Clinical neuroscience 753
Noun/verb distinction in English stress homographs: an ERP study Heechun Moona and Cyrille Magnea,b Sensitivity to speech rhythm, especially the pattern of stressed and unstressed syllables, is an important aspect of language acquisition and comprehension from infancy through adulthood. In English, a strong correlation exists between speech rhythm and grammatical class. This property is well illustrated by a particular group of noun/ verb homographs that are spelled the same but are pronounced with a lexical stress on the first syllable when used as a noun or on the second syllable when used as a verb. The purpose of this study was to further examine the neural markers of speech rhythm and its role in word recognition. To this end, event-related brain potentials were recorded while participants listened to spoken sentences containing a stress homograph either in a noun or a verb position. The rhythmic structure of the stress homographs was manipulated so that they were pronounced with a stress pattern that either matched or mismatched their grammatical class. Results of cluster-based permutation tests on the event-related brain potentials revealed larger negativities over the centrofrontal scalp regions when the
Introduction Prosody plays an important role in speech perception by conveying both emotional and linguistic information. In spoken language, prosodic cues help listeners segment the speech stream into phrases, words, and syllables [1,2] and perform sentence parsing by emphasizing syntactic and semantic information [3]. One particular aspect of prosody known as speech rhythm – created by the pattern of stressed and unstressed syllables – plays an important role in language acquisition during early infancy [4–6] as well as speech segmentation and word recognition in adults [1,7]. A growing literature has investigated the role of rhythm in speech perception using the event-related brain potential (ERP) method to identify the electrophysiological markers underlying the processing of words with incongruous [8–11] or congruous but unexpected stress patterns [12–14]. Overall, these studies found an increased negative ERP component in response to rhythmically incongruous or unexpected words. This effect is generally observed over the frontocentral scalp regions between 200 and 600 ms after the critical-word onset. This negativity has been interpreted as either an N400 component reflecting the lexical retrieval difficulty resulting from the incorrect stress pattern [8,9] or alternatively as reflecting a general nonlinguistic rule-based error-detection mechanism [14,15]. 0959-4965 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
stress homographs were mispronounced, in line with previous studies on lexical ambiguity resolution. In addition, differences between rhythmically unexpected nouns and verbs could be seen as early as 200 ms, suggesting that listeners are sensitive to statistical properties of their language rhythm. Together, these results support the hypothesis that information about speech rhythm is rapidly integrated during speech perception and contributes to lexical retrieval. NeuroReport 26:753–757 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved. NeuroReport 2015, 26:753–757 Keywords: event-related brain potential, frontal negativity, lexical access, speech rhythm, stress homograph a Literacy Studies PhD Program and bPsychology Department, Middle Tennessee State University, Murfreesboro, Tennessee, USA
Correspondence to Heechun Moon, MA, Literacy Studies PhD Program, Middle Tennessee State University, Box 402, Murfreesboro, TN 37132, USA Tel: + 1 615 438 3469; fax: + 1 615 898 5027; e-mail: [email protected] Received 3 June 2015 accepted 12 June 2015
The present study sought to further investigate the electrophysiological basis of speech rhythm in English and its role in lexical ambiguity resolution. A particular property of English rhythm is that it shows a strong correlation between lexical stress and grammatical category. For instance, 93% of bisyllabic nouns are stressed on the first syllable, whereas 76% bisyllabic verbs are stressed on the second syllable [16]. Speech production studies have shown that this distinction is used by speakers who tend to pronounce pseudowords with an initial stress more often when acting as nouns than when acting as verbs in sentence contexts [17]. This property is also well illustrated in noun/verb homographs whose lexical stress shifts syllable position with a change in grammatical class. For instance, the noun ‘permit’ is pronounced with a stress on the first syllable (PERmit), whereas the verb ‘permit’ is pronounced with a stress on the second syllable (perMIT). Thus, these features of noun–verb homographs provide a unique opportunity to study the role of rhythm in word recognition. To that end, participants listened to spoken sentences containing noun/verb stress homographs in either a noun or verb position. We manipulated the stress pattern of the homograph so that it was either expected or unexpected regarding its grammatical role in the sentence (see Table 1 for examples). On the basis of the previous literature, homographs with an unexpected stress pattern DOI: 10.1097/WNR.0000000000000417
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
754 NeuroReport 2015, Vol 26 No 13
Table 1
Examples of stimuli in the four experimental conditions
Rhythmically congruous Rhythmically incongruous
Noun position
Verb position
She has a PERmit to park in the lot She has a perMIT to park in the lot
They do not perMIT public access They do not PERmit public access
Note: Critical words are indicated in italic and stressed syllables are indicated in capital letters.
were expected to elicit an increased negativity between 200 and 500 ms over the frontal electrodes of the scalp.
