1 Introduction

Background noise is considered a stressor and has been shown to negatively affect individuals’ cognitive performance in tasks such as attention, memory and decision-making (Broadbent 1971; Smith 1989b; Szalma and Hancock 2011). In safety critical industries, such as maritime, medicine, aviation, and military, if left unmanaged, noise has the potential to adversely affect individuals’ cognitive performance and increase risk. Even moderate intensity noise, described as unwanted or loud sounds less than < 85 dBA (Smith 1989b) has been found to affect individuals’ focused attention (Smith 1991), working memory (Herweg and Bunzeck 2015), long-term memory (Molesworth et al. 2015) and reading comprehension (Gheewalla et al. 2021). Such noise is commonly present in safety critical industries, such as broadband noise at 65–85 dBA in commercial aircraft cabin (Burgess et al. 2014), or broadband noise at 37–88.6 dBA in indoor health care environment (de Lima Andrade et al. 2021). Therefore, the aim of the present systematic review is to examine the effect of moderate intensity noise (i.e., continuous broadband noise between 55 and 85 dBA) on cognitive performance.

Existing knowledge on the effect of moderate intensity noise on cognitive performance is limited. An intense noise stressor such as loud noise (> 85 dBA; Hockey 1970) or intermittent noise (Broadbent 1971) usually generates a clear pattern of effects on performance. This effect is evident irrespective of any moderator/s. In contrast the effect of moderate intensity noise (i.e., considered mild) is more complex and variable (Smith 1989b). Early studies around 1980s reveal the effect of moderate intensity noise is influenced by a variety of moderators such as personality traits, sex, age, task type, instructions about the noise effect/s, time of day, duration of noise exposure or repeated sessions (Smith 1989b) that may account for the variability. However, Smith’s (1989b) narrative review failed to identify any consistent and reliable moderator. In a more recent and detailed meta-analysis, Szalma and Hancock’s (2011) identified noise parameters (e.g., noise level, noise type) and task type dichotomy based on cognitive demand (perceptual vs. cognitive/communication tasks) to be key moderators in the noise effect. Szalma and Hancock’s meta-analysis, however, remained silent of some of the key moderators relating to moderate intensity noise such as individual difference, and in particular how its effect varies across cognitive processes such as attention, memory or high(er)-order cognition. More recent literature reviews have focused on other topics such as noise, health and wellbeing (e.g., Mucci et al. 2021), the cognitive effect of chronic noise exposure (Thompson et al. 2022), and psychophysiological indicators of noise stress (Grenzebach and Romanus 2022). These areas of research do not directly provide insight to the acute effect of noise on cognitive performance (Smith 1989b), the focus of the present research.

Existing theories that have shown effective in explaining the effect of broader types of noise, including loud noise (i.e., > 85 dBA), varying noise or meaningful speech noise have limitations in accounting for the noise effect with moderate intensity noise. Theories such as Arousal Theory (Broadbent 1971) and resource-based theories (e.g., Maximum Adaptability Theory; Hancock and Warm 1989) primarily focused on the mechanisms underlying how excessive levels of stress impair performance, such as through noise-induced over-arousal (Broadbent 1971) or limited available cognitive resources to cope with the noise stressor (Hancock and Warm 1989; Hockey 1997). However, these theories lacked sufficient details in explaining the characteristics of mild stressors such as moderate intensity noise (i.e., continuous noise).

Another limitation of these theories is their ability to account for the effect of moderators such as individual differences on the noise effect (Smith 2012). Personality traits such as extroversion and neuroticism are two of the few individual difference moderators proposed by theories such as Arousal Theory (Broadbent 1971) and resource-based theories (i.e., Individual Difference Approach; Szalma 2012). These theories for moderators were derived from empirical evidence based on other types of noise such as intermittent noise or loud noise rather than moderate intensity continuous noise (e.g., Broadbent 1971; Szalma 2012). Their generalisability to moderate intensity continuous noise lacks evidence/support. Apart from personality, the role of other individual difference factors/moderators such as sex, age (Smith 1989b) or native language background (Molesworth et al. 2015) remain absent in these theories.

Another gap that remains in both empirical and theoretical research is to account for the effect of noise across a range of cognitive tasks. The empirical literature that shaped the two aforementioned theoretical perspectives was conducted prior to the 1980s and focused on vigilance tasks, a type of signal detection and monitoring dual task that examines sustained attention (for details, see Hancock and Warm 1989). This experiment paradigm helped reveal the moderating effect of cognitive load manipulated by task complexity or duration (e.g., Hockey 1970; Broadbent 1976). In the past two decades, new types of complex cognitive tasks became prevalent in occupational settings where noise is present, such as dynamic decision-making and various problem-solving tasks (Penney et al. 2022; Fu et al. 2019; Kylesten and Nählinder 2011). These tasks have notably different task structure and cognitive processes such as high-order cognition, on which the noise effect and the moderating effects of cognitive load cannot be simply generalised from the previous research and theories.

To fill the gap, this systematic review focuses on two key areas, namely: (1) What is the overall effect of moderate intensity noise on cognitive performance? Understanding this effect will provide insight into how moderate intensity noise acts as a stressor (e.g., over-arousing, cognitively taxing) on human performance. (2) Do moderators reliably explain the variability of the effect of moderate intensity noise?

