Speech dysfluency effects on working memory in otherwise fluent adults

(1)

Speech dysfluency effects on working memory in

otherwise fluent adults

Johan Brage

Department of Computer and Information Science Linköpings Universitet June 8, 2014 LIU-IDA/KOGVET-G–14/012—SE Supervisor Örjan Dahlström Universitetslektor

Department of Behavioural Sciences and Learning Linköpings Universitet

Examiner

Carine Signoret Junior universitetslektor

Department of Behavioural Sciences and Learning Linköpings Universitet

(2)

Upphovsrätt

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare –från publiceringsdatum under förutsättning att inga extraordinära omstän-digheter uppstår. Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och tillgängligheten finns lösningar av teknisk och administrativ art. Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsmannens litterära eller konstnärliga anseende eller egenart. För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/

Copyright

The publishers will keep this document online on the Internet – or its possible replacement –from the date of publication barring exceptional circumstan-ces. The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educatio-nal purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the consent of the copy-right owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility. According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement. For additional information about the Linköping University Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.

c

(3)

Abstract

Using Delayed Auditory Feedback can be used to induce Stutter-like dysfluencies, causing an individual to lose speech fluency. Little is known about the effect of speech dysfluency on working memory and phonological coding. The present study focuses on finding a method that can be used to measure these effects in otherwise fluent adults. 7 adults who normally speak fluently are subjected to Delayed Auditory Feedback during a Reading Span Task. The method proved too weak to induce speech dysfluency in a majority of participants, indicating that the phenomenon is more complex than anticipated.

(4)

Acknowledgments

I would like to thank my supervisor Örjan Dahlström for his guidance and encouragement during this project.

I would also like to direct a special thank you to Sara Flink, without whom this work would never have been finished.

(5)

List of Tables

1 Average RST scores . . . 10 2 Average PASAT scores . . . 11

List of Figures

(7)

1 Introduction

By manipulating the way people hear their own voice using the method of Delayed Auditory Feedback (DAF), speech dysfluencies (SD) can be induced in participants. The present study focuses on creating a method that can be used to assess the effect of SD on Working Memory (WM).

1.1 Background

Developmental Stuttering (DS) is a phenomenon that occurs in a small por-tion of the populapor-tion. It has been reported (e.g. Sudhi et al., 2010) that males are more likely to suffer from DS than females. Finding a method that reliably induces Stutter-like Dysfluencies (SLD) in the majority of partici-pants would allow for exploring the relationship between people who suffer from DS and people who normally speak fluently in working memory or rea-ding comprehension tasks. Previous studies using DAF have focused more on measuring the speech rate of participants (e.g. Corey and Cuddapah, 2008; Fukawa et al., 1988) and thus the effect of DAF on WM or storing of the information read has largely been ignored.

1.2 Purpose

The main purpose of this thesis is to explore if SD can be induced in parti-cipants using a form of DAF. This method is then used in a pilot test that examines if this induced speech dysfluency has an effect on working memory in a Reading Span Task (RST).

1.3 Problem statements

1) Can SD be induced using DAF through a cellphone and headphones? 2) Can we observe with this method the effects of SD on WM in an RST? 3) Are there similarities between performance on the Paced Auditory Serial Addition Task (PASAT) and the RST?

4) Is there a gender difference in susceptibility to DAF in either speech dys-fluency or its effects on the WM?

(8)

2 Theory

2.1 Delayed Auditory Feedback

DAF can be used both to reduce stuttering in people who suffer from DS, and to induce SLD in people who are normally fluent (Corey and Cuddapah, 2008). By manipulating the delay between when a person says a word and when it can be heard by the participant a temporary speech dysfluency is induced. This is achieved by recording the person’s voice and replaying it after a short delay, typically less than 400 milliseconds, preferably through a pair of headphones.

