As a service to our customers we are providing this early version of the manuscript. Vitevitch M. The influence of phonological similarity neighborhoods on speech production. Articulatory initiation was always observed to occur after the signal to respond. de Jong K. The correlation of p-center adjustments with articulatory and acoustic events. (2005) also found shorter reaction times for initial clusters in a delayed naming task. Means and standard error of AAI (articulatory onset at peak velocity to acoustic onset interval, white bars) and DurG1 (constriction formation duration, grey bars) for different initial phonemes. The coupled oscillator planning model also makes the prediction that there should be an intrinsic difference in settling time as a function of syllable structure, even for syllables with an equivalent level of experience. Phone probability was calculated as the sum of the left-to-right position-specific probabilities of the phones that are found in the word, divided by the number of phones in the word. Has language changed? For the lack of a better term, we will refer to specification of temporal control in advance of articulation as a kind of plan, and the deployment of a given control structure as a kind of planning, without implying a commitment to the view that planning and execution are in fact distinct systems. Even though it is not clear how to interpret this result, it underscores the importance of the purely structural variable. Analyzing linguistic data: A practical introduction to statistics. As was expected from the literature (see, e.g. Alternating magnetic fields are generated by three transmitter sensors mounted on a helmet. and transmitted securely. This is in agreement with results from the literature, see e.g. Items with an initial /s/ were produced with the shortest RTac. The existence of a coda does not significantly affect this value. Steriade D. Directional asymmetries of place assimilation: A perceptual account. tea. Goldstein, Pouplier, Chen, Saltzman & Byrd, 2007b, Saltzman, Nam, Krivokapi & Goldstein, 2008, Goldstein, Nam, Saltzman & Chitoran, 2009, Rastle, Croot, Harrington & Coltheart, 2005, Perkell, Cohen, Svirsky, Matthies, Garabieta & Jackson, 1992, Brendel, Erb, Riecker, Grodd, Ackermann & Ziegler (2011), Dilley, Shattuck-Hufnagel & Ostendorf, 1996. In the case of stops, production of voiceless closure causes silence and there will be no energy in the signal until the oral constriction is released, which should take place considerably later than the achievement of its closure. Individual trials were separated by fixed 250 ms intervals. Of further interest here was whether the initiation time differed for varying initial consonants within CV and CVC sequences as was found in Experiment 1 for RTac. eap or tay. A second possible account would be to suggest that the appropriate time point at which to measure RTart is the peak velocity of the glottal constriction (which could not be measured) and if initiation of the vowel constriction gesture (which was actually used to measure RTart) occurs later than the initiation of the glottal gesture, this could account for the apparently slow RTs for VCs. Pinheiro & Bates, 2000; Baayen, 2008, and others) with RTac as the dependent variable, and subject as a random factor. In: Perkell SJ, Klatt HD, editors. To address this question, the data of Experiment 2 were checked for the existence of glottalization. Journal of Experimental Psychology: General. The experiments also illustrate the utility of combining reaction time methods with measurement of articulatory kinematics, a combination that will be necessary for a more complete understanding (and model) of speech production, particularly the regularities of gestural timing. Since the results of the prevocalic and the simple condition are very similar apart from much longer reaction times in the magnitude of 200 ms in the condition without schwa, the results of the simple condition are not reported here. Vowels would be expected to exhibit acoustic consequences much earlier in their articulation than would be the case for voiceless stop consonants, that is, vowels are expected to have a shorter AAI than stop consonants. For LME models for RTart (measured as the interval from the GO signal to the velocity peak of articulatory movement for the initial gesture) 18 of 612 RTart values were excluded as outliers. The interval between the peak of the beep and the articulatory peak velocity was used as the articulatory reaction time, henceforth RTart. Because planning at the level of temporal control is so bound up with the ongoing articulation, it is reasonable to expect that these processes are taking place concurrently and so might be expected to affect delayed naming reaction time. They do not attempt to further specify the details of such a control system, although the gestural coupling graph (Nam & Saltzman, 2003; Goldstein, Byrd & Saltzman, 2006; Nam, 