Disruption of writing by background speech:
The role of speech transmission index
Marijke Keus van de Poll, Robert Ljung, Johan Odelius and Patrik Sörqvist
Linköping University Post Print
N.B.: When citing this work, cite the original article.
Original Publication:
Marijke Keus van de Poll, Robert Ljung, Johan Odelius and Patrik Sörqvist, Disruption of
writing by background speech: The role of speech transmission index, 2014, Applied
Acoustics, (81), 15-18.
http://dx.doi.org/10.1016/j.apacoust.2014.02.005
Copyright: Elsevier
http://www.elsevier.com/
Postprint available at: Linköping University Electronic Press
Disruption of writing by background speech: The role of speech
transmission index
Marijke Keus van de Poll
a,⇑, Robert Ljung
a, Johan Odelius
b, Patrik Sörqvist
a,ca
Department of Building, Energy, and Environmental Engineering, University of Gävle, SE-80176 Gävle, Sweden b
Department of Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, Luleå, Sweden c
Linnaeus Centre HEAD, Swedish Institute for Disability Research, Linköping University, Linköping, Sweden
a r t i c l e
i n f o
Article history: Received 29 April 2013
Received in revised form 14 January 2014 Accepted 6 February 2014
Available online 4 March 2014 Keywords:
Writing Distraction
Speech transmission index Irrelevant speech Open-plan office
a b s t r a c t
Speech transmission index (STI) is an objective measure of the acoustic properties of office environments and is used to specify norms for acceptable acoustic work conditions. Yet, the tasks used to evaluate the effects of varying STIs on work performance have often been focusing on memory (as memory of visually presented words) and reading tasks and may not give a complete view of the severity even of low STI values (i.e., when speech intelligibility is low). Against this background, we used a more typical office-work task in the present study. The participants were asked to write short essays (5 min per essay) in 5 different STI conditions (0.08; 0.23; 0.34; 0.50; and 0.71). Writing fluency dropped drastically and the number of pauses longer than 5 s increased at STI values above 0.23. This study shows that realistic work-related performance drops even at low STI values and has implications for how to evaluate acoustic conditions in school and office environments.
Ó 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).
1. Introduction
Undesired background speech in offices and similar environ-ments potentially impairs work-related skills like writing[1–3], reading comprehension[4,5], proofreading[6–8], and prose mem-ory [9,10]. Occupational noise, and speech in particular, are amongst the most often mentioned sources of annoyance at work
[11–13], decreases satisfaction with the work environment[14], and are stressful [15,16]. Thus, performance and health are challenged when working in open-office environments wherein background sound, like talking colleagues, ringing phones, noise from copying machines, and so on, is common place.
One objective predictor that is used to evaluate the acoustic environment is the Speech transmission index (STI) e.g.,[17]. STI is a physical measure of speech intelligibility (i.e., the possibility to hear what is said) and is standardized by IEC 60268-16[18]. It ranges from perfect speech intelligibility (i.e., 1.00) to no intelligi-bility at all (i.e., 0) and depends mostly on signal-to-noise ratio, reverberation and the amount of early reflections between the source and the receiver. A number of studies have investigated the relation between STI values of background sound and work-related performance e.g.,[8,19–21]. A dominant view developed
by Hongisto[17]is that performance drops most drastically when the background speech has an STI around 0.30 and 0.40, and that the decrement in performance fades out after an STI of 0.50.
Hongisto’s model is based on a generalization across several, different cognitive tasks. More recent studies have investigated the influence of STI on specific tasks and those studies indicate that marked performance decrements are observed with as low STI values as 0.34 [21], and that there is no significant decrease in performance with exceeding STI values. The tasks used in those studies were a short-term memory task denoting recall of visually presented word sequences; an information search task with the instruction to search through a matrix to find answers on certain questions; a math task containing addition of three-digit numbers; and a phonemic and semantic fluency task that requires generation of words from specific categories (i.e. animals or vegetables). In all, the type of tasks used to demonstrate this potentially disruptive effect of relatively low STI values have been rather unrepresenta-tive for office work (e.g., memory of visually presented word sequences) and there is little reason to assume that this type of task is particularly sensitive to disruption from background speech e.g.,[9]. In an attempt to improve upon past studies, we used word processed writing in the present experiment as a tool to investigate whether marked performance decrements kick in at relatively low STI values in a more realistic and common type of office related task—word processed writing—that is known to be particularly susceptible to disruption from background speech.
http://dx.doi.org/10.1016/j.apacoust.2014.02.005
0003-682X/Ó 2014 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).
