JSLHR
Research Article
Bidirectional Interference Between Speech and Nonspeech Tasks in Younger, Middle-Aged, and Older Adults Dallin J. Baileya and Christopher Dromeya
Purpose: The purpose of this study was to examine divided attention over a large age range by looking at the effects of 3 nonspeech tasks on concurrent speech motor performance. The nonspeech tasks were designed to facilitate measurement of bidirectional interference, allowing examination of their sensitivity to speech activity. A cross-sectional design was selected to explore possible changes in divided-attention effects associated with age. Method: Sixty healthy participants were separated into 3 groups of 20: younger (20s), middle-aged (40s), and older (60s) adults. Each participant completed a speech task (sentence repetitions) once in isolation and once concurrently with each of 3 nonspeech tasks: a semantic-
decision linguistic task, a quantitative-comparison cognitive task, and a manual motor task. The nonspeech tasks were also performed in isolation. Results: Data from speech kinematics and nonspeech task performance indicated significant task-specific divided attention interference, with divided attention affecting speech and nonspeech measures in the linguistic and cognitive conditions and affecting speech measures in the manual motor condition. There was also a significant age effect for utterance duration. Conclusions: The results increase what is known about bidirectional interference between speech and other concurrent tasks as well as age effects on speech motor control.
S
Models of Divided Attention
peech is a highly sophisticated feat of fine motor performance. However, research has shown that speech production can be affected in subtle but measureable ways by the concurrent performance of other tasks, even in individuals without speech disorders (Dromey & Bates, 2005; Dromey & Benson, 2003; Dromey & Shim, 2008). As with other activities, when a person speaks while performing another task, one or both tasks can be affected. This may cause a decline in performance known as interference. The purpose of the present study was to explore bidirectional interference involving speech and concurrent nonspeech tasks across groups of younger, middle-aged, and older adults.
a
Brigham Young University, Provo, UT
Correspondence to Dallin Bailey, who is now at the Department of Veterans Affairs and the University of Utah, Salt Lake City: [email protected] Editor: Jody Kreiman Associate Editor: Julie Liss Received March 12, 2014 Revision received September 30, 2014 Accepted July 6, 2015 DOI: 10.1044/2015_JSLHR-S-14-0083
Cognitive psychology models of attention may help explain why interference occurs in dual-task situations. Such models typically make use of one or both of two basic ideas: structural mechanisms and capacities (Wickens, 1980). Although the present study was not designed to explicitly test any of these models, the models are used in interpreting the experimental results. Structural theories view attention as an information processing mechanism that forces serial processing, or a “bottleneck,” at some stage of task performance (Navon & Miller, 2002; Wickens, 1980, 1981); thus, interference in dual-task conditions could be explained by a delay in the second task due to the mechanism being occupied with the first task (Tombu & Jolicœur, 2003). However, a strict interpretation of the structural model fails to predict some aspects of interference (Navon & Miller, 2002), such as mediation by stage and modality of processing (Wickens, 1980). These phenomena are better explained by modeling attention as a resource or capacity. A central capacity could allow parallel processing; however, the finite limits of the pool of attentional resources may be insufficient for the error-free processing of concurrent tasks (Kahneman, 1973; Norman & Bobrow, 1975). Differences
Disclosure: The authors have declared that no competing interests existed at the time of publication.
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1637–1653 • December 2015 • Copyright © 2015 American Speech-Language-Hearing Association
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1637
in interference could result from the amount of the resource each task requires compared with the available capacity (Norman & Bobrow, 1975). Navon and Gopher (1979) rejected the idea of a single, central pool in favor of multiple pools of attentional resources dedicated to different types of cognitive processes. A multiple resources model more accurately explains the effect of task type or modality on interference in dual-task paradigms. Navon and Gopher cited Brooks (1968), who showed that performance of two concurrent visual tasks (or two concurrent verbal tasks) resulted in greater interference than performance of a visual task and a verbal task. A recent revision of the multiple dimensions model of attention (Wickens, 2008) specifies four distinct dimensions, with experimental differences in dual-task interference explained by the separation of the task types along these dimensions. These four dimensions include stages of processing (perceptual and cognitive resources vs. resources for the performance of the action), codes of processing (resources of spatial vs. verbal or linguistic activity), perceptual modalities (auditory vs. visual perception), and visual channels (focal vs. ambient vision; Wickens, 1981, 2008). These dimensions have, as stated by Wickens (2008), neurophysiological plausibility in that they coincide with basic neuroanatomical divisions, such as perceptual activity occurring posterior to the central sulcus, and motor activity more anteriorly.
Studies of Divided Attention and Bidirectional Interference Divided-attention effects have been demonstrated to occur during a wide variety of tasks, including memory (Naveh-Benjamin, Guez, & Marom, 2003; Troyer & Craik, 2000), verbal and visual tasks (Brooks, 1968), speech fluency (Oomen & Postma, 2001), manual tasks (Dromey & Bates, 2005; Dromey & Shim, 2008; Talland, 1962), gait (Camicioli, Howieson, Lehman, & Kaye, 1997; Chen et al., 1996), and postural stability (Dromey et al., 2010). Previous work has also shown that divided attention also affects speech motor performance. For example, Dromey and Benson (2003) found changes in lip kinematics that suggested reduced stability of speech motor control when young adult participants performed various nonspeech distractor tasks. In follow-up studies, Dromey and Bates (2005) and Dromey and Shim (2008) examined performance on the secondary task in addition to performance on the speech task in order to identify the bidirectional interference between speech and nonspeech tasks. It was found that different combinations of tasks caused different patterns of bidirectional interference, with evidence that speech is influenced by the nature of the secondary task, although task difficulty likely also plays a role. Interference between speech and nonspeech tasks has been demonstrated to occur in adults who are neurologically healthy and those with neurological disorders. In a study of individuals with Parkinson’s disease (PD) and a healthy control group, Dromey et al. (2010) found bidirectional
1638
interference between postural stability and diphthong production in a speech task in both the group with PD and the healthy controls. Camicioli, Oken, Sexton, Kaye, and Nutt (1998) also found interference between speech and gross motor tasks in people with PD and in healthy older adults, observing that performing a verbal fluency task led to reduced walking speed and increased number of steps in a concurrent walking task. In another study, Holmes, Jenkins, Johnson, Adams, and Spaulding (2010) found that when speaking distractor tasks were relatively complex, participants both with and without PD experienced measurable changes in postural stability. These findings of interference between speech and nonspeech activities have significant implications for increasing our understanding of the robustness and limitations of speech motor activity both in healthy populations and in populations with speech disorders, as multitasking is a frequent occurrence in everyday life.