Methods Participants
Sixteen native speakers of English (10 female, mean age = 22.2 ± 1.9) were recruited. All participants were right-handed and had no history of hearing, psychological, or neurological problems. Three participants were excluded because of excessive artefacts in the electroencephalography (EEG) signal. The study was approved by the local Institutional Review Board, and all participants gave their written informed consent before the start of the experiment. Stimuli
A total of 128 sentences were built using noun/verb stress homographs as critical words. For half of the sentences, the critical word was used as a noun, whereas it was used as a verb for the other half. Two versions of each sentence were recorded by manipulating the position of the lexical stress in the critical word. In one version, the stress pattern of the critical word was expected (initially stressed noun or noninitially stressed verb). In the other version, the stress pattern of the critical word was unexpected (noninitially stressed noun or initially stressed verb). Each participant only heard the rhythmically expected or unexpected version of each sentence. The presentation of sentences was counterbalanced so that all versions were presented across participants. Sentences were produced by a male voice at a sampling rate of 44 kHz using the Neospeech Text-to-Speech software (Neospeech Inc., Santa Clara, California, USA) to have a consistent intonation and speech rate between the rhythmically expected and rhythmically unexpected versions [18]. In sum, the critical word had either a noun or verb function and was pronounced with either the expected or unexpected stress pattern (Table 1). Participants were presented with 32 stimuli in each experimental condition. Finally, the final word of the sentence was manipulated so that it was semantically expected or unexpected within the sentence context. Participants performed a semantic judgement task to ensure that they read carefully each sentence as well as to focus their attention away from the rhythmic manipulation. Note, however, that as the sentences were always semantically appropriate past the critical words, semantically correct and incorrect sentences were collapsed for data analysis.
Procedure
During the experiment, participants were seated at a comfortable chair in a soundproof room. Participants were given a practice block of 10 trials, followed by four experimental blocks of 32 sentences each. The trials were randomized within each block and the order of the blocks was counterbalanced across participants. Each trial was introduced with a fixation cross displayed in the middle of a computer screen and that remained until the offset of the spoken sentence. Participants were instructed to avoid any eye movement during the presentation of the sentences. At the offset of each sentence, a series of X’s was presented on the screen for 2 s to allow participants to rest their eyes. Participants were asked to listen to each sentence carefully and determine whether it made sense or not by pressing one of two buttons on a response pad.
Data acquisition and analysis
Continuous EEG was recorded from a 64-channel Hydrocel Geodesic Sensor Net (EGI, Eugene, Oregon, USA) connected to a NetAmps 300 amplifier and using the software NetStation 4.3. Six eye channels embedded in the EEG sensor net were used to record the vertical and horizontal electrooculograms, so that blinks and eye movements could be identified. Data were sampled at 500 Hz and filtered with a bandpass of 0.1–100 Hz. Data were reference online to the vertex (Cz) and then rereferenced offline to the averaged mastoids. Electrode impedances were kept below 50 kΩ. EEG epochs timelocked to the presentation of the noun/verb homographs were extracted from − 200 to + 1000 ms relative to the stimulus onset. EEG epochs containing artefacts such as eye blink, eye movement, electrode drifting, or muscle activity were discarded. ERPs were computed by averaging the remaining EEG epochs separately for each participant and condition. Statistical analyses were conducted with MATLAB (Matrix Laboratory; The MathWorks Inc., Natick, Massachusetts, USA) and the FieldTrip toolbox [19]. A cluster-based permutation test was used to perform planned comparison between pairs of conditions. The cluster-based permutation test uses nonparametric statistics to capture ERP effects without prior assumptions about their scalp distribution and latency range [20]. Comparisons were performed between rhythmically unexpected and rhythmically expected conditions, separately for the noun and verb positions.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
Speech rhythm and lexical retrieval Moon and Magne 755
Results The grand-average ERPs for the rhythmically expected and unexpected critical words in noun and verb positions are presented in Fig. 1a. For the noun position, a clusterbased permutation test revealed that the rhythmically unexpected condition was associated with an increased centrofrontal negativity (Fig. 1b) between 344 and 594 ms post critical-word onset (P = 0.003). For the verb position, statistical analyses revealed a significant cluster between 420 and 562 ms post critical-word onset (P = 0.016), resulting from a larger centrofrontal negativity in response to the rhythmically unexpected than rhythmically expected conditions (Fig. 1c). As can be seen in Fig. 1a, early differences appear between nouns and verbs as early as 200 ms post critical-word onset. We thus performed an additional series of analyses to directly compare the unexpected rhythmic effect between homographs in noun versus verb positions. To account for morphological differences between the ERPs for nouns and verbs, difference waves (i.e., rhythmically unexpected condition minus rhythmically expected condition) were
computed separately for the noun and verb positions. Then, the difference waves were analyzed using the same aforementioned cluster-based permutation method. Results revealed that the unexpected rhythmic effects were significantly more negative in the noun than the verb position between 230 and 310 ms post critical-word onset (P = 0.024). This difference was larger over the frontal regions of the scalp (Fig. 1d).
Discussion In the present study, English noun/verb stress homographs were used to investigate neural indices of speech rhythm processing and the role of speech rhythm in lexical retrieval. Participants were presented with rhythmically ambiguous homographs embedded in spoken sentences. Results demonstrated a large negative effect over the centrofrontal regions for both nouns and verbs with an unexpected stress pattern. In addition, rhythmically unexpected nouns were associated with an early left frontal negativity when compared with rhythmically unexpected verbs.