A factor that may be involved in the effects of DAF is age. An early study showed that the effects of DAF on speech fluency increased with age (Chase et al., 1961), while a later study showed the opposite effect, i.e. effects of DAF decreasing with age (Siegel et al., 1980). These opposing finds imply that the effects of DAF may not be tied to age, but rather individual differences. There is some evidence that the specific delay that produces the most significant effect of DAF varies with age (Siegel et al., 1980), however this effect was small enough to once again suggest that individual differences may account for this effect.

A factor that is shown to have a much larger impact on speech fluency using DAF is gender. Among normally fluent people the effect of DAF within males is significantly higher than it is within females (e.g. Fukawa et al., 1988; Bachrach, 1964; Sutton et al., 1964). This finding, along with research on DS, suggests that males are more prone to stuttering, be it naturally occurring or induced.

2.2 Speech Dysfluency

SD is a collective term for conditions that in some way prevent a person from speaking fluently. Inducing SD using this DAF creates a dysfluency that in many ways resembles stuttering. People subjected to DAF often experience difficulty articulating words and often having to repeat syllables or dragging vowels out unnaturally long, compared to their normal speech. Overly simpli-fied this can be seen as h-h-h-ello for repeating or heeeeelloooooo for dragging them out. Speech fluency has been suggested to affect working memory ca-pacity (e.g. Stavrakaki et al., 2012).

(9)

2.3 Working Memory

Perhaps the most influential model of the WM was introduced by Baddeley and Hitch (Baddeley and Hitch, 1974).

Figure 1: The Baddeley model of working memory (Baddeley, 2000)

At the time the model was introduced it consisted of three separate systems that made up the WM. The Visuospatial Sketchpad and the Phonological Loop (PL) are each in charge of specific tasks, while the Central Executive go-verns them both and divides resources between the two subsystems. The PL handles auditory information using its two components, a phonological store and an articulatory rehearsal system. If the information in the phonological store is not rehearsed in the articulatory rehearsal system, the information deteriorates quickly, presumably no more than a couple of seconds (Baddeley, 2002). There are a number of factors that play a role in how long information can be stored in the PL. The factors most important to this experiment are the phonological similarity effect, the word-length effect and the effect of ar-ticulatory suppression. The similarity effect entails that sequences containing letters or words that sound similar are harder to remember than sequences where items sound distinctly different. The word-length effect tells us that it is harder to remember sequences of long words than sequences of shorter words. Articulatory suppression suggests that if a person is not allowed to rehearse the sound and is instead forced to articulate other sounds, such as presenting a new sentence to be read, the performance decreases (Baddeley, 2000). Further studies (e.g. Daneman and Carpenter, 1980) implicated that the PL was not enough to explain sentence recall since it requires keeping

(10)

information stored over a longer period of time than the PL is capable of (Baddeley et al., 2009). Because of this, the original WM model was no long-er adequate and was latlong-er revised (Baddeley, 2000), and extended to also include the Episodic Buffer (EB). The EB is also a limited system in terms of how much information it can contain. It is assumed to be involved in both long-term and short-term memory and can be accessed by the central execu-tive when attention is called to it. Using a form of chunking would explain why people can remember longer sequences, such as sentences, as opposed to single words. Sentences and words are considered a single chunk, making the amount of chunks to be remembered the same, with the only variance being the information in the chunk itself (Baddeley et al., 2009).

2.4 Reading Span Task

The RST is commonly used in testing working memory, cognitive processing and can also be used in reading comprehension. It was first used by Daneman and Carpenter (1980) in an attempt to measure the effect of working memory in reading comprehension. The test has a number of variants and can easily be modified to change the difficulty. In its original appearance, sentences of 13 to 16 words were printed on index cards, the participant read them aloud at their own pace and tried to always remember the final word of each sentence. This was repeated for sequences of 2 to 6 sentences and was ended when a participant failed to recall any word at a the current level, e.g. the 5 sentence level. Performance on a RST has been shown as a good predictor of other cognitive abilities such as language comprehension, logic and reasoning ability (e.g. Daneman, 1996; Kyllonen and Christal, 1990; Engle et al., 1999).