The present review is particularly interested in the effect of moderate intensity continuous broadband noise (moderate broadband noise). On one hand, according to resource-based theories (Szalma and Hancock 2011), moderate broadband noise is theoretically representative of a mild noise stressor. It is milder than varying, intermittent or speech noise at the same noise levels, thus an ideal type of noise for understanding the effect of mild noise stress on performance. On the other hand, moderate broadband noise is prevalent in occupational settings of interest such as commercial civil aviation and road transport. It will still be present when noise control technology has led to reduction of high levels of workplace noise levels to below the noise damage limits (e.g., < 85 dBA). In addition, broadband noise, such as white noise, has been the most commonly used noise type in laboratory settings in the field of noise and performance over the decades. Focusing on broadband noise instead of incorporating other noise types (e.g., traffic noise, office noise) can avoid inducing the influence of noise parameters such as frequency spectrum, intermittency and meaningfulness.

2 Research aim and research questions

This systematic review aims to outline evidence for the effect of moderate broadband noise as well as the factors that influence these effects on indices of cognition such as reaction time, visual perception, visual attention, short-term memory, long-term memory, and high(er)-order cognition such as problem solving and decision-making. In pursuit of this aim, two research questions were proposed.

Research questions:

  1. 1.

    What is the effect of moderate (55–85 dBA) constant broadband noise on indices of cognition (i.e., perception, reaction time, attention, short-term memory, long-term memory and high(er)-order cognition)?

  2. 2.

    What moderating factors contribute to variability in the noise effect?

3 Material and methods

This systematic review follows the Prisma Statement (Moher et al. 2009).

3.1 Search strategy

Empirical studies in the field of broadband noise and its effect on cognition conducted between 1 January 1980 and 1 February 2023 were sourced from three databases: Medline (PubMed), Embase (Scopus), and PsycINFO. The search strategy identified a combination of keywords. Due to the term ‘noise’ being used in relation to many topics that are distinct from the present topic, keywords in the ‘NOT’ group were employed to exclude irrelevant materials. For example, to exclude studies related to hearing loss terms in the NOT group included ‘with hearing loss’, ‘on hearing loss’, ‘effect(s) of hearing loss’. Similarly, studies related to animals, to speech or sound recognition in noise, to all other non-acoustic, and various forms of noise reduction were excluded by including terms in the NOT group as can be seen from the listing below of the search terms. By repeatedly refining the search keywords in order to obtain a balance between comprehensiveness and precision, and considering the characteristics of the search functions provided by each database, the below search string was derived:

noise OR sound OR sounds OR “background noise” OR “white noise” OR “broadband noise” OR “loud noise” OR “auditory distraction” OR “effect of noise” OR “effects of noise” OR “noise effect” OR “noise effects” OR “environmental stressors” OR “environmental stressor” OR “environmental stress”.

AND

Performance OR task OR tasks OR "task performance" OR “information processing” OR cognition OR cognitive OR memory OR recall OR recognition OR "working memory" OR "decision-making" OR "decision making" OR "problem-solving" OR “problem solving” OR attention OR attentional OR vigilance OR "sustained attention" OR skill OR skills OR workplace OR work OR works OR job OR jobs OR “visual perception” OR “perceptual task” OR “perceptual tasks” OR”perception task” OR “perception tasks” OR behaviour OR behavior OR control OR reaction.

NOT

“with hearing loss” OR “on hearing loss” OR “effect of hearing loss” OR “effects of hearing loss” OR animal OR monkey OR monkeys OR rat OR rats OR mice OR rodent OR rodents OR “speech recognition” OR “sound recognition” OR “speech-recognition” OR “music perception” OR “speech perception” OR “speech-in-noise” OR “signal-to-noise ratio” OR “digit-in-noise” OR “digits-in-noise” OR “noise suppression” OR “computation noise” OR “anthropogenic noise” OR “synaptic noise” OR “thermal noise” OR “visual noise” OR “sound memory” OR “sound memories” OR “mental noise” OR “lung sound” OR “lung sounds” OR “market reaction” OR “noise control method” OR “noise control methods” OR “active noise control”

In addition, following the practice of Szalma and Hancock’s (2011) review, a manual search using Google Scholar was employed using similar keywords. Reference lists from existing reviews (Smith 1989b; Szalma and Hancock 2011) were also scanned to identify studies not included in the three databases. Nine additional studies were found from these two avenues. After removal of duplicates, 4544 papers remained.

3.2 Inclusion and exclusion criteria

To be further considered relevant all 4544 papers were assessed via a screening process to see if they met the following criteria:

  1. (a)

    The study should include a group of healthy human adults as participants. The participants should not be children, elderly (i.e., older than 65 years), patients with illness or injuries, or animals. The participants should be above 18 years old and within working age (i.e., less than 65 years). If the study included participants from excluded groups, but results were reported separately for the target group, then the study was included, and only the results for the target groups were included.

  2. (b)

    The study should include tasks with a substantial cognitive component such as visual perception, visual attention, memory or decision-making. Tasks, including real-world tasks, that could not clearly identify a cognitive component were not included. The systematic review focussed on visual tasks as opposed to auditory tasks or speech tasks in order to avoid involvement of the auditory masking effects of noise. Therefore, tasks with auditory or speech components were excluded.

  3. (c)

    Only cognitive performance measured simultaneously with the noise exposure was of interest. That is, any reported after-effect of noise was excluded.