Fig. 1
(b) Unexpected rhythm effect: Noun position
(a) Grand-average ERPs E11
E2 2
−2 E18
E58
(c) Unexpected rhythm effect: Verb position
E11
−5μV
E18
E2 E58
(d) Unexpected rhythm effect: Noun – verb 400
800 ms
Noun expected rhythm
Verb expected rhythm
Noun unexpected rhythm
Verb unexpected rhythm
ERPs elicited by the critical word. (a) Grand-average waveforms recorded at four selected electrodes for rhythmically expected nouns (black solid line) and verbs (gray solid lines) as well as rhythmically unexpected nouns (black dashed line) and verbs (gray dashed line). Waveforms are low-pass filtered at 8 Hz for display only. (b) Topographic maps showing mean differences in scalp amplitudes in the latency range of the significant clusters for the noun position. On this panel and the following, electrodes belonging to the cluster are indicated with a white star. (c) Topographic maps showing mean differences in scalp amplitudes in the latency range of the significant clusters for the verb position. (d) Topographic maps showing mean differences between rhythmically unexpected effects to nouns and verbs in the latency range of the significant clusters. ERPs, event-related brain potentials.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
756 NeuroReport 2015, Vol 26 No 13
The presence of ERP differences for homographs with unexpected stress patterns is in line with previous behavioral findings showing that listeners are sensitive to the rhythmic cues underlying the distinctions between nouns and verbs in English [16,17]. Interestingly, this sensitivity has been found not only for spoken language, but also during silent reading [21]. Thus together, these results suggest that the stress pattern of a word is part of its lexical representation. In addition, the fact that these effects were present in our study even though the task requirement did not direct the participants’ attention toward the rhythmic aspects of the stimuli suggests that the information about the rhythmic structure of a word is automatically reactivated during language comprehension [8]. Our results are also in line with previous ERP studies on speech rhythm showing an increased centrofrontal negativity in responses to words with an unexpected or incongruous stress pattern. Although these studies were performed in German [11,13,14], French [8,9] and Dutch [12], the present study shows that a similar response is triggered in English as well. This negative ERP effect has been interpreted as reflecting either an N400 associated with increased lexico-semantic processing difficulty [8,9] or alternatively a domain-general detection mechanism triggered in response to rule-based violations [14,15]. However, it is important to note that these studies differed in term of language (French vs. German), stimuli (real words vs. nonwords), and task demand (semantic acceptability vs. grammatical correctness). Thus, these two interpretations are not necessarily incompatible and the neurocognitive mechanisms underlying the ERP responses to incongruous/unexpected stress patterns are likely to be influenced by the number and type of linguistic constraints available to the listener. Regarding the possible interpretation of the negative effects found in the present study, it is also important to consider that our rhythmic manipulation led to a violation of grammatical class as the critical words were noun/verb stress homographs. Several studies have examined ERP responses to noun/verb ambiguous words (e.g. a duck vs. to duck) embedded in written sentences with varying semantic constraints. Compared with matched unambiguous words, lexically ambiguous words elicited an increased N400 when presented in semantically-based sentence contexts, but an early sustained frontal negativity when presented in sentences with a weak semantic context [22–24]. This frontal negativity has been interpreted as reflecting top-down processes involved in meaning selection [24]. In line with this finding, several recent neuroimaging studies have found increased activity in the left inferior frontal gyrus in response to lexically ambiguous words presented in semantically lowconstrained contexts, and which has also been interpreted as reflecting the selection of the appropriate
meaning among competitors [25]. These findings thus suggest that the early frontal negativity associated with lexical ambiguity resolution may originate from sources located in frontal brain areas [24]. Interestingly, the findings of a recent study [26] using isolated monosyllabic spoken homophones suggest that prosodic cues (expressed mainly through differences in word duration between nouns and verbs) may play an important role in the resolution of lexical ambiguities. Thus, in the present study, it is possible that the negativity observed in responses to homographs with unexpected stress patterns reflects a conflict between bottom-up information (i.e. acoustical correlates of the word stress pattern) that automatically initiates the lexical search and top-down expectations (both semantic and syntactic) generated from the sentence context. Finally, an early difference was found between the unexpected rhythmic effects in noun and verb positions as early as 230 ms over frontal scalp regions. As previously mentioned, correlations exist between the stress pattern and the lexical status of a word in English [16,17]. However, when considering the entire English lexicon, the trochaic stress pattern (i.e. a stressed syllable followed by an unstressed syllable) is more prevalent (85–90% of spoken English words according to Cutler and Carter [27]) than the iambic stress pattern (i.e. an unstressed syllable followed by a stressed syllable). It is thus possible that this early difference reflects the fact that an unexpected iambic pattern on a noun is more salient than an unexpected trochaic pattern on a verb as the iambic pattern is less frequent in English. Although the exact functional significance of this early difference remains to be determined, it further supports the hypothesis that English listeners are sensitive to the statistical properties of their language rhythm and use this information quickly during spoken language comprehension. Several previous studies have shown that English-learning infants use the ∼ 9 : 1 ratio of trochaic to iambic stress patterns to detect word onset in the continuous speech stream. For instance, infants detect more easily trochaic than iambic words within speech at the age of 7.5 months, and they show a listening preference for trochaic words at the age of 9 months (see Jusczyk [28] for a review). Adult speakers of English show a similar trochaic bias, segmenting the speech signal at the onset of stressed syllables [1].