2007; Saltzman et al., 2008) described below is intended to be a representation at the appropriate level of abstraction. In some cases it has been used as a control condition in combination with simple naming to factor out execution-level effects, so that pre-execution processes can be investigated (e.g., Grainger et al., 2000). When there is a single consonant in the onset (e.g., pot), its constriction gesture onset is approximately synchronous for that of the following vowel (Lfqvist & Gracco, 2002), but in the case of an onset cluster (e.g., spot), the vowel gesture is not initiated synchronously with either consonant, but rather its onset is roughly at the midpoint of the consonant sequence (so-called C-center effect; Browman & Goldstein, 1988, 2000; Goldstein, Nam, Saltzman & Chitoran, 2009; Marin & Pouplier, 2010; Khnert, Hoole & Mooshammer, 2006). Reaction time has generally been measured from the acoustic signal and depending on the nature of the initial gesture to be produced, the time from articulatory initiation (which is what we are interested in from the perspective of speech production) to the acoustic onset (the so-called articulatory to acoustic interval, or AAI, see Kawamoto, Lui, Mura & Sanchez 2008) can be quite variable (see Kessler, Treiman & Mullennix, 2002; Rastle, Croot, Harrington & Coltheart, 2005). pea: 30) have a very low number of occurrences, even though they are very frequent in inflected forms (e.g. Non words are printed in italics. In order to compensate for head movements, the signals of two reference coils on the nasion and the gum above the upper incisors were used to translate and rotate the data to a consistent reference centered on the upper incisors. Browman C, Goldstein L. Tiers in articulatory phonology, with some implications for casual speech. Statistical results of the fixed effects are presented by the estimates of the regression coefficients of the model and the standard error of . Freelance Writing Opportunities, Planned Poems:(See poems with this word. The experimental design and post-hoc analysis make it unlikely that the results are due to differences in speed of lexical access or the fact that VC structures can be articulated with initial glottalization. It is possible though to account for the differences in AAI and in the relation of AAI to DurG1 if we examine the voiceless and voiced segments separately. In our corpus, CV(C) come from a denser onset neighborhood than either V(C) or CCV(C) words. Phonology and phonetic evidence: Papers in laboratory phonology IV. In Experiment 1, the hypothesis that syllable structure affected initiation time was supported. This is true for a couple of reasons: the CV can occur in many more polysyllabic words than can the CVC, and also, the CV words tend to be higher frequency words than the CVCs (reflecting an overall preference for CV). In: Sol M-J, Recasens D, Romero J, editors. Did we divide the syllables correctly? A difference in the AAI potentially could also account for the longer latencies in VC than CV sequences, since vocalic gestures are known to have longer durations (time to reach their constriction targets) than consonantal ones (see e.g. Once the phases stabilize, gestures will be triggered by their corresponding clocks. All figures in this table that pertain to corpus frequency are drawn from the Celex Corpus (Baayen, Piepenbrock & Rijn, 1995), based on the 17.9 million words of the COBUILD corpus. Only words with tense vowels were chosen here because lax vowels do not occur in open syllables. The same procedure as for Experiment 1 was used in Experiment 2, i.e. Tabain M, Breen G, Butcher A. VC vs. CV syllables: a comparison of Aboriginal languages with English. In the coupled oscillator model, the planning oscillators are set into motion, at potentially arbitrary phases, and due to entrainment, they will settle into a pattern of stable relative phases. On the other hand, the phone probability measures do not distinguish CVs and CVCs in this way. Grainger J, Spinelli E, Ferrand L. Effects of baseword frequency and orthographic neighborhood size in pseudohomophone naming. Browman C, Goldstein L. Gestural specification using dynamically-defined articulatory structures. Goldstein L, Byrd D, Saltzman E. The role of vocal tract gestural action units in understanding the evolution of phonology. First and most importantly, the task in this experiment is explicitly not a regular rhythmic one of the sort that has elicited the p-center effect. They explain this asymmetry by hypothesizing a shortening of consonants within clusters. Third, it is also known that temporal properties of the syllable later than the vowel onset also affect the p-center (de Jong, 1994; Villing et al., 2011). Yet the VC-CV(C)difference in RTart was actually larger than for stops or fricatives, so this account is also not likely to be correct. A general problem with using standard reaction time methods to study syllable