Aging and Divided Attention A major complicating factor in the study of dividedattention interference is the effect of aging over the adult life span. Work by Salthouse (2009) indicates that cognitive abilities generally decline, but the trajectories may differ widely. Research has also suggested an age-related multitasking deficit. For example, Talland (1962) found that middle-aged and elderly men were slower at performing two separate motor tasks (one with each hand) than younger men. Chen et al. (1996) found that divided attention negatively affected older adults more than younger adults during an obstacle-avoidance walking task. Evidence from a meta-analysis of 33 dual-task studies (Verhaeghen, Steitz, Sliwinski, & Cerella, 2003) and from experimental research focused on the issue (Salthouse, Fristoe, Lineweaver, & Coon, 1995) showed that even after controlling for singletask performance, task complexity, and age-related declines in the component tasks, occasionally a small but significant age-related effect on dual-task performance remains. These age-related increases in interference have previously been demonstrated during dual-task paradigms including speech motor activity. Dromey et al. (2010) showed that a speech repetition task led to reduced heel height in a concurrent postural stability task in a group of healthy older adults (M = 70.5 years) compared with a group of younger adults (M = 25.5 years). In terms of models of attention, the findings of these studies could be interpreted as a decline in available attentional resources or a decrease in efficiency of the executive control process(es) in older adults compared with younger adults. Besides cognitive factors, another reason divided attention disproportionately affects older individuals may be the aging of the motor control system, including motor speech mechanisms. It is well known that movement control is gradually impaired with increasing age, primarily due to reduced strength, peripheral sensation, balance, and coordination of multijoint movements and increased reaction time (Ketcham, Dounskaia, & Stelmach, 2004; Menz,
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1637–1653 • December 2015
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Lord, & Fitzpatrick, 2003). This motor decline affects both speech and nonspeech movements of the articulators. Ballard, Robin, Woodworth, and Zimba (2001) found that children and older adults were less able than young adults to perform a nonspeech oral visuomotor task and were less accurate in matching biofeedback of the amplitude of the movement of their lower lip, jaw, or fundamental frequency with the movement of a visual target. Smith, Wasowicz, and Preston (1987) found that older adults (M = 70 years) had longer utterance durations than younger adults (M = 25 years) at both habitual and fast speaking rates, measured at the segment, syllable, and sentence levels. Sadagopan and Smith (2013) compared oral movements of younger (M = 20 years 11 months) and older (M = 69 years) adults during nonword repetition; although no age effect was found, they reported that only older speakers with “younglike” nonword repetition ability (six out of 16) were included in the kinematic analyses, as the other 10 were unable to produce a sufficient number of accurate nonword productions for analysis. This suggests that age-related differences in motor speech ability that may be important were not analyzed due to the study’s methodological requirements. Wohlert and Smith (1998) found that the spatiotemporal index (STI; Smith, Goffman, Zelaznik, Ying, & McGillem, 1995), which is a measure of the movement variability across multiple repetitions of the same utterance, was significantly higher in a group of older adults (M = 79 years) than in younger adults (M = 24 years) when the participants spoke at a habitual rate; again, it was found that utterance durations were longer for the older adults at habitual, as well as slow and fast, rates. The authors speculated that their findings could have been at least partially attributable to changes in peripheral sensorimotor mechanisms. This overall decline could undermine the speech motor performance of older adults in dual-task conditions.
Research Questions and Hypotheses Research examining how age influences divided attention for speech has so far been limited (Dromey et al., 2010). Little is known about how bidirectional interference changes when different nonspeech tasks are performed with speech motor activity in dual-task paradigms. In addition, it is not yet known how bidirectional interference in such dual-task paradigms changes with increasing age in adulthood. The present experiment was performed to identify the effects of three nonspeech tasks on speech motor stability as well as possible effects of the speech task on performance of the three nonspeech tasks. In addition, the effect of age was examined, with participants coming from three cross-sectional groups: younger, middle-aged, and older adults. Our specific research questions were as follows: 1.
How is speech motor activity affected by qualitatively different nonspeech tasks performed concurrently?
2.
How is nonspeech task performance affected by concurrent speech?
3.
What effect does adult age have on bidirectional interference between speech and nonspeech tasks?
On the basis of the research reviewed above, it was hypothesized that all participants would experience bidirectional interference, with reduced speech motor stability and reduced nonspeech task performance when divided-attention performance was compared with performance in a singletask condition. Reduced speech motor stability would likely include higher STI values and lower values of lip and jaw displacement and velocity. Reduced nonspeech-task performance would likely include decreases in the rate or accuracy of responses. It was further hypothesized that older adults would demonstrate comparatively greater dual-task costs compared with younger adults. Thus, we predicted that the younger adults would demonstrate the least susceptibility to bidirectional interference compared with the middle-aged and older adults and that the middle-aged adult group would fall somewhere between the other two groups. Differences between the age groups would give preliminary indications of the age at which speech motor control may become more susceptible to divided-attention effects. On the basis of previous work, it was also reasoned that sex differences in interference might exist (Dromey & Benson, 2003). Observing interference of speech on nonspeech performance should give insight into the extent to which speech is automatic, how speech shares attentional resources, and how this sharing may change with age.
Method Participants Thirty men and 30 women participated in the study. They had no history of speech or language disorders with the exception of one participant who reported having typical articulation errors that were treated during childhood. Each of three age groups was divided evenly into 10 male and 10 female participants. The age groups included younger adults (ages 20–28 years, M = 22.95 years, SD = 2.35), middle-aged adults (ages 40–50 years, M = 45.60 years, SD = 3.47), and older adults (ages 58–70 years, M = 63.20 years, SD = 3.55). These age groups were selected based on the assumption that separating the age groups with about a decade between them would allow the observation of potential age-related differences across part of the adult life span in a cross-sectional design. All participants were native speakers of English and signed informed consent documents approved by the institutional review board prior to participation in the study. Hearing status was not formally evaluated, although all participants engaged in conversation with the experimenters without effort and reported no difficulties in everyday communication. No formal cognitive or linguistic screening was administered prior to participation in the study; however, retrospective analysis of the performance on experimental trials of these tasks indicates that no participant scored at or below chance levels for either task in either condition. The relatively high average accuracies (97% and 96% for the linguistic
Bailey & Dromey: Speech and Nonspeech Task Interference in Adults
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1639
task and 93% and 89% for the cognitive task in isolated and dual-task conditions, respectively) suggest levels of ability appropriate for the selected tasks. As an anecdote, the participants in all age groups were active in educational, vocational, social, and civic activities and likely had typical cognitive and linguistic abilities, although this information was not formally collected.