Conclusion The current study extends previous findings on the role of stress patterns in lexical access in English and more broadly on the neural basis of speech rhythm. In addition, our results further support the proposal that listeners are sensitive to the rhythmic cues underlying the distinction between grammatical classes in English.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
Speech rhythm and lexical retrieval Moon and Magne 757
Acknowledgements
12
This study was partially funded by NSF Grant #BCS-1261460 awarded to Cyrille Magne and by the MTSU Faculty Research and Creative Activity Grant Program.
13
14
Conflicts of interest
There are no conflicts of interest.
15
References
16
1 Cutler A, Norris D. The role of strong syllables in segmentation for lexical access. J Exp Psychol 1988; 14:113–131. 2 Quené H, Port RF. Effects of timing regularity and metrical expectancy on spoken-word perception. Phonetica 2005; 62:1–13. 3 Whalley K, Hansen J. The role of prosodic sensitivity in children’s reading development. J Res Read 2006; 29:288–303. 4 Grassmann S, Tomasello M. Prosodic stress on a word directs 24-montholds’ attention to a contextually new referent. J Pragmatics 2010; 42:3098–3105. 5 Jusczyk PW, Cutler A, Redanz NJ. Infants’ preference for the predominant stress patterns of English words. Child Dev 1993; 64:675–687. 6 Nazzi T, Bertoncini J, Mehler J. Language discrimination by newborns: toward an understanding of the role of rhythm. J Exp Psychol Hum Percept Perform 1998; 24:756–766. 7 Small LH, Simon SD, Goldberg JS. Lexical stress and lexical access: homographs versus nonhomographs. Percept Psychophys 1988; 44:272–280. 8 Magne C, Astésano C, Aramaki M, Ystad S, Kronland-Martinet R, Besson M. Influence of syllabic lengthening on semantic processing in spoken French: behavioral and electrophysiological evidence. Cereb Cortex 2007; 17:2659–2668. 9 Marie C, Magne C, Besson M. Musicians and the metric structure of words. J Cogn Neurosci 2011; 23:294–305. 10 McCauley SM, Hestvik A, Vogel I. Perception and bias in the processing of compound versus phrasal stress: evidence from event-related brain potentials. Lang Speech 2013; 56 (Pt 1):23–44. 11 Schmidt-Kassow M, Kotz SA. Event-related brain potentials suggest a late interaction of meter and syntax in the P600. J Cogn Neurosci 2009; 21:1693–1708.
17 18 19
20 21 22 23 24
25
26 27 28
Böcker KB, Bastiaansen MC, Vroomen J, Brunia CH, de Gelder B. An ERP correlate of metrical stress in spoken word recognition. Psychophysiology 1999; 36:706–720. Bohn K, Knaus J, Wiese R, Domahs U. The influence of rhythmic (ir) regularities on speech processing: evidence from an ERP study on German phrases. Neuropsychologia 2013; 51:760–771. Rothermich K, Schmidt-Kassow M, Schwartze M, Kotz SA. Event-related potential responses to metric violations: rules versus meaning. Neuroreport 2010; 21:580–584. Roncaglia-Denissen MP, Schmidt-Kassow M, Kotz SA. Speech rhythm facilitates syntactic ambiguity resolution: ERP evidence. PLoS One 2013; 8: e56000. Sereno JA, Jongman A. Acoustic correlates of grammatical class. Lang Speech 1995; 38:57–76. Kelly MH, Bock JK. Stress in time. J Exp Psychol 1988; 14:389–403. Chen F, Wong LLN, Hu Y. Effects of Lexical Tone Contour on Mandarin Sentence Intelligibility. J Speech, Lang and Hearing Res 2014; 57:338–345. Oostenveld R, Fries P, Maris E, Schoffelen JM. FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosci 2011; 2011:156869. Maris E, Oostenveld R. Nonparametric statistical testing of EEG- and MEGdata. J Neurosci Methods 2007; 164:177–190. Breen M, Clifton C Jr. Stress matters: effects of anticipated lexical stress on silent reading. J Mem Lang 2011; 64:153–170. Federmeier KD, Segal JB, Lombrozo T, Kutas M. Brain responses to nouns, verbs and class-ambiguous words in context. Brain 2000; 123:2552–2566. Lee CL, Federmeier KD. To mind the mind: an event-related potential study of word class and semantic ambiguity. Brain Res 2006; 1081:191–202. Lee CL, Federmeier KD. Wave-ering: An ERP study of syntactic and semantic context effects on ambiguity resolution for noun/verb homographs. J Mem Lang 2009; 61:538–555. Rodd JM, Johnsrude IS, Davis MH. Dissociating frontotemporal contributions to semantic ambiguity resolution in spoken sentences. Cereb Cortex 2012; 22:1761–1773. Conwell E. Neural responses to category ambiguous words. Neuropsychologia 2015; 69:85–92. Cutler A, Carter DM. The predominance of strong initial syllables in the English vocabulary. Comp Speech Lang 1987; 2:133–142. Jusczyk PW. How infants begin to extract words from speech. Trends Cogn Sci 1999; 3:323–328.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