structure is that word forms such as CV vs. VC will necessarily begin with different segments. For CVs with a tense vowel, the frequency of the CV syllable is generally going to be much higher than the CVC that includes that same CV. (2008). Vitevitch M, Luce P. A web-based interface to calculate phonotactic probability for words and nonwords in English. Proceedings of the 15th international congress of phonetic sciences. Because it is impossible to distinguish between multiple glottal stops and glottalization, vowel and lateral initial items were labeled as glottalized if one or more distinct glottal stops were present or if irregular phonation was visible in the spectrogram. (Schwa was elicited in order to avoid articulatory preparation before the imperative signal; see Kawamoto et al., 2008, for a detailed discussion of preparatory movements in delayed naming tasks.) On loops. aHaskins Laboratories, 300 George St., Suite 900, New Haven, CT 06511, U.S.A, bDepartment of Linguistics, Grace Ford Salvatori 301, University of Southern California, Los Angeles, CA 90089-1693, U.S.A, cLinguistics, Yale University, 370 Temple St., New Haven, CT 06520-8366, U.S.A, dDepartment of Physical Therapy & Athletic Training, Boston University, 635 Commonwealth Ave., Boston, MA 02215, U.S.A. They were seated in a quiet room with the computer display in front of them. government site. | Comment on planned. In order to test which of the three variables might be more important in accounting for the RT differences, three LME models were calculated and the improvement of the model evaluated by 2 log-likelihood testing. Sometimes this time-point occurred during irregular voice periods. In order to avoid too many missing values and to distinguish rare words from non-words, 1 was added to the occurrences of all real words and then the logarithm was calculated. The reported significant differences are based on pairwise t-tests with a Bonferroni adjustment for multiple testing. | Words that rhyme with magnanimous. First of all, more repetitions were necessary for the EMMA experiment, and second, an intervening task was included in the EMMA experiment in order to avoid too much boredom. The following lexical and structural properties of the test items were included: word frequency, neighborhood density, onset density, phone and bi-phone probability, and syllable frequency. In: Cole J, Hualde IJ, editors. Perceptual centers in speech production and perception. Marin S. Organization of complex onsets in Romanian. As can be seen, on average, words with CV structure had a higher word frequency than words with VC structure. Investigation of this problem has largely been divided into two separate research areas: studies of internal planning processes (and their timing) that result in a sequence of linguistic units at several levels (phrases, words, feet, syllables, segments), that are ready to be executed, and studies of the articulatory patterning (in space and over time) associated with particular utterance types, observed as they are executed. This study compares the time to initiate words with varying syllable structures (V, VC, CV, CVC, CCV, CCVC). Laterals were approximately in-between stops and fricatives. Response latencies in naming objects. Items in brackets were not included in the corpus for experiment 2. In: Ostry D, Baum S, Menard L, Gracco V, editors. This signal was a sinusoid of 500 Hz, amplitude filtered using a triangular window with a duration of 90 ms. The coupled oscillator planning model (see Nam, 2007) predicts faster oscillator settling time for an initial cluster than for a single C. It is also the case that the biphone probability of the clusters is significantly higher than for any other structural type. For example the frequency of tea is 1062, while that of teak is 23 and teet is 8. There was no significant main effect of VOWEL or interaction between VOWEL and STRUCT. The logic is that effects that show up in simple naming, but not in delayed naming, are assumed to result from pre-execution retrieval and planning. Behavior Research Methods, Instruments, and Computers. Slopes and adjusted R2 for articulatory measured reaction time RTart and several word-based measures. To test whether the exceptional behavior for this item could account for the significant interaction between phone probability and STRUCT a separate regression analysis excluding this item was calculated. Experiment 2 replicates the difference in initiation time between CV(C) and VC syllables, using direct articulatory measurement. Both measures show shorter reaction times for CV(C) than for VC for all consonant types. (2004). One issue that can be raised about the probability effects on planning time is whether experience producing a particular sequence determines the planning time for that sequence, or whether due to some kind of generalization during learning, the probability of a syllable structure class is what controls planning time. Subjects were randomly assigned to start with either condition. Table 3 gives the results of Linear Mixed Effects Models for RTac for each consonant set. The role of speech perception in phonology. Goldstein L, Chitoran I, Selkirk E. Syllable structure as coupled oscillator modes: evidence from Georgian vs. Tashlhiyt Berber.
