Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Neuro-cognitive foundations of word stress processing - evidence from fMRI

Neuro-cognitive foundations of word stress processing - evidence from fMRI Background: To date, the neural correlates of phonological word stress processing are largely unknown. Methods: In the present study, we investigated the processing of word stress and vowel quality using an identity matching task with pseudowords. Results: In line with previous studies, a bilateral fronto-temporal network comprising the superior temporal gyri extending into the sulci as well as the inferior frontal gyri was observed for word stress processing. Moreover, we found differences in the superior temporal gyrus and the superior temporal sulcus, bilaterally, for the processing of different stress patterns. For vowel quality processing, our data reveal a substantial contribution of the left intraparietal cortex. All activations were modulated by task demands, yielding different patterns for same and different pairs of stimuli. Conclusions: Our results suggest that the left superior temporal gyrus represents a basic system underlying stress processing to which additional structures including the homologous cortex site are recruited with increasing difficulty. Introduction The autonomy of vowel quality and word stress It is widely agreed that the processing of spoken words representations comprises acoustic and phonological analysis before in a First evidence for a relative independence of vowel qual- second step lexical and semantic information can be ity and word stress encoding in speech production came retrieved (e.g., [1-3]). With respect to the acoustic-pho- from psycholinguistic research. In particular, speech nological analysis of spoken words, there is general con- errors that involve stress exchange such as “my ‘proso- sensus that the categorical perception of phonetic dic (pro’sodic) colleagues” [11], though occurring rather properties like frequency formants, transitional proper- rarely, specifically demonstrate a separate encoding ties of formants, fundamental frequency, duration, or stage for word stress. Moreover, findings from speech intensity leads to the identification of strings of pho- perception point to a relatively independent processing nemes and - at least in languages with variable stress - of stress and vowel quality information although, of to the identification of word stress patterns. On a course, the metrical feature ‘stress’ inevitably has also its neuro-functional level, phonological processing has been vowel quality correlates such as vowel reduction in attributed to the superior temporal gyrus of both hemi- unstressed syllables ([12,13]). For instance, not only spheres(e.g.,[4-10]).However,sofarnostudyhas minimal stress pairs (i.e., words only differing in their aimed at directly differentiating vowel quality and word stress position) can be successfully discriminated on the stress processing. As a starting point, findings on the basis of their different stress patterns; even isolated syl- processing of both vowel quality as well as stress infor- lables excised from such minimal pairs can be reliably mation will be reviewed briefly. assigned to their source words [12,14]. Isolated syllables bearing a stressed or unstressed pitch contour can influ- ence the processing of subsequently presented targets which have a segmentally identical initial syllable with * Correspondence: klein@neuropsych.rwth-aachen.de Department of Neurology, Section Neuropsychology, University Hospital, congruent pitch [15]. However, while both vowel quality RWTH Aachen University, Aachen, Germany and stress can separately contribute to lexical Full list of author information is available at the end of the article © 2011 Klein et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 2 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 recognition [16], there is evidence that vowel quality imaging (fMRI). In this respect, it has been assumed that information can be exploited earlier than stress informa- the processing of emotional prosody elicits bilateral tion (e.g., vowel duration, pitch height, and amplitude) fronto-temporal patterns (e.g. [37]), while processing of due to coarticulation [16-19]. linguistic prosody has been suggested to be left lateralized Findings from Dupoux and colleagues suggest that in the superior temporal gyrus (for a review see [38]; but also on the level of abstract representation vowel quality see [39] for activation of Broca’s area associated with lin- and stress information may dissociate [20-22]. The so- guistic aspects of prosody). For the processing of linguistic called ‘stress deafness’ investigated by these authors is in aspects of prosody like contrastive stress and intonation, a fact not a difficulty to perceive and distinguish stressed considerable number of studies revealed a consistent and unstressed syllable patterns. Rather, only when involvement of the superior temporal gyrus. However, it is increased memory demands come into play, participants still under debate whether this region is involved left-later- display difficulties to remember stress patterns. More alized or bilaterally. On the one hand, Tong et al. [40] specifically, participants whose native language does not reported significantly stronger left lateralized activation of use stress to distinguish between words (e.g., French) the posterior middle temporal gyrus for the comparison of perform significantly lower in tasks testing memory for stress vs. intonation for Chinese speakers. Furthermore, stress patterns than participants whose language does Ischebeck, Friederici, & Alter [41] compared the proces- contain minimal stress pairs (e.g., Spanish). Crucially, sing of phrase boundaries in natural vs. hummed speech although French participants have particular problems and identified the superior temporal gyrus extending into in remembering stress patterns, their performance in the sulcus to be involved bilaterally in the processing of remembering minimal pairs of pseudowords only distin- natural speech whereas hummed speech revealed only left guishedbyone consonantdid notdifferfromthe per- lateralized activation of this region. On the other hand, formance of Spanish participants [21]. Native speakers when Meyer, Steinhauer, Alter, Friederici, and von Cra- of German have not been tested yet, but they should mon [42] contrasted normal speech (containing vowel obviously belong to the second class of participants, as quality and prosodic information) with degraded speech there are minimal stress pairs like ’Tenor vs. Te’nor. (lacking vowel quality information), they found bilateral Further evidence supporting the autonomy of vowel activation of the posterior superior temporal gyrus even quality and word stress knowledge comes from clinical for the case of degraded speech. Taken together, previous observations on brain-lesioned patients. A classical find- results reported on the processing of linguistic prosody are ing in aphasic word production is that there are more rather heterogeneous as regards possible lateralization. vowel quality errors in unstressed than in stressed sylla- To our knowledge, up to date only one study has bles (e.g., [23-25]). Furthermore, a number of aphasic directly investigated the neuro-anatomical correlates of patients have been described showing a dissociation word stress processing. In an fMRI study, Aleman, For- between spared vowel quality and impaired stress pro- misano, Koppenhagen, Hagoort, de Haan, & Kahn [43] cessing. Typically, their errors have been classified as asked participants to decide whether Dutch bisyllabic regularisation related to the assignment of word stress, i. words were iambic (e.g., salát) or trochaic (e.g., mónat). e. those patients mostly produced the regular or domi- They found areas in the left precentral gyrus, the left nant stress pattern avoiding the irregular or infrequent superior parietal lobule, and in the posterior part of the pattern while preserving syllable and phoneme struc- left superior temporal gyrus extending into the sulcus to tures [26-33]. The reverse pattern, i.e., vowel quality be more active in this stress task compared to a seman- errors with preserved word stress assignment is a stan- tic control condition. However, in their paradigm the dard finding in aphasic patients (e.g., [24]). However, identification of iambic and trochaic stress patterns there is accumulating evidence for an interaction relied on metalinguistic knowledge rather than on nat- between vowel quality and stress processing in German ural language processing. Such a metalinguistic task may speech production. Data from pseudoword reading involvemorethanonlyprosodicprocessing.Most [29,34,35], EEG [36], and patient studies [31] have importantly, contrasting a stress decision task to a shown that the assignment of main stress position in semantic control condition may be not specific enough German words is influenced by their vowel quality. to identify regions involvedinthe processing of word stress (as opposed to phonological processing in gen- Neuronal correlates underlying the processing of eral). In sum, the neural correlates underlying word linguistic prosody stress processing are far from being understood. There is an extensive body of literature on the possible lateralization of processes involved in the comprehension The Present Study of linguistic vs. emotional prosody based on neuro-ima- The current study was conducted to systematically ging methods such as functional magnetic resonance investigate the neuronal correlates underlying word Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 3 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 stress processing. To avoid lexical and semantic con- type. In the non-identical condition the two pseudo- founds on prosodic processing, we conducted an fMRI words either differed in stress or in vowel quality. study on the processing of stress patterns using pseudo- Therefore, activation observed only in the non-identical words. All stimulus items contained only stressable syl- conditions may have most likely reflected stimulus- lables (see [36]) which enabled us to control for vowel related effects, while activations seen in both identical quality in conditions with varying stress patterns. and non-identical pairs might be related to the task manipulation (i.e. particular attention paid to stress or Healthy participants were asked to state whether two vowel differences). auditorily presented bisyllabic pseudowords were the Taken together, the main goal of the present study same or different. In the ‘different’ condition, items dif- fered either in the position of word stress (e.g., Bo’kam was to identify brain regions involved in word stress vs. ‘Bokam) or in the quality of the first vowel. In the processing. Thus, we aimed at directly contrasting stress latter case, vowel quality differences were present both and vowel quality processing. Leaving higher linguistic in stressed and unstressed syllables (e.g., ‘Bekam vs. processing (e.g., lexical or semantic access) aside, our ‘Bokam and Be’kam vs. Bo’kam). Pseudowords only con- study enabled us to evaluate word stress processing in tained two instead of three syllables, as we expected that more detail. Thereby, the research questions motivating the stress pattern of trisyllabic words can already be the current study were twofold: (i) What is the specific inferred after heaving heard the first two syllables activation pattern associated with word stress proces- [14,16,36]. Moreover, the linguistic activity of interest (i. sing? (ii) How are activation patterns influenced by sti- e., the comparison of stress patterns) was contrasted mulus properties (same or different)? (iii) Are there any with a similar phonological activity (i.e., the comparison differences and/or similarities in localization and/or of vowel identity) to allow the investigation of highly intensity of fMRI signal change specifically associated specific activation patterns. In contrast to previous stu- with the metrical processing of different stress patterns dies (e.g., [43]), the word pairs were spoken by two dif- (penultimate vs. final stress)? ferent speakers: one male, one female. This way, in our stimulus-matching task we aimed at investigating the Methods processing of stress patterns at a rather abstract (phono- Participants logical) processing level not allowing for a direct com- Twenty four right-handed native German-speaking parison of phonetic values (see also [21]). Previous fMRI healthy volunteers (12 female; mean age: 28.2 years, SD studies using words and pseudowords revealed that acti- = 7.0 years) participated in this study after having given vations underlying lexical proscessingare notevokedif their written informed consent in accord with the proto- pseudowords are processed in a merely phonological col of the local Ethics Committee of the RWTH Aachen task [44]. Given this finding, the present design should Medical Faculty. be appropriate to investigate phonological processing relatively uncontaminated by lexical or semantic search. Material Building on the above considerations on the proces- A complete overview on all stimulus items used is pro- sing of phonological information the analyses were con- vided in additional file 1. Stimulus material consisted of ducted in two consecutive steps. They started from pairs of bisyllabic pseudowords obeying German phono- examining general activation differences between differ- tactic constraints. All items consisted of an initial open ent tasks addressing stress and vowel quality processing, syllable with a single plosive in onset position followed respectively, to proceed to more specific contrasts inves- by a closed syllable, containing simple consonantal onset tigating the influence of stimulus type (identical and and coda positions, respectively (CV.CVC). Both sylla- non-identical pairs, penultimate and final stress bles were stressable (i.e., excluding schwa-syllables). patterns). Pairs of stimuli were created such that they either dif- Note that all stimuli contained vowels and - given that fered only with respect to word stress (stress condition) they were bisyllabic - they were also marked for stress. or only with respect to vowel quality (vowel condition). Therefore, vowel and stress information were present in Furthermore, each pair consisted of one token spoken both conditions, and presumably participants automati- byafemaleand onetoken spoken byamalevoice, cally processed both types of information irrespective of respectively (see below). condition. Nevertheless, the conditions differ in two cru- In pairs pertaining to the stress condition, two pseu- cial ways: The first difference was task instruction. In dowords containing the same vowels were produced the vowel condition participants were instructed to pay with word-initial and word-final stress. Table 1 gives an attention to vowel information, whereas in the stress overview over phonetic parameters realized by both condition they were told to pay attention to stress infor- speakers to mark stress and Figure 1 exemplifies pho- mation. The second difference was related to stimulus netic information of the stimuli used. The examples of Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 4 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Table 1 Means (standard deviations are given in parentheses) of phonetic parameters duration, fundamental frequency, and intensity of the male and female speakers from a representative sample of 24 quadruples of stimuli (2 speakers × 2 stress patterns). Duration in seconds Fundamental Frequency in Hz Intensity in dB stress pattern PU F PU F PU F syllable type S1 S2 S1 S2 S1 S2 S1 S2 S1 S2 S1 S2 female speaker .37 (.06) .73 (.07) .30 (.06) .80 (.06) 213 (21) 162 (28) 191 (11) 189 (8.5) 80 (1.5) 71 (4.2) 78 (2.1) 77 (2.2) male speaker .47 (.05) .76 (.06) .34 (.06) .83 (.07) 129 (21) 102 (7) 119 (32) 110 (11) 79 (1.1) 73 (2.6) 77 (2.5) 77 (1.4) The values for F0 and intensity are averaged over syllables using PRAAT (Boersma & Weenink; version 5.1.19). Each value is given for each syllable in each stress pattern (numbers indicate the syllable position within a word and stressed syllables are printed in bold). PU: penultimate stress; F: final stress. Figure 1 Spectrograms, pitch contour, and intensity information for both stress patterns and speakers, illustrated with the stimulus quadruple “degis”. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 5 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 spectrograms, pitch, and intensity curves for both speak- speakers. At the same time it shows that stress patterns ers and stress patterns show that phonetic prominence could be clearly distinguished based on a combination was clearly marked in each stress condition. of three relevant phonetic variables (duration, funda- As expected, there was between-speaker variance in mental frequency, and intensity). stress realization. Consequently, Wilcoxon signed-rank In order to control for vowel quality in the stress con- tests for the syllable-wise stressed-unstressed-ratio of dition, all vowels were realized as tensed. Experimental duration, fundamental frequency, and intensity for a pairs contained four different vowels:/u:/,/o:/,/ø:/, and/ representative sample of 24 pairs of tokens revealed sig- e:/. In each pair, the difference in vowel quality invar- iantly affected the nucleus of the first syllable. nificant between-speaker differences for fundamental frequencyand intensityfor thesecondsyllable(Z ≤ For the 4 vowel contrasts differing in one or two fea- -2.342; uncorrected p ≤ .017). We are aware of the pro- tures (difference in 1 feature: between/u:/and/o:/as well blem that non-significant phonetic differences may still as between/e/and/ø:/; in 2 features: between/ø:/and/u:/as influence perception while the mere statistical signifi- well as between/e:/and/o:/, for an overview see Appen- cance of phonetic differences does not grant perceptual dix) 12 item pairs as well as 12 control pairs (with iden- consequences. Nevertheless, we think that presenting tical vowels) were created. Because there was only one tokens by a male and a female speaker should provoke a vowel contrast differing in 3 features (between/e:/and/ strategic shift in auditory processing, disfavoring a u:/), 24 item pairs as well as 24 control pairs (with iden- purely phonetic approach and encouraging a more tical vowels) were created for this vowel contrast. This abstract, phonological type of target comparison. Figure resulted in a total of 24 × 6 = 144 pseudoword pairs. 