Noun/verb distinction in English stress homographs: an ERP study Heechun Moona and Cyrille Magnea,b Sensitivity to speech rhythm, especially the pattern of stressed and unstressed syllables, is an important aspect of language acquisition and comprehension from infancy through adulthood. In English, a strong correlation exists between speech rhythm and grammatical class. This property is well illustrated by a particular group of noun/ verb homographs that are spelled the same but are pronounced with a lexical stress on the first syllable when used as a noun or on the second syllable when used as a verb. The purpose of this study was to further examine the neural markers of speech rhythm and its role in word recognition. To this end, event-related brain potentials were recorded while participants listened to spoken sentences containing a stress homograph either in a noun or a verb position. The rhythmic structure of the stress homographs was manipulated so that they were pronounced with a stress pattern that either matched or mismatched their grammatical class. Results of cluster-based permutation tests on the event-related brain potentials revealed larger negativities over the centrofrontal scalp regions when the
Introduction Prosody plays an important role in speech perception by conveying both emotional and linguistic information. In spoken language, prosodic cues help listeners segment the speech stream into phrases, words, and syllables [1,2] and perform sentence parsing by emphasizing syntactic and semantic information [3]. One particular aspect of prosody known as speech rhythm – created by the pattern of stressed and unstressed syllables – plays an important role in language acquisition during early infancy [4–6] as well as speech segmentation and word recognition in adults [1,7]. A growing literature has investigated the role of rhythm in speech perception using the event-related brain potential (ERP) method to identify the electrophysiological markers underlying the processing of words with incongruous [8–11] or congruous but unexpected stress patterns [12–14]. Overall, these studies found an increased negative ERP component in response to rhythmically incongruous or unexpected words. This effect is generally observed over the frontocentral scalp regions between 200 and 600 ms after the critical-word onset. This negativity has been interpreted as either an N400 component reflecting the lexical retrieval difficulty resulting from the incorrect stress pattern [8,9] or alternatively as reflecting a general nonlinguistic rule-based error-detection mechanism [14,15]. 0959-4965 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
stress homographs were mispronounced, in line with previous studies on lexical ambiguity resolution. In addition, differences between rhythmically unexpected nouns and verbs could be seen as early as 200 ms, suggesting that listeners are sensitive to statistical properties of their language rhythm. Together, these results support the hypothesis that information about speech rhythm is rapidly integrated during speech perception and contributes to lexical retrieval. NeuroReport 26:753–757 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved. NeuroReport 2015, 26:753–757 Keywords: event-related brain potential, frontal negativity, lexical access, speech rhythm, stress homograph a Literacy Studies PhD Program and bPsychology Department, Middle Tennessee State University, Murfreesboro, Tennessee, USA
Correspondence to Heechun Moon, MA, Literacy Studies PhD Program, Middle Tennessee State University, Box 402, Murfreesboro, TN 37132, USA Tel: + 1 615 438 3469; fax: + 1 615 898 5027; e-mail: [email protected] Received 3 June 2015 accepted 12 June 2015
The present study sought to further investigate the electrophysiological basis of speech rhythm in English and its role in lexical ambiguity resolution. A particular property of English rhythm is that it shows a strong correlation between lexical stress and grammatical category. For instance, 93% of bisyllabic nouns are stressed on the first syllable, whereas 76% bisyllabic verbs are stressed on the second syllable [16]. Speech production studies have shown that this distinction is used by speakers who tend to pronounce pseudowords with an initial stress more often when acting as nouns than when acting as verbs in sentence contexts [17]. This property is also well illustrated in noun/verb homographs whose lexical stress shifts syllable position with a change in grammatical class. For instance, the noun ‘permit’ is pronounced with a stress on the first syllable (PERmit), whereas the verb ‘permit’ is pronounced with a stress on the second syllable (perMIT). Thus, these features of noun–verb homographs provide a unique opportunity to study the role of rhythm in word recognition. To that end, participants listened to spoken sentences containing noun/verb stress homographs in either a noun or verb position. We manipulated the stress pattern of the homograph so that it was either expected or unexpected regarding its grammatical role in the sentence (see Table 1 for examples). On the basis of the previous literature, homographs with an unexpected stress pattern DOI: 10.1097/WNR.0000000000000417
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
754 NeuroReport 2015, Vol 26 No 13
Table 1
Examples of stimuli in the four experimental conditions
Rhythmically congruous Rhythmically incongruous
Noun position
Verb position
She has a PERmit to park in the lot She has a perMIT to park in the lot
They do not perMIT public access They do not PERmit public access
Note: Critical words are indicated in italic and stressed syllables are indicated in capital letters.
were expected to elicit an increased negativity between 200 and 500 ms over the frontal electrodes of the scalp.