This alternative explanation to the assembly hypothesis will also be tested with articulatory data in the next section. Data are averaged across speakers. Finally, a third account would be that the glottal constriction is more demanding of planning resources than other gestures, perhaps because it is a prosodically, rather than lexically, controlled gesture for English, and because it is variable in its occurrence. Two of these models that have paid some attention to this issue, however, are the GO-DIVA model (Bohland et al., 2009) and the WEAVER++ model (Levelt et al., 1999). will also be available for a limited time. 
As in Experiment 1, /s/ showed the shortest RTs. After inspecting the data, the simple condition was excluded because the speakers moved their articulators too much during the delay interval, which made it impossible to determine the movement onset for the stimulus production. RTac was affected significantly by STRUCT with reaction times for VC exceeding CV(C) by 39 ms (xVC= 339 ms, xCV(C) = 300 ms). This account is consistent with the coupled oscillator model of planning gestural timing. Unfortunately, given the relative markedness of different syllable structures, structure will necessarily be conflated with syllable frequency and phonotactic probability, particularly in a design that requires a relatively small number of minimal pair stimuli (that is, segmental composition needs to be the same across CV and VC stimuli, for example). Sequences with stops showed only a small, non-significant difference. In general, the assumption has been that in delayed naming of single, syllable-sized units, the time to initiate production, as measured acoustically, depends only on the nature of the initial segment and to a lesser degree the second segment (Kessler et al., 2002). While the effects of factors such as word frequency and neighborhood density can be expected to be mitigated by the use of delayed naming, factors such as syllable frequency and phonotactic probability should reflect the degree to which a gestural pattern is well practiced, and can thus be expected to influence the time required to assemble or deploy a gestural timing plan, and thus affect delayed naming reaction time. Since the participants in Experiment 2 also participated in Experiment 1, their results can be directly compared. Crucially, this is an aspect of planning that has not received attention in the planning literature, and limited investigation in the articulatory patterning literature. Following Vitevitch & Luce (2004), two measures were used for phonotactic probability: phone probability and biphone probability. The reason for this is that for vowels, the end of voicing caused by glottal constriction sometimes preceded the onset of formant transitions, so that no formant transitions were visible. No participant reported any history of speech or hearing impairment. 8600 Rockville Pike The closer the initial phases are to the stable attractor values, the faster the settling process will be. The bold line corresponds to the regression line for stimuli of all different syllables types. Proceedings of the 9th International Seminar on Speech Production. The peak of the beep was identified as the reference point for the onset. Kawamoto A, Kello C. Effect of onset cluster complexity in speeded naming: A test of rule-based approaches. Labeling of acoustic (upper figure) and articulatory data (lower figure). Even though there are problems in identifying the speech event(s) that align most closely with the p-center across different conditions (for an extensive and recent review see Villing, Repp, Ward & Timoney, 2011), it has been found that the acoustic vowel onset (see Janker, 1996), or better, the peak velocity of the vowel gesture (see de Jong, 1994) seem to be reasonable approximations. A higher probability is associated with faster RTs. Thus, data of 12 female and 8 male speakers were used. The stepwise regression analyses and the within-syllable structure category regressions performed on the output of Experiment 2 slightly favor the generalized structural effects, but those results are not particularly strong, and a new experiment with wider sampling of within-category probabilities would be needed to resolve the issue. The results reported here bring to mind the p-center as a possible alternative explanation. This would predict a somewhat smaller VC-CV(C) difference for /l/ words, than for stops or fricatives. Thus, the schwa offset landmark also does not reflect some fixed point in the articulatory trajectory. There has been no clear evidence for these effects