Equipment Each participant was seated in a sound booth to allow high-quality audio recordings and to reduce potential auditory distractions. A lightweight head-mounted strain gauge system (Barlow, Cole, & Abbs, 1983) was used to measure the vertical movements of each participant’s lips and jaw. Three cantilever beams were attached to the skin using double-sided tape at the midpoint of the vermillion border of the upper lips (UL) and lower lips (LL) and under the chin to track the jaw. These cantilevers were displaced incrementally with a sliding micrometer to allow calibration of the displacement and velocity signals in the softwarederived computations. The cantilevers were connected to bridge amplifiers (model 215, Biocommunication Electronics, Madison, WI) that produced analog voltage signals. A microphone (model C420, AKG, Vienna, Austria) was attached to the headset and placed about 10 cm away from the participant’s mouth in order to record speech. A sound level meter (model 712, Larson Davis, Depew, NY, or model 407736, Extech Instruments, Nashua, NH; C-weighting, fast response) was placed 100 cm away from the participant’s mouth and measured speech intensity in dB SPL. The analog signals from these instruments were digitized using a Windaq 720 (DATAQ Instruments, Akron, OH) analog–digital converter. The sampling rate was 1 kHz for the kinematic and sound level meter channels and 25 kHz (following 12-kHz low-pass filtering; model 9002, Frequency Devices, Ottawa, IL) for the audio channel. A computer with a 24-in. flat panel monitor located outside the window of the sound booth and a computer mouse located inside the sound booth were used for the cognitive and linguistic tasks.
Procedures All participants completed four types of tasks selected to represent broader domains of behavior: a speech task, a linguistic task, a cognitive task, and a manual motor task. Each participant completed one 30-s practice trial of each task before beginning the experimental trials to mitigate any possible practice effects. The practice trials were identical to the experimental trials except for the reduced duration. The practice speech task consisted of the same phrase to be repeated as the experimental trials. Practice items in the cognitive and linguistic tasks were similar to those in experimental trials; no item was ever repeated for a given participant. Following the practice trials, all four tasks were performed in the isolated condition. Last, each of the nonspeech tasks was performed concurrently with repetition of the target sentence. Specific feedback on accuracy of
1640
performance was not given after practice trials or any of the experimental trials, although it should be noted that visual feedback was available, by nature of the activity, during manual motor task performance. No feedback, visual or otherwise, was available during the other tasks. Order of the tasks within the isolated and dual-task blocks was counterbalanced across participants. The total time to complete the tasks, including equipment setup and training, was about 45 min per participant.
Tasks Individual tasks were chosen to represent separate domains that were believed to be behaviorally and functionally distinct from one another. They included activities reflecting a focus on speaking, cognitive, and linguistic abilities and manual dexterity. This was done to examine the relative effects of nonspeech task type on speech variables during dual-task conditions. In this article we refer to them by their focus domains for the sake of convenience. The tasks were designed to be mildly challenging in order to elicit purposeful attention but without being so difficult that they would discourage participants from attempting them. The speech task entailed speaking the phrase “I saw Patrick pull a wagon packed with apples” every time the participant heard a tone. The tone was repeated at regular intervals of about 4 s. The vowels and consonants in this sentence require large jaw and lip movements and were chosen to facilitate signal segmentation during analysis. During divided-attention conditions, participants repeated the phrase 14 to 15 times while performing one of the three nonspeech tasks. For the linguistic task, participants performed a semantic-decision activity similar to the one described in a lexical semantics neuroimaging study by Müller, Kleinhans, and Courchesne (2003). Their semantic-decision task involved viewing a computer screen that showed two words, a noun and a verb, and responding whether the two were semantically related. Novel stimuli were created for this study and were based on high-frequency nouns and verbs in the Corpus of Contemporary American English (Davies, 2012). For example, joke–laughing made a semantic pair in the present study, whereas dough–interviewing did not make a semantic pair. All participants were given the same practice and experimental lists, although the condition of the two experimental lists (isolated vs. dual task) was counterbalanced across participants. The participants were given 60 s to categorize as many pairs of words as possible. During the divided-attention condition, participants completed the semantic-decision task while they were repeating the target sentence. For the cognitive task, participants performed a quantity-comparison activity. This task was designed to be analogous to the semantic-decision task. This task involved viewing a computer screen that showed two numerical values in fraction notation with an equal sign in between them; the participants used a computer mouse to select one button if the quantity comparison was correct (e.g.,
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1637–1653 • December 2015
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
2/3 = 4/6) or another button if the quantity comparison was not correct (e.g., 1/4 = 4/8). The participants were given 60 s to evaluate as many quantity comparisons as possible. During the divided-attention condition, participants completed the quantity-comparison task while they were repeating the target sentence. The manual motor task consisted of the Purdue Pegboard Test (Tiffin, 1948). Participants were instructed to perform the test with both hands in order to eliminate handedness and dominance effects across participants. Manual motor performance was quantified as the number of pegs placed in 60 s. During the divided-attention condition, participants placed pegs in the pegboard while they completed the speech task.
Data Analysis Speech-dependent measures were taken from 10 tokens of the target phrase in each of the conditions (the speechonly condition and the three divided-attention conditions). The speech measures were selected from those that have shown sensitivity to changes due to age and divided-attention conditions in previous research (Dromey & Benson, 2003; Tingley & Dromey, 2000; Wohlert & Smith, 1998). The first 10 perceptually correct tokens for each participant for each condition were segmented for analysis. Tokens judged as perceptually incorrect by an experimenter (the first author) were excluded at this time. Tokens that were judged as correct contained all of the words of the target phrase in the correct sequence without significant disfluencies (i.e., articulatory errors, changes in word order, and substantial inconsistencies or differences in prosody). The digital signals associated with these tokens were segmented with custom MATLAB (The Mathworks, 2009) routines on the basis of consistent markers in the velocity record, as shown in the lower panel of Figure 1. The markers used for signal segmentation were the downward peak during the LL opening from the /p/ to the /æ/ in Patrick and the upward peak during the LL closing from the /æ/ to the /p/ in apples. Figure 1. Sample displacement and velocity record for one utterance, showing segmentation points.