2.5 Paced Auditory Serial Addition Task

The PASAT was originally developed to measure information processing in patients who suffered from brain injury (Tombaugh, 2006). Since then it has been widely used to measure abilities related to attention rather than information processing. In the original appearance, the PASAT contained four different intervals (2.4 s, 2.0 s, 1.6 s, 1.2 s). The test gets progressively harder as the interval gets shorter. The mean score for the 2.4 second interval was ~76 percent correct while at the 1.2 second interval, the mean score was less than half at ~36 percent correct (Gronwall, 1977). The test can be prone to learning in test-retest situations, and should as such be used with some care Tombaugh (2006). The reliability of the PASAT in terms of measuring attention combined with its high internal validity makes is a good test for

(11)

comparison to the RST, as both tasks require an individual to focus their attention while inhibiting irrelevant information.

3 Method

3.1 Participants

Participants were recruited using a convenience sample by directly contacting possible participants. The majority of participants were undergraduate stu-dents at Linköpings University. In total, 7 people participated, ranging from 22 to 60 years of age (M =33.86, SD=17.55). Of the 7 participants, 5 were female and 2 were male.

3.2 Ethics

At the introduction of the experiment, participants were informed of their rights, and that the result of their participation would be used for the purpose of this thesis. This was done both verbally and through a consent form (see Appendix C). Participants were to be entirely anonymous and none of their personal information, participation, or results could ever be connected to them, and they were given the option to cancel the experiment at any time and to have their results excluded from the study. Participant recordings were also deleted immediately after having been analyzed.

3.3 Material

The PASAT audio was played through a Toshiba R850 15.6” laptop PC with an attached Razer Tiamat headset. The reading tasks were conducted on the same Toshiba laptop PC. The audio delay was produced using the Speech Jammer app for iOS running on an iPhone 4S with an attached Razer Tiamat headset. The video clips used to present the sentences were made using Windows Movie Maker.

3.4 Procedure

The experiments were conducted with one participant at the time. The entire length of the experiment was between 20 and 30 minutes for each partici-pant. All of the experiments were conducted in a closed room with little to no visual distractions present. All participants had the same general order of tests, that is, they all did the PASAT first and the RST second. The different

(12)

conditions were counter-balanced across participants, explained in more de-tail below. At the end of each session, an informal conversation was held with each participant, in order to collect their thoughts and experience during the experiment. This qualitative data was not recorded as participant experience was not the primary objective of the experiment. This data was however of interest while analyzing participant recordings and is as such included in the results.

PASAT

Participants first received instructions that they would be hearing a sequence of numbers in the headphones and that they should add the two most re-cent numbers together. If the participant was to receive the sequence 1, 2, 3, they should give the answer 3 (2+1) and 5 (3+2). If the participant in-dicated she or he understood the task the participant was asked to put on the headphones. They were then presented with a training phase. This trai-ning sequence was a part of the audio file that contained the entire test, and as such it was almost identical to the task used in the real test. After this training phase the participant was offered two choices, they could either repeat the trial if they were not entirely comfortable with the procedure, or they could move on to the real test. In each sequence there was a total of 61 numbers presented and the duration for the longer interval test was just over 3 minutes and 8 seconds, while the shorter interval was just over 2 minutes and 5 seconds. The interval in the longer test was roughly 2000 milliseconds while the shorter interval was roughly 1500 ms. The duration of the spoken numbers spanned from ~500ms to ~800ms. In an attempt to procure results as unbiased as possible, the order in which the two tests were presented was alternated between each participant. Participant A started with the shorter interval test while participant B started with the longer interval test, and so forth. In total, 4 participants started with the shorter interval test and 3 with the longer interval test.