in a delayed naming paradigm, in which there is adequate time to access the item and to activate its segments (e.g., Monsell, 1990; Goldinger et al., 1997; Grainger et al., 2000; Mulatti et al., 2006). Linear mixed model for RTac split by consonant groups, and the results from pairwise t-test comparisons. For each consonant type the subset consists of VC syllables with identical coda consonants and of C(C)V(C) syllables with identical initial onset consonants. Regularities in patterning have nonetheless been uncovered, and two examples are given here. For tense vowels this restriction does not exist. A possible account for this could be found in the excess degrees of articulator freedom found in V as opposed to CV syllables. As can be seen in Table 2 items starting with clusters have come from sparser neighborhoods than V(C) items, which in turn have significantly fewer neighbors than CV items. Cohn A. This suggestion is much in the spirit of exemplar models of phonological representation (Pierrehumbert, 2001, 2003), and particularly the dynamical field model of experience on assembling phonological representations (Gafos & Kirov, 2009), which makes explicit connections between experience and the time to assemble phonetic control structures. For each consonant type the subset consisted of VC syllables with identical coda consonants and of CV(C) syllables with identical initial onset consonants. In: Pellegrino F, Marsico E, Chitoran I, Coup C, editors. Planned comparison showed that adding a coda did not have a significant effect on RTac. The upper figure exemplifies the acoustic labeling procedure for the item ape (acoustic signal, spectrogram, RMS for detecting the beep). These examples also illustrate the central role of syllable structure, transparently so in the case of velum-oral coordination. In order to test the hypothesis that different syllable structures require different amounts of time to prepare their temporal controls, or plans, two delayed naming experiments were carried out. Planned comparisons were carried out for significant multilevel factors, such as STRUCT, based on models with uncentered factors and p adjustments for multiple comparisons using the Holm algorithm. Due to technical problems, only 7 repetitions could be recorded for the male subject. A case of surface constraint violation. Characteristics of items used in Experiment 1. Adding Phone and biphone probability did not improve the model significantly (2 =2.44, p> 0.05) compared to the model with only STRUCT. The filled triangles correspond to the peak velocity during the first gesture. With long delays and standard instructions, which do not control the subjects articulation during the delay interval, subjects can effectively begin execution during the delay interval, by completing much of the articulation of the first segment during the delay interval but withholding phonation until the response signal. Fujimura O. Setting up the temporal plan for gestures that are coupled in-phase would be expected to be faster than for those coupled in anti-phase, either because of accessibility of the in-phase mode during assembly of the graph, or due to reduced time required for in-phase oscillators to settle. Honorof D, Browman CP. Here we consider why this might be the case. A linear regression analysis was calculated for the averaged RTart values with phone probability as a numerical predictor and STRUCT as a factorial predictor. In other words, this number gives the average probability of the phones that appear in the word in a given position. Only minor effects of the initial segment were found. For each initial segment, the upper (striped) bar shows the interval between the two acoustic landmarks, end of schwa and acoustic onset. Therefore, the EMMA experiment took more time than the acoustic experiment. (Rastle et al. Lukatela G, Turvey M. An evaluation of the Two-Cycles model of phonology assembly. Hermes A, Ridouane R, Mcke D, Grice M. Kinematics of syllable structure in Tashlhiyt Berber: The case of vocalic and consonantal nuclei. (2008) s analysis, the difference in RTac between stops and fricatives is observed here, where we can be sure early execution did not occur. These changes in material affected the characteristics of word frequency, neighborhood density, onset density, phone and biphone probability and syllable frequency only slightly for words with CVC, CV and VC structures (see Table 6). The relationship is weaker and only marginally significant for biphone probability, unlike Experiment 1, where it was the strongest. Fujimura O. Temporal organization of articulatory movements as a multidimensional phrasal structure. Because it is impractical to run a large group of subjects in an articulatory kinematic experiment, we decided to undertake two versions of a naming task; audio-only (experiment 1), and audio recorded concurrently with articulatory motion (experiment 2). However, in producing a sequence of vowel gestures with no intervening consonant, speakers of English will often produce a glottal constriction, resulting in either silence or glottalization (Pierrehumbert & Talkin, 1992), and reducing or eliminating the diphthongal percept that would otherwise result. In Table 2 lexical characteristics of the items used for this experiment are presented grouped by syllable structure. For the lateral it was more difficult to detect the acoustic onset in a consistent manner.