These segments were then analyzed for duration, STI for the LL, and SPL. Utterance duration (ms) was selected as a representation of speaking rate. STI, which is a measure of the variability of speech movements over multiple repetitions, was calculated by time and amplitude normalizing the 10 displacement waveforms for each condition and calculating the sum of the displacement standard deviations at 50 equally spaced points throughout the movement record (Smith et al., 1995). Previous research has shown increases in STI for older adults (Wohlert & Smith, 1998) during speech and for younger adults in divided-attention conditions (Dromey & Bates, 2005; Dromey & Benson, 2003; Dromey & Shim, 2008). Vocal intensity was included because a previous study in our laboratory (Dromey & Bates, 2005) reported that vocal intensity increases in younger adults speaking under dual-task conditions, including tasks similar to those used in the present study. In addition, a single closing gesture (the closing gesture after the low, front vowel of packed) was segmented from the displacement record as shown in the upper panel of Figure 1 and used to make measurements of LL displacement, LL peak velocity, and UL–LL correlation. LL displacement and peak velocity are measures of the magnitude and rate of LL movement for a selected articulatory gesture in the utterance. UL–LL correlation is an index reflecting the coordination of lip movements, with a value of –1 indicating that the lips are moving in opposite directions, as would be expected during a bilabial opening or closing movement. This correlation measure has previously been used to show differences between typical speakers and speakers with a disorder (Tingley & Dromey, 2000). It was included in the present study to examine how lip coordination during speech was affected by the various divided-attention conditions. Measurements of the nonspeech tasks (linguistic, cognitive, and manual motor) in the isolated condition were taken to compare with measurements of the same tasks in the divided-attention condition in order to quantify the impact of speech activity on these nonspeech tasks. The cognitive and linguistic tasks were timed at 60 s and were scored as the total number of responses, the number of correct responses (total responses minus incorrect responses), and the accuracy of responses (number of correct responses divided by number of total responses). The manual motor task was scored as the number of pegs placed in 60 s.
Results Speech dependent measures were based on the average of the 10 selected tokens of the target phrase per condition. Participants typically produced about 14 tokens per trial. Two participants (one middle-aged woman and one middle-aged man) failed to produce 10 correct utterances during the concurrent cognitive task; measures for those participants in that condition were excluded from analysis. Equipment malfunction prevented the collection of intensity data for one of the participants (a middle-aged man) in the speech-only and cognitive dual-task conditions.
Bailey & Dromey: Speech and Nonspeech Task Interference in Adults
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1641
In addition, concurrent manual motor task data for one participant (an older man) were not collected due to a procedural error. The descriptive statistics for each dependent variable were calculated for each of the three age groups and separated within age group by sex. Descriptive statistics for all measurements are available in the Appendix.
Repeated Measures Analyses of Variance The effect of condition on the kinematic and intensity measures for the speech task in isolation and during each divided-attention condition was tested with a withinparticipant repeated measures analysis of variance (ANOVA), with age and sex as between-participants factors. In addition, Tukey’s honestly significant difference post hoc analyses were performed to more closely examine age group and sex contrasts. The effect of condition on response counts and accuracy rates for the linguistic, cognitive, and manual motor nonspeech tasks was also tested with a separate but identical within-participant repeated measures ANOVA, again with age and sex as between-participants factors. If a case had missing data for any of the variables in the repeated measures model, the entire case was excluded from the model for that particular variable. When Mauchly’s test of a sphericity violation was significant, the corrected HuynhFeldt values were used. Because proportion and correlation coefficient data are typically not normally distributed, they were transformed (arcsine and Fisher z, respectively) prior to analysis.
Condition Effects on Speech and Nonspeech Task Performance There was a significant main effect of condition on all speech variables except LL velocity. There were also many significant contrasts. The F ratios and p values for the speech measures are displayed in Table 1. Means and error bars for utterance duration, LL displacement, and velocity
are reported in Figure 2. There was a main effect of condition on duration, F(2.56, 138.11) = 12.00, p < .001. Contrasts indicate that duration was significantly longer during the linguistic and cognitive concurrent tasks compared with the speech-only condition. There was also a main effect of condition on LL displacement, F(2.59, 139.99) = 3.25, p = .03. Contrasts indicate that LL displacement decreased during the concurrent performance of the manual motor task. Means and error bars for UL–LL correlation, LL STI, and intensity are shown in Figure 3. Correlation between the UL and LL became more strongly negative during concurrent task performance, F(2.78, 150.17) = 3.38, p = .023, with contrasts showing significant differences occurring between the speech-only condition and the concurrent linguistic and cognitive conditions. There was also a main effect of condition on the STI of the LL, F(3, 162) = 29.40, p < .001; contrasts showed significant increases occurring from the speech-only condition to the linguistic and cognitive conditions. Last, there was also a main effect of condition on SPL, F(2.76, 146.20) = 33.00, p < .001, with contrasts showing an increase in SPL from baseline isolated speech performance for the manual motor condition only. There was also a significant condition main effect on some of the nonspeech task measures. The F ratios and p values for the nonspeech concurrent tasks are displayed in Table 2. The means and standard deviations of the nonspeech task response counts are shown in Figure 4. The linguistic and cognitive task accuracy rates and their standard deviations are shown in Figure 5. Linguistic task measures that significantly differed between conditions included the number of correct responses, F(1, 56) = 210.83, p < .001, and the number of total responses, F(1, 56) = 211.93, p < .001, with both decreasing compared with the single-task condition. One cognitive task measure also decreased significantly between conditions: the number of correct responses, F(1, 56) = 5.88, p = .019. Condition did not significantly affect the number of pegs the participants placed in the board.
