Background speech -disparate impact on job performance, depending on the language?

AKADEMIN FÖR TEKNIK OCH MILJÖ

Avdelningen för bygg-, energi- och miljöteknik

Background speech

-disparate impact on job performance, depending on the language?

Bakgrundsprat -olika inverkan på arbetsprestation beroende på språk?

Mira Rutanen 2015

År

Uppsats, Avancerad nivå (magisterexamen), 15 hp Miljöpsykologi

Masterprogram i miljöpsykologi Handledare: Patrik Sörqvist

Examinator: Robert Ljung

3

A

Background speech is annoying and distracting when working on tasks that require focus, and according to previous research, background speech is a common cause of reduced work performance. According to the interference-by-process theory, distraction is a function of the similarity between the processes involved in the involuntary analysis of the background speech and the voluntary processes involved in the task. In view of this theory, a similarity in language—between the produced language and the language that is listened to—may increase the magnitude of distraction by background chatter in comparison with when the language which is produced is different from the language that is perceived. The purpose of the present experiment was to investigate whether work performance—as indexed by writing fluency (WF)—varies depending on the similarity between the language that is heard in the background and the language that is produced. The experiment had a within-participants design with two factors: language to-be-produced (Swedish vs. English) and language of the background speech (Swedish vs. English). The sample constituted of 43 university students, with Swedish as native language. The result showed a main effect of language to-be-

produced: WF increased when the participants produced text in their native language compared to text production in their second language. No main effect of language of background speech was found, and no interaction between these two factors was revealed.

Key words: Background speech, language production, speech intelligibility, interference-by- process, distraction, working memory.

4

1 T ABLE OF C ONTENT

2 Introduction ... 5

2.1 Attentional capture ... 6

2.2 Interference-by-process ... 6

2.3 Language production ... 7

2.4 Speech perception ... 7

2.5 Problem statements ... 8

2.6 Hypothesis ... 8

3 Method ... 9

3.1 Participants ... 9

3.2 Design ... 9

Dependent variables ... 9

3.3 Materials and apparatus ... 10

Background Speech ... 10

Writing Task... 10

3.4 Procedure ... 11

3.5 Statistics ... 11

4 Results ... 12

4.1 Language knowledge ... 12

4.2 Writing fluency ... 12

5 Discussion ... 13

5.1 Discussion of hypothesis ... 13

5.2 Explanations ... 13

5.3 Discussion of method ... 14

5.4 Conclusion ... 14

References ... 16

5

2 I NTRODUCTION

With an increasing world population and an excessive consumption of various noise-creating appliances and other technicalities, total silence will become a scarce commodity in the future. As noise exposure increases, it will have consequences that can affect many aspects of people's daily lives, such as the availability of private space and noise-free environments, which in turn may inhibit the possibilities of the daily recovery. According to World Health Organization (2011) exposure to noise is a major global environmental problem and

considered as a threat to public health. Students in school environments and people in working life, is already complaining that they occasionally are disturbed by noise from e.g.

technical appliances and by people talking when they perform their work tasks. This kind of interrupting disturbance can in the long-term become a health issue which in turn can lead to mental overload and furthermore to long-term sick leave which leaves major costs to the society (Seddigh, Berntson, Bodin Danielson, & Westerlund, 2014; Smith-Jackson & Klein, 2009). Despite this knowledge, several companies and municipalities still apply open-plan layouts.

Noise survey is in general difficult to perform in authentic surroundings, of the reason that the noise sources usually are multiple. Many researchers therefore prefer conducting

laboratory studies where every noise source can be controlled and examined individually.

Several researchers have investigated the effects of noise as impact on work performance and previous studies show that speech is the most disturbing noise (Hongisto, 2005; Schlittmeier, Hellbrück, Thaden, & Vorländer, 2008; Sörqvist, Nöstl, & Halin, 2012). A study by Hongisto (2005) tested the effects of background speech on performance, showed that the background speech´s intelligibility is the prime cause of disruption and reduced work performance. In another study, Schlittmeier, Hellbrück, Thaden, and Vorländer (2008) tested different speech variables; good and poor intelligibility at 35 dB and 55 dB in comparison to silence, which pointed out that background speech with high intelligibility is more disturbing and

interruptive than speech of low intelligibility. Sörqvist, Nöstl and Halin (2012) investigated the semanticity of background speech as distractor of writing process and found out that normal speech impaired the writing process to a greater extent than spectrally-rotated speech did, and that spectrally-rotated speech had no effect on the writing process. Speech

