemispheres are to some extent involved in processing Akan grammatical tone. While it is clear that the LHDs’ problems with Akan grammatical tone emanates from their language deficits and not tone perception per se, the observed pattern in the RHD group seem rather nuanced. However, these results underscore the need to further investigate the contributions of the right hemisphere to language processing.

Hemispheric lateralisation of linguistic and non-linguistic tone processing

In the light of the two prominent models that account for the hemispheric lateralisation of tone, the observed pattern of impairment in grammatical tone perception in the current study lends support to the function-dependent brain asymmetry model (Van Lanker, 1980). According to this model, lateralisation is dependent on the function of pitch. That is, on one hand, when pitch variations are processed primarily as lower-level acoustic (non-linguistic) units, their processing is lateralised to the right. However, when pitch is used to serve a higher-level (linguistic) purposes, the processing is left lateralised (Liberman & Whalen, 2000; Whalen & Liberman, 1987). Function-dependent brain asymmetry predicts that damage to the left hemisphere results in worse performance in lexical/grammatical tone perception than damage to the right hemisphere. This asymmetry did not prove to be significant in the current study. However, the fact that the individuals with RHD in turn performed worse than the individuals with no brain damage is inconsistent with the assumptions of this model. One potential explanation for the reduced performance of the individuals with RHD could be attributed to attention and short-term memory deficit, commonly reported deficit for individuals with RHD (Blake et al., 2002; C. A. Tompkins & Flowers, 1985; C. Tompkins, 2012). Since discrimination tasks require the involvement of such cognitive control abilities (Miller & Cohen, 2001), it is possible that the subdued performance of the RHD individuals in the grammatical tone task is a result of an increased cognitive load. Another possible explanation can be attributed to the inability of the RHD individuals to perceive tone in general, and, thus, leading to the failure to perceive linguistically relevant (grammatical) tone (more on this in the previous subsection).

Turning now to the question of whether Akan individuals with LHD and RHD show different and/or similar impairment patterns when perceiving linguistic and non-linguistic tones, our predictions were threefold. The first prediction was that individuals with RHD would show more impairments in non-linguistic tone perception than the individuals with LHD. Our results indicated that this was the case: non-linguistic tone perception was significantly better in the individuals with LHD, than the individuals with RHD. Overall, this pattern of results observed for the individuals with RHD and LHD is in-line with the assumptions of the function-dependent brain asymmetry model. Thus, the hemispheric lateralisation of pitch processing is dependent on the function pitch serves: non-linguistic and linguistic functions tend to be right and left lateralised, respectively (Liberman & Whalen, 2000; Whalen & Liberman, 1987). Therefore, damage to the right (rather than the left) hemisphere is more likely to result in difficulties in perceiving pure tone variations. Nevertheless, this is contrary to the findings of Kadyamusuma et al. (2011b) who showed that Shona speaking individuals with LHD were still more impaired than the RHD individuals in the lowpass filtered (supposedly non-linguistic) task. An explanation for this discrepancy can be attributed to the differences in the nature of non-linguistic stimuli used in the current study and that of Kadyamusuma et al. (2011b). While the current study used pure tone stimuli, Kadyamusuma et al. (2011b) used low-pass filtered minimal pairs homologous to the bisyllabic Shona words, which may have still remained linguistic in nature, and hence, were processed like the Shona words rather than non-linguistic units. Another possible explanation for the differences between the Shona lowpass filtered stimuli used in the study of Kadyamusuma et al. (2011b) and the pure tone stimuli employed in the present study pertains to the pitch differences between them. The lowpass filtered stimuli in the aforementioned study have similar pitch differences within the same range as the native lexical tone stimuli. In contrast, the pure tones used in the current study have pitch variations that ranged from 325 to 1994 Hz, which lies beyond the tonal range in adult Akan speech. Therefore, it is likely that these differences, coupled with the cognitive load required for discrimination tasks affected the perceptual abilities of the LHD and the RHD groups.

The second prediction was that the LHD group would perform better on pure non-linguistic rather than linguistic tones. As shown by our results, all individuals with LHD showed a better performance in pure (non-linguistic) tone perception than when pitch was used for grammatical (linguistic) purposes. Again, this is consistent with the assumptions of the function-dependent brain asymmetry model. That is, since pure tone variation signals no grammatical processing, individuals with LHD should show fewer problems in perceiving them than the grammatical tones.

Thirdly, regarding the ability of the RHD group in grammatical tone and non-linguistic tone perception, we predicted that for this group, there should be no difference in performance whether pitch is processed as speech or as non-speech. Our results showed that this is the case. The RHD group showed no difference in their ability to perceive the linguistic and the non-linguistic tones. Meaning, for this group of individuals, both linguistic and non-linguistic tone perception is problematic. This pattern is consistent with the acoustic-dependent brain asymmetry model (rather than the function-dependent brain asymmetry model), which proposes that pitch patterns are processed on the basis of their acoustic structures, regardless of their functions, and are, therefore, processed only by the right hemisphere (Sidtis & Van Lancker-Sidtis, 2003; Van Lancker & Sidtis, 1992; Zatorre & Belin, 2001). Hence, damage to the right hemisphere should result in difficulties in both linguistic and non-linguistic tone perception.