Table 1. Significant repeated measures analysis of variance effects of condition on speech measures, with the concurrent task contrasted against the isolated-speech condition. Condition main effect Variable
df
Duration (ms) Displacement (mm) Lower lip velocity (mm/s) Upper lip–lower lip correlation Lower lip spatiotemporal index dB SPL at 100 cm
F
p
Linguistic contrast
Cognitive contrast
df
df
F
p
F
p
df
F
p
2.56, 138.11 12.00
Research Article
Bidirectional Interference Between Speech and Nonspeech Tasks in Younger, Middle-Aged, and Older Adults Dallin J. Baileya and Christopher Dromeya
Purpose: The purpose of this study was to examine divided attention over a large age range by looking at the effects of 3 nonspeech tasks on concurrent speech motor performance. The nonspeech tasks were designed to facilitate measurement of bidirectional interference, allowing examination of their sensitivity to speech activity. A cross-sectional design was selected to explore possible changes in divided-attention effects associated with age. Method: Sixty healthy participants were separated into 3 groups of 20: younger (20s), middle-aged (40s), and older (60s) adults. Each participant completed a speech task (sentence repetitions) once in isolation and once concurrently with each of 3 nonspeech tasks: a semantic-
decision linguistic task, a quantitative-comparison cognitive task, and a manual motor task. The nonspeech tasks were also performed in isolation. Results: Data from speech kinematics and nonspeech task performance indicated significant task-specific divided attention interference, with divided attention affecting speech and nonspeech measures in the linguistic and cognitive conditions and affecting speech measures in the manual motor condition. There was also a significant age effect for utterance duration. Conclusions: The results increase what is known about bidirectional interference between speech and other concurrent tasks as well as age effects on speech motor control.
S
Models of Divided Attention
peech is a highly sophisticated feat of fine motor performance. However, research has shown that speech production can be affected in subtle but measureable ways by the concurrent performance of other tasks, even in individuals without speech disorders (Dromey & Bates, 2005; Dromey & Benson, 2003; Dromey & Shim, 2008). As with other activities, when a person speaks while performing another task, one or both tasks can be affected. This may cause a decline in performance known as interference. The purpose of the present study was to explore bidirectional interference involving speech and concurrent nonspeech tasks across groups of younger, middle-aged, and older adults.
a
Brigham Young University, Provo, UT
Correspondence to Dallin Bailey, who is now at the Department of Veterans Affairs and the University of Utah, Salt Lake City: [email protected] Editor: Jody Kreiman Associate Editor: Julie Liss Received March 12, 2014 Revision received September 30, 2014 Accepted July 6, 2015 DOI: 10.1044/2015_JSLHR-S-14-0083
Cognitive psychology models of attention may help explain why interference occurs in dual-task situations. Such models typically make use of one or both of two basic ideas: structural mechanisms and capacities (Wickens, 1980). Although the present study was not designed to explicitly test any of these models, the models are used in interpreting the experimental results. Structural theories view attention as an information processing mechanism that forces serial processing, or a “bottleneck,” at some stage of task performance (Navon & Miller, 2002; Wickens, 1980, 1981); thus, interference in dual-task conditions could be explained by a delay in the second task due to the mechanism being occupied with the first task (Tombu & Jolicœur, 2003). However, a strict interpretation of the structural model fails to predict some aspects of interference (Navon & Miller, 2002), such as mediation by stage and modality of processing (Wickens, 1980). These phenomena are better explained by modeling attention as a resource or capacity. A central capacity could allow parallel processing; however, the finite limits of the pool of attentional resources may be insufficient for the error-free processing of concurrent tasks (Kahneman, 1973; Norman & Bobrow, 1975). Differences
Disclosure: The authors have declared that no competing interests existed at the time of publication.
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1637–1653 • December 2015 • Copyright © 2015 American Speech-Language-Hearing Association
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1637
in interference could result from the amount of the resource each task requires compared with the available capacity (Norman & Bobrow, 1975). Navon and Gopher (1979) rejected the idea of a single, central pool in favor of multiple pools of attentional resources dedicated to different types of cognitive processes. A multiple resources model more accurately explains the effect of task type or modality on interference in dual-task paradigms. Navon and Gopher cited Brooks (1968), who showed that performance of two concurrent visual tasks (or two concurrent verbal tasks) resulted in greater interference than performance of a visual task and a verbal task. A recent revision of the multiple dimensions model of attention (Wickens, 2008) specifies four distinct dimensions, with experimental differences in dual-task interference explained by the separation of the task types along these dimensions. These four dimensions include stages of processing (perceptual and cognitive resources vs. resources for the performance of the action), codes of processing (resources of spatial vs. verbal or linguistic activity), perceptual modalities (auditory vs. visual perception), and visual channels (focal vs. ambient vision; Wickens, 1981, 2008). These dimensions have, as stated by Wickens (2008), neurophysiological plausibility in that they coincide with basic neuroanatomical divisions, such as perceptual activity occurring posterior to the central sulcus, and motor activity more anteriorly.
Studies of Divided Attention and Bidirectional Interference Divided-attention effects have been demonstrated to occur during a wide variety of tasks, including memory (Naveh-Benjamin, Guez, & Marom, 2003; Troyer & Craik, 2000), verbal and visual tasks (Brooks, 1968), speech fluency (Oomen & Postma, 2001), manual tasks (Dromey & Bates, 2005; Dromey & Shim, 2008; Talland, 1962), gait (Camicioli, Howieson, Lehman, & Kaye, 1997; Chen et al., 1996), and postural stability (Dromey et al., 2010). Previous work has also shown that divided attention also affects speech motor performance. For example, Dromey and Benson (2003) found changes in lip kinematics that suggested reduced stability of speech motor control when young adult participants performed various nonspeech distractor tasks. In follow-up studies, Dromey and Bates (2005) and Dromey and Shim (2008) examined performance on the secondary task in addition to performance on the speech task in order to identify the bidirectional interference between speech and nonspeech tasks. It was found that different combinations of tasks caused different patterns of bidirectional interference, with evidence that speech is influenced by the nature of the secondary task, although task difficulty likely also plays a role. Interference between speech and nonspeech tasks has been demonstrated to occur in adults who are neurologically healthy and those with neurological disorders. In a study of individuals with Parkinson’s disease (PD) and a healthy control group, Dromey et al. (2010) found bidirectional
1638
interference between postural stability and diphthong production in a speech task in both the group with PD and the healthy controls. Camicioli, Oken, Sexton, Kaye, and Nutt (1998) also found interference between speech and gross motor tasks in people with PD and in healthy older adults, observing that performing a verbal fluency task led to reduced walking speed and increased number of steps in a concurrent walking task. In another study, Holmes, Jenkins, Johnson, Adams, and Spaulding (2010) found that when speaking distractor tasks were relatively complex, participants both with and without PD experienced measurable changes in postural stability. These findings of interference between speech and nonspeech activities have significant implications for increasing our understanding of the robustness and limitations of speech motor activity both in healthy populations and in populations with speech disorders, as multitasking is a frequent occurrence in everyday life.