Methods Participants
Sixteen native speakers of English (10 female, mean age = 22.2 ± 1.9) were recruited. All participants were right-handed and had no history of hearing, psychological, or neurological problems. Three participants were excluded because of excessive artefacts in the electroencephalography (EEG) signal. The study was approved by the local Institutional Review Board, and all participants gave their written informed consent before the start of the experiment. Stimuli
A total of 128 sentences were built using noun/verb stress homographs as critical words. For half of the sentences, the critical word was used as a noun, whereas it was used as a verb for the other half. Two versions of each sentence were recorded by manipulating the position of the lexical stress in the critical word. In one version, the stress pattern of the critical word was expected (initially stressed noun or noninitially stressed verb). In the other version, the stress pattern of the critical word was unexpected (noninitially stressed noun or initially stressed verb). Each participant only heard the rhythmically expected or unexpected version of each sentence. The presentation of sentences was counterbalanced so that all versions were presented across participants. Sentences were produced by a male voice at a sampling rate of 44 kHz using the Neospeech Text-to-Speech software (Neospeech Inc., Santa Clara, California, USA) to have a consistent intonation and speech rate between the rhythmically expected and rhythmically unexpected versions [18]. In sum, the critical word had either a noun or verb function and was pronounced with either the expected or unexpected stress pattern (Table 1). Participants were presented with 32 stimuli in each experimental condition. Finally, the final word of the sentence was manipulated so that it was semantically expected or unexpected within the sentence context. Participants performed a semantic judgement task to ensure that they read carefully each sentence as well as to focus their attention away from the rhythmic manipulation. Note, however, that as the sentences were always semantically appropriate past the critical words, semantically correct and incorrect sentences were collapsed for data analysis.
Procedure
During the experiment, participants were seated at a comfortable chair in a soundproof room. Participants were given a practice block of 10 trials, followed by four experimental blocks of 32 sentences each. The trials were randomized within each block and the order of the blocks was counterbalanced across participants. Each trial was introduced with a fixation cross displayed in the middle of a computer screen and that remained until the offset of the spoken sentence. Participants were instructed to avoid any eye movement during the presentation of the sentences. At the offset of each sentence, a series of X’s was presented on the screen for 2 s to allow participants to rest their eyes. Participants were asked to listen to each sentence carefully and determine whether it made sense or not by pressing one of two buttons on a response pad.
Data acquisition and analysis
Continuous EEG was recorded from a 64-channel Hydrocel Geodesic Sensor Net (EGI, Eugene, Oregon, USA) connected to a NetAmps 300 amplifier and using the software NetStation 4.3. Six eye channels embedded in the EEG sensor net were used to record the vertical and horizontal electrooculograms, so that blinks and eye movements could be identified. Data were sampled at 500 Hz and filtered with a bandpass of 0.1–100 Hz. Data were reference online to the vertex (Cz) and then rereferenced offline to the averaged mastoids. Electrode impedances were kept below 50 kΩ. EEG epochs timelocked to the presentation of the noun/verb homographs were extracted from − 200 to + 1000 ms relative to the stimulus onset. EEG epochs containing artefacts such as eye blink, eye movement, electrode drifting, or muscle activity were discarded. ERPs were computed by averaging the remaining EEG epochs separately for each participant and condition. Statistical analyses were conducted with MATLAB (Matrix Laboratory; The MathWorks Inc., Natick, Massachusetts, USA) and the FieldTrip toolbox [19]. A cluster-based permutation test was used to perform planned comparison between pairs of conditions. The cluster-based permutation test uses nonparametric statistics to capture ERP effects without prior assumptions about their scalp distribution and latency range [20]. Comparisons were performed between rhythmically unexpected and rhythmically expected conditions, separately for the noun and verb positions.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
Speech rhythm and lexical retrieval Moon and Magne 755
Results The grand-average ERPs for the rhythmically expected and unexpected critical words in noun and verb positions are presented in Fig. 1a. For the noun position, a clusterbased permutation test revealed that the rhythmically unexpected condition was associated with an increased centrofrontal negativity (Fig. 1b) between 344 and 594 ms post critical-word onset (P = 0.003). For the verb position, statistical analyses revealed a significant cluster between 420 and 562 ms post critical-word onset (P = 0.016), resulting from a larger centrofrontal negativity in response to the rhythmically unexpected than rhythmically expected conditions (Fig. 1c). As can be seen in Fig. 1a, early differences appear between nouns and verbs as early as 200 ms post critical-word onset. We thus performed an additional series of analyses to directly compare the unexpected rhythmic effect between homographs in noun versus verb positions. To account for morphological differences between the ERPs for nouns and verbs, difference waves (i.e., rhythmically unexpected condition minus rhythmically expected condition) were
computed separately for the noun and verb positions. Then, the difference waves were analyzed using the same aforementioned cluster-based permutation method. Results revealed that the unexpected rhythmic effects were significantly more negative in the noun than the verb position between 230 and 310 ms post critical-word onset (P = 0.024). This difference was larger over the frontal regions of the scalp (Fig. 1d).
Discussion In the present study, English noun/verb stress homographs were used to investigate neural indices of speech rhythm processing and the role of speech rhythm in lexical retrieval. Participants were presented with rhythmically ambiguous homographs embedded in spoken sentences. Results demonstrated a large negative effect over the centrofrontal regions for both nouns and verbs with an unexpected stress pattern. In addition, rhythmically unexpected nouns were associated with an early left frontal negativity when compared with rhythmically unexpected verbs.