  4. (d)

    The experiment design could be either between-group design or within-subject design, with at least one condition exposed to a given level of noise stimulus (noise condition), and another condition exposed to either a lower level of noise stimulus or ambient noise (control condition). The noise condition or the control condition should not be confounded with another variable such as sleep loss, shift work, intake of drugs, caffeine, alcohol, vibration or heat, or any prior exposure to the stressor (noise).

  5. (e)

    The noise types tested in the study should include continuous broadband noise. Broadband noise is synonymous with white noise and pink noise. The noise level should be at a moderate level range, which was defined as between 55 and 85 dBA.

  6. (f)

    The duration of noise exposure should be no longer than 7 h, which is traditionally the limit for examination of acute effects of noise (Smith and Miles 1985; Smith 1989b).

  7. (g)

    The screening first involved checking the title and the abstract of each study. When it was not possible to include/exclude a paper based on the title or abstract, the second part of the screening involved reviewing the full text of the paper.

3.3 Quality assessment

Available tools for assessing the quality of studies for inclusion in a systematic review (e.g., PEDro scale, Verhagen et al. 1998; Jadad scale, Jadad et al. 1996; Cochrane Collaboration tool, Higgins et al. 2011) were considered for their suitability in the context of the present topic (i.e., noise and cognition). No tool was judged to be suitable. A majority of the quality assessment tools were designed for a clinical setting. Important differences exist between clinical settings and the noise and performance field. For example, double-blind design is important in the clinical setting but not in the noise and performance field, largely because the noise treatment is difficult to hide. Duration, in terms of noise exposure is not relevant to the clinical setting, and therefore does not appear in the quality assessment tools, however is vital to the noise and performance field. Likewise is the aftereffect of noise, a carry-over effect of noise that can last for several hours to days after the noise is switched off (Broadbent 1978; Poulton 1979).

Therefore, in the absence of a suitable (and existing) quality assessment tool, the assessment was based on two factors: Design and Reported. ‘Design’ relates to the experimental design, such as between versus within-subjects design (ED) as well as balancing sex between groups (SB). ‘Reported’ includes the duration of the noise exposure (DU), stating the age range of participants (A) as well as their sex (S). Below outlines these five key criteria in detail.

  1. 1.

    Experimental Design (ED): appropriate experiment design should be either between-group design or within-subject design with the test sessions separated by more than one day to control for the aftereffect of noise (Broadbent 1978; Poulton 1979);

  2. 2.

    Sex Balanced (SB): sex should be balanced (± 5%) between groups;

  3. 3.

    Duration (DU): the duration of noise exposure should be reported;

  4. 4.

    Age (A): the distribution of the participants’ age range should be reported;

  5. 5.

    Sex (S): the sex of the participants should be reported.

4 Results

4.1 Description of studies included

As can be seen in Fig. 1, a total of 43 relevant publications were identified, which included 60 relevant experiments. Performance on a total of 84 cognitive tasks presented simultaneously with moderate continuous broadband noise were reported.

Fig. 1
figure 1

Flowchart of research article selection

Full size image

Cognitive tasks under examination were categorised into one of six categories: (1) reaction time tasks, (2) attentional tasks, (3) short-term memory tasks, (4) long-term memory tasks, (5) high(er)-order cognitive tasks, and (6) other. If a task type was cognitive in nature but had an index or illustration showing that the task had multiple components, it was categorised into ‘other’ task type. Working memory tasks that only involve short-term recall of stimuli (e.g., N-back) were categorised into short-term memory tasks. Working memory tasks that involve executive functions such as reasoning (e.g., grammatical reasoning), mathematical problem-solving (e.g., additions), and decision-making based on understanding of the context in semantic tasks (e.g., semantic processing), are categorised into high(er)-order cognitive tasks. Those studies with recall tasks which did not mention short-term or long-term memory were categorised based on the descriptions of the experiment design. If the recall phase started within half a minute after the presentation of stimuli, the task was categorised as a short-term memory task (Atkinson and Shiffrin 1971). If the recall phase started after half a minute, the task was categorised as a long-term memory task. Recognition memory tasks and semantic memory tasks without mentioning short-term memory were categorised into long-term memory.

Information related to the basis for categorising the tasks for long-term memory tasks has been shown in the column ‘Cognitive task’ in Tables 6 and 7. If the category of the task has been specified as long-term memory by the authors, the task was labelled ‘SbA’ (Specified by Author); if the category was not specified by the authors, the task was categorized based on the description of the task in the text, labelled ‘CbD’ (Categorised by Description). For each long-term memory task in Tables 6 and 7, the interval between the presentation of the stimuli and recall or recognition (labelled as ‘EtR’; Encoding-to-recall Interval) has been recorded or calculated based on the description in the text. There were eight long-term memory tasks that only involved recognition of the semantic information without an encoding phase, such as detecting typographical errors in the four versions of proofreading tasks [56] [18], and recognition whether a word item belongs to a previously shown category [38] [39] [40] [49], these tasks are deemed not applicable for reporting their EtRs, thus labelled ‘EtR: NA’.

Among the 84 tasks, four were reaction time tasks; 13 were attentional tasks; 22 were short-term memory tasks; 21 were long-term memory tasks; 24 were high(er)-order cognitive tasks. These high(er)-order cognitive tasks included four creativity tasks, four versions of proofreading tasks, three semantic processing tasks, three addition tasks, two grammatical reasoning tasks, two experience-based decision-making tasks, two Norinder mental arithmetic tasks, one reading comprehension task, one syntactic reasoning task, one digit symbol substitution task, and one dynamic decision-making task. Four tasks were classified as other task types because they could not be simply classified into any of the above task categories.