Aging and Divided Attention A major complicating factor in the study of dividedattention interference is the effect of aging over the adult life span. Work by Salthouse (2009) indicates that cognitive abilities generally decline, but the trajectories may differ widely. Research has also suggested an age-related multitasking deficit. For example, Talland (1962) found that middle-aged and elderly men were slower at performing two separate motor tasks (one with each hand) than younger men. Chen et al. (1996) found that divided attention negatively affected older adults more than younger adults during an obstacle-avoidance walking task. Evidence from a meta-analysis of 33 dual-task studies (Verhaeghen, Steitz, Sliwinski, & Cerella, 2003) and from experimental research focused on the issue (Salthouse, Fristoe, Lineweaver, & Coon, 1995) showed that even after controlling for singletask performance, task complexity, and age-related declines in the component tasks, occasionally a small but significant age-related effect on dual-task performance remains. These age-related increases in interference have previously been demonstrated during dual-task paradigms including speech motor activity. Dromey et al. (2010) showed that a speech repetition task led to reduced heel height in a concurrent postural stability task in a group of healthy older adults (M = 70.5 years) compared with a group of younger adults (M = 25.5 years). In terms of models of attention, the findings of these studies could be interpreted as a decline in available attentional resources or a decrease in efficiency of the executive control process(es) in older adults compared with younger adults. Besides cognitive factors, another reason divided attention disproportionately affects older individuals may be the aging of the motor control system, including motor speech mechanisms. It is well known that movement control is gradually impaired with increasing age, primarily due to reduced strength, peripheral sensation, balance, and coordination of multijoint movements and increased reaction time (Ketcham, Dounskaia, & Stelmach, 2004; Menz,
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1637–1653 • December 2015
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
Lord, & Fitzpatrick, 2003). This motor decline affects both speech and nonspeech movements of the articulators. Ballard, Robin, Woodworth, and Zimba (2001) found that children and older adults were less able than young adults to perform a nonspeech oral visuomotor task and were less accurate in matching biofeedback of the amplitude of the movement of their lower lip, jaw, or fundamental frequency with the movement of a visual target. Smith, Wasowicz, and Preston (1987) found that older adults (M = 70 years) had longer utterance durations than younger adults (M = 25 years) at both habitual and fast speaking rates, measured at the segment, syllable, and sentence levels. Sadagopan and Smith (2013) compared oral movements of younger (M = 20 years 11 months) and older (M = 69 years) adults during nonword repetition; although no age effect was found, they reported that only older speakers with “younglike” nonword repetition ability (six out of 16) were included in the kinematic analyses, as the other 10 were unable to produce a sufficient number of accurate nonword productions for analysis. This suggests that age-related differences in motor speech ability that may be important were not analyzed due to the study’s methodological requirements. Wohlert and Smith (1998) found that the spatiotemporal index (STI; Smith, Goffman, Zelaznik, Ying, & McGillem, 1995), which is a measure of the movement variability across multiple repetitions of the same utterance, was significantly higher in a group of older adults (M = 79 years) than in younger adults (M = 24 years) when the participants spoke at a habitual rate; again, it was found that utterance durations were longer for the older adults at habitual, as well as slow and fast, rates. The authors speculated that their findings could have been at least partially attributable to changes in peripheral sensorimotor mechanisms. This overall decline could undermine the speech motor performance of older adults in dual-task conditions.
Research Questions and Hypotheses Research examining how age influences divided attention for speech has so far been limited (Dromey et al., 2010). Little is known about how bidirectional interference changes when different nonspeech tasks are performed with speech motor activity in dual-task paradigms. In addition, it is not yet known how bidirectional interference in such dual-task paradigms changes with increasing age in adulthood. The present experiment was performed to identify the effects of three nonspeech tasks on speech motor stability as well as possible effects of the speech task on performance of the three nonspeech tasks. In addition, the effect of age was examined, with participants coming from three cross-sectional groups: younger, middle-aged, and older adults. Our specific research questions were as follows: 1.
How is speech motor activity affected by qualitatively different nonspeech tasks performed concurrently?
2.
How is nonspeech task performance affected by concurrent speech?
3.
What effect does adult age have on bidirectional interference between speech and nonspeech tasks?
On the basis of the research reviewed above, it was hypothesized that all participants would experience bidirectional interference, with reduced speech motor stability and reduced nonspeech task performance when divided-attention performance was compared with performance in a singletask condition. Reduced speech motor stability would likely include higher STI values and lower values of lip and jaw displacement and velocity. Reduced nonspeech-task performance would likely include decreases in the rate or accuracy of responses. It was further hypothesized that older adults would demonstrate comparatively greater dual-task costs compared with younger adults. Thus, we predicted that the younger adults would demonstrate the least susceptibility to bidirectional interference compared with the middle-aged and older adults and that the middle-aged adult group would fall somewhere between the other two groups. Differences between the age groups would give preliminary indications of the age at which speech motor control may become more susceptible to divided-attention effects. On the basis of previous work, it was also reasoned that sex differences in interference might exist (Dromey & Benson, 2003). Observing interference of speech on nonspeech performance should give insight into the extent to which speech is automatic, how speech shares attentional resources, and how this sharing may change with age.
Method Participants Thirty men and 30 women participated in the study. They had no history of speech or language disorders with the exception of one participant who reported having typical articulation errors that were treated during childhood. Each of three age groups was divided evenly into 10 male and 10 female participants. The age groups included younger adults (ages 20–28 years, M = 22.95 years, SD = 2.35), middle-aged adults (ages 40–50 years, M = 45.60 years, SD = 3.47), and older adults (ages 58–70 years, M = 63.20 years, SD = 3.55). These age groups were selected based on the assumption that separating the age groups with about a decade between them would allow the observation of potential age-related differences across part of the adult life span in a cross-sectional design. All participants were native speakers of English and signed informed consent documents approved by the institutional review board prior to participation in the study. Hearing status was not formally evaluated, although all participants engaged in conversation with the experimenters without effort and reported no difficulties in everyday communication. No formal cognitive or linguistic screening was administered prior to participation in the study; however, retrospective analysis of the performance on experimental trials of these tasks indicates that no participant scored at or below chance levels for either task in either condition. The relatively high average accuracies (97% and 96% for the linguistic
Bailey & Dromey: Speech and Nonspeech Task Interference in Adults
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1639
task and 93% and 89% for the cognitive task in isolated and dual-task conditions, respectively) suggest levels of ability appropriate for the selected tasks. As an anecdote, the participants in all age groups were active in educational, vocational, social, and civic activities and likely had typical cognitive and linguistic abilities, although this information was not formally collected.