⇑ Corresponding author. Tel.: +46 26648851.
E-mail address:marijke.keus@hig.se(M. Keus van de Poll).
Contents lists available atScienceDirect
Applied Acoustics
Several cognitive processes are involved in writing such as idea generation, retrieval from long-term memory, organizing ideas and transforming thoughts and ideas into orthographic representations
[22]. Most essentially, writing requires processing of meaning, a situation that makes writing easily disrupted by the presence of background speech[26,27]. For instance, Sörqvist et al.[3]found that writing performance is impaired by background speech (pre-dominantly an impairment of quantitative aspects of the writing process such as writing fluency, but also an increase in pauses dur-ing the writdur-ing process) in comparison with a silent condition and with a condition wherein the background sound consisted of a spectrally-rotated version of the speech signal. The impairment of writing fluency (the sum obtained when adding deleted charac-ters to the total number of characcharac-ters in the final edited text) and an increase in the number of pauses (longer than 5 s) by background speech have also been confirmed in related studies
[1,2]. Thus, writing (especially writing fluency) appears to be impaired specifically by the meaning of background speech, whereas acoustic properties of the sound are not especially disrup-tive. This finding fits well with the general view that deliberate semantic processes (e.g., interpreting the meaning of a word) is disrupted by meaningful background speech, because the meaning of the background speech is semantically analyzed, and this involuntary analysis activates cognitive representations in seman-tic memory that interfere with the execution of the deliberate semantic processes[23–25].
In accordance with this interference-by-process view, writing processes should be increasingly impaired as STI values increases, because higher STI value are associated with higher speech intelli-gibility. In low STI conditions, the semanticity of the background speech is hardly noticeable, meaning there will be no (or at least only weak) conflict with the deliberate semantic processes. In the present study, the function between STI and writing performance was investigated by using five different STI conditions, giving a more fine-tuned manipulation of background speech intelligibility than in the study by Sörqvist et al.[3]. The expectations in the present study were that writing fluency would decrease, and the number of pauses above 5 s would increase, as a function of background speech intelligibility. Specifically, writing fluency should drop as STI value increases and the largest drop should be observed at values around 0.34. Moreover, the number of pauses should increase with increasing STI and the largest increase should be found around 0.34.
2. Methods 2.1. Participants
A total of 33 students (mean age = 25.36 years, SD = 5.99) at the University of Gävle participated in the study. All participants had completed Swedish compulsory school and high school and all par-ticipants had normal or corrected to normal vision. One participant reported hearing loss. As this person’s data were not markedly dif-ferent from the sample means, and control analyses without this person’s data excluded were entirely consistent with analyses with those data included, data from this person were included in the reported analyses. The participants received a cinema ticket for participation.
2.2. Apparatus and materials 2.2.1. Sound
The irrelevant speech consisted of five different stories. The stories were about different topics (e.g., frogs’ and fish’s ability to predict weather and history of poems). They were spoken in a male
voice and binaurally recorded at 44.1 kHz using an artificial head (Head Acoustics HMS IV) in an anechoic chamber at 0 degrees azi-muth. Octave levels for the five stories are presented inFig. 1. Each story was masked by binaurally uncorrelated pink noise (i.e., equal level at every third octave band) to produce five different STI val-ues (0.08, 0.23, 0.34, 0.50, and 0.71). The pink noise was band-pass filtered between 22 Hz and 18 kHz. The better-ear STI value was used, as recommended in [18]; however, the experimental setup caused insignificant binaural differences. The equivalent A-weighted levels (summed for left and right channel) were nor-malized between sound files. SeeFig. 2for an overview of how the sound stimuli were composed. In all, a total of 25 sound files were used. All sound stimuli were presented through headphones (Sennheiser HD 202) at approximately 60 dBA, corresponding to the sound level of a conversation within 1–2 m. The noise level is typically a little lower (45–55 dBA) in open office environments
[18].