RST

In order to be able to test the effects of SD on WM, the experiment had to be designed from this premise. Sentences had to be of a length that made it possible for SD to occur. Reading a short sentence of 3-4 words was unlikely to produce an effect, due to the nature of induced SD. Having too long sentences would make it nigh impossible to remember the words that began or ended the sentence. A middle ground was decided upon, where the length

(13)

of a sentence should be enough to induce SD while not being too long, making remembering words unlikely. Sentences were 8-10 words in length, although with a fairly large variance in the number of syllables.

The manner in which the material is presented in the RST used in this experiment has in its entirety been constructed for the purpose of this thesis. Sentences of between 8 and 10 words with different starting and ending words were used. Sentences were shown in their entirety for 6000ms with a 1000ms blank screen after each sentence. Each clip contained three sequences of sen-tences, 2, 3 and 4 sentences long and the length of the clips was 1 minute and 48 seconds. All sequences started with a 5000ms white screen and ended with the Swedish word for first or last displayed in all-caps indicating which word was to be given in answer. This screen was displayed for 10 seconds while the participant gave their answer. The same procedure was repeated for the second and third sequence. The three different video clips were always shown in the same order, however, the condition used rotated among partici-pants. These conditions will be known as NoDAF for no auditory feedback, ShDAF for the shorter interval, (~200ms), auditory feedback and LoDAF for the longer, (~300ms), interval. This means that the order for the first parti-cipant was NoDAF, ShDAF and LoDAF, while the second partiparti-cipant had the order of ShDAF, LoDAF and NoDAF, and so forth. In total, this was spread out so that 3 participants started with NoDAF, 2 participants with ShDAF and 2 participants with LoDAF.

Instructions were given to the participants that they would see sequences of sentences on the screen, and that they should read them aloud. For the NoDAF they were informed that they would not be hearing themselves in the headphones, although they should still wear them to make the experience as similar as possible. For the parts with DAF, they were informed that they would be hearing themselves with a certain delay. They were asked to try to keep reading the sentences at a normal pace even if they were experiencing difficulties doing it. They were informed that sequences would increase in length and that the first trial would consist of 2 sentences, the second of 3 sentences and the third of 4 sentences. They were also informed that they should attempt to recall the correct words in the order they were presented, although remembering more words was of greater importance than the order. When they saw that the sequence had ended, as indicated by the word first or last on the screen, they should give their answer verbally. Following the first and second sequence, they would know a new sequence was about to start after the end screen containing first or last turned in to a blank white screen. After each trial they were given a short break as the next video clip was prepared.

(14)

The main experiment, i.e. the two parts using DAF were recorded using the built-in recorder in the Speech Jammer app. This was done in order to be able to analyze the effects of DAF during the two reading tasks. A fairly simple method was used to assess the impact of DAF on speech fluency. Stuttering or otherwise losing fluency, such as dragging out a word or a long pause where there should be none, was counted as 1 Speech Error (SE). The number of speech errors for each sentence was then noted down. Because of the maximum length of 6 seconds for every sentence, the number of SEs was in a sense limited. For simplicity’s sake, if a participant failed to utter the entire sentence in the allotted time, this number was set to an amount greater than the highest number of SEs for a fully completed sentence.

3.5 Scoring

PASAT

The correct sequence of sums was printed on a piece of paper for the expe-rimenter (see Appendix A). As the participant gave an answer, that answer was checked against the list. During the test, the experimenter sat behind the laptop monitor in order to be able to follow the audio file and determine when an answer should be given. If the participant gave the correct answer, the number was underlined by the experimenter. In the answer given was incorrect, the experimenter did nothing and proceeded to the next number in the sequence. The same procedure was also used if the participant gave no answer, or gave the correct answer late, meaning after the following number in the sequence had been presented. As such the only two possibilities for each number was correct or incorrect, making the maximum score 60, with sequences being 61 numbers long and the first answer was given upon hearing the second number.