2 illustrates that, indeed, phonetic means to mark word Another 144 pairs of different items were used in the stress varied considerably both within and between stress condition. These 288 pairs of different items were Figure 2 Combined groups plot of a linear discriminant analysis on syllable-wise ratios (stressed: unstressed) of duration, fundamental frequency, and intensity for 24 representative pseudoword quadruples, revealing three discriminant functions. Function 1 2 2 explained 98.5% of the variance, canonical R = .89, whereas Function 2 explained only 1.1%, canonical R = .08 and Function 3 only .4%, 2 2 canonical R = .03. In combination, all three discriminant functions significantly differentiated the conditions, Λ = .10, c (9) = 207.91, p < .001. After removing Function 1, the remaining functions still differentiated the conditions significantly, Λ = .90, c (4) = 10.17, p = .038. However, Function 3 alone did not differentiate the conditions significantly Λ = .97, c (1) = 2.58, p = .108. Note that Function 1 clearly differentiates between both stress patterns. All three phonetic variables (duration, fundamental frequency, and intensity) loaded on Function 1 (r = .93, r = -.48, and r = -.34, respectively). Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 6 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 opposed to 288 pairs of identical items. Thus, each Each block consisted of 12 trials (6 pairs of identical experimental pseudoword appeared four times in second and 6 pairs of non-identical pseudowords, see Material), position of a pair: (i) in the identical and (ii) in the non- lasting 3700 ms per trial. Participants had to decide, identical stress condition, as well as (iii) in the identical whether the two items of a given pair were phonologi- and (iv) in the non-identical vowel condition. From this cally identical or not by pressing a button with the left overall set blocks were determined consisting of 12 item (non-identical) or the right (identical) hand. The dura- pairs which contained each six pairs of different items tion of the pseudowords ranged between 1000 and 1200 ms. Presentation rate of the trials was kept constant (all stemming from one cell of the experimental design) irrespective of the participants’ response speed. There- and six pairs of identical items. All initial items of a given block had the same stress pattern, such that in fore, each block invariantly lasted 44.4 seconds. Order every trial the decision could only be based on the sec- of trials, blocks, and speakers (male or female) was ond item of a pair. pseudo-randomized such that systematic confounds All stimuli were spoken by two experienced native between condition (e.g. identical vs. non-identical) and speakers of German - one female and one male - and stimulus order were avoided. Each participant was recorded using Amadeus Pro sound editing software exposed to the same sequence of trials. (Version 1.5.1, HairerSoft). In each pair presented, one item was spoken by the female and the other one by the Scanning procedure and imaging data acquisition male speaker - order being counterbalanced across con- For each participant, a high-resolution T1-weighted ana- ditions. Thus, strictly speaking, even ‘identical pairs’ tomical scan was acquired with a 3T Siemens Magne- were not identical on a (phonetic) ‘token’ level, but only tom TrioTim MRI system using the standard head coil on a more abstract (phonological) ‘type’ level of repre- (TR = 19 s, matrix = 256 × 256 mm, 190 slices, voxel sentation. This approach was chosen to increase pho- size = 1 × 1 × 1 mm; FOV = 256 mm, TE = 4.9 ms; flip netic variation and, in consequence, to highlight angle = 25°). Moreover, one functional imaging block processes at the level of abstract phonological represen- sensitive to blood oxygenation level-dependent (BOLD) tations (see also [21]). contrast was recorded for each participant (T2*- weighted echo-planar sequence, TR = 2400 ms; TE = 30 Task and Procedure ms; flip angle = 90°; FOV = 220 mm, 88 × 88 matrix; 42 The experiment was a combined functional magnetic slices, voxel size = 2.5 × 2.5 × 2.5 mm, gap = 10%). resonance imaging (fMRI) and reaction time (RT) study. Trials were presented at a rate of 3700 ms. Participants were lying in the scanner and listening to the word pairs presented auditorily via headphones. Analyses Head movements were prevented by using soft foam Reaction time (RT) analysis was based on correct trials pads. Participants were instructed to respond as quickly only. Furthermore, response latencies faster than 200 ms and accurately as possible avoiding unnecessary move- were not considered and in a second step responses out- ments. To familiarize participants with the task and to side the interval of +/-3 standard deviations around the reduce potential training effects during fMRI data acqui- individual mean were excluded. This resulted in a total sition, all volunteers were given the opportunity to prac- loss of 12.0% of the data. Error rates were arcsine-trans- tice on 16 pairs in a separate room before they entered formed prior to statistical analyses. RT and error rates the scanner. None of these practice items was repeated (ER) were analyzed using a 2 × 2 × 2 within-participant during the fMRI experiment. repeated measures ANOVA comprising the factors iden- The experiment was conducted in a box-car design tity (identical vs. non-identical pairs), phonological comprising 48 blocks. Two seconds prior to the start of manipulation (stress vs. vowel condition), and stress pat- each block one of two specific warning sounds was pre- tern of the second item (penultimate vs. final stress). sented, indicating whether the following block belonged The anatomical scans were normalized and averaged to the stress or to the vowel condition. The assignment in SPM8 http://www.fil.ion.ucl.ac.uk/spm. The fMRI of warning sounds varied over participants (e.g., for half time series was corrected for movement and unwarped of the participants a ringing sound indicated the vowel in SPM8. Images were motion corrected and realigned condition and a smashing sound the word stress condi- to each participant’s first image. Data were normalized tion, whereas for the other half of the participants the into standard stereotaxic MNI space. Images were opposite assignment was chosen). In the off-phase resampled every 2.5 mm using trilinear interpolation between blocks (duration 11.1 seconds) no audio signal and smoothed with a 5 mm FWHM Gaussian kernel to was presented until the onset of the next warning accommodate inter-subject variation in brain anatomy sound. and to increase signal-to-noise ratio in the images. The Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 7 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 data were high-pass filtered (128 s) to remove low-fre- stress led to more errors than second items with penulti- quency signal drifts and corrected for autocorrelation mate stress (13.2% vs. 8.6%), and decisions in the stress assuming an AR(1) process. Brain activity was convolved condition tended to be less accurate than those in the over all experimental trials with the canonical haemody- vowel condition (11.2% vs. 10.5%). While there was no sig- namic response function (HRF). For activation, which nificant three-way interaction, all two-way interactions was evaluated at an uncorrected p-value of < .001, clus- were significant or marginally significant [F(1, 23) ≥ 3.77, ter threshold correction was applied as a threshold lar- p ≤ .064]. Specifically, the increase in error rates from ger than 12 voxels corresponded to a corrected alpha identical to non-identical was particularly pronounced for second items with final compared to penultimate stress level < .05 with our parameters given. Localization of activation peaks was determined using the anatomic (final stress: 10.7% to 15.7%, penultimate stress: 7.7% to automatic labling tool (AAL, http://www.cyceron.fr/web/ 9.4%) and for the stress condition compared to the vowel aal__anatomical_automatic_labeling.html) as well as the condition (stress condition: 8.9% to 13.5%, vowel condi- SPM Anatomy Toolbox [45], available with all published tion: 9.5% to 11.6%). The advantage for penultimate com- cytoarchitectonic maps from http://www2.fz-juelich.de/ pared to final stress was more pronounced in the vowel inm/index.php?index=194). Complex contrasts were condition than in the stress condition (vowel condition: masked inclusively to prevent that e.g. subtraction of a 7.6% to 13.4%, stress condition: 9.5% to 12.9%). strong from a less strong deactivation suggests activa- tion while in fact there is an underactivation. fMRI data Analysis of fMRI data was based on all trials. In a first Results step, a conjunction over the contrasts of stress vs. base- Behavioral data line and vowel quality vs. baseline was conducted to A descriptive overview of the results is provided in show the largely overlapping cortical areas, which were Table 2. The ANOVA of RT data revealed main effects activated in both contrasts. The baseline covered the of identity and phonological manipulation [F(1, 23) ≥ rest periods between the blocks, in which no stimulus 39.50, p ≤ .001], indicating that decisions were faster on material was presented. non-identical pairs than on identical pairs (1100 ms vs. Conjunction over the contrasts stress vs. baseline and 1173 ms) and faster for vowel contrasts than for stress vowel quality vs. baseline (see Figure 3, Table 3) contrasts (1081 ms vs. 1192 ms). There was no main The conjunction revealed large common clusters of acti- effect of stress pattern [F(1, 23) = 2.77, p = .109]. More- vated voxels, bilaterally, in the superior temporal gyri over, there was a significant interaction of identity and (BA 22), the insula, the putamen as well as the cerebel- phonological manipulation [F(1, 23) ≥ 26.45, p ≤ .001], lum (p < .05, FWE-corrected, k = 12 voxels). meaning that the disadvantage for identical as compared In a second step, contrasts between vowel quality and to non-identical pairs was more pronounced in the stress processing were calculated to evaluate the regions vowel condition (1140 ms vs. 1021 ms) than in the found to be specifically active in word stress processing stress condition (1206 ms vs. 1179 ms). None of the by Aleman et al. [43]. other two- or three-way interactions reached statistical Stress vs. vowel quality (see Figure 4A, Table 3) significance. Word stress was contrasted to vowel quality at an The ANOVA of arcsine-transformed error-rates yielded uncorrected voxelwise p < .001 and a cluster size of 12 significant main effects of both identity and stress pattern voxels. This comparison indicated activation in the left [F(1, 23) ≥ 7.52, p ≤ .012], while the main effect of phono- inferior frontal gyrus [Brodmann Area (BA) 47], the logical manipulation only approached the conventional right superior temporal gyrus (STG, BA 22), the right level of significance [F(1, 23) ≥ 3.12, p = .091]. Specifically, inferior frontal gyrus (BA 44, BA 45), the right middle non-identical trials were somewhat more error prone than temporal gyrus (BA 21), the left fusiform gyrus (BA 19), identical trials (12.6% vs. 9.2%), second items with final and the right supplementary motor area (BA 6). Please Table 2 Overview of behavioral results. identical non-identical Penultimate Final Penultimate Final phonological manipulation stress 1215.7 (203.9) 1195.5 (204.5) 1177.9 (218.2) 1179.3 (206.1) 9.5 (4.4) 8.3 (5.8) 9.5 (5.4) 17.6 (6.7) vowel 1147.9 (237.6) 1131.8 (220.6) 1022.5 (220.0) 1020.2 (216.5) 5.8 (5.5) 13.1 (5.3) 9.4 (5.3) 13.8 (6.1) Mean RT (SD) in ms given in the first line and mean error rates (SD) in % given in the second line of each cell. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 8 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Figure 3 Conjunction over the contrasts stress - baseline and vowel quality - baseline: widespread activation in the bilateral superior temporal cortices (FWE-corrected voxelwise p < .05, cluster size k = 12). Table 3 Comparison of stress- and vowel quality-related activation. Contrast Brain region (BA) TC (x, y, z) Cluster size z score Conjunction* LH superior temporal gyrus (BA22) -48 -20 8 264 6.61 Stress - baseline and RH superior temporal gyrus (BA 22) 63 -25 0 236 6.54 Vowel quality vs. Baseline RH insula 30 23 5 20 5.87 LH insula -33 20 3 27 5.84 LH putamen -25 3 -3 60 5.71 RH putamen 25 3 5 17 5.38 LH supplementary motor area (BA 6) 0 0 65 114 6.14 RH cerebellum 30 -65 -25 24 6.17 LH cerebellum -28 -68 -25 21 6.04 Stress - vowel quality LH inferior frontal gyrus (BA 47) -48 18 -8 17 4.36 RH inferior frontal gyrus (BA 44) 53 15 20 36 4.33 RH inferior frontal gyrus (BA 45) 50 23 -5 15 4.06 RH superior temporal gyrus (BA 22) 58 -35 18 14 3.73 RH middle temporal gyrus (BA 21) 50 -23 -5 25 3.80 RH superior frontal gyrus (BA 6) 8 0 68 39 4.89 Non-identical pairs: RH middle temporal gyrus (BA 21) 55 -43 8 93 5.41 Stress - vowel quality LH inferior temporal gyrus (BA 21) -43 -3 -20 23 4.47 LH inferior frontal gyrus (BA 47) -48 18 -8 36 4.66 RH inferior frontal gyrus (BA 45) 53 23 -5 43 4.52 RH inferior frontal gyrus (BA 44) 50 13 20 60 4.02 RH superior temporal gyrus (BA 22) 55 -25 -3 26 3.81 LH superior temporal gyrus (BA 22) -50 -40 20 25 3.71 LH insula -33 20 8 34 4.32 RH intraparietal sulcus (BA 7) 43 -45 60 18 3.91 RH intraparietal sulcus (BA 7) 43 -40 50 14 3.64 RH postcentral gyrus (BA 1) 48 -23 50 54 4.10 RH superior frontal gyrus (BA 6) 8 3 68 50 5.34 LH middle frontal gyrus (BA 6) -40 0 55 14 3.95 RH middle frontal gyrus (BA 8) 45 8 40 19 3.71 LH cingulate gyrus (BA 32) -5 18 40 87 5.07 Identical pairs: LH intraparietal sulcus (BA 7) -35 -65 43 16 3.74 Vowel quality - stress *p < .05, FWE-corrected; p < .001, uncorrected; cluster size = 12 voxels; masks were created at uncorrected p < .05; MNI: Montreal Neurological Institute coordinates. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 9 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Figure 4 Comparisons of stress and vowel quality. A: Stress - vowel quality at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively with stress: Activation specific for prosodic processing in the right superior temporal gyri as well as in Broca’s area. B: Stress vs. vowel quality in non-identical pairs at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively with stress in non- identical pairs reveals a widespread right-lateralized temporo-parieto-frontal network. C: Vowel quality vs. stress in identical pairs at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively vowel quality in identical pairs: Activation of the left intraparietal cortex. note that the activation in the STG is not lateralized on uncorrected, k = 12 voxels), trials with non-identical the right hemisphere for the main effect of stress vs. pairs yielded a large network of activation when com- vowel quality. Flipping another sample of the contrast paring stress to vowel quality: images along the y-axis (from left to right orientation) Stress vs. vowel quality in non-identical pairs (Figure 4B, and calculating a paired t-test between the left STG in Table 3) the original data set and the left STG in the flipped data Contrasting stress and vowel quality in non-identical set (formerly the right STG) revealed that even at a very word pairs (p < .001, uncorrected, k =12voxels) liberal p-value < .05 there was no significant difference revealed activation in the bilateral superior temporal of activation in the STG between the two hemispheres. gyrus (BA 22), the bilateral middle temporal gyrus (BA More in-depth examination of this main effect 21), the left inferior frontal gyrus (BA 45), and the right revealed that the activation observed was mainly driven inferior frontal gyrus (BA 45 and 44). Further clusters of by differences between stress and vowel quality in non- activated voxels were found in the left insula, the right identical trials: Whereas in trials with identical pairs intraparietal sulcus (BA 7), the right superior parietal the comparison of stress with vowel quality revealed lobule (BA 7), the right postcentral gyrus (BA 2), the no activated voxels at the threshold chosen (p < .001, right supplementary motor area (BA 6), the left Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 10 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 precentral gyrus (BA 6), the right middle frontal gyri processing. Neural correlates underlying vowel quality (BA 8), and the left superior medial gyrus (BA 32). processing could best be identified comparing identical Inspection of the inverse contrast (vowel quality vs. pairs. An increase of the fMRI signal with vowel qual- stress) revealed no clusters of activated voxels at the ity processing in the difficult condition was found in threshold chosen (p < .001, uncorrected, k =12voxels). the left intraparietal cortex. However, closer inspection of the data indicated activation In a second step, we aimed at comparing activation when identical pairs were presented. In contrast, there was patterns for targets with different stress patterns. no activation observed for non-identical pairs. Penultimate vs. final stress in the stress condition (Figure Vowel quality vs. stress in identical pairs 5A, Table 4) In identical pairs, vowel quality was contrasted with Stronger activation was found in a large bilateral tem- stress at an uncorrected voxelwise p <.001and cluster poro-frontal network (FWE-corrected at p < .05, k =12 size of 12 voxels (Figure 4C, Table 3). Activated voxels voxels). The network comprised the bilateral superior were observed in the left intraparietal sulcus (BA 7). frontal gyri (BA 22), the bilateral putamen, the bilateral Taken together, there was a temporo-frontal activa- insula, the left supplementary motor area (BA 6) as well tion pattern specifically associated with word stress as the bilateral cerebellum. Figure 5 Comparisons of penultimate and final stress. A: Main effect of penultimate vs. final stress (FWE-corrected, cluster size k = 12 voxels, masked inclusively): Activation of a bilateral temporo-frontal network. B: Penultimate vs. final stress when comparing identical pairs according their stress at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively: Activation of the bilateral superior temporal gyri. C: Penultimate vs. final stress when comparing non-identical pairs regarding their stress at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively: Left-lateralized activation of the superior temporal gyrus. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 11 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Table 4 Penultimate and final stress: Main effect and contrasts depending on the identity of the word pair presented. Contrast Brain region (BA) MNI (x, y, z) Cluster size z value Stress pattern: LH superior temporal gyrus (BA 22) -48 -20 8 261 6.49 Penultimate - final stress* RH superior temporal gyrus (BA 22) 63 -25 0 231 6.36 LH putamen -23 5 3 72 6.13 RH putamen 25 0 5 20 5.67 RH insula lobe 30 25 3 33 5.79 LH insula lobe -30 20 3 23 5.66 LH supplementary motor area (BA 6) -3 0 65 86 6.31 RH cerebellum 28 -63 -25 32 6.55 LH cerebellum -28 -68 -25 50 6.02 LH cerebellum 3 -63 -25 18 5.87 Stress in identical pairs: LH superior temporal gyrus (BA 22) -55 -25 8 192 4.73 Penultimate - final RH superior temporal gyrus (BA 22) 65 -20 3 208 5.80 Stress in non-identical pairs: LH superior temporal gyrus (BA 22) -50 -18 5 23 4.10 Penultimate - final * p < .05, FWE-corrected; p < .001, uncorrected; cluster size = 12 voxels; masks were created at uncorrected p < .05; MNI: Montreal Neurological Institute coordinates. Penultimate vs. final stress when comparing identical pairs processing involved a network of bilateral fronto-tem- in the stress condition (Figure 5B, Table 4) poral activation, resembling patterns previously Activation specific for penultimate stress in identical described to subserve auditory processing of bisyllabic pairs was present in large clusters in both superior tem- pseudo-words [46]. However, while the general contrast poral gyri (BA 22) extending along the superior tem- between stress and vowel conditions showed only right- poral sulcus (uncorrected p < .001, k = 12 voxels). hemispheric activation of the superior temporal gyrus, Penultimate vs. final stress when comparing non-identical the more fine-grained analysis over non-identical pseu- pairs in the stress condition (Figure 5C, Table 4) doword pairs showed that a bilateral fronto-temporal Comparing conditions with penultimate stress to condi- network was specifically associated with word stress pro- tions with final stress in non-identical word pairs, only cessing. In particular, we were able to identify task-spe- activation in the left superior temporal gyrus was cific differences of stress processing in the superior observed(BA 22)atanuncorrected p <.001 and voxel temporal gyrus and the superior temporal sulcus. size of k = 12 voxels. Finally, our data suggested higher cognitive demands for For the opposite comparison (final vs. penultimate the processing of penultimate stress compared to final stress) no activation was observed at the threshold cho- stress in the experimental design chosen. sen neither for identical nor for non-identical pairs. Taken together, in the superior temporal gyrus as well Stress vs. vowel quality processing as the superior temporal sulcus differential effects of Main purpose of the present study was to evaluate the stress processing were found dependent on both the fac- neural correlates of phonological word stress proces- tors identity (non-identical vs. identical auditory word sing by comparing these correlates to those related to pairs) and stress pattern (penultimate stress vs. final a similar task - vowel qualityprocessing. Thediffer- stress). ence of stress and vowel quality processing in general was corroborated by the effect of phonological manip- Discussion ulation which was significant in the ANOVA on RT The current study aimed at investigating the processing and marginally significant in the ANOVA on ER, of word stress information. For this purpose, behavioral meaning that reaction times were faster for vowel con- and neuro-imaging data of word stress and vowel quality trasts than for stress contrasts. In line with previous processing were contrasted directly. In general, the studies investigating activation related to prosodic sen- neural networks associated with word stress and vowel tence processing, the comparison of stress and vowel quality processing were observed to be largely overlap- quality processing revealed a network of activation ping. In particular, the conjunction of stress and vowel comprising the right superior temporal gyrus. This tasks revealed that both aspects of phonological brain region has been identified repeatedly to be Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 12 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 associated with prosodic processing (e.g., [38,40-42]). non-identical pairs). For stress activation in the context Moreover, activation was evidenced in Broca’sarea, of non-identical pairs, a widespread pattern of temporo- which has also been found to be associated with lin- frontal activation was observed, while the processing of guistic aspects of prosody [39]. Finally, increased occi- vowel quality information vs. stress processing in the pital activation extending into the left fusiform gyrus context of identical pseudoword pairs seems to be asso- was observed, where the visual identification area for ciated with the intraparietal cortex as already reported by Meyer et al. [42]. However, as already suggested by word formsissupposedtobelocated [47].Thisactiva- the behavioral data, it is important to note that both the tion may indicate that participants also searched for effect of word stress and the effect of vowel quality associations with familiar word forms and their stress patterns whenever they had to process stress informa- information have to be evaluated in the context of the tion in pseudowords (see also [48]). stimulus type (identical or different). The contrasts Comparing vowel quality and stress processing between vowel quality and stress processing seem to revealed no super-threshold activation in the whole reflect qualitative differences rather than being only brain. However, as the behavioral analysis revealed a related to different degrees of difficulty. strong impact of the factor identity, a more fine-grained analysis which takes this factor into account seems to be Stimulus specific effects on prosodic processing more adequate. Indeed, breaking down the task-specific The present study revealed that the type of stimulus pair interaction between stress/vowel quality processing and (identical vs. different) influenced stress processing. The the factor identity into its constituting conditions effect of identity was significant in the ANOVA on both revealed that the activation observed in the comparison RT and ER, with identical pairs being classified more of stress to vowel quality processing was mainly driven slowly than non-identical pairs. Moreover, the present by trials with non-identical pairs. In such a comparison, neuro-imaging data clearly indicated the importance of not only the same areas were observed which were stimulus-specific effects for the above described network found to be active in the main effect of stress vs. vowel of activation for stress processing: Whenever a pair of quality processing, but also the superior temporal gyrus pseudowords with non-identical stress patterns had to was activated bilaterally. In addition, the right intrapar- be decided on, a large bilateral network in the superior ietal cortex was activated. This cortex site has been sug- temporal gyrus was activated, which has repeatedly been gested to underlie the processing of proximity relations identified to be vitally involved in processing prosodic [49] as well as mental imagery (e.g., [50]). Thus, the information (e.g., [38,41]). However, when the stress non-identical stress patterns may have been evaluated pattern in the pair of pseudowords was identical, no with respect to the relation and extent of their differ- activation was observed. ences; moreover, participants may have tried to intern- Mean RTs in the present study were faster for non- ally memorize and compare the stress patterns they had identical than for identical stimulus pairs, whereas in been presented with. It should be pointed out, that some behavioral experiments reported in the literature stress is an inherently relational property, i.e., its recog- involving same-different decisions on vowel-consonant nition requires the comparison of phonetic measures (e. syllables, faster mean RTs were obtained for the proces- g., duration, pitch, and intensity) between stressed and sing of same syllables compared to different ones (e.g., unstressed syllables and this relation may even be differ- [51,52]). However, the difference between “same” and ent within and between different speakers as in our task “different” responses is subject to specific task demands (see Figures 1 and 2 and Table 1). (e.g., [51-53]. In the present investigation even in the In contrast, for identical pairs no activation was “same” condition items were actually not identical but observed for stress vs. vowel quality processing. How- realized by different speakers. Listeners therefore could ever, the opposite contrast showed that within identical not rely on superficial phonetic deviations in the “differ- pairs, vowel quality compared to stress processing was ent” condition but had to derive abstract representations related to stronger intraparietal activation in the left to perform the evaluation task. Since the phonetic devia- hemisphere.Thisisinlinewithpreviousfindingscom- tions in the “same” condition were more fine-grained paring vowel quality (flattened without prosody) and compared to the “different” condition, the latencies for natural speech [42] as these data already suggested that “same"-decisions were higher. However, the asymmetri- the left intraparietal cortex may be associated with cal neurophysiological effect of the matching task on vowel quality processing. stress vs. vowel processing indicates qualitatively differ- To sum up, our imaging data indicate different activa- ent demands on positive or negative responses. tion patterns for vowel quality and stress processing The finding of stress processing being influenced by when contrasting these two aspects of phonological pro- stimulus specific effects is relevant regarding the possi- cessing directly in different stimulus context (identical/ ble lateralization of processes which subserve the Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 13 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 comprehension of linguistic prosody. As already out- Regarding the main effect of stress patterns, our fMRI lined above, a consistent involvement of the superior data yielded different results than the behavioral data. In temporal gyrus has been shown frequently for the pro- particular, no activation was found for final stress as cessing of linguistic aspects of sentence prosody like compared to penultimate stress. However, the inverse contrastive stress and intonation [40]. However, it still contrast revealed a bihemispheric activation of the remains debatable whether this region is involved only superior temporal gyrus, which has been repeatedly in the left hemisphere or rather bilaterally. On the one reported to be associated with prosodic processing (e.g., hand, a considerable number of studies reported signif- [40,41]). This finding suggests that the processing of icantly stronger left lateralized activation of the poster- penultimate stress may have involved a more detailed ior middle temporal gyrus for processing stress auditory analysis than the processing of final stress. On information (e.g., [38,40]). On the other hand, bilateral a phonetic level of explanation, this may have been due activation of the posterior superior temporal gyrus has to the different perceptual saliency of both patterns. On also been reported repeatedly for processing prosodic a phonological level, this activation pattern may indicate information in natural (e.g., [41]) and degraded [42] that penultimate stress has not a general default status speech. in German as already argued by Janßen, Domahs, and The current study may add to the understanding of colleagues [29,31,36]. This is a challenge to approaches such apparently heterogeneous findings. When only assuming that given the fact that penultimate stress (or comparing main effects such as the main effect of stress in bisyllabic words: initial stress) is the most frequent to the main effect of vowel quality processing, only German stress pattern it forms some kind of default lateralized activation of the superior temporal gyrus was stress pattern which- in contrast to final stress - has not found. However, as outlined above, our behavioral data to be lexically specified (e.g., [55,56]). However, Janßen indicated that the identity or non-identity of stress pat- [29], Janßen & Domahs [31] and Domahs et al. [36] terns may be relevant. Indeed, when the processing of report behavioral and electrophysiological evidence that stress information was evaluated in the context of the the “regularity” of word stress is strongly influenced by stimulus type (identical or non-identical), bilateral acti- the structure of the final and penultimate syllable vation of the superior temporal gyrus was found, which [30,36,57,58]. In particular, penultimate stress occurs seems to correspond well to the findings of Ischebeck et predominantly in words with an open final syllable (e.g., al. [41] as well as Meyer and colleagues [42]. In contrast, Pánda, [panda]), but not in words with a closed final the processing of identical stress patterns as well as a syllable (e.g. Spinát, [spinach]), casting doubts on a comparable contrast in the vowel task within non-iden- structure-independent default status of penultimate tical items did not reveal such an activation pattern. stress. Since the pseudowords presented consist of an Taken together, diverging previous results regarding open penultimateand aclosed final syllable, the higher the lateralization of prosodic processing may have possi- processing costs for items with penultimate stress may bly been due to stimulus- or task-specific properties (see reflect the fact that this pattern is not preferred in also [54] for task specific effects on neural activation words with a closed final syllable(seealso[34]). Again, patterns in two language groups requiring different amorefine-grainedanalysisofthe imagingdata efforts in the processing of stress properties). Taking revealed that the factor identity differentially influenced these properties into account, our data suggest that the results. Activation observed for identical stress pat- whenever more fine-grained decisions have to be made terns was found bilaterally, whereas the processing of at an increasingly abstract level, bilateral activation of non-identical stress patterns was only associated with the superior temporal gyri is needed. This view fits well left-lateralized activation. This distribution of activation with previous observations on bilateral processing of maybeexplained by thefollowing arguments. Most stress comparison [41,42]. probably, it may have been easier to decide that two stress patterns are different than to decide that two Effects of stress patterns tokens of the same stress pattern, produced by different The present study also revealed different behavioural speakers, are indeed identical at a phonological level. and imaging results for different stress patterns. The This assumption is supported by our behavioural data effect of stress pattern was significant in the ANOVA showing that responses for non-identical pairs were sig- on error rates, with final stress in the second item nificantly faster than for identical pairs. However, it is being more difficult to be processed than penultimate important to consider that the difference in the neuro- stress. Stimulus-specific effects again influenced perfor- functional data is restricted to the superior temporal gyrus, while it does not seem to involve areas associated mance as the increase in error rates was particularly with generally higher levels of working memory or more pronounced for pairs with non-identical stress attentional load (e.g., the dorsolateral prefrontal cortex patterns. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 14 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 and/or the intraparietal cortex) where activation would fine-grained analysis showed that the activation be expected if the different performance on identical observed for stress vs. vowel quality processing was in and non-identical stimuli is purely ascribed to higher fact driven by the comparison of stress and vowel memory load. Thus, the greater activation for identical quality processing in trials with non-identical pairs, pairs seems to be rather specific to the processing of while for identical pairs no activation was observed for stress information itself than to reflect more general stress vs. vowel quality processing. On the contrary, processes associated with a higher level of working within identical pairs stronger activation was observed memory and/or attentional demands. Thereby, the for vowel quality as compared to stress processing. To sum up, our imaging data indicate that both the effect increased activation may reflect most likely the extended auditory evaluation of the more fine-grained phonetic of word stress as well as the effect of vowel quality differences in pairs with identical stress patterns. information have to be evaluated in the context of the Taken together, diverging previous results regarding stimulus type (identical or non-identical), as was the lateralization of prosodic evaluation may have pos- already suggested by the behavioral effect of stimulus sibly been due to stimulus- or task-specific properties. type (identical or non-identical word pairs). Thereby, Taking these properties into account, our data support the differences between vowel quality and stress pro- the view that the left posterior superior temporal gyrus cessingseemtobequalitativeratherthanonlybeing is a kind of basic system mainly involved in the evalua- related to different degrees of difficulty. This interpre- tion of prosodic properties as outlined in part of the tation is further supported by the facts that areas typi- previous literature (e.g., [40]). However, once more cally associated with higher cognitive demands (e.g., fine-grained decisions have to be made at an increas- left dorsolateral prefrontal gyrus, anterior cingulate or ingly abstract level, the right superior temporal gyrus intraparietal cortices) were not observed for the com- seems to be called for assistance (e.g., [54]). This view parison of stress and vowel quality processing. Quite fits well with previous observations of bilateral proces- the contrary, in the identity condition, the intraparietal sing related to rather abstract stress comparison, e.g., cortex was in fact significantly stronger involved in the in degraded speech [42]. Thus, the present finding processing of vowel quality than of stress information. again underlines the impact of task and stimulus-speci- This