Thus far, our results have demonstrated consistent involvement of the left hemisphere in grammatical (linguistic) tone perception, whereas this hemisphere is relatively less involved in non-linguistic tone perception, denoting a predominance of the left hemisphere in processing linguistically-relevant prosodic cues, as predicted by the function-dependent brain asymmetry model. However, the pattern of impairment for the individuals with RHD suggests an involvement of the right hemisphere in both grammatical tone and non-linguistic tone perception, as predicted by the acoustic-dependent brain asymmetry model. What is clear from the current study is that a strict dichotomy in hemispheric lateralisation for either linguistic or non-linguistic tone perception is not well-founded. In fact, both the LHD and the RHD groups show reduced performance in tone perception, as compared to the non-brain damaged group, suggesting that to some extent, both hemispheres are involved in (lexical or grammatical) tone perception. It is clear that claiming one or the other as the sole explanatory model on the basis of the current data is impossible. Consequently, as proposed in a review study by Zatorre and Gandour (2008), these two models should not be treated as mutually exclusive, and thus, aspects of each model would have to be integrated into a more comprehensive model of bilateral hemispheric involvement in tone processing. The nature and the components of such a model is beyond the scope of the current study.

Conclusions and recommendations for future research

To summarise, our study demonstrates that both Akan individuals with left hemisphere damage (LHD) and those with right hemisphere damage (RHD) exhibit difficulties in perceiving grammatical tone, which is consistent with previous findings on lexical tone in other languages. However, it’s important to note that our study did not examine lexical tone perception, and thus, a direct comparison is not warranted. Our results indicate that for the LHD group, grammatical tone perception is impaired while non-linguistic tone perception remains relatively intact. However, this asymmetry was not clearly observed for the RHD group. Our prediction was that this group would be better at perceiving grammatical tone in Akan, whereas non-linguistic tone would be affected due to the presence of a brain damage in the right hemisphere. Surprisingly, the RHD group showed difficulties in past verb perception compared to the present habitual counterpart, which suggests that a simplistic dichotomy based on brain lateralisation is insufficient to explain our findings. Despite some potential confounding factors identified in our study design, we argue that our data calls for a more nuanced understanding of the relationship between brain lateralisation and tonal perception.

An important recommendation for future studies would be to include a tone discrimination task with Akan lexical tone. This will make a good comparative study by comparing data lexical tone in Asian tone languages, Akan lexical and grammatical tone, as well as non-linguistic pitch.

Another possible area of future research would be to investigate the production of grammatical tone in similar clinical groups. Such a study should adopt a well-defined acoustic and phonetic analysis to investigate the extent to which phonetic and acoustic features may differ for individuals with LHD and RHD. Also, since linguistic tone perception varies depending both on the site and severity of the damage (Gandour et al., 1992), future studies should consider examining these effects in relation to both linguistic and non-linguistic tone processing. Although time post onset of stroke did not prove to be significant in the current study, this measure cannot be used as a substitute for severity of impairment.

Acknowledgments

We would like to thank all the participants of this study, as well as the medical staff at the speech therapy and the physiotherapy centers, at the Korle Bu Teaching Hospital, Accra, Ghana, who put us into contact with the individuals with brain damage.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This project was supported by the Center for Language and Cognition Groningen.

Demographic data for the clinical group and the none-brain-damaged speakers

Table

Participants	Age	Gender	Handed-ness	Education (years)	Months POS	Etiology/Lesion site	Hemi- plegia
Agrammatic speakers
LHD1	69	M	Right	13	24	CVA Left temporo-parietal	none
LHD2	37	M	Right	16	7	Left fronto-parietal infarct	right
LHD3	19	F	Right	12	7	iCVA Acute left-fronto-parietal-occipital infarct	right
LHD4	49	M	Right	10	25	CVA Left middle cerebral artery	none
LHD5	76	F	Right	16	12	hCVA left	right
LHD6	67	F	Right	10	7	hCVA left	right
RHD1	53	M	Right	12	3	hCVA right temporoparietal	left
RHD2	34	M	Right	9	3	Acute right temporoparietal	left
RHD3	40	F	Right	10	36	hCVA right	left
RHD4	61	M	Right	12	5	hCVA right	left
RHD5	67	M	Ambidex	12	24	Right left middle cerebral artery	left
RHD6	63	M	Right	18	29	CVA right temporo-occipital region	none
Non-brain-damaged participants
NBD1	45	F	Right	9
NBD2	35	F	Right	9
NBD3	52	F	Right	16
NBD4	44	M	Right	16
NBD5	71	F	Right	16
NBD6	37	M	Right	12
NBD7	63	M	Right	16
NBD8	48	M	Right	16
NBD9	58	M	Right	12
NBD10	58	F	Right	9

POS = post onset of stroke.