2.2.2. Writing task
The participants were asked to write five stories associated with different words displayed on a computer screen. The target-words were the names of different nature scenes (i.e., forest, des-ert, sea, field, and mountains). A different name was used for each story. They were presented in the same sequential order to all par-ticipants. The onset and the offset of the target-word and the sound were synchronized. The time limit for each story was set to five minutes. After five minutes, a warning signal was played in the headphones and the participants received spoken instructions that told them to delete all written text and press a button to pass onto the next condition. The computer software ScriptLog was used to obtain data. This program is developed for real-time analysis of the writing process and it registers all keyboard activity. This makes it possible to replay the writing sequence for real-time anal-ysis and to extract relevant statistics automatically by using the built-in functions.
2.3. Dependent variables
Two dependent variables were extracted using ScriptLog: writ-ing fluency (i.e., the number of characters in the final edited text plus the number of characters deleted during the writing process) and the number of pauses longer than 5 s. Another set of poten-tially relevant dependent variables (the same as in Sörqvist et al.
[3]) was also considered but did not reveal anything valuable be-yond the two reported.
Fig. 1. The sound pressure levels in octave bands for the 5 different speech stories plotted for the ear with maximum A-weighted level.
2.4. Design and procedure
A within-subjects design was used. The participants sat alone in a sound attenuated room in front of a stationary computer. They wore headphones during the whole experiment. The writing task was introduced by a practice phase of 30 s so the participants would get acquainted with the task and the procedure. The tar-get-word ‘city’ was presented in this practice phase. The practice phase was followed by the five STI conditions. The order of the five conditions was counterbalanced across participants for STI value and the story using a Latin Square design. Participants were instructed to write as much, as fast and as accurate as they could, to write stories associated with the presented word and to ignore the sounds in the headphones. The experiment took about 40 min to complete.
3. Results 3.1. Writing fluency
Fig. 3shows that writing fluency drops as a function of STI, with the highest mean writing fluency in the lowest STI condition. A repeated measures analysis of variance (ANOVA) across the five STI conditions revealed a significant effect of condition, F(4, 128) = 3.84, MSE = 5531, p = 0.006,
g
2p¼ 0:11. Moreover,
con-trast analyses showed that the linear trend was significant, F(1, 32) = 13.04, MSE = 5854, p < .001,
g
2p¼ 0:29. A follow-up t-test
showed that writing fluency was, most importantly, significantly lower in the STI = 0.34 condition in comparison with the STI = 0.23 condition, t(32) = 1.90, p = 0.033 (one-tailed). Further t-test analyses showed significant differences (one-tailed) between STI = 0.08 and STI = 0.34, t(32) = 1.79, p = 0.04, between STI = 0.08 and STI = 0.50, t(32) = 2.41, p = 0.01, between STI = 0.08 and STI = 0.71, t(32) = 3.07, p = 0.002, between STI = 0.23 and STI = 0.50, t(32) = 2.68, p = 0.005, and between STI = 0.23 and STI = 0.71, t(32) = 2.79, p = 0.005. All other comparisons were non-significant.
3.2. Number of pauses
Fig. 4shows that the number of 5 s long pauses increases with increasing STI. This was supported by an ANOVA, F(4, 128) = 3.32, MSE = 2.95, p = 0.013,
g
2p¼ 0:09. A follow-up t-test showed that
the difference between the STI = 0.23 condition and the STI = 0.34 condition was significant, t(32) = 2.63, p = 0.013. Further t-test analyses showed significant differences (one tailed) between STI = 0.08 and STI = 0.34, t(32) = 2.30, p = 0.014, between STI = 0.08 and STI = 0.50, t(32) = 1.90, p = .033, between STI = 0.08 and STI = 0.71, t(32) = 2.77, p = 0.005, between STI = 0.23 and STI = 0.50, t(32) = 2.03, p = 0.025, and between STI = 0.23 and STI = 0.71, t(32) = 2.37, p = 0.012. All other comparisons were non-significant.
3.3. Relation between writing fluency and number of pauses As the function between STI and writing fluency and the function between STI and pause frequency are quite similar, one likely possibility is that writing fluency depends on the number of pauses the writer takes during the writing process. To further investigate this matter, Pearson product-moment correlation coefficients between writing fluency and number of pauses, within each STI condition respectively, were analyzed. All correlation coefficients were significant, r(31) = 0.55, r(31) = 0.58, r(31) = 0.63, r(31) = 0.78, and r(31) = 0.60 (all p < 0.01). These results further reinforce the assumption that writing fluency ultimately depends on the number of longer pauses (i.e., above 5 s) taken by the writer.