finding leads us to the question why response fic effects. latencies were generally longer when evaluating stress patterns. In the literature, there is evidence that vowel Evaluation and perspectives quality information can be exploited earlier than stress We believe that the current study is a first step towards information due to coarticulation [16-19]. Neverthe- less, in order to explain the extent of these differences, a more comprehensive understanding of the underlying processes subserving word stress processing. However, it may be helpful to consider that our design enabled there are still a lot more steps to go. Therefore, in the participants to decide on the vowel quality structure as remainder of this Discussion some points requiring soon as the first syllable of the second item was further investigation will be addressed. encountered. In contrast, for decisions on stress infor- Consider firstthatthe responsestoword stresseva- mation, the second syllable of the second item had to luation were significantly slower and tended to be more be perceived before a confident judgement was possi- error prone than the evaluation of vowel quality infor- ble. This explanation may account for a general differ- mation. The question may arise whether the stress con- ence of 100-200 ms in response latencies. Indeed, dition was generally more difficult than the vowel inspection of the behavioral data revealed that all quality condition - a methodological artefact potentially stress conditions were evaluated systematically slower fateful for the validity of our data and the conclusions than the vowel quality conditions (see Table 2). Taking we have drawn. all these arguments into account, the current paradigm However, support for the validity of our data comes seems to be a valid approach to further investigate the from several different aspects. First, the above men- neural correlates of processing word stress and vowel tioned RT-findings neglect that the pattern observed is quality information. driven by a speed-accuracy trade-off as the slower con- Consider next the effect of stress patterns. The com- dition also tended to be less error prone. Second, parison between stress patterns revealed a bihemispheric inspection of the imaging data provides helpful activation of the superior temporal gyrus for penulti- insights. Indeed, the comparison of stress with vowel mate stress compared to final stress. This finding sug- gests that the processing of penultimate stress produced quality processing revealed a bilateral network of acti- higher costs than the processing of final stress. At which vation, whereas the contrast of vowel quality vs. stress level of processing may pseudowords with penultimate showed no voxel in the whole brain activated signifi- cantly stronger at the threshold used. However, a more stress have been harder to process than pseudowords Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 15 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 with final stress? The activation differences in the iden- Summary and Conclusion tity condition may just reflect higher efforts at the level The current study addressed two main research issues: of phonetic analysis. Unfortunately, knowledge about First, we were interested in the activation pattern asso- the perceptual consequences of specific phonetic fea- ciated with stress processing. By controlling stimulus tures in word stress processing is still lacking. In conse- material for vowel quality in conditions with varying quence, a perceptual account of the activation stress patterns and by varying phonetic realizations we differences cannot be excluded with the data at hand. intended to provoke a matching of stress patterns on a At a more abstract phonological level of processing it rather abstract, phonological level. may be speculated that penultimate stress is generally We observed a fronto-temporal network basically more difficult to be processed or represented than final comprising the right superior temporal gyrus extending stress. However, such an interpretation would not be into the sulci as well as the inferior frontal gyri, bilater- warranted as penultimate stress is the statistically predo- ally, to be specifically associated with stress processing. minant pattern in German words and, thus, is not likely However, when the contrast was evaluated more specifi- to evoke higher costs in processing than the less fre- cally in the context of the stimulus type (identical/non- quent pattern. However, note that all stimuli presented identical pairs), the data became clearer and revealed contained a heavy final syllable and therefore do not fit that stress was processed in the bilateral superior tem- the typical pattern of German words with penultimate poral gyri and sulci in the more difficult non-identical stress, namely bisyllabic words with a light or reduced trials. For vowel quality processing, our data emphasize final syllable. Thus, the higher processing costs may a substantial contribution of the left intraparietal cortex. support quantity sensitive approaches on German stress Second, our data suggest that higher cognitive assignment [30,36,57,58] and show that for words with a demands were needed for processing penultimate com- heavy final syllable penultimate stress is not the pared to final stress possibly suggesting that penultimate unmarked pattern (see also [34]). Further neuro-func- stress has not a default status in German. Thereby, our tional examinations with varying syllable structures results support the view that the left superior temporal should bring more light into this debate. gyrus represents a kind of basic system underlying stress Even if we assume that a difference in phonetic para- processing to which additional structures including the meters may have affected our results, it is important to homologous cortex site are recruited with increasing note that this is clearly not the case for our behavioral difficulty. data.First,there wasnomaineffect of stress patternin the ANOVA on RT. The analysis of error rates even pro- Additional material vided evidence in favour of the assumption that final stress may have been more difficult to be processed than Additional file 1: Lists of pseudowords used. penultimate stress as final stress in the second word of a pair was associated with significantly more errors than penultimate stress. Taking into account the phonetic Acknowledgements parameters we do not claim that the differential imaging The research and the preparation of this article were supported by a START- programme grant (AZ 37/07) of the Faculty of Medicine at the RWTH effects found for specific stress patterns (penultimate vs. Aachen University supporting M. Grande, a DFG priority programme grant final stress) in our study can be generalized to studies (DO 1433/1-1) supporting F. Domahs, as well as a DFG priority programme using other stimuli, presentation formats or tasks. How- grant (WI 853/7-2) supporting U. Domahs. We wish to thank Katja Halm, Stefanie Jung, and Timo Roettger for their help in stimulus preparation and ever, the present study definitely shows that these specific data collection as well as Ralph Schnitker and Georg Eder from the stress patterns may be processed differently and should “Functional Imaging Unit” of the Interdisciplinary Centre for Clinical Research be target of further investigations, for instance, with IZKF ‘BioMAT.’ for their precious help in conducting this study. more precisely controlled phonetic parameters and dif- Author details ferent syllable structures. Imaging studies on different Department of Neurology, Section Neuropsychology, University Hospital, stress patterns may be a crucial source of evidence feed- RWTH Aachen University, Aachen, Germany. Interdisciplinary Center for Clinical Research Aachen, University Hospital, RWTH Aachen University, ing phonological theories on stress systems. Aachen, Germany. Institute of Psychology, Eberhard Karls University, Taken together, even though there are still a number Tuebingen, Germany. Institute of Germanic Linguistics, University of of questions to be answered, the present results provide Marburg, Germany. Department of Neurology, Section Clinical Cognition Research, University Hospital, RWTH Aachen University, Aachen, Germany. first evidence not only on the neural correlates subser- ving stress processing, but also for the impact of stimu- Authors’ contributions lus-dependent effects (e.g., whether the stress/vowel FD, UD, EK, and MG conceived the study. All authors participated in its design. EK performed data collection, processing and statistical analyses. EK, quality decision has to be made within identical or non- FD, and UD drafted the manuscript. All authors contributed to the identical stimuli). Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 16 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 interpretation of the data. All authors read and approved the final 26. Cappa SF, Nespor M, Ielasi W, Miozzo A: The representation of stress: manuscript. evidence from an aphasic patient. Cognition 1997, 65:1-13. 27. Colombo L, Brivio C, Benaglio I, Siri S, Cappa SF: Alzheimer patients’ ability Competing interests to read words with irregular stress. Cortex 2000, 36:703-714. The authors declare that they have no competing interests. 28. Galante E, Tralli A, Zuffi M, Avanzi S: Primary progressive aphasia: a patient with stress assignment impairment in reading aloud. Neurol Sci Received: 1 December 2010 Accepted: 16 May 2011 2000, 21:39-48. Published: 16 May 2011 29. Janssen U: Stress assignment in German patients with surface dyslexia. Brain Lang 2003, 87:114-115. 30. Janssen U: Untersuchungen zum Wortakzent im Deutschen und References Niederländischen. PhD thesis, University of Düsseldorf; 2003 [http:// 1. McClelland JL, Elman JL: The TRACE model of speech perception. Cognit docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=2911]. Psychol 1986, 18:1-86. 31. Janssen U, Domahs F: Going on with optimised feet: Evidence for the 2. Marslen-Wilson WD: Functional parallelism in spoken word-recognition. interaction between segmental and metrical structure in phonological Cognition 1987, 25:71-102. encoding from a case of primary progressive aphasia. Aphasiology 2008, 3. Luce PA, Pisoni DB: Recognizing spoken words: the neighborhood 22:1157-1175. activation model. Ear Hear 1998, 19:1-36. 32. Laganaro M, Vacheresse F, Frauenfelder UH: Selective impairment of 4. Rauschecker JP: Cortical processing of complex sounds. Curr Opin lexical stress assignment in an Italian-speaking aphasic patient. Brain Neurobiol 1998, 8:516-521. Lang 2002, 81:601-609. 5. Scott SK, Blank CC, Rosen S, Wise RJS: Identification of a pathway for 33. Miceli G, Caramazza A: The Assignment of Word Stress in Oral Reading: intelligible speech in the left temporal lobe. Brain 2000, 123:2400-2406. Evidence from a Case of Acquired Dyslexia. Cogn Neuropsychol 1993, 6. Narain C, Scott SK, Wise RJ, Rosen S, Leff A, Iversen SD, Matthews PM: 10:273-296. Defining a left-lateralized response specific to intelligible speech using 34. Roettger T, Domahs U, Grande M, Domahs F: Structural factors affecting fMRI. Cereb Cortex 2003, 13:1362-1368. the assignment of word stress in German.. 7. Liebenthal E, Binder JR, Spitzer SM, Possing ET, Medler DA: Neural 35. Tappeiner E, Domahs U, Domahs F: Wortakzent im Sprachkontakt substrates of phonemic perception. Cereb Cortex 2005, 15:1621-1631. Deutsch-Italienisch. [Word stress in German-Italian Language Contact.]. 8. Obleser J, Zimmermann J, Van Meter J, Rauschecker JP: Multiple stages of Zeitschrift für Dialektologie und Linguistik 2007, 74:266-291. auditory speech perception reflected in event-related fMRI. Cereb Cortex 36. Domahs U, Wiese R, Bornkessel-Schlesewsky I, Schlesewsky M: The processing of German word stress: evidence for the prosodic hierarchy. 9. Hickok G, Poeppel D: The cortical organization of speech processing. Phonology 2008, 25:1-36. Nature Reviews 2007, 8:393-402. 37. Kotz SA, Meyer M, Alter K, Besson M, von Cramon DY, Friedericia AD: On 10. Specht K, Huber W, Willmes K, Shah J, Jäncke L: Tracing the ventral stream the lateralization of emotional prosody: An event-related functional MR for auditory speech processing in the temporal lobe by using a investigation. Brain Lang 2003, 86:366-376. combined time series and independent component analysis. Neurosci 38. Dogil G: Understanding Prosody.Edited by: Rickheit H, Herrmann T, Lett 2008, 442:180-185. Deutsch W. Psycholinguistik: Ein Internationales Handbuch; 2003:544-565, 11. Cutler A: Speakers’ conceptions of the function of prosody. In Models and Berlin: de Gruyter. measurements. Edited by: Cutler A, Lad Prosody DR. Heidelberg: Springer; 39. Wildgruber D, Hertrich I, Riecker A, Erb M, Anders S, Grodd W, 1983:79-92. Ackermann H: Distinct frontal regions subserve evaluation of linguistic 12. Cutler A, van Donselaar W: Voornaam is not (really) a homophone: Lexical and emotional aspects of speech intonation. Cerebr Cortex 2004, prosody and lexical access in Dutch. Lang Speech 2001, 44:171-195. 14(12):1384-1389. 13. Boecker KBE, Bastiaansen MCM, Vroomen J, Brunia CHM, De Gelder B: An 40. Tong Y, Gandour J, Talavage T, Wong D, Dzemidzic M, Xu Y, Li X, Lowed M: ERP correlate of metrical stress in spoken word recognition. Neural circuitry underlying sentence-level linguistic prosody. NeuroImage Psychophysiology 1999, 36:706-720. 2005, 28:417-428. 14. van Heuven VJ: Effects of stress and accent on the human recognition of 41. Ischebeck A, Friederici AD, Alter K: Processing prosodic boundaries in word fragments in spoken context: Gating and shadowing. 7th FASE natural and hummed speech. An fMRI study. Cerebr Cortex 2008, Symposium, Edinburgh 1988, 811-818. 18:541-552. 15. Friedrich CK, Kotz SA, Friederici AD, Alter K: Pitch modulates lexical 42. Meyer M, Steinhauer K, Alter K, Friederici AD, von Cramon DY: Brain activity identification in spoken word recognition: ERP and behavioral evidence. varies with modulation of dynamic pitch variance in sentence melody. Brain Res Cogn Brain Res 2004, 300-308. Brain Lang 2004, 89:277-289. 16. van Donselaar W, Koster M, Cutler A: Exploring the role of lexical stress in 43. Aleman A, Formisano E, Koppenhagen H, Hagoort P, de Haan EHF, Kahn RS: lexical recognition. Q J Exp Psychol A 2005, 58:251-273. The Functional Neuroanatomy of Metrical Stress Evaluation of Perceived 17. Cutler A, Chen HC: Lexical tone in Cantonese spoken-word processing. and Imagined Spoken Words. Cereb Cortex 2005, 15:221-228. Percept Psychophys 1997, 59:165-179. 44. Heim S, Alter K, Ischebeck AK, Amunts K, Eickhoff SB, Mohlberg H, Zilles K, 18. Robinson K, Patterson RD: The stimulus duration required to identify von Cramon DY, Friederici AD: The role of the left Brodmann’s areas 44 vowels, their octave, and their pitch chroma. J Acoust Soc Am 1995, and 45 in reading words and pseudowords. Brain Res Cogn Brain Res 98:1858-1865. 2005, 982-993. 19. Strange W: Dynamic specification of coarticulated vowels spoken in 45. Eickhoff SB, Stephan KE, Mohlberg H, Grefkes C, Fink GR, Amunts K, Zilles K: sentence context. J Acoust Soc Am 1989, 85:2135-2153. A new SPM toolbox for combining probabilistic cytoarchitectonic maps 20. Dupoux E, Pallier C, Sebastian N, Mehler J: A destressing ‘’deafness’’ in and functional imaging data. Neuroimage 2005, 25:1325-1335. French? J Mem Lang 1997, 36:406-421. 46. Tervaniemi M, Kruck S, De Baene W, Schröger E, Alter K, Friederici AD: Top- 21. Dupoux E, Peperkamp S, Sebastian-Galles N: A robust method to study down modulation of auditory processing: effects of sound context, stress “deafness”. J Acoust Soc Am 2001, 110:1606-1618. musical expertise and attentional focus. Eur J Neurosci 2009, 1636-42. 22. Dupoux E, Sebastian-Galles N, Navarrete E, Peperkamp S: Persistent stress 47. Allison T, McCarthy G, Nobre A, Puce A, Belger A: Human extrastriate ‘deafness’: The case of French learners of Spanish. Cognition 2008, visual cortex and the perception of faces, words, numbers, and colors. 106:682-706. Cereb Cortex 2006, 4:544-554. 23. Howard D, Smith K: The effects of lexical stress in aphasic word 48. Guion SG, Clark JJ, Harada T, Wayland R: Factors affecting stress production. Aphasiology 2002, 16:198-237. placement for English nonwords include syllabic structure, lexical class, 24. Nickels L, Howard D: Effects of lexical stress on aphasic word production. and stress patterns of phonologically similar words. Lang Speech 2003, Clinical Linguistics & Phonetics 1999, 13:269-294. 46(4):403-427. 25. Niemi J, Koivuselka-Sallinen P, Hanninen R: Phoneme errors in Broca’s 49. Dehaene S, Piazza M, Pinel P, Cohen L: Three parietal circuits for number aphasia: Three Finnish cases. Brain Lang 1985, 26:28-48. processing. Cogn Neuropsychol 2003, 20:487-506. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 17 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 50. Just MA, Newman SD, Keller TA, McEleny A, Carpenter P: Imagery in sentence comprehension: an fMRI study. Neuroimage 2004, 21:112-124. 51. Reed C: Reaction times for a same-different discrimination of vowel- consonant syllables. Perception, & Psychophysics 1975, 18(2):65-70. 52. Ratcliff R, Hacker MJ: Speed and accuracy of same and different responses in perceptual matching. Perception and Psychophysics 1981, 30:303-307. 53. Ratcliff R: Theoretical interpretations of the speed and accuracy of positive and negative responses. Psychological Review 1985, 92:212-225. 54. Gandour J, Wong D, Lowe M, Dzemidzic M, Satthamnuwong N, Iong Y, Lurito J: Neural circuity underlying perception of duration depends on language experience. Brain Lang 2002, 83:268-290. 55. Eisenberg P: Syllabische Struktur und Wortakzent: Prinzipien der Prosodik deutscher Wörter. Zeitschrift für Sprachwissenschaft 1991, 37-64. 56. Levelt WJM: Trends Cogn Sci 1999, 3:223-232. 57. Giegerich H: Metrical phonology and phonological structure: German and English. Cambridge: Cambridge University Press; 1985. 58. Féry C: German word stress in Optimality Theory. Journal of Comparative Germanic Linguistics 1998, 2:101-142. doi:10.1186/1744-9081-7-15 Cite this article as: Klein et al.: Neuro-cognitive foundations of word stress processing - evidence from fMRI. Behavioral and Brain Functions 2011 7:15. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral and Brain Functions Springer Journals

Neuro-cognitive foundations of word stress processing - evidence from fMRI

Loading next page...
 
/lp/springer-journals/neuro-cognitive-foundations-of-word-stress-processing-evidence-from-kdFYvlamQI

References (110)

Publisher
Springer Journals
Copyright
Copyright © 2011 by Klein et al; licensee BioMed Central Ltd.