Equipment Each participant was seated in a sound booth to allow high-quality audio recordings and to reduce potential auditory distractions. A lightweight head-mounted strain gauge system (Barlow, Cole, & Abbs, 1983) was used to measure the vertical movements of each participant’s lips and jaw. Three cantilever beams were attached to the skin using double-sided tape at the midpoint of the vermillion border of the upper lips (UL) and lower lips (LL) and under the chin to track the jaw. These cantilevers were displaced incrementally with a sliding micrometer to allow calibration of the displacement and velocity signals in the softwarederived computations. The cantilevers were connected to bridge amplifiers (model 215, Biocommunication Electronics, Madison, WI) that produced analog voltage signals. A microphone (model C420, AKG, Vienna, Austria) was attached to the headset and placed about 10 cm away from the participant’s mouth in order to record speech. A sound level meter (model 712, Larson Davis, Depew, NY, or model 407736, Extech Instruments, Nashua, NH; C-weighting, fast response) was placed 100 cm away from the participant’s mouth and measured speech intensity in dB SPL. The analog signals from these instruments were digitized using a Windaq 720 (DATAQ Instruments, Akron, OH) analog–digital converter. The sampling rate was 1 kHz for the kinematic and sound level meter channels and 25 kHz (following 12-kHz low-pass filtering; model 9002, Frequency Devices, Ottawa, IL) for the audio channel. A computer with a 24-in. flat panel monitor located outside the window of the sound booth and a computer mouse located inside the sound booth were used for the cognitive and linguistic tasks.
Procedures All participants completed four types of tasks selected to represent broader domains of behavior: a speech task, a linguistic task, a cognitive task, and a manual motor task. Each participant completed one 30-s practice trial of each task before beginning the experimental trials to mitigate any possible practice effects. The practice trials were identical to the experimental trials except for the reduced duration. The practice speech task consisted of the same phrase to be repeated as the experimental trials. Practice items in the cognitive and linguistic tasks were similar to those in experimental trials; no item was ever repeated for a given participant. Following the practice trials, all four tasks were performed in the isolated condition. Last, each of the nonspeech tasks was performed concurrently with repetition of the target sentence. Specific feedback on accuracy of
1640
performance was not given after practice trials or any of the experimental trials, although it should be noted that visual feedback was available, by nature of the activity, during manual motor task performance. No feedback, visual or otherwise, was available during the other tasks. Order of the tasks within the isolated and dual-task blocks was counterbalanced across participants. The total time to complete the tasks, including equipment setup and training, was about 45 min per participant.
Tasks Individual tasks were chosen to represent separate domains that were believed to be behaviorally and functionally distinct from one another. They included activities reflecting a focus on speaking, cognitive, and linguistic abilities and manual dexterity. This was done to examine the relative effects of nonspeech task type on speech variables during dual-task conditions. In this article we refer to them by their focus domains for the sake of convenience. The tasks were designed to be mildly challenging in order to elicit purposeful attention but without being so difficult that they would discourage participants from attempting them. The speech task entailed speaking the phrase “I saw Patrick pull a wagon packed with apples” every time the participant heard a tone. The tone was repeated at regular intervals of about 4 s. The vowels and consonants in this sentence require large jaw and lip movements and were chosen to facilitate signal segmentation during analysis. During divided-attention conditions, participants repeated the phrase 14 to 15 times while performing one of the three nonspeech tasks. For the linguistic task, participants performed a semantic-decision activity similar to the one described in a lexical semantics neuroimaging study by Müller, Kleinhans, and Courchesne (2003). Their semantic-decision task involved viewing a computer screen that showed two words, a noun and a verb, and responding whether the two were semantically related. Novel stimuli were created for this study and were based on high-frequency nouns and verbs in the Corpus of Contemporary American English (Davies, 2012). For example, joke–laughing made a semantic pair in the present study, whereas dough–interviewing did not make a semantic pair. All participants were given the same practice and experimental lists, although the condition of the two experimental lists (isolated vs. dual task) was counterbalanced across participants. The participants were given 60 s to categorize as many pairs of words as possible. During the divided-attention condition, participants completed the semantic-decision task while they were repeating the target sentence. For the cognitive task, participants performed a quantity-comparison activity. This task was designed to be analogous to the semantic-decision task. This task involved viewing a computer screen that showed two numerical values in fraction notation with an equal sign in between them; the participants used a computer mouse to select one button if the quantity comparison was correct (e.g.,
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1637–1653 • December 2015
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
2/3 = 4/6) or another button if the quantity comparison was not correct (e.g., 1/4 = 4/8). The participants were given 60 s to evaluate as many quantity comparisons as possible. During the divided-attention condition, participants completed the quantity-comparison task while they were repeating the target sentence. The manual motor task consisted of the Purdue Pegboard Test (Tiffin, 1948). Participants were instructed to perform the test with both hands in order to eliminate handedness and dominance effects across participants. Manual motor performance was quantified as the number of pegs placed in 60 s. During the divided-attention condition, participants placed pegs in the pegboard while they completed the speech task.
Data Analysis Speech-dependent measures were taken from 10 tokens of the target phrase in each of the conditions (the speechonly condition and the three divided-attention conditions). The speech measures were selected from those that have shown sensitivity to changes due to age and divided-attention conditions in previous research (Dromey & Benson, 2003; Tingley & Dromey, 2000; Wohlert & Smith, 1998). The first 10 perceptually correct tokens for each participant for each condition were segmented for analysis. Tokens judged as perceptually incorrect by an experimenter (the first author) were excluded at this time. Tokens that were judged as correct contained all of the words of the target phrase in the correct sequence without significant disfluencies (i.e., articulatory errors, changes in word order, and substantial inconsistencies or differences in prosody). The digital signals associated with these tokens were segmented with custom MATLAB (The Mathworks, 2009) routines on the basis of consistent markers in the velocity record, as shown in the lower panel of Figure 1. The markers used for signal segmentation were the downward peak during the LL opening from the /p/ to the /æ/ in Patrick and the upward peak during the LL closing from the /æ/ to the /p/ in apples. Figure 1. Sample displacement and velocity record for one utterance, showing segmentation points.