4. Discussion
By using a realistic task, that is highly representative of typical office work, this study reveals that cognitive performance is impaired by speech even at relatively low STI values. More specif-ically, word processed writing is impaired (writing quantity
Fig. 2. The design of target STI values by changing the speech and pink noise levels while the total A-weighted level is kept constant. The levels shown are with regard to the octave levels of one of speech stories; however the level difference due to the different spectral content of the five stories was less than 0.5 dBA.
780 795 810 825 840 855 870 885 0.00 0.20 0.40 0.60 0.80 1.00
Mean wring fluency
Speech Transmission Index
Fig. 3. Mean writing fluency (i.e., the number of characters in the final edited text plus the number of characters deleted during the writing process) in the five experimental speech transmission index conditions (0.08; 0.23; 0.34; 0.50; and 0.71). 0 1 2 3 4 5 0.00 0.20 0.40 0.60 0.80 1.00
Mean number of pauses
Speech Transmission Index
Fig. 4. Mean number of pauses longer than 5 s in the five experimental Speech Transmission Index conditions (0.08; 0.23; 0.34; 0.50; and 0.71).
suppressed and the number of pauses longer than 5 s increased) as STI exceeds 0.23. This confirms Hongisto’s model[17]wherein per-formance starts to drop at an STI of about 0.20.
The present study extends past work in several important direc-tions. By using more categorically distinguished sound conditions (e.g., silence vs. normal speech vs. spectrally rotated speech) than the present study, Sörqvist et al.[3]found that a fully intelligible speech signal impairs writing performance whereas a spectrally-rotated speech signal—which is completely incomprehensible to the unpracticed ear—has no effect whatsoever on writing perfor-mance as compared to silence. Here, using a more fine-tuned speech intelligibility manipulation, we show that there appears to be a cut-off at around STI 0.23 after which performance drops drastically. This conceptually replicates and specifies the finding that writing is susceptible to disruption from background speech, and that this disruption is specifically caused by the speech’s semanticity rather than the sound’s acoustic properties.
The findings reported here seems to suggest that the semantic-ity of speech suppresses the semantic output processes needed to produce text materials. Hence, the disruption of the writing process seems to be a result of a conflict between the deliberate semantic processes involved in writing and the obligatory process-ing of semantic information conveyed by irrelevant background speech[3]. As such, the present experiment supports the interfer-ence-by-process view of auditory distraction[26,27].
In the present study, only quantitative variables were analyzed, no qualitative variables such as the complexity of the written texts. Qualitative aspects (e.g., the number of propositions in the final edited texts, and readability index) were analyzed in Sörqvist et al.[3], but no differences between conditions were found. A ma-jor reason for this appeared to be the relatively low reliability of those data. Because of this, qualitative analyses are not reported in the present study.
The study reported here is an important complement to Hongisto’s model[17]in the sense that the model is an average of several cognitive tasks (writing not included), neither of which that is known to be particularly susceptible to disruption from background speech. By presenting a specific, relevant office task, this study contributes to a more detailed view on how STI and cog-nitive performance are related in an applied setting. We show that the strongest decrease in performance occurs between STI 0.23 and 0.34 in the context of word processed writing, with no more signif-icant decrease in performance when STI exceeds 0.34. This finding is roughly consistent with the results reported by Jahncke et al.
[21]who used a wider range of cognitive tasks.
In conclusion, this study shows that relatively low speech intel-ligibility can have negative/disturbing effects on tasks representing office work. The nature of the task that we employed—word pro-cessed writing—is one of the dominant activities in classrooms and open-office settings wherein background speech is relatively common. These negative consequences of background speech on writing should be taken into account in the evaluation of acoustical environments in classrooms and open-office settings.
Acknowledgements
This investigation was part of Keus van de Poll’s doctoral thesis and it was financially supported by a grant from the Swedish Research Council for Environment, Agricultural Sciences and Spa-tial Planning. We would like to thank Helena Jahncke for assistance with designing the sound material.