Subject
Biomedicine; Neurosciences; Neurology; Behavioral Therapy; Psychiatry
eISSN
1744-9081
DOI
10.1186/1744-9081-7-15
pmid
21575209
Publisher site
See Article on Publisher Site

Abstract

Background: To date, the neural correlates of phonological word stress processing are largely unknown. Methods: In the present study, we investigated the processing of word stress and vowel quality using an identity matching task with pseudowords. Results: In line with previous studies, a bilateral fronto-temporal network comprising the superior temporal gyri extending into the sulci as well as the inferior frontal gyri was observed for word stress processing. Moreover, we found differences in the superior temporal gyrus and the superior temporal sulcus, bilaterally, for the processing of different stress patterns. For vowel quality processing, our data reveal a substantial contribution of the left intraparietal cortex. All activations were modulated by task demands, yielding different patterns for same and different pairs of stimuli. Conclusions: Our results suggest that the left superior temporal gyrus represents a basic system underlying stress processing to which additional structures including the homologous cortex site are recruited with increasing difficulty. Introduction The autonomy of vowel quality and word stress It is widely agreed that the processing of spoken words representations comprises acoustic and phonological analysis before in a First evidence for a relative independence of vowel qual- second step lexical and semantic information can be ity and word stress encoding in speech production came retrieved (e.g., [1-3]). With respect to the acoustic-pho- from psycholinguistic research. In particular, speech nological analysis of spoken words, there is general con- errors that involve stress exchange such as “my ‘proso- sensus that the categorical perception of phonetic dic (pro’sodic) colleagues” [11], though occurring rather properties like frequency formants, transitional proper- rarely, specifically demonstrate a separate encoding ties of formants, fundamental frequency, duration, or stage for word stress. Moreover, findings from speech intensity leads to the identification of strings of pho- perception point to a relatively independent processing nemes and - at least in languages with variable stress - of stress and vowel quality information although, of to the identification of word stress patterns. On a course, the metrical feature ‘stress’ inevitably has also its neuro-functional level, phonological processing has been vowel quality correlates such as vowel reduction in attributed to the superior temporal gyrus of both hemi- unstressed syllables ([12,13]). For instance, not only spheres(e.g.,[4-10]).However,sofarnostudyhas minimal stress pairs (i.e., words only differing in their aimed at directly differentiating vowel quality and word stress position) can be successfully discriminated on the stress processing. As a starting point, findings on the basis of their different stress patterns; even isolated syl- processing of both vowel quality as well as stress infor- lables excised from such minimal pairs can be reliably mation will be reviewed briefly. assigned to their source words [12,14]. Isolated syllables bearing a stressed or unstressed pitch contour can influ- ence the processing of subsequently presented targets which have a segmentally identical initial syllable with * Correspondence: klein@neuropsych.rwth-aachen.de Department of Neurology, Section Neuropsychology, University Hospital, congruent pitch [15]. However, while both vowel quality RWTH Aachen University, Aachen, Germany and stress can separately contribute to lexical Full list of author information is available at the end of the article © 2011 Klein et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 2 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 recognition [16], there is evidence that vowel quality imaging (fMRI). In this respect, it has been assumed that information can be exploited earlier than stress informa- the processing of emotional prosody elicits bilateral tion (e.g., vowel duration, pitch height, and amplitude) fronto-temporal patterns (e.g. [37]), while processing of due to coarticulation [16-19]. linguistic prosody has been suggested to be left lateralized Findings from Dupoux and colleagues suggest that in the superior temporal gyrus (for a review see [38]; but also on the level of abstract representation vowel quality see [39] for activation of Broca’s area associated with lin- and stress information may dissociate [20-22]. The so- guistic aspects of prosody). For the processing of linguistic called ‘stress deafness’ investigated by these authors is in aspects of prosody like contrastive stress and intonation, a fact not a difficulty to perceive and distinguish stressed considerable number of studies revealed a consistent and unstressed syllable patterns. Rather, only when involvement of the superior temporal gyrus. However, it is increased memory demands come into play, participants still under debate whether this region is involved left-later- display difficulties to remember stress patterns. More alized or bilaterally. On the one hand, Tong et al. [40] specifically, participants whose native language does not reported significantly stronger left lateralized activation of use stress to distinguish between words (e.g., French) the posterior middle temporal gyrus for the comparison of perform significantly lower in tasks testing memory for stress vs. intonation for Chinese speakers. Furthermore, stress patterns than participants whose language does Ischebeck, Friederici, & Alter [41] compared the proces- contain minimal stress pairs (e.g., Spanish). Crucially, sing of phrase boundaries in natural vs. hummed speech although French participants have particular problems and identified the superior temporal gyrus extending into in remembering stress patterns, their performance in the sulcus to be involved bilaterally in the processing of remembering minimal pairs of pseudowords only distin- natural speech whereas hummed speech revealed only left guishedbyone consonantdid notdifferfromthe per- lateralized activation of this region. On the other hand, formance of Spanish participants [21]. Native speakers when Meyer, Steinhauer, Alter, Friederici, and von Cra- of German have not been tested yet, but they should mon [42] contrasted normal speech (containing vowel obviously belong to the second class of participants, as quality and prosodic information) with degraded speech there are minimal stress pairs like ’Tenor vs. Te’nor. (lacking vowel quality information), they found bilateral Further evidence supporting the autonomy of vowel activation of the posterior superior temporal gyrus even quality and word stress knowledge comes from clinical for the case of degraded speech. Taken together, previous observations on brain-lesioned patients. A classical find- results reported on the processing of linguistic prosody are ing in aphasic word production is that there are more rather heterogeneous as regards possible lateralization. vowel quality errors in unstressed than in stressed sylla- To our knowledge, up to date only one study has bles (e.g., [23-25]). Furthermore, a number of aphasic directly investigated the neuro-anatomical correlates of patients have been described showing a dissociation word stress processing. In an fMRI study, Aleman, For- between spared vowel quality and impaired stress pro- misano, Koppenhagen, Hagoort, de Haan, & Kahn [43] cessing. Typically, their errors have been classified as asked participants to decide whether Dutch bisyllabic regularisation related to the assignment of word stress, i. words were iambic (e.g., salát) or trochaic (e.g., mónat). e. those patients mostly produced the regular or domi- They found areas in the left precentral gyrus, the left nant stress pattern avoiding the irregular or infrequent superior parietal lobule, and in the posterior part of the pattern while preserving syllable and phoneme struc- left superior temporal gyrus extending into the sulcus to tures [26-33]. The reverse pattern, i.e., vowel quality be more active in this stress task compared to a seman- errors with preserved word stress assignment is a stan- tic control condition. However, in their paradigm the dard finding in aphasic patients (e.g., [24]). However, identification of iambic and trochaic stress patterns there is accumulating evidence for an interaction relied on metalinguistic knowledge rather than on nat- between vowel quality and stress processing in German ural language processing. Such a metalinguistic task may speech production. Data from pseudoword reading involvemorethanonlyprosodicprocessing.Most [29,34,35], EEG [36], and patient studies [31] have importantly, contrasting a stress decision task to a shown that the assignment of main stress position in semantic control condition may be not specific enough German words is influenced by their vowel quality. to identify regions involvedinthe processing of word stress (as opposed to phonological processing in gen- Neuronal correlates underlying the processing of eral). In sum, the neural correlates underlying word linguistic prosody stress processing are far from being understood. There is an extensive body of literature on the possible lateralization of processes involved in the comprehension The Present Study of linguistic vs. emotional prosody based on neuro-ima- The current study was conducted to systematically ging methods such as functional magnetic resonance investigate the neuronal correlates underlying word Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 3 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 stress processing. To avoid lexical and semantic con- type. In the non-identical condition the two pseudo- founds on prosodic processing, we conducted an fMRI words either differed in stress or in vowel quality. study on the processing of stress patterns using pseudo- Therefore, activation observed only in the non-identical words. All stimulus items contained only stressable syl- conditions may have most likely reflected stimulus- lables (see [36]) which enabled us to control for vowel related effects, while activations seen in both identical quality in conditions with varying stress patterns. and non-identical pairs might be related to the task manipulation (i.e. particular attention paid to stress or Healthy participants were asked to state whether two vowel differences). auditorily presented bisyllabic pseudowords were the Taken together, the main goal of the present study same or different. In the ‘different’ condition, items dif- fered either in the position of word stress (e.g., Bo’kam was to identify brain regions involved in word stress vs. ‘Bokam) or in the quality of the first vowel. In the processing. Thus, we aimed at directly contrasting stress latter case, vowel quality differences were present both and vowel quality processing. Leaving higher linguistic in stressed and unstressed syllables (e.g., ‘Bekam vs. processing (e.g., lexical or semantic access) aside, our ‘Bokam and Be’kam vs. Bo’kam). Pseudowords only con- study enabled us to evaluate word stress processing in tained two instead of three syllables, as we expected that more detail. Thereby, the research questions motivating the stress pattern of trisyllabic words can already be the current study were twofold: (i) What is the specific inferred after heaving heard the first two syllables activation pattern associated with word stress proces- [14,16,36]. Moreover, the linguistic activity of interest (i. sing? (ii) How are activation patterns influenced by sti- e., the comparison of stress patterns) was contrasted mulus properties (same or different)? (iii) Are there any with a similar phonological activity (i.e., the comparison differences and/or similarities in localization and/or of vowel identity) to allow the investigation of highly intensity of fMRI signal change specifically associated specific activation patterns. In contrast to previous stu- with the metrical processing of different stress patterns dies (e.g., [43]), the word pairs were spoken by two dif- (penultimate vs. final stress)? ferent speakers: one male, one female. This way, in our stimulus-matching task we aimed at investigating the Methods processing of stress patterns at a rather abstract (phono- Participants logical) processing level not allowing for a direct com- Twenty four right-handed native German-speaking parison of phonetic values (see also [21]). Previous fMRI healthy volunteers (12 female; mean age: 28.2 years, SD studies using words and pseudowords revealed that acti- = 7.0 years) participated in this study after having given vations underlying lexical proscessingare notevokedif their written informed consent in accord with the proto- pseudowords are processed in a merely phonological col of the local Ethics Committee of the RWTH Aachen task [44]. Given this finding, the present design should Medical Faculty. be appropriate to investigate phonological processing relatively uncontaminated by lexical or semantic search. Material Building on the above considerations on the proces- A complete overview on all stimulus items used is pro- sing of phonological information the analyses were con- vided in additional file 1. Stimulus material consisted of ducted in two consecutive steps. They started from pairs of bisyllabic pseudowords obeying German phono- examining general activation differences between differ- tactic constraints. All items consisted of an initial open ent tasks addressing stress and vowel quality processing, syllable with a single plosive in onset position followed respectively, to proceed to more specific contrasts inves- by a closed syllable, containing simple consonantal onset tigating the influence of stimulus type (identical and and coda positions, respectively (CV.CVC). Both sylla- non-identical pairs, penultimate and final stress bles were stressable (i.e., excluding schwa-syllables). patterns). Pairs of stimuli were created such that they either dif- Note that all stimuli contained vowels and - given that fered only with respect to word stress (stress condition) they were bisyllabic - they were also marked for stress. or only with respect to vowel quality (vowel condition). Therefore, vowel and stress information were present in Furthermore, each pair consisted of one token spoken both conditions, and presumably participants automati- byafemaleand onetoken spoken byamalevoice, cally processed both types of information irrespective of respectively (see below). condition. Nevertheless, the conditions differ in two cru- In pairs pertaining to the stress condition, two pseu- cial ways: The first difference was task instruction. In dowords containing the same vowels were produced the vowel condition participants were instructed to pay with word-initial and word-final stress. Table 1 gives an attention to vowel information, whereas in the stress overview over phonetic parameters realized by both condition they were told to pay attention to stress infor- speakers to mark stress and Figure 1 exemplifies pho- mation. The second difference was related to stimulus netic information of the stimuli used. The examples of Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 4 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Table 1 Means (standard deviations are given in parentheses) of phonetic parameters duration, fundamental frequency, and intensity of the male and female speakers from a representative sample of 24 quadruples of stimuli (2 speakers × 2 stress patterns). Duration in seconds Fundamental Frequency in Hz Intensity in dB stress pattern PU F PU F PU F syllable type S1 S2 S1 S2 S1 S2 S1 S2 S1 S2 S1 S2 female speaker .37 (.06) .73 (.07) .30 (.06) .80 (.06) 213 (21) 162 (28) 191 (11) 189 (8.5) 80 (1.5) 71 (4.2) 78 (2.1) 77 (2.2) male speaker .47 (.05) .76 (.06) .34 (.06) .83 (.07) 129 (21) 102 (7) 119 (32) 110 (11) 79 (1.1) 73 (2.6) 77 (2.5) 77 (1.4) The values for F0 and intensity are averaged over syllables using PRAAT (Boersma & Weenink; version 5.1.19). Each value is given for each syllable in each stress pattern (numbers indicate the syllable position within a word and stressed syllables are printed in bold). PU: penultimate stress; F: final stress. Figure 1 Spectrograms, pitch contour, and intensity information for both stress patterns and speakers, illustrated with the stimulus quadruple “degis”. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 5 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 spectrograms, pitch, and intensity curves for both speak- speakers. At the same time it shows that stress patterns ers and stress patterns show that phonetic prominence could be clearly distinguished based on a combination was clearly marked in each stress condition. of three relevant phonetic variables (duration, funda- As expected, there was between-speaker variance in mental frequency, and intensity). stress realization. Consequently, Wilcoxon signed-rank In order to control for vowel quality in the stress con- tests for the syllable-wise stressed-unstressed-ratio of dition, all vowels were realized as tensed. Experimental duration, fundamental frequency, and intensity for a pairs contained four different vowels:/u:/,/o:/,/ø:/, and/ representative sample of 24 pairs of tokens revealed sig- e:/. In each pair, the difference in vowel quality invar- iantly affected the nucleus of the first syllable. nificant between-speaker differences for fundamental frequencyand intensityfor thesecondsyllable(Z ≤ For the 4 vowel contrasts differing in one or two fea- -2.342; uncorrected p ≤ .017). We are aware of the pro- tures (difference in 1 feature: between/u:/and/o:/as well blem that non-significant phonetic differences may still as between/e/and/ø:/; in 2 features: between/ø:/and/u:/as influence perception while the mere statistical signifi- well as between/e:/and/o:/, for an overview see Appen- cance of phonetic differences does not grant perceptual dix) 12 item pairs as well as 12 control pairs (with iden- consequences. Nevertheless, we think that presenting tical vowels) were created. Because there was only one tokens by a male and a female speaker should provoke a vowel contrast differing in 3 features (between/e:/and/ strategic shift in auditory processing, disfavoring a u:/), 24 item pairs as well as 24 control pairs (with iden- purely phonetic approach and encouraging a more tical vowels) were created for this vowel contrast. This abstract, phonological type of target comparison. Figure resulted in a total of 24 × 6 = 144 pseudoword pairs. 