As to the participant sex information, over half of the experiments did not balance females and males in their participant group. Of the 60 experiments, 24 had both female and male participants with numbers of each sex reported, among which five had females and males balanced, and another eight with females and males approximately but not strictly balanced; five had participants of both sexes but numbers were not given; 23 had female participants only; one had male participants only; four did not report participant sex information.

In the selected studies, 42 of 60 experiments used a within-subject design. Twelve of them presented noise and control conditions one week apart; for one experiment it was one day apart; four experiments did not mention the gap duration.

Among the 60 experiments, five experiments used pink noise which has lower pitch than white noise. The remaining 55 experiments used white noise or broadband noise which had equal intensity per octave (e.g., Smith 1991).

As to noise level, as the included studies didn’t use a consistent metric for the noise level (e.g., dBA, dBC or dB), the following sections refer to the metric that these studies originally reported. There has been a convention in the field that many early researchers only stated dB when they actually meant dBC (e.g., Poulton 1978; Broadbent 1978). So far there is a lack of available index for direct comparing dBA and dBC for a given noise spectrum. Based on this convention, the present review assumed dBC for dB in the included studies. Typically, the value based on dBA is generally smaller than that of dBC for a given type of noise (Poulton 1978; Broadbent 1978). For broadband noise, it is known that this difference between the two metrics were relatively small (Poulton 1978). In the reviewed literature, it was mentioned that broadband noise at 85 dBC and 60 dBC equals 78 dBA and 50 dBA, respectively (Smith 1991).

Noise levels of the broadband noise condition among the 60 experiments ranged from 55 dBA to 85 dB SPL. For 17 experiments the noise level of the control conditions was not reported. Among the remaining 43 experiments, the noise levels for the control conditions ranged from 25 dBA to 65 dBC. Only 36 experiments reported the duration of noise exposure, which ranged from a few minutes to an hour and could be seen as investigating acute effects of noise.

The present review considered the influence of moderators that may be broadly relevant. Moderators that were specific to a particular task or experiment paradigm were not considered. To clarify, moderators included in the present review were task factors such as repeated session (e.g., trial, weeks, session, block), cognitive load (e.g., lists, workload, task complexity, task difficulty, serial position, memory load, word length), time of day, instructions about the noise effect, and individual difference factors such as sex, native language background, extraversion, trait anxiety, subjective rating of activation or cognitive failure.

Of the 84 tasks, 46 tasks included consideration of moderators. The moderators, the frequency they appeared in the included studies and their meanings are presented in Table 1, grouped according to their interrelation. For example, moderator termed ‘trial’, ‘weeks’ ‘session’ and ‘block’ were just different names for the task factor of repeated task session, thus grouped into ‘repeated session’. Repeated session, cognitive load, time of day, sex, and native language background appeared to be the most commonly considered moderators.

Table 1 The moderators in the included studies, their frequency and explanation
Full size table

The results of the systematic review are presented in Tables 2, 3, 4, 5, 6, 7, 8 and 9, categorised according to the type of cognitive task under examination because noise effects may differ for different types of cognitive task. To make it convenient to refer to the studies in the tables in the text, each of the 60 experiments has been assigned a unique number and referred to in the following sections as [1] through to [60]. Within each table the studies are presented in reverse chronological order.

Table 2 The effect of moderate (55–85 dBA) broadband noise on reaction time tasks
Full size table
Table 3 The effect of moderate (55–85 dBA) broadband noise on attention tasks
Full size table
Table 4 The effect of moderate (55–85 dBA) broadband noise on short-term memory with noise presented during the whole task
Full size table
Table 5 The effect of moderate (55–85 dBA) broadband noise on short-term memory with noise presented during encoding, or delayed period, or encoding + delayed period
Full size table
Table 6 The effect of moderate (55–85 dBA) broadband noise on long-term memory with noise presented during the whole task
Full size table
Table 7 The effect of moderate (55–85 dBA) broadband noise on long-term memory with noise presented during encoding only
Full size table

4.2 Noise effect on reaction time

Table 2 outlines three studies (with four experiments) that examined the noise effect on reaction time tasks. Overall, moderate broadband noise had little negative effect on reaction time. Among the four experiments, one reported a beneficial effect and three reported no effect. Reaction time tasks were not affected by broadband noise at 55 dBA [3], 69 dBA [56] and 78 dBA [29], while 80 dBA broadband noise benefited (reduced) reaction time [18].

Among the four experiments, two experiments tested moderation effects. The effect of trial (the first trial vs. the second trial) was examined. Both experiments showed trial had no significant moderating effects on reaction time.

4.3 Noise effect on attention tasks

Table 3 outlines 11 experiments (13 tasks) that examined the noise effect on attention. The 13 included attention tasks were Posner task [11], Star Counting test [54], focused attention task [17], categoric search [17], simple attention task [21], dual monitoring task [25], passive forms of attention [31] [32], Stroop task [33] [60], attentional selectivity task [51] [52], sustained attention task [60].

Among the 13 attention tasks, two revealed harmful effects, one revealed beneficial effect, while the other ten revealed no effects. Moderate broadband noise only impaired focused attention and dual monitoring [17] [25], but not categoric search task [17]. Moderate broadband noise had a facilitating effect on a simple visual attention in a light detection task [21]. Detection performance improved when continuous white noise increased from 50 to 70 dB, and then declined when noise increased through 80 and 90 dB [21]. It should be noted that performance at 80 dB or 90 dB did not fall below that of 50 dB. No other attention task demonstrated a significant effect of moderate broadband noise.