RST

During the trial, the experimenter had a sheet of paper with the correct words written on it (see Appendix B). If the participant gave the right answer, a number was written next to the word, indicating the order of answers given. If the participant gave the wrong word, the experimenter indicated that the wrong word was given by writing down the number corresponding to the order in which the word was given in the margin on the sheet of paper. If the participant gave an incorrect word, that word was not noted down, as only correct answers were of interest. Every correct answer was awarded with 1 point. Giving the correct answer in the correct order was still only considered

(15)

one point. No difference was made between giving the wrong word in answer, or not giving an answer at all. For each sequence of 2, 3 and sentences, the maximum number of points were 2, 3 and 4 respectively. The maximum combined score for a condition was 9.

3.6 Data Analysis

The study contained a within-subjects design using a quantitative method of analysis. This analysis was done using IBM SPSS Statistics 22. Scores on both the RST and the PASAT were analyzed using a paired sample t-test to examine differences across the different intervals, i.e. amount of delay and interval between numbers, respectively. To explore the relationship between performance on the PASAT and the RST a correlation analysis using Pear-son’s correlation coefficient was used. In order to examine gender differences for the reading span task an independent samples t-test was conducted.

The qualitative data was analyzed using a simplified phenomenographic approach, where user experience of the experiment was the primary focus.

4 Results

4.1 Problem statement 1: Speech fluency effects of DAF

Only one of the participants was exhibiting a significant effect of the DAF, i.e. losing speech fluency. A significant effect in this setting was a subjective judgment by the experimenter during the test. All participants were allowed to continue the experiment uninterrupted in order to see if there was an effect on recall using DAF even if their speech fluency was not affected. Further analysis of participant recordings showed a very slight impact of the DAF in all participants over the first new sentence for both delays. After the first sentence, a slight effect of the DAF was shown in 3 additional participants, while the remaining participants did not get noticeably affected from the second sentence and on. Out of the affected participants, two employed the same coping mechanism where they raised their voice and spoke louder than they did during the NoDAF test. One of these two started the sentence in a normal speaking volume, but increased the volume over the course of the sentence and ended up close to shouting toward the end. The other participant did not raise speech volume to this extreme degree, however it was noticeably louder than the trial not using DAF. Due to this circumstance, SE rate was not calculated for any of the participants, as the sample size would have only consisted of 1 participant.

(16)

4.2 Problem statement 2: Working memory effects of

DAF

Scores for the participant who was noticeably affected by DAF, henceforth known as P, were calculated against the score of the remaining participants. The results of the RST showed that P scored lower than the average for the other participants for each individual task, as well as the combined score (NoDAF 5 (M = 6.33, SD = 1.21, n = 6), ShDAF 4 (M = 5.17, SD = 1.17, n = 6), LoDAF 5 (M = 5.17, SD = 1.17, n = 6), Combined 13 (M = 16.67,

SD = 2.5, n = 6)). This finding makes it hard to assume that performance

was impacted by the DAF as P also scored lower than the average of the remaining participants on the NoDAF test. The performance of P did not differ noticeably between the conditions, with scores being 5 for the NoDAF, 4 for ShDAF and 5 for LoDAF. For all participants, the average score of the NoDAF test is higher than the other conditions indicating that there may still be some effect of the DAF on performance for this task.

Table 1: Average RST scores Mean Std. Dev. NoDAF 6.14 1.22 ShDAF 5.0 1.16 LoDAF 5.14 1.07

There was a significant difference in scores between the NoDAF (M = 6.14,

SD = 1.22) and ShDAF (M = 5.0, SD = 1.155) conditions t(6) = 2.83, p =

.03, r = .76. There was no significant difference in scores between the NoDAF (M = 6.14, S D = 1.22) and LoDAF (M = 5.14, SD = 1.07) conditions t(6) = 1.87, p = .11. There was no significant difference in scores between the ShDAF (M = 5.0, SD = 1.16) and LoDAF (M = 5.14, SD = 1.07) conditions

t(6) = -.24, p = .82.