Fig. 1
(b) Unexpected rhythm effect: Noun position
(a) Grand-average ERPs E11
E2 2
−2 E18
E58
(c) Unexpected rhythm effect: Verb position
E11
−5μV
E18
E2 E58
(d) Unexpected rhythm effect: Noun – verb 400
800 ms
Noun expected rhythm
Verb expected rhythm
Noun unexpected rhythm
Verb unexpected rhythm
ERPs elicited by the critical word. (a) Grand-average waveforms recorded at four selected electrodes for rhythmically expected nouns (black solid line) and verbs (gray solid lines) as well as rhythmically unexpected nouns (black dashed line) and verbs (gray dashed line). Waveforms are low-pass filtered at 8 Hz for display only. (b) Topographic maps showing mean differences in scalp amplitudes in the latency range of the significant clusters for the noun position. On this panel and the following, electrodes belonging to the cluster are indicated with a white star. (c) Topographic maps showing mean differences in scalp amplitudes in the latency range of the significant clusters for the verb position. (d) Topographic maps showing mean differences between rhythmically unexpected effects to nouns and verbs in the latency range of the significant clusters. ERPs, event-related brain potentials.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
756 NeuroReport 2015, Vol 26 No 13
The presence of ERP differences for homographs with unexpected stress patterns is in line with previous behavioral findings showing that listeners are sensitive to the rhythmic cues underlying the distinctions between nouns and verbs in English [16,17]. Interestingly, this sensitivity has been found not only for spoken language, but also during silent reading [21]. Thus together, these results suggest that the stress pattern of a word is part of its lexical representation. In addition, the fact that these effects were present in our study even though the task requirement did not direct the participants’ attention toward the rhythmic aspects of the stimuli suggests that the information about the rhythmic structure of a word is automatically reactivated during language comprehension [8]. Our results are also in line with previous ERP studies on speech rhythm showing an increased centrofrontal negativity in responses to words with an unexpected or incongruous stress pattern. Although these studies were performed in German [11,13,14], French [8,9] and Dutch [12], the present study shows that a similar response is triggered in English as well. This negative ERP effect has been interpreted as reflecting either an N400 associated with increased lexico-semantic processing difficulty [8,9] or alternatively a domain-general detection mechanism triggered in response to rule-based violations [14,15]. However, it is important to note that these studies differed in term of language (French vs. German), stimuli (real words vs. nonwords), and task demand (semantic acceptability vs. grammatical correctness). Thus, these two interpretations are not necessarily incompatible and the neurocognitive mechanisms underlying the ERP responses to incongruous/unexpected stress patterns are likely to be influenced by the number and type of linguistic constraints available to the listener. Regarding the possible interpretation of the negative effects found in the present study, it is also important to consider that our rhythmic manipulation led to a violation of grammatical class as the critical words were noun/verb stress homographs. Several studies have examined ERP responses to noun/verb ambiguous words (e.g. a duck vs. to duck) embedded in written sentences with varying semantic constraints. Compared with matched unambiguous words, lexically ambiguous words elicited an increased N400 when presented in semantically-based sentence contexts, but an early sustained frontal negativity when presented in sentences with a weak semantic context [22–24]. This frontal negativity has been interpreted as reflecting top-down processes involved in meaning selection [24]. In line with this finding, several recent neuroimaging studies have found increased activity in the left inferior frontal gyrus in response to lexically ambiguous words presented in semantically lowconstrained contexts, and which has also been interpreted as reflecting the selection of the appropriate
meaning among competitors [25]. These findings thus suggest that the early frontal negativity associated with lexical ambiguity resolution may originate from sources located in frontal brain areas [24]. Interestingly, the findings of a recent study [26] using isolated monosyllabic spoken homophones suggest that prosodic cues (expressed mainly through differences in word duration between nouns and verbs) may play an important role in the resolution of lexical ambiguities. Thus, in the present study, it is possible that the negativity observed in responses to homographs with unexpected stress patterns reflects a conflict between bottom-up information (i.e. acoustical correlates of the word stress pattern) that automatically initiates the lexical search and top-down expectations (both semantic and syntactic) generated from the sentence context. Finally, an early difference was found between the unexpected rhythmic effects in noun and verb positions as early as 230 ms over frontal scalp regions. As previously mentioned, correlations exist between the stress pattern and the lexical status of a word in English [16,17]. However, when considering the entire English lexicon, the trochaic stress pattern (i.e. a stressed syllable followed by an unstressed syllable) is more prevalent (85–90% of spoken English words according to Cutler and Carter [27]) than the iambic stress pattern (i.e. an unstressed syllable followed by a stressed syllable). It is thus possible that this early difference reflects the fact that an unexpected iambic pattern on a noun is more salient than an unexpected trochaic pattern on a verb as the iambic pattern is less frequent in English. Although the exact functional significance of this early difference remains to be determined, it further supports the hypothesis that English listeners are sensitive to the statistical properties of their language rhythm and use this information quickly during spoken language comprehension. Several previous studies have shown that English-learning infants use the ∼ 9 : 1 ratio of trochaic to iambic stress patterns to detect word onset in the continuous speech stream. For instance, infants detect more easily trochaic than iambic words within speech at the age of 7.5 months, and they show a listening preference for trochaic words at the age of 9 months (see Jusczyk [28] for a review). Adult speakers of English show a similar trochaic bias, segmenting the speech signal at the onset of stressed syllables [1].