2 illustrates that, indeed, phonetic means to mark word Another 144 pairs of different items were used in the stress varied considerably both within and between stress condition. These 288 pairs of different items were Figure 2 Combined groups plot of a linear discriminant analysis on syllable-wise ratios (stressed: unstressed) of duration, fundamental frequency, and intensity for 24 representative pseudoword quadruples, revealing three discriminant functions. Function 1 2 2 explained 98.5% of the variance, canonical R = .89, whereas Function 2 explained only 1.1%, canonical R = .08 and Function 3 only .4%, 2 2 canonical R = .03. In combination, all three discriminant functions significantly differentiated the conditions, Λ = .10, c (9) = 207.91, p < .001. After removing Function 1, the remaining functions still differentiated the conditions significantly, Λ = .90, c (4) = 10.17, p = .038. However, Function 3 alone did not differentiate the conditions significantly Λ = .97, c (1) = 2.58, p = .108. Note that Function 1 clearly differentiates between both stress patterns. All three phonetic variables (duration, fundamental frequency, and intensity) loaded on Function 1 (r = .93, r = -.48, and r = -.34, respectively). Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 6 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 opposed to 288 pairs of identical items. Thus, each Each block consisted of 12 trials (6 pairs of identical experimental pseudoword appeared four times in second and 6 pairs of non-identical pseudowords, see Material), position of a pair: (i) in the identical and (ii) in the non- lasting 3700 ms per trial. Participants had to decide, identical stress condition, as well as (iii) in the identical whether the two items of a given pair were phonologi- and (iv) in the non-identical vowel condition. From this cally identical or not by pressing a button with the left overall set blocks were determined consisting of 12 item (non-identical) or the right (identical) hand. The dura- pairs which contained each six pairs of different items tion of the pseudowords ranged between 1000 and 1200 ms. Presentation rate of the trials was kept constant (all stemming from one cell of the experimental design) irrespective of the participants’ response speed. There- and six pairs of identical items. All initial items of a given block had the same stress pattern, such that in fore, each block invariantly lasted 44.4 seconds. Order every trial the decision could only be based on the sec- of trials, blocks, and speakers (male or female) was ond item of a pair. pseudo-randomized such that systematic confounds All stimuli were spoken by two experienced native between condition (e.g. identical vs. non-identical) and speakers of German - one female and one male - and stimulus order were avoided. Each participant was recorded using Amadeus Pro sound editing software exposed to the same sequence of trials. (Version 1.5.1, HairerSoft). In each pair presented, one item was spoken by the female and the other one by the Scanning procedure and imaging data acquisition male speaker - order being counterbalanced across con- For each participant, a high-resolution T1-weighted ana- ditions. Thus, strictly speaking, even ‘identical pairs’ tomical scan was acquired with a 3T Siemens Magne- were not identical on a (phonetic) ‘token’ level, but only tom TrioTim MRI system using the standard head coil on a more abstract (phonological) ‘type’ level of repre- (TR = 19 s, matrix = 256 × 256 mm, 190 slices, voxel sentation. This approach was chosen to increase pho- size = 1 × 1 × 1 mm; FOV = 256 mm, TE = 4.9 ms; flip netic variation and, in consequence, to highlight angle = 25°). Moreover, one functional imaging block processes at the level of abstract phonological represen- sensitive to blood oxygenation level-dependent (BOLD) tations (see also [21]). contrast was recorded for each participant (T2*- weighted echo-planar sequence, TR = 2400 ms; TE = 30 Task and Procedure ms; flip angle = 90°; FOV = 220 mm, 88 × 88 matrix; 42 The experiment was a combined functional magnetic slices, voxel size = 2.5 × 2.5 × 2.5 mm, gap = 10%). resonance imaging (fMRI) and reaction time (RT) study. Trials were presented at a rate of 3700 ms. Participants were lying in the scanner and listening to the word pairs presented auditorily via headphones. Analyses Head movements were prevented by using soft foam Reaction time (RT) analysis was based on correct trials pads. Participants were instructed to respond as quickly only. Furthermore, response latencies faster than 200 ms and accurately as possible avoiding unnecessary move- were not considered and in a second step responses out- ments. To familiarize participants with the task and to side the interval of +/-3 standard deviations around the reduce potential training effects during fMRI data acqui- individual mean were excluded. This resulted in a total sition, all volunteers were given the opportunity to prac- loss of 12.0% of the data. Error rates were arcsine-trans- tice on 16 pairs in a separate room before they entered formed prior to statistical analyses. RT and error rates the scanner. None of these practice items was repeated (ER) were analyzed using a 2 × 2 × 2 within-participant during the fMRI experiment. repeated measures ANOVA comprising the factors iden- The experiment was conducted in a box-car design tity (identical vs. non-identical pairs), phonological comprising 48 blocks. Two seconds prior to the start of manipulation (stress vs. vowel condition), and stress pat- each block one of two specific warning sounds was pre- tern of the second item (penultimate vs. final stress). sented, indicating whether the following block belonged The anatomical scans were normalized and averaged to the stress or to the vowel condition. The assignment in SPM8 http://www.fil.ion.ucl.ac.uk/spm. The fMRI of warning sounds varied over participants (e.g., for half time series was corrected for movement and unwarped of the participants a ringing sound indicated the vowel in SPM8. Images were motion corrected and realigned condition and a smashing sound the word stress condi- to each participant’s first image. Data were normalized tion, whereas for the other half of the participants the into standard stereotaxic MNI space. Images were opposite assignment was chosen). In the off-phase resampled every 2.5 mm using trilinear interpolation between blocks (duration 11.1 seconds) no audio signal and smoothed with a 5 mm FWHM Gaussian kernel to was presented until the onset of the next warning accommodate inter-subject variation in brain anatomy sound. and to increase signal-to-noise ratio in the images. The Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 7 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 data were high-pass filtered (128 s) to remove low-fre- stress led to more errors than second items with penulti- quency signal drifts and corrected for autocorrelation mate stress (13.2% vs. 8.6%), and decisions in the stress assuming an AR(1) process. Brain activity was convolved condition tended to be less accurate than those in the over all experimental trials with the canonical haemody- vowel condition (11.2% vs. 10.5%). While there was no sig- namic response function (HRF). For activation, which nificant three-way interaction, all two-way interactions was evaluated at an uncorrected p-value of < .001, clus- were significant or marginally significant [F(1, 23) ≥ 3.77, ter threshold correction was applied as a threshold lar- p ≤ .064]. Specifically, the increase in error rates from ger than 12 voxels corresponded to a corrected alpha identical to non-identical was particularly pronounced for second items with final compared to penultimate stress level < .05 with our parameters given. Localization of activation peaks was determined using the anatomic (final stress: 10.7% to 15.7%, penultimate stress: 7.7% to automatic labling tool (AAL, http://www.cyceron.fr/web/ 9.4%) and for the stress condition compared to the vowel aal__anatomical_automatic_labeling.html) as well as the condition (stress condition: 8.9% to 13.5%, vowel condi- SPM Anatomy Toolbox [45], available with all published tion: 9.5% to 11.6%). The advantage for penultimate com- cytoarchitectonic maps from http://www2.fz-juelich.de/ pared to final stress was more pronounced in the vowel inm/index.php?index=194). Complex contrasts were condition than in the stress condition (vowel condition: masked inclusively to prevent that e.g. subtraction of a 7.6% to 13.4%, stress condition: 9.5% to 12.9%). strong from a less strong deactivation suggests activa- tion while in fact there is an underactivation. fMRI data Analysis of fMRI data was based on all trials. In a first Results step, a conjunction over the contrasts of stress vs. base- Behavioral data line and vowel quality vs. baseline was conducted to A descriptive overview of the results is provided in show the largely overlapping cortical areas, which were Table 2. The ANOVA of RT data revealed main effects activated in both contrasts. The baseline covered the of identity and phonological manipulation [F(1, 23) ≥ rest periods between the blocks, in which no stimulus 39.50, p ≤ .001], indicating that decisions were faster on material was presented. non-identical pairs than on identical pairs (1100 ms vs. Conjunction over the contrasts stress vs. baseline and 1173 ms) and faster for vowel contrasts than for stress vowel quality vs. baseline (see Figure 3, Table 3) contrasts (1081 ms vs. 1192 ms). There was no main The conjunction revealed large common clusters of acti- effect of stress pattern [F(1, 23) = 2.77, p = .109]. More- vated voxels, bilaterally, in the superior temporal gyri over, there was a significant interaction of identity and (BA 22), the insula, the putamen as well as the cerebel- phonological manipulation [F(1, 23) ≥ 26.45, p ≤ .001], lum (p < .05, FWE-corrected, k = 12 voxels). meaning that the disadvantage for identical as compared In a second step, contrasts between vowel quality and to non-identical pairs was more pronounced in the stress processing were calculated to evaluate the regions vowel condition (1140 ms vs. 1021 ms) than in the found to be specifically active in word stress processing stress condition (1206 ms vs. 1179 ms). None of the by Aleman et al. [43]. other two- or three-way interactions reached statistical Stress vs. vowel quality (see Figure 4A, Table 3) significance. Word stress was contrasted to vowel quality at an The ANOVA of arcsine-transformed error-rates yielded uncorrected voxelwise p < .001 and a cluster size of 12 significant main effects of both identity and stress pattern voxels. This comparison indicated activation in the left [F(1, 23) ≥ 7.52, p ≤ .012], while the main effect of phono- inferior frontal gyrus [Brodmann Area (BA) 47], the logical manipulation only approached the conventional right superior temporal gyrus (STG, BA 22), the right level of significance [F(1, 23) ≥ 3.12, p = .091]. Specifically, inferior frontal gyrus (BA 44, BA 45), the right middle non-identical trials were somewhat more error prone than temporal gyrus (BA 21), the left fusiform gyrus (BA 19), identical trials (12.6% vs. 9.2%), second items with final and the right supplementary motor area (BA 6). Please Table 2 Overview of behavioral results. identical non-identical Penultimate Final Penultimate Final phonological manipulation stress 1215.7 (203.9) 1195.5 (204.5) 1177.9 (218.2) 1179.3 (206.1) 9.5 (4.4) 8.3 (5.8) 9.5 (5.4) 17.6 (6.7) vowel 1147.9 (237.6) 1131.8 (220.6) 1022.5 (220.0) 1020.2 (216.5) 5.8 (5.5) 13.1 (5.3) 9.4 (5.3) 13.8 (6.1) Mean RT (SD) in ms given in the first line and mean error rates (SD) in % given in the second line of each cell. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 8 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Figure 3 Conjunction over the contrasts stress - baseline and vowel quality - baseline: widespread activation in the bilateral superior temporal cortices (FWE-corrected voxelwise p < .05, cluster size k = 12). Table 3 Comparison of stress- and vowel quality-related activation. Contrast Brain region (BA) TC (x, y, z) Cluster size z score Conjunction* LH superior temporal gyrus (BA22) -48 -20 8 264 6.61 Stress - baseline and RH superior temporal gyrus (BA 22) 63 -25 0 236 6.54 Vowel quality vs. Baseline RH insula 30 23 5 20 5.87 LH insula -33 20 3 27 5.84 LH putamen -25 3 -3 60 5.71 RH putamen 25 3 5 17 5.38 LH supplementary motor area (BA 6) 0 0 65 114 6.14 RH cerebellum 30 -65 -25 24 6.17 LH cerebellum -28 -68 -25 21 6.04 Stress - vowel quality LH inferior frontal gyrus (BA 47) -48 18 -8 17 4.36 RH inferior frontal gyrus (BA 44) 53 15 20 36 4.33 RH inferior frontal gyrus (BA 45) 50 23 -5 15 4.06 RH superior temporal gyrus (BA 22) 58 -35 18 14 3.73 RH middle temporal gyrus (BA 21) 50 -23 -5 25 3.80 RH superior frontal gyrus (BA 6) 8 0 68 39 4.89 Non-identical pairs: RH middle temporal gyrus (BA 21) 55 -43 8 93 5.41 Stress - vowel quality LH inferior temporal gyrus (BA 21) -43 -3 -20 23 4.47 LH inferior frontal gyrus (BA 47) -48 18 -8 36 4.66 RH inferior frontal gyrus (BA 45) 53 23 -5 43 4.52 RH inferior frontal gyrus (BA 44) 50 13 20 60 4.02 RH superior temporal gyrus (BA 22) 55 -25 -3 26 3.81 LH superior temporal gyrus (BA 22) -50 -40 20 25 3.71 LH insula -33 20 8 34 4.32 RH intraparietal sulcus (BA 7) 43 -45 60 18 3.91 RH intraparietal sulcus (BA 7) 43 -40 50 14 3.64 RH postcentral gyrus (BA 1) 48 -23 50 54 4.10 RH superior frontal gyrus (BA 6) 8 3 68 50 5.34 LH middle frontal gyrus (BA 6) -40 0 55 14 3.95 RH middle frontal gyrus (BA 8) 45 8 40 19 3.71 LH cingulate gyrus (BA 32) -5 18 40 87 5.07 Identical pairs: LH intraparietal sulcus (BA 7) -35 -65 43 16 3.74 Vowel quality - stress *p < .05, FWE-corrected; p < .001, uncorrected; cluster size = 12 voxels; masks were created at uncorrected p < .05; MNI: Montreal Neurological Institute coordinates. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 9 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Figure 4 Comparisons of stress and vowel quality. A: Stress - vowel quality at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively with stress: Activation specific for prosodic processing in the right superior temporal gyri as well as in Broca’s area. B: Stress vs. vowel quality in non-identical pairs at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively with stress in non- identical pairs reveals a widespread right-lateralized temporo-parieto-frontal network. C: Vowel quality vs. stress in identical pairs at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively vowel quality in identical pairs: Activation of the left intraparietal cortex. note that the activation in the STG is not lateralized on uncorrected, k = 12 voxels), trials with non-identical the right hemisphere for the main effect of stress vs. pairs yielded a large network of activation when com- vowel quality. Flipping another sample of the contrast paring stress to vowel quality: images along the y-axis (from left to right orientation) Stress vs. vowel quality in non-identical pairs (Figure 4B, and calculating a paired t-test between the left STG in Table 3) the original data set and the left STG in the flipped data Contrasting stress and vowel quality in non-identical set (formerly the right STG) revealed that even at a very word pairs (p < .001, uncorrected, k =12voxels) liberal p-value < .05 there was no significant difference revealed activation in the bilateral superior temporal of activation in the STG between the two hemispheres. gyrus (BA 22), the bilateral middle temporal gyrus (BA More in-depth examination of this main effect 21), the left inferior frontal gyrus (BA 45), and the right revealed that the activation observed was mainly driven inferior frontal gyrus (BA 45 and 44). Further clusters of by differences between stress and vowel quality in non- activated voxels were found in the left insula, the right identical trials: Whereas in trials with identical pairs intraparietal sulcus (BA 7), the right superior parietal the comparison of stress with vowel quality revealed lobule (BA 7), the right postcentral gyrus (BA 2), the no activated voxels at the threshold chosen (p < .001, right supplementary motor area (BA 6), the left Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 10 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 precentral gyrus (BA 6), the right middle frontal gyri processing. Neural correlates underlying vowel quality (BA 8), and the left superior medial gyrus (BA 32). processing could best be identified comparing identical Inspection of the inverse contrast (vowel quality vs. pairs. An increase of the fMRI signal with vowel qual- stress) revealed no clusters of activated voxels at the ity processing in the difficult condition was found in threshold chosen (p < .001, uncorrected, k =12voxels). the left intraparietal cortex. However, closer inspection of the data indicated activation In a second step, we aimed at comparing activation when identical pairs were presented. In contrast, there was patterns for targets with different stress patterns. no activation observed for non-identical pairs. Penultimate vs. final stress in the stress condition (Figure Vowel quality vs. stress in identical pairs 5A, Table 4) In identical pairs, vowel quality was contrasted with Stronger activation was found in a large bilateral tem- stress at an uncorrected voxelwise p <.001and cluster poro-frontal network (FWE-corrected at p < .05, k =12 size of 12 voxels (Figure 4C, Table 3). Activated voxels voxels). The network comprised the bilateral superior were observed in the left intraparietal sulcus (BA 7). frontal gyri (BA 22), the bilateral putamen, the bilateral Taken together, there was a temporo-frontal activa- insula, the left supplementary motor area (BA 6) as well tion pattern specifically associated with word stress as the bilateral cerebellum. Figure 5 Comparisons of penultimate and final stress. A: Main effect of penultimate vs. final stress (FWE-corrected, cluster size k = 12 voxels, masked inclusively): Activation of a bilateral temporo-frontal network. B: Penultimate vs. final stress when comparing identical pairs according their stress at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively: Activation of the bilateral superior temporal gyri. C: Penultimate vs. final stress when comparing non-identical pairs regarding their stress at an uncorrected voxelwise p < .001 and cluster size k = 12 voxels, masked inclusively: Left-lateralized activation of the superior temporal gyrus. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 11 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 Table 4 Penultimate and final stress: Main effect and contrasts depending on the identity of the word pair presented. Contrast Brain region (BA) MNI (x, y, z) Cluster size z value Stress pattern: LH superior temporal gyrus (BA 22) -48 -20 8 261 6.49 Penultimate - final stress* RH superior temporal gyrus (BA 22) 63 -25 0 231 6.36 LH putamen -23 5 3 72 6.13 RH putamen 25 0 5 20 5.67 RH insula lobe 30 25 3 33 5.79 LH insula lobe -30 20 3 23 5.66 LH supplementary motor area (BA 6) -3 0 65 86 6.31 RH cerebellum 28 -63 -25 32 6.55 LH cerebellum -28 -68 -25 50 6.02 LH cerebellum 3 -63 -25 18 5.87 Stress in identical pairs: LH superior temporal gyrus (BA 22) -55 -25 8 192 4.73 Penultimate - final RH superior temporal gyrus (BA 22) 65 -20 3 208 5.80 Stress in non-identical pairs: LH superior temporal gyrus (BA 22) -50 -18 5 23 4.10 Penultimate - final * p < .05, FWE-corrected; p < .001, uncorrected; cluster size = 12 voxels; masks were created at uncorrected p < .05; MNI: Montreal Neurological Institute coordinates. Penultimate vs. final stress when comparing identical pairs processing involved a network of bilateral fronto-tem- in the stress condition (Figure 5B, Table 4) poral activation, resembling patterns previously Activation specific for penultimate stress in identical described to subserve auditory processing of bisyllabic pairs was present in large clusters in both superior tem- pseudo-words [46]. However, while the general contrast poral gyri (BA 22) extending along the superior tem- between stress and vowel conditions showed only right- poral sulcus (uncorrected p < .001, k = 12 voxels). hemispheric activation of the superior temporal gyrus, Penultimate vs. final stress when comparing non-identical the more fine-grained analysis over non-identical pseu- pairs in the stress condition (Figure 5C, Table 4) doword pairs showed that a bilateral fronto-temporal Comparing conditions with penultimate stress to condi- network was specifically associated with word stress pro- tions with final stress in non-identical word pairs, only cessing. In particular, we were able to identify task-spe- activation in the left superior temporal gyrus was cific differences of stress processing in the superior observed(BA 22)atanuncorrected p <.001 and voxel temporal gyrus and the superior temporal sulcus. size of k = 12 voxels. Finally, our data suggested higher cognitive demands for For the opposite comparison (final vs. penultimate the processing of penultimate stress compared to final stress) no activation was observed at the threshold cho- stress in the experimental design chosen. sen neither for identical nor for non-identical pairs. Taken together, in the superior temporal gyrus as well Stress vs. vowel quality processing as the superior temporal sulcus differential effects of Main purpose of the present study was to evaluate the stress processing were found dependent on both the fac- neural correlates of phonological word stress proces- tors identity (non-identical vs. identical auditory word sing by comparing these correlates to those related to pairs) and stress pattern (penultimate stress vs. final a similar task - vowel qualityprocessing. Thediffer- stress). ence of stress and vowel quality processing in general was corroborated by the effect of phonological manip- Discussion ulation which was significant in the ANOVA on RT The current study aimed at investigating the processing and marginally significant in the ANOVA on ER, of word stress information. For this purpose, behavioral meaning that reaction times were faster for vowel con- and neuro-imaging data of word stress and vowel quality trasts than for stress contrasts. In line with previous processing were contrasted directly. In general, the studies investigating activation related to prosodic sen- neural networks associated with word stress and vowel tence processing, the comparison of stress and vowel quality processing were observed to be largely overlap- quality processing revealed a network of activation ping. In particular, the conjunction of stress and vowel comprising the right superior temporal gyrus. This tasks revealed that both aspects of phonological brain region has been identified repeatedly to be Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 12 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 associated with prosodic processing (e.g., [38,40-42]). non-identical pairs). For stress activation in the context Moreover, activation was evidenced in Broca’sarea, of non-identical pairs, a widespread pattern of temporo- which has also been found to be associated with lin- frontal activation was observed, while the processing of guistic aspects of prosody [39]. Finally, increased occi- vowel quality information vs. stress processing in the pital activation extending into the left fusiform gyrus context of identical pseudoword pairs seems to be asso- was observed, where the visual identification area for ciated with the intraparietal cortex as already reported by Meyer et al. [42]. However, as already suggested by word formsissupposedtobelocated [47].Thisactiva- the behavioral data, it is important to note that both the tion may indicate that participants also searched for effect of word stress and the effect of vowel quality associations with familiar word forms and their stress patterns whenever they had to process stress informa- information have to be evaluated in the context of the tion in pseudowords (see also [48]). stimulus type (identical or different). The contrasts Comparing vowel quality and stress processing between vowel quality and stress processing seem to revealed no super-threshold activation in the whole reflect qualitative differences rather than being only brain. However, as the behavioral analysis revealed a related to different degrees of difficulty. strong impact of the factor identity, a more fine-grained analysis which takes this factor into account seems to be Stimulus specific effects on prosodic processing more adequate. Indeed, breaking down the task-specific The present study revealed that the type of stimulus pair interaction between stress/vowel quality processing and (identical vs. different) influenced stress processing. The the factor identity into its constituting conditions effect of identity was significant in the ANOVA on both revealed that the activation observed in the comparison RT and ER, with identical pairs being classified more of stress to vowel quality processing was mainly driven slowly than non-identical pairs. Moreover, the present by trials with non-identical pairs. In such a comparison, neuro-imaging data clearly indicated the importance of not only the same areas were observed which were stimulus-specific effects for the above described network found to be active in the main effect of stress vs. vowel of activation for stress processing: Whenever a pair of quality processing, but also the superior temporal gyrus pseudowords with non-identical stress patterns had to was activated bilaterally. In addition, the right intrapar- be decided on, a large bilateral network in the superior ietal cortex was activated. This cortex site has been sug- temporal gyrus was activated, which has repeatedly been gested to underlie the processing of proximity relations identified to be vitally involved in processing prosodic [49] as well as mental imagery (e.g., [50]). Thus, the information (e.g., [38,41]). However, when the stress non-identical stress patterns may have been evaluated pattern in the pair of pseudowords was identical, no with respect to the relation and extent of their differ- activation was observed. ences; moreover, participants may have tried to intern- Mean RTs in the present study were faster for non- ally memorize and compare the stress patterns they had identical than for identical stimulus pairs, whereas in been presented with. It should be pointed out, that some behavioral experiments reported in the literature stress is an inherently relational property, i.e., its recog- involving same-different decisions on vowel-consonant nition requires the comparison of phonetic measures (e. syllables, faster mean RTs were obtained for the proces- g., duration, pitch, and intensity) between stressed and sing of same syllables compared to different ones (e.g., unstressed syllables and this relation may even be differ- [51,52]). However, the difference between “same” and ent within and between different speakers as in our task “different” responses is subject to specific task demands (see Figures 1 and 2 and Table 1). (e.g., [51-53]. In the present investigation even in the In contrast, for identical pairs no activation was “same” condition items were actually not identical but observed for stress vs. vowel quality processing. How- realized by different speakers. Listeners therefore could ever, the opposite contrast showed that within identical not rely on superficial phonetic deviations in the “differ- pairs, vowel quality compared to stress processing was ent” condition but had to derive abstract representations related to stronger intraparietal activation in the left to perform the evaluation task. Since the phonetic devia- hemisphere.Thisisinlinewithpreviousfindingscom- tions in the “same” condition were more fine-grained paring vowel quality (flattened without prosody) and compared to the “different” condition, the latencies for natural speech [42] as these data already suggested that “same"-decisions were higher. However, the asymmetri- the left intraparietal cortex may be associated with cal neurophysiological effect of the matching task on vowel quality processing. stress vs. vowel processing indicates qualitatively differ- To sum up, our imaging data indicate different activa- ent demands on positive or negative responses. tion patterns for vowel quality and stress processing The finding of stress processing being influenced by when contrasting these two aspects of phonological pro- stimulus specific effects is relevant regarding the possi- cessing directly in different stimulus context (identical/ ble lateralization of processes which subserve the Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 13 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 comprehension of linguistic prosody. As already out- Regarding the main effect of stress patterns, our fMRI lined above, a consistent involvement of the superior data yielded different results than the behavioral data. In temporal gyrus has been shown frequently for the pro- particular, no activation was found for final stress as cessing of linguistic aspects of sentence prosody like compared to penultimate stress. However, the inverse contrastive stress and intonation [40]. However, it still contrast revealed a bihemispheric activation of the remains debatable whether this region is involved only superior temporal gyrus, which has been repeatedly in the left hemisphere or rather bilaterally. On the one reported to be associated with prosodic processing (e.g., hand, a considerable number of studies reported signif- [40,41]). This finding suggests that the processing of icantly stronger left lateralized activation of the poster- penultimate stress may have involved a more detailed ior middle temporal gyrus for processing stress auditory analysis than the processing of final stress. On information (e.g., [38,40]). On the other hand, bilateral a phonetic level of explanation, this may have been due activation of the posterior superior temporal gyrus has to the different perceptual saliency of both patterns. On also been reported repeatedly for processing prosodic a phonological level, this activation pattern may indicate information in natural (e.g., [41]) and degraded [42] that penultimate stress has not a general default status speech. in German as already argued by Janßen, Domahs, and The current study may add to the understanding of colleagues [29,31,36]. This is a challenge to approaches such apparently heterogeneous findings. When only assuming that given the fact that penultimate stress (or comparing main effects such as the main effect of stress in bisyllabic words: initial stress) is the most frequent to the main effect of vowel quality processing, only German stress pattern it forms some kind of default lateralized activation of the superior temporal gyrus was stress pattern which- in contrast to final stress - has not found. However, as outlined above, our behavioral data to be lexically specified (e.g., [55,56]). However, Janßen indicated that the identity or non-identity of stress pat- [29], Janßen & Domahs [31] and Domahs et al. [36] terns may be relevant. Indeed, when the processing of report behavioral and electrophysiological evidence that stress information was evaluated in the context of the the “regularity” of word stress is strongly influenced by stimulus type (identical or non-identical), bilateral acti- the structure of the final and penultimate syllable vation of the superior temporal gyrus was found, which [30,36,57,58]. In particular, penultimate stress occurs seems to correspond well to the findings of Ischebeck et predominantly in words with an open final syllable (e.g., al. [41] as well as Meyer and colleagues [42]. In contrast, Pánda, [panda]), but not in words with a closed final the processing of identical stress patterns as well as a syllable (e.g. Spinát, [spinach]), casting doubts on a comparable contrast in the vowel task within non-iden- structure-independent default status of penultimate tical items did not reveal such an activation pattern. stress. Since the pseudowords presented consist of an Taken together, diverging previous results regarding open penultimateand aclosed final syllable, the higher the lateralization of prosodic processing may have possi- processing costs for items with penultimate stress may bly been due to stimulus- or task-specific properties (see reflect the fact that this pattern is not preferred in also [54] for task specific effects on neural activation words with a closed final syllable(seealso[34]). Again, patterns in two language groups requiring different amorefine-grainedanalysisofthe imagingdata efforts in the processing of stress properties). Taking revealed that the factor identity differentially influenced these properties into account, our data suggest that the results. Activation observed for identical stress pat- whenever more fine-grained decisions have to be made terns was found bilaterally, whereas the processing of at an increasingly abstract level, bilateral activation of non-identical stress patterns was only associated with the superior temporal gyri is needed. This view fits well left-lateralized activation. This distribution of activation with previous observations on bilateral processing of maybeexplained by thefollowing arguments. Most stress comparison [41,42]. probably, it may have been easier to decide that two stress patterns are different than to decide that two Effects of stress patterns tokens of the same stress pattern, produced by different The present study also revealed different behavioural speakers, are indeed identical at a phonological level. and imaging results for different stress patterns. The This assumption is supported by our behavioural data effect of stress pattern was significant in the ANOVA showing that responses for non-identical pairs were sig- on error rates, with final stress in the second item nificantly faster than for identical pairs. However, it is being more difficult to be processed than penultimate important to consider that the difference in the neuro- stress. Stimulus-specific effects again influenced perfor- functional data is restricted to the superior temporal gyrus, while it does not seem to involve areas associated mance as the increase in error rates was particularly with generally higher levels of working memory or more pronounced for pairs with non-identical stress attentional load (e.g., the dorsolateral prefrontal cortex patterns. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 14 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 and/or the intraparietal cortex) where activation would fine-grained analysis showed that the activation be expected if the different performance on identical observed for stress vs. vowel quality processing was in and non-identical stimuli is purely ascribed to higher fact driven by the comparison of stress and vowel memory load. Thus, the greater activation for identical quality processing in trials with non-identical pairs, pairs seems to be rather specific to the processing of while for identical pairs no activation was observed for stress information itself than to reflect more general stress vs. vowel quality processing. On the contrary, processes associated with a higher level of working within identical pairs stronger activation was observed memory and/or attentional demands. Thereby, the for vowel quality as compared to stress processing. To sum up, our imaging data indicate that both the effect increased activation may reflect most likely the extended auditory evaluation of the more fine-grained phonetic of word stress as well as the effect of vowel quality differences in pairs with identical stress patterns. information have to be evaluated in the context of the Taken together, diverging previous results regarding stimulus type (identical or non-identical), as was the lateralization of prosodic evaluation may have pos- already suggested by the behavioral effect of stimulus sibly been due to stimulus- or task-specific properties. type (identical or non-identical word pairs). Thereby, Taking these properties into account, our data support the differences between vowel quality and stress pro- the view that the left posterior superior temporal gyrus cessingseemtobequalitativeratherthanonlybeing is a kind of basic system mainly involved in the evalua- related to different degrees of difficulty. This interpre- tion of prosodic properties as outlined in part of the tation is further supported by the facts that areas typi- previous literature (e.g., [40]). However, once more cally associated with higher cognitive demands (e.g., fine-grained decisions have to be made at an increas- left dorsolateral prefrontal gyrus, anterior cingulate or ingly abstract level, the right superior temporal gyrus intraparietal cortices) were not observed for the com- seems to be called for assistance (e.g., [54]). This view parison of stress and vowel quality processing. Quite fits well with previous observations of bilateral proces- the contrary, in the identity condition, the intraparietal sing related to rather abstract stress comparison, e.g., cortex was in fact significantly stronger involved in the in degraded speech [42]. Thus, the present finding processing of vowel quality than of stress information. again underlines the impact of task and stimulus-speci- This finding leads us to the question why response fic effects. latencies were generally longer when evaluating stress patterns. In the literature, there is evidence that vowel Evaluation and perspectives quality information can be exploited earlier than stress We believe that the current study is a first step towards information due to coarticulation [16-19]. Neverthe- less, in order to explain the extent of these differences, a more comprehensive understanding of the underlying processes subserving word stress processing. However, it may be helpful to consider that our design enabled there are still a lot more steps to go. Therefore, in the participants to decide on the vowel quality structure as remainder of this Discussion some points requiring soon as the first syllable of the second item was further investigation will be addressed. encountered. In contrast, for decisions on stress infor- Consider firstthatthe responsestoword stresseva- mation, the second syllable of the second item had to luation were significantly slower and tended to be more be perceived before a confident judgement was possi- error prone than the evaluation of vowel quality infor- ble. This explanation may account for a general differ- mation. The question may arise whether the stress con- ence of 100-200 ms in response latencies. Indeed, dition was generally more difficult than the vowel inspection of the behavioral data revealed that all quality condition - a methodological artefact potentially stress conditions were evaluated systematically slower fateful for the validity of our data and the conclusions than the vowel quality conditions (see Table 2). Taking we have drawn. all these arguments into account, the current paradigm However, support for the validity of our data comes seems to be a valid approach to further investigate the from several different aspects. First, the above men- neural correlates of processing word stress and vowel tioned RT-findings neglect that the pattern observed is quality information. driven by a speed-accuracy trade-off as the slower con- Consider next the effect of stress patterns. The com- dition also tended to be less error prone. Second, parison between stress patterns revealed a bihemispheric inspection of the imaging data provides helpful activation of the superior temporal gyrus for penulti- insights. Indeed, the comparison of stress with vowel mate stress compared to final stress. This finding sug- gests that the processing of penultimate stress produced quality processing revealed a bilateral network of acti- higher costs than the processing of final stress. At which vation, whereas the contrast of vowel quality vs. stress level of processing may pseudowords with penultimate showed no voxel in the whole brain activated signifi- cantly stronger at the threshold used. However, a more stress have been harder to process than pseudowords Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 15 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 with final stress? The activation differences in the iden- Summary and Conclusion tity condition may just reflect higher efforts at the level The current study addressed two main research issues: of phonetic analysis. Unfortunately, knowledge about First, we were interested in the activation pattern asso- the perceptual consequences of specific phonetic fea- ciated with stress processing. By controlling stimulus tures in word stress processing is still lacking. In conse- material for vowel quality in conditions with varying quence, a perceptual account of the activation stress patterns and by varying phonetic realizations we differences cannot be excluded with the data at hand. intended to provoke a matching of stress patterns on a At a more abstract phonological level of processing it rather abstract, phonological level. may be speculated that penultimate stress is generally We observed a fronto-temporal network basically more difficult to be processed or represented than final comprising the right superior temporal gyrus extending stress. However, such an interpretation would not be into the sulci as well as the inferior frontal gyri, bilater- warranted as penultimate stress is the statistically predo- ally, to be specifically associated with stress processing. minant pattern in German words and, thus, is not likely However, when the contrast was evaluated more specifi- to evoke higher costs in processing than the less fre- cally in the context of the stimulus type (identical/non- quent pattern. However, note that all stimuli presented identical pairs), the data became clearer and revealed contained a heavy final syllable and therefore do not fit that stress was processed in the bilateral superior tem- the typical pattern of German words with penultimate poral gyri and sulci in the more difficult non-identical stress, namely bisyllabic words with a light or reduced trials. For vowel quality processing, our data emphasize final syllable. Thus, the higher processing costs may a substantial contribution of the left intraparietal cortex. support quantity sensitive approaches on German stress Second, our data suggest that higher cognitive assignment [30,36,57,58] and show that for words with a demands were needed for processing penultimate com- heavy final syllable penultimate stress is not the pared to final stress possibly suggesting that penultimate unmarked pattern (see also [34]). Further neuro-func- stress has not a default status in German. Thereby, our tional examinations with varying syllable structures results support the view that the left superior temporal should bring more light into this debate. gyrus represents a kind of basic system underlying stress Even if we assume that a difference in phonetic para- processing to which additional structures including the meters may have affected our results, it is important to homologous cortex site are recruited with increasing note that this is clearly not the case for our behavioral difficulty. data.First,there wasnomaineffect of stress patternin the ANOVA on RT. The analysis of error rates even pro- Additional material vided evidence in favour of the assumption that final stress may have been more difficult to be processed than Additional file 1: Lists of pseudowords used. penultimate stress as final stress in the second word of a pair was associated with significantly more errors than penultimate stress. Taking into account the phonetic Acknowledgements parameters we do not claim that the differential imaging The research and the preparation of this article were supported by a START- programme grant (AZ 37/07) of the Faculty of Medicine at the RWTH effects found for specific stress patterns (penultimate vs. Aachen University supporting M. Grande, a DFG priority programme grant final stress) in our study can be generalized to studies (DO 1433/1-1) supporting F. Domahs, as well as a DFG priority programme using other stimuli, presentation formats or tasks. How- grant (WI 853/7-2) supporting U. Domahs. We wish to thank Katja Halm, Stefanie Jung, and Timo Roettger for their help in stimulus preparation and ever, the present study definitely shows that these specific data collection as well as Ralph Schnitker and Georg Eder from the stress patterns may be processed differently and should “Functional Imaging Unit” of the Interdisciplinary Centre for Clinical Research be target of further investigations, for instance, with IZKF ‘BioMAT.’ for their precious help in conducting this study. more precisely controlled phonetic parameters and dif- Author details ferent syllable structures. Imaging studies on different Department of Neurology, Section Neuropsychology, University Hospital, stress patterns may be a crucial source of evidence feed- RWTH Aachen University, Aachen, Germany. Interdisciplinary Center for Clinical Research Aachen, University Hospital, RWTH Aachen University, ing phonological theories on stress systems. Aachen, Germany. Institute of Psychology, Eberhard Karls University, Taken together, even though there are still a number Tuebingen, Germany. Institute of Germanic Linguistics, University of of questions to be answered, the present results provide Marburg, Germany. Department of Neurology, Section Clinical Cognition Research, University Hospital, RWTH Aachen University, Aachen, Germany. first evidence not only on the neural correlates subser- ving stress processing, but also for the impact of stimu- Authors’ contributions lus-dependent effects (e.g., whether the stress/vowel FD, UD, EK, and MG conceived the study. All authors participated in its design. EK performed data collection, processing and statistical analyses. EK, quality decision has to be made within identical or non- FD, and UD drafted the manuscript. All authors contributed to the identical stimuli). Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 16 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 interpretation of the data. All authors read and approved the final 26. Cappa SF, Nespor M, Ielasi W, Miozzo A: The representation of stress: manuscript. evidence from an aphasic patient. Cognition 1997, 65:1-13. 27. Colombo L, Brivio C, Benaglio I, Siri S, Cappa SF: Alzheimer patients’ ability Competing interests to read words with irregular stress. Cortex 2000, 36:703-714. The authors declare that they have no competing interests. 28. Galante E, Tralli A, Zuffi M, Avanzi S: Primary progressive aphasia: a patient with stress assignment impairment in reading aloud. Neurol Sci Received: 1 December 2010 Accepted: 16 May 2011 2000, 21:39-48. Published: 16 May 2011 29. Janssen U: Stress assignment in German patients with surface dyslexia. Brain Lang 2003, 87:114-115. 30. Janssen U: Untersuchungen zum Wortakzent im Deutschen und References Niederländischen. PhD thesis, University of Düsseldorf; 2003 [http:// 1. McClelland JL, Elman JL: The TRACE model of speech perception. Cognit docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=2911]. Psychol 1986, 18:1-86. 31. Janssen U, Domahs F: Going on with optimised feet: Evidence for the 2. Marslen-Wilson WD: Functional parallelism in spoken word-recognition. interaction between segmental and metrical structure in phonological Cognition 1987, 25:71-102. encoding from a case of primary progressive aphasia. Aphasiology 2008, 3. Luce PA, Pisoni DB: Recognizing spoken words: the neighborhood 22:1157-1175. activation model. Ear Hear 1998, 19:1-36. 32. Laganaro M, Vacheresse F, Frauenfelder UH: Selective impairment of 4. Rauschecker JP: Cortical processing of complex sounds. Curr Opin lexical stress assignment in an Italian-speaking aphasic patient. Brain Neurobiol 1998, 8:516-521. Lang 2002, 81:601-609. 5. Scott SK, Blank CC, Rosen S, Wise RJS: Identification of a pathway for 33. Miceli G, Caramazza A: The Assignment of Word Stress in Oral Reading: intelligible speech in the left temporal lobe. Brain 2000, 123:2400-2406. Evidence from a Case of Acquired Dyslexia. Cogn Neuropsychol 1993, 6. Narain C, Scott SK, Wise RJ, Rosen S, Leff A, Iversen SD, Matthews PM: 10:273-296. Defining a left-lateralized response specific to intelligible speech using 34. Roettger T, Domahs U, Grande M, Domahs F: Structural factors affecting fMRI. Cereb Cortex 2003, 13:1362-1368. the assignment of word stress in German.. 7. Liebenthal E, Binder JR, Spitzer SM, Possing ET, Medler DA: Neural 35. Tappeiner E, Domahs U, Domahs F: Wortakzent im Sprachkontakt substrates of phonemic perception. Cereb Cortex 2005, 15:1621-1631. Deutsch-Italienisch. [Word stress in German-Italian Language Contact.]. 8. Obleser J, Zimmermann J, Van Meter J, Rauschecker JP: Multiple stages of Zeitschrift für Dialektologie und Linguistik 2007, 74:266-291. auditory speech perception reflected in event-related fMRI. Cereb Cortex 36. Domahs U, Wiese R, Bornkessel-Schlesewsky I, Schlesewsky M: The processing of German word stress: evidence for the prosodic hierarchy. 9. Hickok G, Poeppel D: The cortical organization of speech processing. Phonology 2008, 25:1-36. Nature Reviews 2007, 8:393-402. 37. Kotz SA, Meyer M, Alter K, Besson M, von Cramon DY, Friedericia AD: On 10. Specht K, Huber W, Willmes K, Shah J, Jäncke L: Tracing the ventral stream the lateralization of emotional prosody: An event-related functional MR for auditory speech processing in the temporal lobe by using a investigation. Brain Lang 2003, 86:366-376. combined time series and independent component analysis. Neurosci 38. Dogil G: Understanding Prosody.Edited by: Rickheit H, Herrmann T, Lett 2008, 442:180-185. Deutsch W. Psycholinguistik: Ein Internationales Handbuch; 2003:544-565, 11. Cutler A: Speakers’ conceptions of the function of prosody. In Models and Berlin: de Gruyter. measurements. Edited by: Cutler A, Lad Prosody DR. Heidelberg: Springer; 39. Wildgruber D, Hertrich I, Riecker A, Erb M, Anders S, Grodd W, 1983:79-92. Ackermann H: Distinct frontal regions subserve evaluation of linguistic 12. Cutler A, van Donselaar W: Voornaam is not (really) a homophone: Lexical and emotional aspects of speech intonation. Cerebr Cortex 2004, prosody and lexical access in Dutch. Lang Speech 2001, 44:171-195. 14(12):1384-1389. 13. Boecker KBE, Bastiaansen MCM, Vroomen J, Brunia CHM, De Gelder B: An 40. Tong Y, Gandour J, Talavage T, Wong D, Dzemidzic M, Xu Y, Li X, Lowed M: ERP correlate of metrical stress in spoken word recognition. Neural circuitry underlying sentence-level linguistic prosody. NeuroImage Psychophysiology 1999, 36:706-720. 2005, 28:417-428. 14. van Heuven VJ: Effects of stress and accent on the human recognition of 41. Ischebeck A, Friederici AD, Alter K: Processing prosodic boundaries in word fragments in spoken context: Gating and shadowing. 7th FASE natural and hummed speech. An fMRI study. Cerebr Cortex 2008, Symposium, Edinburgh 1988, 811-818. 18:541-552. 15. Friedrich CK, Kotz SA, Friederici AD, Alter K: Pitch modulates lexical 42. Meyer M, Steinhauer K, Alter K, Friederici AD, von Cramon DY: Brain activity identification in spoken word recognition: ERP and behavioral evidence. varies with modulation of dynamic pitch variance in sentence melody. Brain Res Cogn Brain Res 2004, 300-308. Brain Lang 2004, 89:277-289. 16. van Donselaar W, Koster M, Cutler A: Exploring the role of lexical stress in 43. Aleman A, Formisano E, Koppenhagen H, Hagoort P, de Haan EHF, Kahn RS: lexical recognition. Q J Exp Psychol A 2005, 58:251-273. The Functional Neuroanatomy of Metrical Stress Evaluation of Perceived 17. Cutler A, Chen HC: Lexical tone in Cantonese spoken-word processing. and Imagined Spoken Words. Cereb Cortex 2005, 15:221-228. Percept Psychophys 1997, 59:165-179. 44. Heim S, Alter K, Ischebeck AK, Amunts K, Eickhoff SB, Mohlberg H, Zilles K, 18. Robinson K, Patterson RD: The stimulus duration required to identify von Cramon DY, Friederici AD: The role of the left Brodmann’s areas 44 vowels, their octave, and their pitch chroma. J Acoust Soc Am 1995, and 45 in reading words and pseudowords. Brain Res Cogn Brain Res 98:1858-1865. 2005, 982-993. 19. Strange W: Dynamic specification of coarticulated vowels spoken in 45. Eickhoff SB, Stephan KE, Mohlberg H, Grefkes C, Fink GR, Amunts K, Zilles K: sentence context. J Acoust Soc Am 1989, 85:2135-2153. A new SPM toolbox for combining probabilistic cytoarchitectonic maps 20. Dupoux E, Pallier C, Sebastian N, Mehler J: A destressing ‘’deafness’’ in and functional imaging data. Neuroimage 2005, 25:1325-1335. French? J Mem Lang 1997, 36:406-421. 46. Tervaniemi M, Kruck S, De Baene W, Schröger E, Alter K, Friederici AD: Top- 21. Dupoux E, Peperkamp S, Sebastian-Galles N: A robust method to study down modulation of auditory processing: effects of sound context, stress “deafness”. J Acoust Soc Am 2001, 110:1606-1618. musical expertise and attentional focus. Eur J Neurosci 2009, 1636-42. 22. Dupoux E, Sebastian-Galles N, Navarrete E, Peperkamp S: Persistent stress 47. Allison T, McCarthy G, Nobre A, Puce A, Belger A: Human extrastriate ‘deafness’: The case of French learners of Spanish. Cognition 2008, visual cortex and the perception of faces, words, numbers, and colors. 106:682-706. Cereb Cortex 2006, 4:544-554. 23. Howard D, Smith K: The effects of lexical stress in aphasic word 48. Guion SG, Clark JJ, Harada T, Wayland R: Factors affecting stress production. Aphasiology 2002, 16:198-237. placement for English nonwords include syllabic structure, lexical class, 24. Nickels L, Howard D: Effects of lexical stress on aphasic word production. and stress patterns of phonologically similar words. Lang Speech 2003, Clinical Linguistics & Phonetics 1999, 13:269-294. 46(4):403-427. 25. Niemi J, Koivuselka-Sallinen P, Hanninen R: Phoneme errors in Broca’s 49. Dehaene S, Piazza M, Pinel P, Cohen L: Three parietal circuits for number aphasia: Three Finnish cases. Brain Lang 1985, 26:28-48. processing. Cogn Neuropsychol 2003, 20:487-506. Klein et al. Behavioral and Brain Functions 2011, 7:15 Page 17 of 17 http://www.behavioralandbrainfunctions.com/content/7/1/15 50. Just MA, Newman SD, Keller TA, McEleny A, Carpenter P: Imagery in sentence comprehension: an fMRI study. Neuroimage 2004, 21:112-124. 51. Reed C: Reaction times for a same-different discrimination of vowel- consonant syllables. Perception, & Psychophysics 1975, 18(2):65-70. 52. Ratcliff R, Hacker MJ: Speed and accuracy of same and different responses in perceptual matching. Perception and Psychophysics 1981, 30:303-307. 53. Ratcliff R: Theoretical interpretations of the speed and accuracy of positive and negative responses. Psychological Review 1985, 92:212-225. 54. Gandour J, Wong D, Lowe M, Dzemidzic M, Satthamnuwong N, Iong Y, Lurito J: Neural circuity underlying perception of duration depends on language experience. Brain Lang 2002, 83:268-290. 55. Eisenberg P: Syllabische Struktur und Wortakzent: Prinzipien der Prosodik deutscher Wörter. Zeitschrift für Sprachwissenschaft 1991, 37-64. 56. Levelt WJM: Trends Cogn Sci 1999, 3:223-232. 57. Giegerich H: Metrical phonology and phonological structure: German and English. Cambridge: Cambridge University Press; 1985. 58. Féry C: German word stress in Optimality Theory. Journal of Comparative Germanic Linguistics 1998, 2:101-142. doi:10.1186/1744-9081-7-15 Cite this article as: Klein et al.: Neuro-cognitive foundations of word stress processing - evidence from fMRI. Behavioral and Brain Functions 2011 7:15. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit

Journal

Behavioral and Brain FunctionsSpringer Journals

Published: May 16, 2011

There are no references for this article.