4.3 Problem statement 3: Performance on the PASAT

In the table below short and long stands for the 1.5 second interval and 2 second interval tests, respectively.

(17)

Table 2: Average PASAT scores Mean Std. Dev. Short 40.0 6.68 Long 50.71 3.86 Combined 90.71 9.91

There was a significant difference in scores between the short (M = 40.0, SD = 3.68) and the long (M = 50.71, SD = 3.87) PASAT interval tests t(6) = -6.2, p = .01, r = .93

There was no significant correlation in performance on the PASAT and the combined RST (Pearson r = .02, N = 7, p = .97).

P scored 91 on the combined PASAT score, while the average of the remai-ning participants was 90.67 (SD = 10.86), implying that performance on the PASAT test is in no way related to the effect of DAF.

4.4 Problem statement 4: Gender differences

As only one participant was noticeably affected by DAF it is hard to do anything but speculate in gender differences of DAF. The participant who was affected, P, was male. There was a significant difference between the genders on combined RST performance t(5)=5.71, p = .002, r = .93. However, for each individual condition, this difference was significant only in the NoDAF condition (t(5) = 5.39, p = .003, r = .92).

5 Discussion

5.1 Result discussion

The fact that only one of the participants had the intended effect from the DAF poses some interesting questions. Was there something different between that participant and the others in terms of execution during the experiment? Does DAF as a concept not work on the other participants? The answer to the first question is there shouldn’t have been. The volume was the same over all participants and the equipment was the same. The one thing that differed between participants was the setting, however it is unlikely that it had an impact as no significant background noise or visual distractions were present in either scenario, and P was not the only one who performed the task in that particular setting.

(18)

Scores on the RST only showed a significance between the NoDAF and ShDAF conditions, which was fairly surprising, even in such a limited sample. Presumably the additional cognitive load when having to inhibit the DAF would have put enough strain on the WM to account for larger differences. However, even if the results were not significant, there was still a visible difference between the different conditions, which does still indicate that the DAF made it harder to recall words. The significant differences between the shorter and longer interval PASAT was entirely unsurprising, as there is plenty of evidence in the literature that a shorter interval PASAT is much harder than a longer interval test. One of the biggest factors in the PASAT is psychological pressure. As time between numbers decreases, participants experience more and more stress. This in turn causes both incorrect additions and the inability to sort out the relevant information from the irrelevant, making a participant less likely to give an answer in time. The absence of a significant correlation between the combined PASAT scores and RST scores was unexpected. Both tests are commonly used in WM capacity experiments, and performance in one was expected to translate to performance in the other.’

Two of the participants attempted to overcome the effects of DAF by raising their voice to a level higher than in the NoDAF condition. When asked, they explained that they did so in order to overpower the sound coming from the headphones, thus decreasing its effects. This would suggest that the better you can hear your own voice while speaking, such as the sound that inevitably comes through the body, does reduce the effects of DAF.

That the only participant who experienced speech dysfluency from the DAF was male does not confirm previous research on gender differences, even though it is in line with it.

5.2 Method discussion

Using this form of set delay to produce SD within participants may not have been the optimal choice for this task. As the focus was on the effects of SD on working memory rather than at which interval this was most prominent the approach could have been altered. Instead of using the set intervals, the delay could have been changed for each individual, perhaps by having all participants read a passage of text with varying intervals to determine a level that produced sufficient SD. Using a different pair of headphones may have made a difference, as many of the participants indicated that it was easier to ignore the sound coming from the headphones if they heard themselves while they were speaking. Bypassing this particular issue could

(19)

circumvent this phenomenon. What could have made a difference was to make each session more individual, in an attempt to find the particular delay and volume combination with which the greatest speech dysfluency effect was accomplished for each participant. Deciding which intervals to use beforehand does have its own merits, namely to explore at which interval the effects are most common.