Conclusion The current study extends previous findings on the role of stress patterns in lexical access in English and more broadly on the neural basis of speech rhythm. In addition, our results further support the proposal that listeners are sensitive to the rhythmic cues underlying the distinction between grammatical classes in English.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
Speech rhythm and lexical retrieval Moon and Magne 757
Acknowledgements
12
This study was partially funded by NSF Grant #BCS-1261460 awarded to Cyrille Magne and by the MTSU Faculty Research and Creative Activity Grant Program.
13
14
Conflicts of interest
There are no conflicts of interest.
15
References
16
1 Cutler A, Norris D. The role of strong syllables in segmentation for lexical access. J Exp Psychol 1988; 14:113–131. 2 Quené H, Port RF. Effects of timing regularity and metrical expectancy on spoken-word perception. Phonetica 2005; 62:1–13. 3 Whalley K, Hansen J. The role of prosodic sensitivity in children’s reading development. J Res Read 2006; 29:288–303. 4 Grassmann S, Tomasello M. Prosodic stress on a word directs 24-montholds’ attention to a contextually new referent. J Pragmatics 2010; 42:3098–3105. 5 Jusczyk PW, Cutler A, Redanz NJ. Infants’ preference for the predominant stress patterns of English words. Child Dev 1993; 64:675–687. 6 Nazzi T, Bertoncini J, Mehler J. Language discrimination by newborns: toward an understanding of the role of rhythm. J Exp Psychol Hum Percept Perform 1998; 24:756–766. 7 Small LH, Simon SD, Goldberg JS. Lexical stress and lexical access: homographs versus nonhomographs. Percept Psychophys 1988; 44:272–280. 8 Magne C, Astésano C, Aramaki M, Ystad S, Kronland-Martinet R, Besson M. Influence of syllabic lengthening on semantic processing in spoken French: behavioral and electrophysiological evidence. Cereb Cortex 2007; 17:2659–2668. 9 Marie C, Magne C, Besson M. Musicians and the metric structure of words. J Cogn Neurosci 2011; 23:294–305. 10 McCauley SM, Hestvik A, Vogel I. Perception and bias in the processing of compound versus phrasal stress: evidence from event-related brain potentials. Lang Speech 2013; 56 (Pt 1):23–44. 11 Schmidt-Kassow M, Kotz SA. Event-related brain potentials suggest a late interaction of meter and syntax in the P600. J Cogn Neurosci 2009; 21:1693–1708.
17 18 19
20 21 22 23 24
25
26 27 28
Böcker KB, Bastiaansen MC, Vroomen J, Brunia CH, de Gelder B. An ERP correlate of metrical stress in spoken word recognition. Psychophysiology 1999; 36:706–720. Bohn K, Knaus J, Wiese R, Domahs U. The influence of rhythmic (ir) regularities on speech processing: evidence from an ERP study on German phrases. Neuropsychologia 2013; 51:760–771. Rothermich K, Schmidt-Kassow M, Schwartze M, Kotz SA. Event-related potential responses to metric violations: rules versus meaning. Neuroreport 2010; 21:580–584. Roncaglia-Denissen MP, Schmidt-Kassow M, Kotz SA. Speech rhythm facilitates syntactic ambiguity resolution: ERP evidence. PLoS One 2013; 8: e56000. Sereno JA, Jongman A. Acoustic correlates of grammatical class. Lang Speech 1995; 38:57–76. Kelly MH, Bock JK. Stress in time. J Exp Psychol 1988; 14:389–403. Chen F, Wong LLN, Hu Y. Effects of Lexical Tone Contour on Mandarin Sentence Intelligibility. J Speech, Lang and Hearing Res 2014; 57:338–345. Oostenveld R, Fries P, Maris E, Schoffelen JM. FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosci 2011; 2011:156869. Maris E, Oostenveld R. Nonparametric statistical testing of EEG- and MEGdata. J Neurosci Methods 2007; 164:177–190. Breen M, Clifton C Jr. Stress matters: effects of anticipated lexical stress on silent reading. J Mem Lang 2011; 64:153–170. Federmeier KD, Segal JB, Lombrozo T, Kutas M. Brain responses to nouns, verbs and class-ambiguous words in context. Brain 2000; 123:2552–2566. Lee CL, Federmeier KD. To mind the mind: an event-related potential study of word class and semantic ambiguity. Brain Res 2006; 1081:191–202. Lee CL, Federmeier KD. Wave-ering: An ERP study of syntactic and semantic context effects on ambiguity resolution for noun/verb homographs. J Mem Lang 2009; 61:538–555. Rodd JM, Johnsrude IS, Davis MH. Dissociating frontotemporal contributions to semantic ambiguity resolution in spoken sentences. Cereb Cortex 2012; 22:1761–1773. Conwell E. Neural responses to category ambiguous words. Neuropsychologia 2015; 69:85–92. Cutler A, Carter DM. The predominance of strong initial syllables in the English vocabulary. Comp Speech Lang 1987; 2:133–142. Jusczyk PW. How infants begin to extract words from speech. Trends Cogn Sci 1999; 3:323–328.
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