As in Experiment 1, /s/ showed the shortest RTs. After inspecting the data, the simple condition was excluded because the speakers moved their articulators too much during the delay interval, which made it impossible to determine the movement onset for the stimulus production. RTac was affected significantly by STRUCT with reaction times for VC exceeding CV(C) by 39 ms (xVC= 339 ms, xCV(C) = 300 ms). This account is consistent with the coupled oscillator model of planning gestural timing. Unfortunately, given the relative markedness of different syllable structures, structure will necessarily be conflated with syllable frequency and phonotactic probability, particularly in a design that requires a relatively small number of minimal pair stimuli (that is, segmental composition needs to be the same across CV and VC stimuli, for example). Sequences with stops showed only a small, non-significant difference. In general, the assumption has been that in delayed naming of single, syllable-sized units, the time to initiate production, as measured acoustically, depends only on the nature of the initial segment and to a lesser degree the second segment (Kessler et al., 2002). While the effects of factors such as word frequency and neighborhood density can be expected to be mitigated by the use of delayed naming, factors such as syllable frequency and phonotactic probability should reflect the degree to which a gestural pattern is well practiced, and can thus be expected to influence the time required to assemble or deploy a gestural timing plan, and thus affect delayed naming reaction time. Since the participants in Experiment 2 also participated in Experiment 1, their results can be directly compared. Crucially, this is an aspect of planning that has not received attention in the planning literature, and limited investigation in the articulatory patterning literature. Following Vitevitch & Luce (2004), two measures were used for phonotactic probability: phone probability and biphone probability. The reason for this is that for vowels, the end of voicing caused by glottal constriction sometimes preceded the onset of formant transitions, so that no formant transitions were visible. No participant reported any history of speech or hearing impairment. 8600 Rockville Pike The closer the initial phases are to the stable attractor values, the faster the settling process will be. The bold line corresponds to the regression line for stimuli of all different syllables types. Proceedings of the 9th International Seminar on Speech Production. The peak of the beep was identified as the reference point for the onset. Kawamoto A, Kello C. Effect of onset cluster complexity in speeded naming: A test of rule-based approaches. Labeling of acoustic (upper figure) and articulatory data (lower figure). Even though there are problems in identifying the speech event(s) that align most closely with the p-center across different conditions (for an extensive and recent review see Villing, Repp, Ward & Timoney, 2011), it has been found that the acoustic vowel onset (see Janker, 1996), or better, the peak velocity of the vowel gesture (see de Jong, 1994) seem to be reasonable approximations. A higher probability is associated with faster RTs. Thus, data of 12 female and 8 male speakers were used. The stepwise regression analyses and the within-syllable structure category regressions performed on the output of Experiment 2 slightly favor the generalized structural effects, but those results are not particularly strong, and a new experiment with wider sampling of within-category probabilities would be needed to resolve the issue. The results reported here bring to mind the p-center as a possible alternative explanation. This would predict a somewhat smaller VC-CV(C) difference for /l/ words, than for stops or fricatives. Thus, the schwa offset landmark also does not reflect some fixed point in the articulatory trajectory. There has been no clear evidence for these effects in a delayed naming paradigm, in which there is adequate time to access the item and to activate its segments (e.g., Monsell, 1990; Goldinger et al., 1997; Grainger et al., 2000; Mulatti et al., 2006). Linear mixed model for RTac split by consonant groups, and the results from pairwise t-test comparisons. For each consonant type the subset consists of VC syllables with identical coda consonants and of C(C)V(C) syllables with identical initial onset consonants. Regularities in