Using this length for each sentence may also have influenced the results. Because of the nature of the reading span task as getting progressively more difficult as sentences increase in length paired with the fact that little to no DAF effects are present if sentences are too short, it would appear that a reading span task is not the optimal choice for this sort of test.

The number of participants is likely to have had an effect on the ex-periment. It’s always hard to generalize results from a limited sample size. As gender differences in DAF effects have previously been reported, a more heterogeneous participant group would have been prefered.

6 Conclusion

As there was no significant effects of DAF for more than one participant it would appear that the approach was less than optimal. As the main purpo-se of the study was to examine the effect of induced speech dysfluency on working memory, this method was not the most fitting for this experiment.

The effect of the PASAT proved to be unrelated to the performance on the RST, and even though both tests are commonly used for measuring similar phenomena, they may do so in different enough manners that they cannot be used to predict each other’s results.

There is still work to be done in the area, as a deeper understanding of how induced SD affects WM would be of use to help understand DS and the additional cognitive load this puts on individuals suffering from it.

Using a method as simple as a cellphone with headphones was not po-werful enough of a method to reliably induce SD. A more solid method, that could easily be adapted for each participant would likely have been a better approach.

Based on the findings in this experiment, there is much room for further exploration of the phenomenon that is inducedSD. Some ideas and outlines are described in the following chapter. Overall the effects of DAF could be of importance in terms of exploring possibilities and limitations of the WM.

(20)

7 Future research

The next logical step is to develop a more reliable method of inducing speech dysfluency using DAF. Such as keeping a participant in a soundproof room with no windows or other visual distractions and using a DAF software that allows for more control in terms of volume and delay to find a specific point where maximum speech dysfluency is achieved. A different approach may be to have some form of noise in the headphones that would cause participants to hear their own voice as they are speaking to a lesser extent. Over the course of the experiment, the effect of induced SD on reading comprehension and how it translated to performance on the RST was given greater consi-deration. Based on this, two new ideas were formed. Measuring the effect of DAF on reading comprehension for passage-based reading could show if in-formation is lost due to DAF. This could be achieved by using a standardized reading comprehension test (e.g. the SAT for American English, Nationella läsförståelseprov for Swedish) and having participants perform a number of these tests in conditions not using DAF and at a maximum speech dysfluency point.

The other approach is to add a common aspect of RST, that participants must decide whether a sentence makes sense or not before moving on to the next sentence. This added task ensures that participants read the entire sen-tence, process it and could show if comprehension is affected by the increased cognitive load of DAF.

Due to previous research being unclear as to how age relates to the effects of DAF, this would also be of interest, and could likely be a part of any experiment that involves using DAF. Measuring both the effect of DAF on speech fluency and its effects on coding in to memory could be done by the same task, as the present study showed slight effects on recall even if SD was not achieved. Different age groups, such as teenagers, young adults and adults could give insight in to how reliant WM capacity is on phonological coding.

As gender is one of the few factors that is known to affect susceptibility to DAF, this could be explored further by having a more heterogeneous group of participants. Exploring if there’s a difference in susceptibility within each gender would also contribute to understanding WM capacity.

(21)

References

D.L. Bachrach. Sex differences in reactions to delayed auditory feedback.

Perceptual and motor skills, 19:81–82, 1964.

A. D. Baddeley. The episodic buffer: A new component of working memory? 4(11):417–423, 2000.

A.D. Baddeley. Is working memory still working? European Psychologist, 7: 85–97, 2002.

A.D. Baddeley and G. Hitch. Working memory. volume 8 of Psychology of

Learning and Motivation, pages 47 – 89. Academic Press, 1974.

A.D. Baddeley, G.J. Hitch, and R.J. Allen. Working memory and binding in sentence recall. Journal of Memory and Language, 61(3):438 – 456, 2009.