Among the 11 experiments (with 13 attention tasks), four experiments tested moderation effects. As to task factors, two experiments tested ‘session’; one ‘task difficulty’; one ‘time of day’. As to individual difference factors, one experiment tested ‘subjective cognitive failure’; one ‘trait anxiety’. None showed a significant influence on the effect of noise on attention.

4.4 Effect of noise on memory

The timing of noise presentation in relation to phase of memory process varied across studies. Noise was presented during the whole task (from encoding to recall or recognition), during the encoding phase (stimuli presentation), during the delay period after stimulus offset (maintenance phase between encoding and recall or recognition), or during encoding plus the delay period. For both short-term and long-term memory, results are reported separately for tasks with noise presented throughout and for tasks with noise presented during only a part of the task.

4.4.1 Short-term memory

As shown in Tables 4 and 5, 18 selected studies with 21 experiments were included for short-term memory tasks. Table 4 outlines 12 studies (with 14 experiments and 15 tasks) that examined the noise effect on short-term memory with noise presented during the whole task. Table 5 outlines four experiments that examined the noise effect on short-term memory with noise presented during the encoding only, and two experiments that had noise presented during encoding + delayed period; and one experiment that had noise presented during the delayed period only.

Across the 21 experiments, there was little support for a negative effect of moderate broadband noise on overall short-term memory performance, except for when noise was presented during the delay period only [7]. What deserves notice is some factors that led to performance being impaired or improved (e.g., variability in the effect) under specific circumstances. When noise was presented during the whole task, white noise at the same level (65 dB) improved the 2-back task performance [60], but not pink noise [1]. Another factor which was not examined as a moderator, the specific type of information to be recalled, appeared to account for the variability in the effects of moderate broadband noise on short-term recall ([41] [42] [43] [30] [35]). In addition, moderate broadband noise appeared to increase individuals’ selectivity of stimuli to be processed in short-term memory, which led to impairment on some of the items to be recalled [19] [34].

Among the 21 experiments, eight experiments tested moderation effects. As to task factors, one experiment tested ‘session’; one ‘trial’; one ‘serial position’; one ‘word length x serial position’; one ‘memory load x serial position’. As to individual difference factors, two experiments tested ‘language background’; one ‘sex’. Overall, the effect of moderate broadband noise on short-term memory tasks was significantly moderated by repeated session (‘session’ [35], ‘trial’ [5]) and cognitive load (‘serial position’ [26], ‘word length x serial position’ [36], memory load x serial position’ [34]), while individual difference factors such as language background [4] [6] and sex [14] had little moderation effect. In other words, impairment of short-term memory in moderate broadband noise was most likely to occur when the cognitive load on memory capacity was large [34], such as when more items were to be recalled [36], on items which were rehearsed earlier [26], and over repeated sessions [35]. However, the moderating effect of repeated session was opposite when noise was presented during encoding, reflecting a learning effect over time [5].

4.4.2 Long-term memory

Thirteen studies with 19 experiments were included for long-term memory tasks. As shown in Table 6, eight studies with 13 experiments (15 tasks) had noise presented during the whole task. As shown in Table 7, five studies with six experiments had noise presented during encoding only.

Overall, harmful effects of moderate broadband noise were only found when the noise was presented during the whole task, while little effect was found when noise was presented during encoding only [10] [12] [13] [5] [44] [37]. Long-term memory seemed vulnerable to high(er) noise level, such as 65 dBA/80 dBA compared to 55 dBA [4] [6], or 85 dBC [48] compared to 80 dBC [38] [46] [47]. White noise at 70 dB sometimes improved the formation of long-term memory during encoding [10] [12]. In addition, some factors that were not examined as a moderator seemed to have a further influence, and caused variability in the noise effect, such as previous task experience [39], and task settings related to being machine-paced or self-paced [18].

Among the 19 experiments, nine experiments tested moderation effects. As to task factors, two experiments tested ‘trial’; two ‘weeks’; two ‘lists’. As to individual difference factors, two studies tested ‘language background’; one ‘sex x time of day’; one ‘subjective rating of activation’. Task factors such as repeated session (‘trial’ [56] [18]) and cognitive load (‘lists’ [46] [47]) had no significant moderating effect. ‘Weeks’ had inconsistent moderating effects, which appeared to be influenced by noise levels (80 dBC [46] vs. 85 dBC [48]). Broadband noise at 80 dBC was only harmful on clustering performance for the first week session, while 85 dBC broadband noise was harmful for both weeks’ sessions. Individual differences factors had significant moderating effects. Language background had no moderating effect in 55 dBA and 65 dBA broadband noise [4], but was a significant moderator for 80 dBA [6]. ‘Sex x time of day’ was significant moderators for clustering in free recall [23].

4.5 Effect of noise on high(er)-order cognition

Table 8 outlines the studies that examined the noise effect on high(er)-order cognition. The 24 included high(er)-order cognitive tasks came from a total of 12 studies with 16 experiments. Tasks were a reading comprehension task [2], three grammatical / syntactic reasoning tasks [4] [6] [23], two mathematical problem solving tasks [4] [6], one mathematical addition task [20], two Norinder mental arithmetic tasks [22] [28], three semantic processing tasks [23] [24] [27], one digit symbol substitution task [23], two sets of creativity tasks [16] [60], four proofreading tasks [18] [56], two experience-based decision-making tasks [58] [59] and one dynamic decision-making task [57].