Transmission Index (STI), a measure used to evaluate the acoustic properties of environments, has been used in several studies where noise impact on work related

6 performance has been investigated (Venetjoki, Hongisto, Kaarlela-Tuomaala, & Keskinen, 2006; Haapakangas, Hongisto, Hyönä, Kokko, & Keränen, 2014; Keus van de Poll, Ljung, Odelius, & Sörqvist, 2014). According to the work of Keus van de Poll et al. (2014) background speech, even with low intelligibility, can have distracting effects on task performance and show furthermore that STI values over 0.23 radically decreases writing fluency.

2.1 A

One reason as to why noise has a negative effect on cognitive performance is its ability to catch peoples attention, which, regardless of the source of the noise, almost always have unpleasent effects on humans, such as mental exhaustion. However, each individual’s ability to cope with noise exposure is depending on the capacity of working memory. Lavie, Hirst, de Fockert and Viding (2004) argues that there are two mechanisms; one passive and one active, that is involved in how irrelevant distractors are processed. The passive mechanism's main task is simply to ignore the distractors, which only becomes possible when sufficient capacity is available, while the other mechanism, more active, is controlled by higher cognitive

functions such as working memory. This second mechanism works by actively identifying irrelevant distractors in order to store or reject them and in this way maintain the processing of current priorities, to ensure that low-priority stimuli (eg. distractors) do not take over the behavior. There are several reseachers that support the thesis of Lavie et al. (Bell, Röer, Dentale, & Buchner, 2012; Higgins & Johnson, 2013 ; Zeamer & Fox Tree, 2013), but other rasearchers argue that there are more to evaluate, such as the impact of selective attention and dilution effects that also impairs on working memory capacity and indicates that attention capture can be adapted and modulated by attention control (Wilson, Muroi, & MacLeod, 2011; Yeh & Lin, 2013; Zhang & Luck, 2015).

2.2 I

-

-

Interference-by-process, an alternative explanation to the attention capture theory, is based on the notion that tasks of the same character cannot be performed simultaneously. When a target (e.g. actual task) is forced to compete with a non-target (distractor) a so-called interference arises. The interference can occur in both low-level (when only parts of the target signal is masked) or upper-level (when the target signal is driven out entirely), which in both cases are disruptive for the individual (Mattys, Davis, Bradlow, & Scott, 2012; Jones, Madden, &

7 Miles, 1992). A study made by Marsh, Hughes and Jones (2009) who investigated impact of interference on performance in diffrent noise types (silence, pink noise, speech; English vs.Welsh), showed that performance was best in the silent condition, while noise with high intelligibility, in this case English, had more negative affect on performance. The potential explanation to the results in Sörqvist et al. (2012) can be described by inteference-by-process;

as an involuntary analysis of the meaning of the background chatter cames into conflict with the ability to perform the language-based writing task.

2.3 L

Language production (e.g. speaking and writing) is a complex process involving a number of cognitive abilities that for each individual is different and there are limitations of producing language when it comes to working memory (McLachlan & Wilson, 2010; Saito & Baddeley, 2004; McCutchen, 2000; Venetjoki, Hongisto, Kaarlela-Tuomaala, & Keskinen, 2006). The system of language production is an interwoven and inevitable part of neurocognitive processing that involves cognitive abilities such as retrieval, parsing, and interpretation, and speech recognition consist of identifying short and long strings of phonemes that in the next stage is grouped, to form understandable words (Buchwald, 2015; James & Gauthier, 2009).

McCutchen (2000) explains further, that language producing is individual for each human being and it is depending of the richness of the individuals’ language knowledge. The language production comes also more difficult when distractors are involved.

2.4 S

The registration of speech intelligibility takes place automatically, which means that the brain involuntarily register both primary and secondary audible sources, which in turn hold back the ability of sorting relevant information from the irrelevant (Beaman , Broar, & Scott, 2007).