References
[1]Ransdell S, Gilroy L. The effects of background music on word processed writing. Comput Hum Behav 2001;17:141–8.
[2]Ransdell S, Levy CM, Kellogg RT. The structure of writing processes as revealed by secondary task demands. L1-Educ Stud Lang Lit 2002;2:141–63. [3]Sörqvist P, Nöstl A, Halin N. Disruption of writing processes by the semanticity
of background speech. Scand J Psychol 2012;53:97–102.
[4]Martin RC, Wogalter MS, Forlano JG. Reading comprehension in the presence of unattended speech and music. J Mem Lang 1988;27:382–98.
[5]Sörqvist P, Halin N, Hygge S. Individual differences in susceptibility to the effects of speech on reading comprehension. Appl Cogn Psychol 2010;24:67–76.
[6]Jones DM, Miles C, Page J. Disruption of proofreading by irrelevant speech: Effects of attention, arousal or memory? Appl Cogn Psychol 1990;4:89–108. [7]Smith-Jackson TL, Klein KW. Open-plan offices: task performance and mental
workload. J Environ Psychol 2009;29:279–89.
[8]Venetjoki N, Kaarlela-Tuomaala A, Keskinen E, Hongisto V. The effect of speech and speech intelligibility on task performance. Ergonomics 2006;49:1068–91. [9]Banbury S, Berry DC. Disruption of office-related tasks by speech and office
noise. Br J Psychol 1998;89:499–517.
[10]Sörqvist P. Effects of aircraft noise and speech on prose memory: what role for working memory capacity? J Environ Psychol 2010;30:112–8.
[11]Banbury S, Berry DC. Office noise and employee concentration: identifying causes of disruption and potential improvements. Ergonomics 2005;48:25–37. [12]Boyce PR. User’s assessments of a landscaped office. J Archit Res 1974;3:44–62. [13]Young H, Berry GL. The impact of environment on the productivity attitudes of intellectually challenged office workers. Hum Fact: J Hum Fact Ergonom Soc 1979;21:399–407.
[14]Sundstrom E, Town JP, Rice RW, Osborn DP, Brill M. Office noise, satisfaction, and performance. Environ Behav 1994;26:195–222.
[15]Babisch W. Stress hormones in the research on cardiovascular effects of noise. Noise Health 2003;5:1–11.
[16]Smith A. A review of the non-auditory effects of noise on health. Work Stress 1991;5:49–62.
[17]Hongisto V. A model predicting the effect of speech of varying intelligibility on work performance. Indoor Air 2005;15:458–68.
[18] IEC 60268-16. Sound system equipment – Part 16: objective rating of speech intelligibility by speech transmission index, 4th ed. Switzerland: IEC; 2011. [19]Haka M, Haapakangas A, Keränen J, Hakala J, Keskinen E, Hongisto V.
Performance effects and subjective disturbance of speech in acoustically different office types–a laboratory experiment. Indoor Air 2009;19:454–67. [20]Hongisto V, Haapakangas A, Helenius R, Keränen J, Oliva D. Acoustic
satisfaction in an open-plan office before and after the renovation. Euronoise 2012:654–9. Prague, Check Republic.
[21]Jahncke H, Hongisto V, Virjonen P. Cognitive performance during irrelevant speech: effects of speech intelligibility and office-task characteristics. Appl Acoust 2013;74:307–16.
[22]Flower L, Hayes JR. A cognitive process theory of writing. Coll Compos Commun 1981;32:365–87.
[23]Marsh JE, Hughes RW, Jones DM. Auditory distraction in semantic memory: a process-based approach. J Mem Lang 2008;58:682–700.
[24]Marsh JE, Hughes RW, Jones DM. Interference by process, not content, determines semantic auditory distraction. Cognition 2009;110:23–38. [25]Sörqvist P, Marsh JE, Jahncke H. Hemispheric asymmetries in auditory
distraction. Brain Cogn 2010;74:79–87.
[26]Marsh JE, Jones DM. Cross-modal distraction by background speech: what role for meaning? Noise Health 2010;12:210–6.
[27]Jones DM, Marsh JE, Hughes RW. Retrieval from memory: vulnerable or inviolable? J Exp Psychol Learn Mem Cogn 2012;38:905–22.