Table 8 The effect of moderate (55–85 dBA) broadband noise on high(er)-order cognitive tasks
Full size table

High(er)-order cognitive tasks such as mathematical problem solving [28] [4] [6] [20] and processing of semantic information [2] [18] [4] [6] [27] appeared to be only occasionally impaired by moderate broadband noise. The effects of moderate broadband noise on decision-making tasks yielded mixed results. A simple type (Gonzalez and Dutt 2011; Gonzalez et al. 2017) of experience-based decision-making task was impaired by white noise at 65 dBA [58] [59]. In contrast, broadband noise at 75 dBA had no main effect on a complex dynamic decision-making task [57]. The known difference between the two decision-making tasks in complexity and dynamic complexity (Gonzalez, Vanyukov and Martin, 2005; Gonzalez and Dutt 2011) could explain the variability of the noise effects.

Among the 16 experiments with 24 high(er)-order cognitive tasks, twelve experiments (12 tasks) tested moderation effects. As to task factors, three experiments tested ‘time of day’; two ‘trial’; two ‘block’; one ‘session’; one ‘workload’; one ‘workload x session’. As to individual difference factors, two experiments tested ‘language background’; two ‘sex’; two ‘sex x time of day’; one ‘sex x session’; one ‘sex x workload’; one ‘sex x instructions about the noise effect’; one ‘extraversion’. Task factors such as ‘time of day’ [22] [23] and ‘workload’ or ‘workload x session’ [57] had no significant moderating effects. Repeated session (‘trial’ [56] [18], ‘block’ [58] [59], ‘session’ [57]) revealed mixed findings, which seemed to be determined by task parameters and the duration span of the task measured across these experiments (2-min vs. 15-min vs. 3-day sessions). Language background [4] [6] and extraversion [2] had no moderating effects. Performance on tasks requiring high(er)-order cognition (i.e., mental arithmetic) were moderated by instructions, however this effect varied based on sex (i.e., sex x instruction interaction; [28]). The moderating effects of sex interacting with another two task factors such as workload and session, respectively, were significant for the effect of moderate broadband noise on dynamic decision-making [57]. Sex interacting with time of day received mixed findings [23] [24].

4.6 Effect of noise on other cognitive tasks

Table 9 outlines the one study that examined the noise effect on tasks not otherwise noted. Gawron (1982) used a battery of four tasks that involved multiple information processing components, which could not be categorised into any one of the above indicators of cognition. Among the results, no harmful effect of moderate broadband noise was found in any of the four tasks. Moderate broadband noise in some circumstances benefited performance. Specifically, when individuals were told that the effect of noise was ‘facilitating’ or had ‘no effect’, moderate broadband noise was beneficial to (shortened) reaction time. High task complexity eliminated the beneficial effect of moderate broadband noise on accuracy in low task complexity However, this relation between task complexity and the noise effect was altered by individuals’ anticipation of the noise effects, as when individuals received no instructions about the noise effect, high task complexity became beneficial to accuracy.

Table 9 The effect of moderate (55–85 dBA) broadband noise on other tasks
Full size table

5 Discussion

The aim of this systematic review was to outline evidence concerning the effect of moderate broadband noise on indicators of cognition, such as reaction time, attentional tasks, short-term memory, long-term memory, and high(er)-order cognitive tasks such as problem solving, as well as the factors that influence these effects. Reviewed results offered minimal evidence for a negative effect of moderate broadband noise on cognitive performance. Out of the 84 tasks testing for an effect of moderate broadband noise on task performance, 17 tasks (20.2% of all the tasks) showed negative effects and nine tasks (10.7% of all the tasks) showed beneficial effects. Fifty-eight (69.0%) of the tasks showed no beneficial or harmful main effects. This is likely due to the reason that individuals’ ability to overcome the effect of moderate broadband noise is strong (Hockey 1997). Negative effects have been found for relatively few cognitive processes, and sometimes only under specific circumstances. In terms of moderators, no consistent moderators were identified, as few of the moderators (e.g., sex, trial, language background, etc.) were tested in multiple studies, therefore, making any firm conclusions difficult. However, those few that did, and showed promise or consistent evidence as non-significant, are summarised below.

Promising moderators include repeated session for short-term memory tasks and high[er]-order cognitive tasks, and ‘serial position’ for short-term memory tasks. Variables such as time of day, cognitive load—‘lists’ [46] [47], ‘subjective cognitive failure’ and ‘anxiety’ [17] showed consistent non-significant effects.

The effects of other moderators were not sufficient to draw firm conclusions on their effects. Specifically, ‘language background’ [4] [6] received mixed findings on the same long-term memory task when being tested in different noise levels (65 dBA vs. 80 dBA), and thus needs more replications. ‘Task difficulty’, ‘task complexity × instructions about the noise effect’, ‘subjective rating of activation’, ‘extraversion’, ‘instructions about the noise effect’ each were tested only once across all the studies and the findings were too patchy to draw a firm conclusion on their effects. In addition, type of information to be recalled in the short-term memory task [30] [35] appeared to explain the variability in the effects of moderate broadband noise across different experiments, but they were not formally tested as a moderator in any of the included studies, thus requiring more studies.

In summary, this review revealed three key findings:

  1. 1.

    The moderating effect of cognitive load in the effect of moderate broadband noise was not the same across task types;

  2. 2.