Speech perception is partly determined by the way it is presented i.e. structure of syllables, intonation, speed, etc., which allows speech intelligibility directly affected by pronunciation and accent (Mattys, Davis, Bradlow, & Scott, 2012). Lecumberri, Cooke and Cutler (2010), investigated non-native speech perception in adverse conditions and pointed out that

individuals' linguistic vocabulary memory can not make a difference on language differences.

Hearing spoken words in adverse conditions makes it hard to identify what is heard, which in turn starts a fill-in-process by producing words that are similar to the word that was heard, without having the same meaning. For example, if the heard word was lake, it can easily be

8 confused with take, cake, fake etc. There are also differences in terms of intelligibility of different languages. This seems to be because the languages is structured differently in terms of intonation and pronunciation, which is significant for the language heard as a mother tongue or as the second language. An individual’s listening experience in different languages is thus determined by the speech intelligibility, which in turn determines the understanding of a language.

2.5 P

The problem discussed and analysed in this paper is based upon previous environmental psychology research. Several researchers have pointed out that noise in our environment have impact on human behaviour and interferes with human’s attention, performance and other cognitive abilities (Lavie, Hirst, de Fockert, & Viding, 2004; Haapakangas, Hongisto, Hyönä, Kokko, & Keränen, 2014; Marsh, Hughes, & Jones, 2009; James & Gauthier, 2009).

The perspective of interference-by-process gives good reason to believe that the similarities in language (between producing language and listening to language) should do the distraction of background chatter larger. This is tested in the current study, and it is done in the context of a highly applied task (translating/writing), which makes this study unique and especially valuable from an applied perspective.

In the present experiment, participants performed tasks consisting translation of text from native language to second language and vice versa, while also hearing one of the languages as irrelevant background speech through headphones. Background speech, generally is

considered annoying and disruptive when working with tasks that requires focus, but does it differ in working performance, as in writing fluency depending of what language is spoken in the background?

2.6 H

Writing fluency (WF) decreases when hearing native language (Swedish) compared to hearing second language (English) as background speech through headphones when producing Swedish text (translating English text into Swedish).

9

3 M ETHOD

The experimental study, which this essay is based on, took place in a laboratory at the University of Gävle in Sweden.

3.1 P

Of fifty recruited university students, forty-three (mean age = 23.46 years, 28 female and 15 male) were included in the final analysis. All participants included had Swedish as the native language. Language knowledge in both Swedish and English where self-estimated on a Likert scale 1-7, where one is very bad and seven is very good.

None of the participants reported hearing difficulties or other dysfunctions. Data from seven persons were excluded because of technical errors.

3.2 D

The experiment had a within-participants design with two factors: language to-be-produced (Swedish vs. English) and language of background-speech (Swedish vs. English).

There was hence four conditions (Table 1.)

Table 1. The four conditions of the current experiment

Background speech Language to-be-produced

Swedish

Swedish English

English

Swedish English

The order between the conditions was counterbalanced across participants with a Latin Square design. Each participant listened to two English and two Swedish sound files and translated two English and two Swedish texts.

Dependent variables

One dependent variable was extracted by means of the built-in functions in the ScriptLog:

Writing fluency (number of characters in the final edited text summarized with number of deleted characters during the writing process).

10

3.3 M

Background Speech

The background sound consisted of two different audio files, one in English and one in Swedish, both cut into two pieces of 5 minutes/bit. Additionally, a third slice of 60 seconds was cut from the English audio file and this sound file was played back during the test phase.

In total, there were five sound files, two unique English audio files stemming from the same English story and the training sequence, as well as two unique Swedish audio files stemming from the same Swedish story.

Two women, with similar voice character, recited each story. The content of the stories were completely irrelevant to the main task and dealt with various women's lives.

The sound files, with a sound level of approximately 60 dB, was presented through headphones (Sennheiser HD 202). Presentation order of the audio files determined by

counterbalancing within- and between the conditions, so that each participant listened to two English and two Swedish sound files.

Writing Task

The writing task containing eight different texts consisted of four texts to be translated from Swedish to English, and four to be translated from English to Swedish. The texts were

adapted from the Swedish and the English version of the children's book Matilda (authored by Roald Dahl) and were selected from four different chapters. A piece of text in Swedish was selected to the test phase, also from the same book. In total, there were five different texts, all from the same book but from different chapters. Each text was approximately one A4 page long. All texts were presented on the computer screen through the computer program Script Log.