patterning have nonetheless been uncovered, and two examples are given here. For tense vowels this restriction does not exist. A possible account for this could be found in the excess degrees of articulator freedom found in V as opposed to CV syllables. As can be seen in Table 2 items starting with clusters have come from sparser neighborhoods than V(C) items, which in turn have significantly fewer neighbors than CV items. Cohn A. This suggestion is much in the spirit of exemplar models of phonological representation (Pierrehumbert, 2001, 2003), and particularly the dynamical field model of experience on assembling phonological representations (Gafos & Kirov, 2009), which makes explicit connections between experience and the time to assemble phonetic control structures. For each consonant type the subset consisted of VC syllables with identical coda consonants and of CV(C) syllables with identical initial onset consonants. In: Pellegrino F, Marsico E, Chitoran I, Coup C, editors. Planned comparison showed that adding a coda did not have a significant effect on RTac. The upper figure exemplifies the acoustic labeling procedure for the item ape (acoustic signal, spectrogram, RMS for detecting the beep). These examples also illustrate the central role of syllable structure, transparently so in the case of velum-oral coordination. In order to test the hypothesis that different syllable structures require different amounts of time to prepare their temporal controls, or plans, two delayed naming experiments were carried out. Planned comparisons were carried out for significant multilevel factors, such as STRUCT, based on models with uncentered factors and p adjustments for multiple comparisons using the Holm algorithm. Due to technical problems, only 7 repetitions could be recorded for the male subject. A case of surface constraint violation. Characteristics of items used in Experiment 1. Adding Phone and biphone probability did not improve the model significantly (2 =2.44, p> 0.05) compared to the model with only STRUCT. The filled triangles correspond to the peak velocity during the first gesture. With long delays and standard instructions, which do not control the subjects articulation during the delay interval, subjects can effectively begin execution during the delay interval, by completing much of the articulation of the first segment during the delay interval but withholding phonation until the response signal. Fujimura O. Setting up the temporal plan for gestures that are coupled in-phase would be expected to be faster than for those coupled in anti-phase, either because of accessibility of the in-phase mode during assembly of the graph, or due to reduced time required for in-phase oscillators to settle. Honorof D, Browman CP. Here we consider why this might be the case. A linear regression analysis was calculated for the averaged RTart values with phone probability as a numerical predictor and STRUCT as a factorial predictor. In other words, this number gives the average probability of the phones that appear in the word in a given position. Only minor effects of the initial segment were found. For each initial segment, the upper (striped) bar shows the interval between the two acoustic landmarks, end of schwa and acoustic onset. Therefore, the EMMA experiment took more time than the acoustic experiment. (Rastle et al. Lukatela G, Turvey M. An evaluation of the Two-Cycles model of phonology assembly. Hermes A, Ridouane R, Mcke D, Grice M. Kinematics of syllable structure in Tashlhiyt Berber: The case of vocalic and consonantal nuclei. (2008) s analysis, the difference in RTac between stops and fricatives is observed here, where we can be sure early execution did not occur. These changes in material affected the characteristics of word frequency, neighborhood density, onset density, phone and biphone probability and syllable frequency only slightly for words with CVC, CV and VC structures (see Table 6). The relationship is weaker and only marginally significant for biphone probability, unlike Experiment 1, where it was the strongest. Fujimura O. Temporal organization of articulatory movements as a multidimensional phrasal structure. Because it is impractical to run a large group of subjects in an articulatory kinematic experiment, we decided to undertake two versions of a naming task; audio-only (experiment 1), and audio recorded concurrently with articulatory motion (experiment 2). However, in producing a sequence of vowel gestures with no intervening consonant, speakers of English will often produce a glottal constriction, resulting in either silence or glottalization (Pierrehumbert & Talkin, 1992), and reducing or eliminating the diphthongal percept that would otherwise result. In Table 2 lexical characteristics of the items used for this experiment are presented grouped by syllable structure. For the lateral it was more difficult to detect the acoustic onset in a consistent manner.