    The moderating effect of repeated session is related to the test session duration; and

  3. 3.

    Theoretical constructs are needed to account for individual difference variables such as sex, extraversion, or language background.

5.1 Mixed effects of cognitive load across task types

It is generally accepted that the impact of noise on performance is related to the cognitive load of the task engaged in. Theories, such as resource-based theories try to capture this impact (Hancock and Warm 1989; Szalma and Hancock 2011). However, such theories lack detail on the precise impact, nor articulate which tasks impose lower or higher cognitive loads. The results of this systematic review revealed that with moderate broadband noise, this noise effect is evident with tasks involving reaction time, attention, and short-term memory, but not long-term memory or high(er)-order cognitive tasks. For example, reaction time tasks are perceptual in nature, which were deemed cognitively ‘simple’, and thus affected little by noise (Hancock and Warm 1989; Szalma and Hancock 2011). Among attention tasks, specific effortful attention tasks such as a focused attention task [17] and a dual monitoring task [25] that required sustained monitoring on stimuli were particularly susceptible to moderate level noise. In short-term memory tasks, noise-induced performance impairment was found on the more cognitively taxing 3-back task, as opposed to the less cognitively taxing 2-back, 1-back, 0-back [1]; some items were missing in recall when the total number of items to be recalled (memory load) increased [19] [34] [26] [36]. A similar effect was evident with the ‘other task’ category—delayed digit cancelling task which involved short-term memory as one of its cognitive components.

In contrast, the effect of cognitive load was not obvious in long-term memory tasks. What primarily determined the negative impact of moderate broadband noise among long-term memory tasks appeared to be the noise level and variations in the version of the task. This could be due to the fact that long-term memory tasks generally have longer period for storing information in the memory, thus not suffering from limited cognitive capacity for information processing at a moment as it was for short-term memory tasks.

In high(er)-order cognitive tasks, cognitive load appeared to have little effect or even opposite effects against theoretical predictions that high cognitive load should exaggerate the harmful noise effect (Broadbent 1971; Hancock and Warm 1989; Szalma and Hancock 2011). For example, high cognitive workload facilitated individuals to combat the noise effect in dynamic decision-making as compared to low cognitive workload [57]. This was evident with both the dynamic decision-making task [57] and experience-based decision-making task [59]. The former task was more complex in its task structure than the latter. However, it was the simplified task [59] impaired by even a lower noise level (65 dBA), while the complex task [57] overall unaffected by even a higher noise level (75 dBA).

Together these findings seem to lead to two conclusions. First, the noise effect actually depends on the cognitive load placed on a specific cognitive component/process (e.g., sustained attention, short-term memory) which the task primarily measures, rather than the summated cognitive load imposed across all the cognitive components/processes involved in the task. Though high(er)-order cognitive tasks (or a complex cognitive task) involved utilisation of multiple low(er)-order cognitive processes, individuals could utilize strategies to reduce the load placed on any single cognitive component/process that is exposed under the stressor, thus circumstantially avoiding performance impairment in noise due to high cognitive load. Second, cognitive load is only one of many factors that acting or interacting together to influence the performance of high(er)-order cognitive tasks in noise. These factors could include emotional response, effort investment, strategy used and other task-specific factors such as task instructions, which is an area to be focused on for high(er)-order cognitive tasks in future research.

5.2 Repeated session related to the test session duration

The moderating effect of repeated task session on the noise effect is one of theoretically proposed moderators. This moderating effect for moderate broadband noise was mixed, with two beneficial [5] [46], two harmful [35] [58] and the rest non-significant [18] [48] [56] [57] [59]. Among those significant, the exact effect of repeated session swayed between a harmful effect of cumulative stress over time (Smith 1989b; Szalma and Hancock 2011) and a beneficial effect of practicing (Broadbent 1971) or habituation (Szalma and Hancock 2011) over sessions. Due to differences across these studies in the experiment design, such as the timing of noise presentation, the noise level applied and the duration spans of the task measured, no further firm conclusions could be derived, and thus a need for more consistent replications.

What deserves notice is that the duration span that these significant repeated session effects occurred in was longer than that of intense noise stressors. Except for one study where the duration of the session was not given [5], the experiment design of these studies [35] [46] [58] showed either beneficial or harmful effect of noise occurred at a session longer than 1.5 min, ranging from 5 min [35] [46] to 15 min [58]. This temporal scale was different from the picture outlined based on broader noise types. Szalma and Hancock’s (2011) meta-analytic review showed that a mixture of noise types including loud noise, intermittent noise and speech noise had the most harmful effect only at the beginning of the task performance, which was subsequently mitigated through habituation after 1.5 min of noise exposure. This makes sense as Broadbent (1971) suggested that the beginning of loud noise bursts was stimulant and arousing, thus harmful to performance. Moderate broadband noise effect is mild, so that individuals tend not to perceive it as a stressor immediately, thus gradually impaired by the cumulative stress of noise over time without perceiving this negative effect as a threat or developing a strategy to overcome it. What remained absent in the existing theories and a need for future empirical research is to detail the competing roles between the harmful impact of cumulative stress, and beneficial effect of habituation over sessions. Furthermore, future research should investigate the temporal curve of their combined effect, along the time dimension considering different task types.