The text remained on the screen throughout condition length (5 minutes).

Next to the text file, there was an empty text box in which participants were requested to write the translation. Presentation order of the texts were determined by counterbalancing within- and between conditions so that each participant in the end translated two English and two Swedish texts, all from individual chapters.

11

3.4 P

Three soundproof rooms were prepared with a simple office set consisting a chair, a table, a computer and headphones. Before entering the test room, the participants received

instructions according the test, which told them to ignore the background speech and to try to write as fast and accurately as possible. The participants sat alone in each room, instructed to sit in front of the computer, with the headphones on, and then start the test by pressing the

“start” button. Before the actual test, the participants completed a training phase of 60

seconds to be acquainted with the test procedure. The instruction was not to translate the texts word by word; instead, they were told to create as natural sentences as possible without changing the content of origin. Each test condition lasted for 5 minutes. The text to be translated appeared on one side of the screen and the writing space on the other side of the screen. After every condition (5 minutes text section) a voice in the headphones said to the participants to stop writing and move on to the next condition. By pressing “Next”, the

participants could continue to the next condition. When each participant was finished with the translation task, shared test leader out the list of questions regarding the language skills.

3.5 S

All data from the experimental trials collected with Computer software Script Log that is a program designed to record all the keystrokes during testing. With a built-in program Script Log extracted relevant variables. The statistical analyses completed with software SPSS 22.0.

12

4 R ESULTS

4.1 L

Language knowledge was self-estimated on a Likert scale (1-7).

Twenty participants estimated their knowledge in Swedish language as very good (7). None of the participants estimated their knowledge in English language as very good (7). None of the participants estimated their knowledge in either of the languages as very bad (0). The mean value of the test group in Swedish language was 6.27 and in English language was 4.58.

4.2 W

Number of keystrokes (total number of characters produced summarized with the total number of deleted characters) measured the writing fluency.

As can be seen in Figure 1, the participants had higher writing fluency when they produced Swedish text compared to when they produced English text, but there was no significant effect of background speech.

These conclusions received support from a 2 × 2 repeated measures analysis of variance:

Text to-be-produced (Swedish vs. English) × Language of background speech (Swedish vs.

English). The analysis of variance revealed a main effect of text to-be-produced, F (1, 42) = 18.80, MSE = 4642.03, p < .001, ηp2 = .31, but no main effect of language of background speech, F (1, 42) = 0.50, MSE = 2814.00, p = .482, ηp2 = .01, and no interaction between the two factors, F (1, 42) = 0.17, MSE = 3378.83, p = .681, ηp2 < .01.

Figure 1. Mean values of writing fluency among all participants in each condition.

560 580 600 620 640 660

Swedish text production

English text production

Swedish text production

English text production Swedish Background Speech English Background Speech

W riting flue n cy

13

5 D ISCUSSION

The aim of this essay was to investigate if writing fluency depend on the similarity between the languages (Swedish as native vs. English as second) spoken in the background and the language of the text that is produced in the writing task.

The following hypothesis was tested:

Writing fluency (WF) decreases when hearing native language (Swedish) compared to hearing second language (English) as background speech through headphones when producing Swedish text (translating English text into Swedish).

5.1 D

The results of this study revealed two main finding: There was no evidence for the assumption that background speech causes interference. The second finding is that language knowledge, in terms of understanding and writing, seem to be an important factor that influences writing fluency, because the participants produced more text when they wrote in Swedish (their native language) in comparison with writing in English. More specifically, the result shows that writing fluency increased when native language (Swedish) was heard as background speech and native language (Swedish) was produced. Regardless of what language was heard in the background, writing fluency increased more when the language to-be-produced was the native language (Swedish) compared to when the language to-be-produced was the second language (English). Hence, the results cannot support the hypothesis.