These segments were then analyzed for duration, STI for the LL, and SPL. Utterance duration (ms) was selected as a representation of speaking rate. STI, which is a measure of the variability of speech movements over multiple repetitions, was calculated by time and amplitude normalizing the 10 displacement waveforms for each condition and calculating the sum of the displacement standard deviations at 50 equally spaced points throughout the movement record (Smith et al., 1995). Previous research has shown increases in STI for older adults (Wohlert & Smith, 1998) during speech and for younger adults in divided-attention conditions (Dromey & Bates, 2005; Dromey & Benson, 2003; Dromey & Shim, 2008). Vocal intensity was included because a previous study in our laboratory (Dromey & Bates, 2005) reported that vocal intensity increases in younger adults speaking under dual-task conditions, including tasks similar to those used in the present study. In addition, a single closing gesture (the closing gesture after the low, front vowel of packed) was segmented from the displacement record as shown in the upper panel of Figure 1 and used to make measurements of LL displacement, LL peak velocity, and UL–LL correlation. LL displacement and peak velocity are measures of the magnitude and rate of LL movement for a selected articulatory gesture in the utterance. UL–LL correlation is an index reflecting the coordination of lip movements, with a value of –1 indicating that the lips are moving in opposite directions, as would be expected during a bilabial opening or closing movement. This correlation measure has previously been used to show differences between typical speakers and speakers with a disorder (Tingley & Dromey, 2000). It was included in the present study to examine how lip coordination during speech was affected by the various divided-attention conditions. Measurements of the nonspeech tasks (linguistic, cognitive, and manual motor) in the isolated condition were taken to compare with measurements of the same tasks in the divided-attention condition in order to quantify the impact of speech activity on these nonspeech tasks. The cognitive and linguistic tasks were timed at 60 s and were scored as the total number of responses, the number of correct responses (total responses minus incorrect responses), and the accuracy of responses (number of correct responses divided by number of total responses). The manual motor task was scored as the number of pegs placed in 60 s.
Results Speech dependent measures were based on the average of the 10 selected tokens of the target phrase per condition. Participants typically produced about 14 tokens per trial. Two participants (one middle-aged woman and one middle-aged man) failed to produce 10 correct utterances during the concurrent cognitive task; measures for those participants in that condition were excluded from analysis. Equipment malfunction prevented the collection of intensity data for one of the participants (a middle-aged man) in the speech-only and cognitive dual-task conditions.
Bailey & Dromey: Speech and Nonspeech Task Interference in Adults
Downloaded From: http://jslhr.pubs.asha.org/ by a Univ of York-England User on 02/18/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1641
In addition, concurrent manual motor task data for one participant (an older man) were not collected due to a procedural error. The descriptive statistics for each dependent variable were calculated for each of the three age groups and separated within age group by sex. Descriptive statistics for all measurements are available in the Appendix.
Repeated Measures Analyses of Variance The effect of condition on the kinematic and intensity measures for the speech task in isolation and during each divided-attention condition was tested with a withinparticipant repeated measures analysis of variance (ANOVA), with age and sex as between-participants factors. In addition, Tukey’s honestly significant difference post hoc analyses were performed to more closely examine age group and sex contrasts. The effect of condition on response counts and accuracy rates for the linguistic, cognitive, and manual motor nonspeech tasks was also tested with a separate but identical within-participant repeated measures ANOVA, again with age and sex as between-participants factors. If a case had missing data for any of the variables in the repeated measures model, the entire case was excluded from the model for that particular variable. When Mauchly’s test of a sphericity violation was significant, the corrected HuynhFeldt values were used. Because proportion and correlation coefficient data are typically not normally distributed, they were transformed (arcsine and Fisher z, respectively) prior to analysis.
Condition Effects on Speech and Nonspeech Task Performance There was a significant main effect of condition on all speech variables except LL velocity. There were also many significant contrasts. The F ratios and p values for the speech measures are displayed in Table 1. Means and error bars for utterance duration, LL displacement, and velocity
are reported in Figure 2. There was a main effect of condition on duration, F(2.56, 138.11) = 12.00, p < .001. Contrasts indicate that duration was significantly longer during the linguistic and cognitive concurrent tasks compared with the speech-only condition. There was also a main effect of condition on LL displacement, F(2.59, 139.99) = 3.25, p = .03. Contrasts indicate that LL displacement decreased during the concurrent performance of the manual motor task. Means and error bars for UL–LL correlation, LL STI, and intensity are shown in Figure 3. Correlation between the UL and LL became more strongly negative during concurrent task performance, F(2.78, 150.17) = 3.38, p = .023, with contrasts showing significant differences occurring between the speech-only condition and the concurrent linguistic and cognitive conditions. There was also a main effect of condition on the STI of the LL, F(3, 162) = 29.40, p < .001; contrasts showed significant increases occurring from the speech-only condition to the linguistic and cognitive conditions. Last, there was also a main effect of condition on SPL, F(2.76, 146.20) = 33.00, p < .001, with contrasts showing an increase in SPL from baseline isolated speech performance for the manual motor condition only. There was also a significant condition main effect on some of the nonspeech task measures. The F ratios and p values for the nonspeech concurrent tasks are displayed in Table 2. The means and standard deviations of the nonspeech task response counts are shown in Figure 4. The linguistic and cognitive task accuracy rates and their standard deviations are shown in Figure 5. Linguistic task measures that significantly differed between conditions included the number of correct responses, F(1, 56) = 210.83, p < .001, and the number of total responses, F(1, 56) = 211.93, p < .001, with both decreasing compared with the single-task condition. One cognitive task measure also decreased significantly between conditions: the number of correct responses, F(1, 56) = 5.88, p = .019. Condition did not significantly affect the number of pegs the participants placed in the board.
Table 1. Significant repeated measures analysis of variance effects of condition on speech measures, with the concurrent task contrasted against the isolated-speech condition. Condition main effect Variable
df
Duration (ms) Displacement (mm) Lower lip velocity (mm/s) Upper lip–lower lip correlation Lower lip spatiotemporal index dB SPL at 100 cm
F
p
Linguistic contrast
Cognitive contrast
df
df
F
p
F
p
df
F
p
2.56, 138.11 12.00