R.A. Chase, S. Sutton, D. First, and J. Zubin. A developmental study of changes in behavior under delayed auditory feedback. The Journal of

ge-netic psychology ; child behavior, animal behavior, and comparative psycho-logy, 99:101–112, 1961.

D.M. Corey and V.A. Cuddapah. Delayed auditory feedback effects during reading and conversation tasks: Gender differences in fluent adults. Journal

of Fluency Disorders, 33(4):291 – 305, 2008.

M. Daneman. Working memory and language comprehension: A meta-analysis. Psychonomic Bulletin & Review, 3(4):422–433, 1996.

M. Daneman and P.A. Carpenter. Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19(4):450– 466, 1980.

R.W. Engle, J.E. Laughlin, S.W. Tuholski, and A.R.A. Conway. Working me-mory, short-term meme-mory, and general fluid intelligence: A latent-variable approach. Journal of Experimental Psychology: General, 128(3):309–331, 1999.

T. Fukawa, H. Yoshioka, E. Ozawa, and S. Yoshida. Difference of suscepti-bility to delayed auditory feedback between stutterers and nonstutterers.

Journal of Speech and Hearing Research, 31(3):475–479, 1988.

P.C. Kyllonen and R.E. Christal. Reasoning ability is (little more than) working-memory capacity?! Intelligence, 14(4):389 – 433, 1990.

(22)

G.M. Siegel, C.A. Fehst, S.R. Garber, and H.L. Pick Jr. Delayed auditory feedback with children. Journal of Speech and Hearing Research, 23(4): 802–813, 1980.

S. Stavrakaki, K. Megari, M.H. Kosmidis, M. Apostolidou, and E. Takou. Working memory and verbal fluency in simultaneous interpreters. Journal

of Clinical & Experimental Neuropsychology, 34(6):624 – 633, 2012.

N. Sudhi, J. Merin, and Y. V. Geetha. Age and gender differences in persons with stuttering. Journal of the All India Institute of Speech & Hearing, 29 (2):131 – 138, 2010.

S. Sutton, W.C. Roehrig, and J. Kramer. Delayed auditory feedback of speech in schizophrenics and normals*. Annals of the New York Academy

of Sciences, 105(15):832–844, 1964.

Tom N Tombaugh. A comprehensive review of the paced auditory serial addition test (pasat). Archives of Clinical Neuropsychology, 21(1):53 – 76, 2006.

(23)

Appendices

(24)

(25)

MEDGIVANDEBLANKETT

Tack för att du valt att delta i detta experiment.

Resultatet av experimentet kommer att användas som underlag för en Kandidatuppsats på det Kognitionsvetenskapliga programmet vid Linköpings Universitet.

Detta experiment går ut på att du ska utföra två (2) olika tester, där de olika testerna innehåller två (2), respektive tre (3) deltester. Hela experimentet beräknas ta cirka 30 minuter.

Av dessa fem delmoment kommer två att spelas in med en mikrofon för att senare kunna analyseras.

- Jag kan när som helst under testets utförande välja att avbryta och få mina resultat exkluderade ur studien. Efter testtillfällets slut kan jag inte längre välja att exkludera mina resultat.

- Jag förstår att mina resultat kommer att behandlas anonymt och att ingen annan än testledaren kommer att kunna para ihop mig med mitt deltagande.

- Jag accepterar att min inspelning används för detta projekt.

- Jag kan få en kopia på denna blankett genom att ange min e-mailadress nedan. - Jag kan få ta del av resultatet av denna studie genom att ange min e-mailadress nedan.

Jag bekräftar att jag förstått informationen ovan.



Ja



Nej

Namn

Ort och datum

E-mailadress (  Jag vill få en kopia på denna blankett.  Jag vill ta del av resultatet av denna studie)

Speech dysfluency effects on working memory in otherwise fluent adults