5.3 More research needed for individual differences

Moderators, and especially individual variables is a necessary issue to be considered, and could be another factor underlying the absence of consistent negative effect of moderate intensity noise (Smith 1989b). Overall, individual differences were more likely to be found among long-term memory and high(er)-order cognitive tasks than other task types. Little moderating effect of individual difference variables was found on the noise effects on reaction time tasks, attention tasks, or short-term memory tasks. Extraversion, sex and language background were the three individual difference variables that received the largest interest in existing literature. The mechanisms underlying their moderating effects were largely absent in existing theories. This is, however, not ideal and needs future (theoretical) research.

As to extraversion, only one study [2] tested the moderating effect of extraversion, which showed no significant effect. This finding was in contrast to theoretical suggestions that extraversion might explain the individual differences in the effect of noise (e.g., Broadbent 1971; Szalma 2012). It has been generally thought that extraverts are less vulnerable to noise than introverts. However, it is noticed that all these existing theoretical suggestions aimed at investigating the effect of broader types of noise instead of moderate intensity noise in particular. As a tentative postulation, personality traits such as extraversion are more significant as a moderator for louder noise (> 90 dB; Smith 1989b) or intermittent (Szalma 2012) noise than for moderate broadband noise. Moderate broadband noise is mild as a stressor and may not be sufficiently arousing or stimulating to increase extraverts’ arousal level and benefit extraverts’ performance to a degree that makes the noise-extraversion interaction significant. Besides this noise factor, the authors of the study [2] postulated that a high-complexity cognitive task such as reading comprehension could be the reason that extraversion was not a significant moderator, while existing theories about personality were largely based on simple cognitive tasks data. It is noticed that another personality trait ‘anxiety’ also had non-significant moderating effect [17]. Taken together, more detailed research is needed to clarify the moderating role of personality in moderate broadband noise.

The moderating effects of sex for moderate broadband noise showed mixed findings. Sex alone was not significant in any of the three studies [14] [28] [57], while sex was significant when interacting with another variable, such as instructions about the noise effect [28], workload or session [57], and this was only found in high(er)-order cognitive tasks only. Based on these limited findings available, it seemed that females and males in some circumstances behave differently in the reactions to the stressor, such as in the pace of solving problems [28] or an initial inefficiency [57]. Due to the rather small sample size these studies used for examining these individual difference variables and the lack of sufficient replications, it is hard to draw firm conclusions on the real moderating effect of sex. However, there is still a lack of theories to account for or explain the effects of sex. Whether the difference between sexes relates to any physiological or psychological mechanisms such as arousal, motivation state (Smith 1989b) or affective reactions remains a theoretical question to be answered.

The other individual difference variable, language background, was not clearly proposed as a moderator based on Arousal Theory (Broadbent 1971) or Resource-based theories (e.g., Hockey 1997; Szalma and Hancock 2011). It was found that language background significantly moderated the effect of 80 dBA (but not 65 dBA) broadband noise on long-term memory (Molesworth et al. 2017, 2018) [6] [4]. It is plausible that the act of searching one’s language memory store (schemata) for a word or the meaning of a word is cognitively taxing. The cumulative effect of this cognitive load, along with the task cognitive load, and noise stressor may result in performance detriments, which is a theory that is yet to be tested.

It should be noticed that the robustness of these results may not be sufficient to unveil reliable effects of these individual difference moderators. Specifically, among the studies that examined personality and sex, the number of participants involved within each study (e.g., [2] [14] [28] [57]) only had as many as 12 participants for each condition group, which might still be too small to assess the real moderating effects of these individual difference variables.

5.4 Limitations and future research

This review has some limitations which provide scope for future reviews. First, it has not investigated whether and how within-subject design contributed to inconsistencies in the effects of moderate broadband noise. Thirty-three of the 43 studies used within-subject design with the noise and control conditions tested in the same day, a type of design known as having the risk of inducing the aftereffect of noise (especially for loud noise) in the results (Broadbent 1976; Poulton 1979; Smith 1989b). By far there is a lack of literature documenting whether the risk of inducing the aftereffects of noise similarly exists in the experiment using moderate broadband noise. Due to the rather large proportion of the included studies that used a within-subject design with noise and control conditions in the same day, it was deemed not practicable to address this issue in the present systematic review. It is suggested that future studies test the influence of different types of experiment design on the results quantitatively.

Second, the present results only included studies published in English language. It is suggested that future study may include studies from other languages.

Third, due to the scope of the present review, the effect of moderate broadband noise was reviewed in isolation from other noise types. A comparative approach (e.g., moderate broadband noise vs. loud noise > 90 dBC or moderate intensity intermittent noise) may have the benefit of outlining noise effects in a more comprehensive way; hence a possible scope for a future review.

6 Conclusions

The present systematic review outlined evidence concerning the effect of moderate broadband noise on indicators of cognition, such as reaction time, attentional tasks, short-term memory, long-term memory, and high(er)-order cognitive tasks, as well as the moderators that influence these effects. In summary, the review showed that there is limited evidence for a negative impact of moderate broadband noise on cognitive performance. Existing empirical literature revealed inconsistent selection and testing of moderators as well as a lack of sufficient replications. Based on the available findings, the moderating effects of cognitive load, repeated session, and individual difference factors were found to differ from theoretical predictions based on evidence from a broader range of noise, which highlights the distinctive characteristics of the mild stressor, moderate broadband noise. Given a lack of sufficient theoretical knowledge account for it, it is suggested future theoretical and empirical research should pay more attention to the distinctive characteristics of mild stressor(s) such as moderate broadband noise as well as its effect among complex task scenarios such as high(er)-order cognitive tasks.