5.2 E

Noise, such as background speech, is a well-known factor that interferes with the attention control and individuals with weaker working memory have generally less capacity to maintain attention on cognitive tasks, according to previous studies (Bell, Röer, et al. 2012; Jones, et al.1992). According to McLachlan et al. (2010; McCutchen, 2000; Venetjoki et al., 2006) working memory capacity is an important factor when it comes to implementing language activities such as reading and writing. Saito and Baddeley (2004) believes that the linguistic performance is linked to working memory capacity, and that individual differences in working memory creates differences in the ability to divide attention. High working memory capacity makes it easier to ignore distractors, such as e.g. background speech. This explanation could be plausible for the present study as the participants of this study were university students

14 who may be more accustomed to performing tasks in an environment where background speech is common, such as universities, which may explain why they appeared to be immune to disruption. University students are also relatively skilled in both Swedish and English language, which could have facilitated the implementation of the task, whereby performance may have become relatively immune to disruption. Another explanation of why the

participants were not affected by background speech may be that the task may have been too simple in the sense that it did not require the involvement of a large portion of working memory capacity, which in turn could have made it easier to ignore the sound (Lavie et al., 2004). Venetjoki, et al. (2006) argue that when university students are used as a test group, it is possible to use more demanding reading comprehension tests. The perspective of

interference-by-process, in the expression of that the similarities in the task (language-to-be- produced) and the listening activity (language to-be-heard) would inhibit job performance, does not work in the context of present study because opposite effects were revealed in the results. The more probable explanation for the results in the present study could be the theories of working memory load and its function when performing linguistic tasks.

5.3 D

The task used in this experiment is similar to the tasks used in previous studies (Keus van de Poll, Ljung, Odelius, & Sörqvist, 2014; Marsh, Hughes, & Jones, 2009; Sörqvist, Nöstl, &

Halin, 2012) except for the fact that the task used in the present experiment was a translation task instead of a story writing task. In the present study, there also was no silent condition, without which it is impossible to conclude that the background noise had no effect at all. This means that we do not know how each individual performs in writing fluency in silent

condition, which limits the possibilities to draw conclusions about the effects of the

background speech. Regarding the participants, there is a certain skewed sex ratio (28 females vs. 15 males), which may have affected the results. A larger and a more gender-balanced group would have made it easier to draw reliable, general conclusions.

5.4 C

Noise effects are typically investigated with larger samples; however, this study suggests that the language of the background speech does not interact with the language of the to-be- produced text in their effects on writing fluency. The differences found in language to-be- produced indicates that language skills probably is the key of higher writing fluency and that

15 the individual's working memory determines how distractors are received. Previous studies also show that it is possible to ignore distractors such as background speech when there is sufficiently of capacity in the working memory (Bell, et al. 2012; Jones, et al.1992). The absence of silent condition makes it difficult to do any general conclusions about the present study, but still allows believing (according to previous research) that background speech may have some impact on performance in other circumstances. However, further research will be needed to determine whether distracting background speech in various languages have impact on work performance or not.

16

R EFERENCES

Beaman , P. C., Broar, A. M., & Scott, S. K. ( 2007). From Dichotic Listening to the

Irrelevant Sound Effect: a Behavioural and Neuroimaging Analysis of the Processing of Unattended Speech. Cortex, Volume 43, Issue 1,, 124–134.

Bell, R., Röer, J. P., Dentale, S., & Buchner, A. (2012). Habituation of the Irrelevant Sound Effect: Evidence for an Attentional Theory of Short-Term Memory Disruption.

Journal of Experimental Psychology: Learning, Memory, and Cognition, Vol.38, No.6, 1542–1557.

Buchwald, A. (2015). Cognitive and neural constraints on theories of language production.

Language, Cognition and Neuroscience, Volume 30, Issue 3, 235-237.

Haapakangas, A., Hongisto, V., Hyönä, J., Kokko, J., & Keränen, J. (2014). Effects of unattended speech on performance and subjective distraction: The role of acoustic design in open-plan offices. Applied Acoustics 86, 1-16.

Higgins, J. A., & Johnson, M. K. (2013 ). Lost Thoughts: Implicit Semantic Interference Impairs Reflective Access to Currently Active Information. Journal of Experimental Psychology: General, Vol. 142, No. 1, 6-11.

Hongisto, V. (2005). A model predicting the effect of speech of varying intelligibility on work performance. Indoor Air, 15, 458-468.

James, K. H., & Gauthier, I. (2009). When Writing Impairs Reading: Letter Perception’s Susceptibility to Motor Interference. Journal of Experimental Psychology: General, Vol. 138, No 3, 416 – 431.

Jones, D., Madden, C., & Miles, C. (1992). Priviliged Accsess by irrelevant Speech to Short- Term Memory: The Roleof Changing State. The Quartely Journal of Experimental Psychology, 654-669.

Keus van de Poll, M., Ljung, R., Odelius, J., & Sörqvist, P. (2014). Disruption of writing by background speech: The role of speech transmission index. Applied Acoustics, 81, 15- 18.

17 Lavie, N., Hirst, A., de Fockert, J. W., & Viding, E. (2004). Load Theory of Selective

Attention and Cognitive Control. Journal of Experimental Psychology, Vol. 133, No.

3, 339–354.

Lecumberri, M. L., Cooke, M., & Cutler, A. (2010). Non-native speech perception in adverse conditions: A review. Speech Communication, Volume 52, Issues 11–12, 864-886.

Marsh, J. E., Hughes, R. W., & Jones, D. M. (2009). Interference by process, not content, determines semantic auditory distraction. Cognition 110 , 23-38.

Mattys, S. L., Davis, M. H., Bradlow, A. R., & Scott, S. K. (2012). Speech recognition in adverse conditions: A review. Language and Cognitive Processes, Volume 27, Issue 7- 8, 953-978.

McCutchen, D. (2000). Knowledge, Processing, and Working Memory: Implications for a Theory of Writing. Educational Psychologist, 13-23.

McLachlan, N., & Wilson, S. (2010). The Central Role of Recognition in Auditory

Perception: A Neurobiological Model. Psychological Review, Vol. 117, No. 1,, 175- 196.

Saito, S., & Baddeley, A. D. (2004). Irrelevant sound disrupts speech production: Exploring the relationship between short-term memory and experimentally induced slips of the tongue. The Quarterly Journal of Experimental Psychology, 1309-1340.

Schlittmeier, S. J., Hellbrück, J., Thaden, R., & Vorländer, M. (2008). The impact of background speech varying in intelligibility: Effects on cognitive performance and perceived disturbance. Ergonomics, Vol. 51, No. 5, 719–736.

Seddigh, A., Berntson, E., Bodin Danielson, C., & Westerlund, H. (2014). Concentration requirements modify the effect of office type on indicators of health and performance.

Journal of Environmental Psychology, Volume 38, 167–174.

Smith-Jackson, T. L., & Klein, K. W. (2009). Open-plan offices: Task performance and mental workload. Journal of Environmental Psychology, Volume 29, Issue 2, 279–289.

Sörqvist, P., Nöstl, A., & Halin, N. (2012). Disruption of writing processes by the semanticity of background speech. Scandinavian Journal of Psychology, 97-102.

18 Venetjoki, N., Hongisto, V., Kaarlela-Tuomaala, A., & Keskinen, E. (2006). The effect of

speech and speech intelligibility on task performance. Ergonomics, Vol. 49, No. 11, 1068–1091.

Wilson, D. E., Muroi, M., & MacLeod, C. M. (2011). Dilution, Not Load, Affects Distractor Processing. Journal of Experimental Psychology: Human Perception and

Performance, Vol. 37, No. 2, 319 –335.

World Health Organization. (2011). Burden of Disease from Environmental Noise - Quantification of healthy life years lost in Europe. Copenhagen: World Health Organization, Regional Office for Europe.

Yeh, Y.-Y., & Lin, S.-H. (2013). Two Mechanisms of Distractor Dilution: Visual Selection in a Continuous Flow. Journal of Experimental Psychology: Human Perception and Performance, Vol. 39, Issue 3, 872– 892.

Zeamer, C., & Fox Tree, J. E. (2013). The Process of Auditory Distraction: Disrupted Attention and Impaired Recall in a Simulated Lecture Environment. Journal of Experimental Psychology: Learning, Memory, and Cognition, Vol. 39, No 5, 1463–

1472.

Zhang, W., & Luck, S. J. (2015). Opposite effects of capacity load and resolution load on distractor processing. Journal of Experimental Psychology: Human Perception and Performance, Vol 41(1), 22-27.