Irrelevant sound effect describes the empirical finding that short-term memory is disturbed by background sounds that are irrelevant to the memory task at hand, even though participants are told to ignore them (Beaman & Jones, 1997). To date, accumulating research evidence suggests that the irrelevant sound effect is a highly robust and reliable finding, as demonstrated by which it has been replicated in numerous studies using a wide range of materials, designs and procedures (Colle, 1980; Colle & Welsh, 1976; Ellermeier & Zimmer, 1997; Hanley, 1997; Macken, Mosdell, & Jones, 1999; Jones, & Macken, 1993; Salame & Baddeley, 1982, 1986).
This effect is of considerable theoretical interest as it provides a means to reveal the interplay between perceptual and memorial functions (Banbury, Macken, Tremblay, & Jones, 2001). Over the years, different theories have been proposed in an attempt to account for this phenomenon and research data has largely been channeled towards theory development and assessment. For instance, the phonological loop hypothesis suggests that speech information is the source of disruption to verbal short-term memory (Salame & Baddeley, 1982). Jones, Maden and Miles (1992), however, proposed the changing state hypothesis which posited that the degree of acoustic change is the key determinant of disruption. Although no consensus has been reached over the primary source of disruption, these two theories together suggest that some sounds are particularly more disruptive to verbal short-term memory performance.
Despite its theoretical importance, the practical implications of the irrelevant sound effect are presumed to be substantial (Beaman, 2005) as it addresses a pervasive part of everyday life — environment sound distraction. Literature exploring the practical implication of irrelevant sound effect is, however, limited. In the existing literature, only a scarce amount of studies have examined the potential application of the effect in a broader context, such as music. Today, music is almost a common facet of everyday life. Background music is not only played routinely in the car, but also in many public places such as stores and supermarkets, to name a few. Development of portable audio devices has also led music for accompaniment to become increasingly common. Studying with music is widely popular among students and background music is also being used in the workplace to reduce stress.
Providing that background music bears a strong resemblance to the irrelevant sound that both are unrelated to the task at hand, whethe0r background music would impact our verbal short-term memory performance and produce a ‘musical irrelevant sound effect’ is thus a significant applied question. Moreover, as the phonological loop hypothesis and changing-state hypothesis suggest that some sounds are potentially more disruptive, applying these two theories to the context of background music could help to delineate which type of music would have a higher disruptive potency to our verbal short-term memory.
In addition, short-term memory, the ability to hold information in a sufficiently brief period of time, is thought to be a fundamental component of proficient cognitive functioning. According to the information processing approach (Atkinson & Shiffrin, 1968), short-term memory, together with sensory memory and long term memory form the three stages of information processing. Information in the short-term memory store is retained through rehearsal process before further processing or transferring into a longer-term and more permanent store. The importance of short-term memory is that it helps maintain information in a highly active state and is thus an essential step towards higher level information processing.
Confirming this notion, research evidence suggests that verbal short-term memory, the verbal component of short-term memory, is an integral component of several complex cognitive tasks, such as reading comprehension and complex mental calculation. For instance, verbal short-term memory is shown to be important in assisting the integration of between-sentence ideas (Coltheart, Avons & Trollope, 1990) and comprehension of parsed sentence (Hanson, Goodwall, & Perfetti, 1991). In the case of solving complex mental calculation, verbal short-term memory is regarded as essential for holding intermediate product for later computation (Hitch, 1980; Noel, Desert, Aubrun & Seron, 2001).
Given that music for accompaniment has become increasingly common and short-term memory is considered to be fundamental to proficient cognitive functioning, examining the effects of background music in the irrelevant sound effect paradigm thus provides significant implications. If it proves to be the case that certain kinds of music are particularly disruptive to verbal short-term memory, it could help us to make an informed choice on choosing the right type of background music for accompaniment.
Conceivably, an informed choice of background music might help to reduce plausible errors in verbal short-term memory, and hence any complex cognitive functioning that relies on this short-term storage function could also be expected to improve in performance. Therefore, the present study sought to explore the irrelevant sound effect in the context of music and applied the phonological loop hypothesis and the changing state hypothesis to understand how different kinds of background music might impact our verbal short-term memory performance.
The irrelevant sound effect
Present work on the irrelevant sound effect has predominately concerned the effect of extraneous speech and tone on serial recall performance. In a typical irrelevant sound effect experiment, verbal-based items, such as digits or words, are presented successively while irrelevant sound is played in the background. Participants are asked to ignore any sound to the best of their ability and are required to report back exactly in the order of presentation. In the presence of irrelevant background sounds, serial recall performance is disrupted. Compared to the quiet condition, the disruptive effect of irrelevant speech and tone has been consistently shown to increase in errors by a rate of 30% to as large as 50% (Neath, 2000). To date, the irrelevant sound effect produced by speech and tones are extremely robust and reliable, as demonstrated by the evidence that it has been replicated by a large number of studies (Colle & Welsh, 1976; Ellermeier & Zimmer, 1997; Macken, Mosdell, & Jones, 1999; Jones, & Macken, 1993; Salame & Baddeley, 1982).
For many years, researchers have sought to tease out the primary source of disruption and to deduce an interference mechanism that best explain the irrelevant sound effect. As a result, different theories have been proposed to account for this phenomenon and the research data has largely been channeled to theory development and assessment (eg., Nairne, 1988; Neath, 2000; Jones, 1993; Salame & Baddeley, 1987). Of which, two theories have attracted the most attention and they are the phonological loop hypothesis (Salame & Baddeley, 1976) and the changing-state hypothesis (Jones, Maden & Miles, 1992).
Phonological loop hypothesis
An early influential account of the irrelevant sound effect is provided by Baddeley’s working memory framework (Baddeley and Hitch, 1974), which holds that speech is the key to interference. A fundamental assumption of the working memory framework is that the phonological loop translates visually presented material, such as items in the to-be-remembered list, into speech-based code and enters the phonological store via an active sub-vocal rehearsal process. While the to-be-remembered list must be rehearsed to enter the phonological store, speech-like irrelevant sound could, on the other hand, pass through a speech sensitive filter and gain obligatory access to this store.
According to the phonological loop hypothesis (Salame & Baddeley, 1987), irrelevant sound effect occurs because spoken materials interfere with the to-be-remembered list also in the store. As spoken materials and the traces of to-be-remembered list are similar in the sense that they are both phonologically-based, confusion is resulted from the co-existence of these two phonological materials, and errors in recalling the list are thus likely to make. Given this explanation, an important prediction of the phonological store hypothesis is hence speech reduces serial recall performance, whereas speech-unlike sound does not.
Consistent with this view, human speech has been repeatedly shown to impact serial recall performance when it is played during the list presentation or during the subsequent rehearsal period, (Hanley & Broadbent, 1987; Jones, 1994; Salame & Baddeley, 1982, 1987) whereas speech-unlike sound, such as Gaussian noise and white noise, is found to have no significant disruptive effect (Colle & Welsh, 1976; Salame & Baddeley, 1982).
A few more recent findings have challenged the speech sensitive filter assumption of the phonological loop hypothesis, as they found that non-speech tones could also significantly disrupt serial recall performance. For example, Jones and Macken (1993) found that sounds with an equal amount of variation as speech could produce a similar degree of disruption on serial recall performance. Based on this finding, they concluded that both speech and non-speech sound could be equally disruptive. The authors posited that the important prerequisite of disruption is the degree of acoustic variation, and this is known as the changing-state hypothesis.
The changing-state hypothesis (Jones, 1993), on the other hand, argue that the interference occurs because of the similarity of the organization process rather than the content between the irrelevant sound to the items being rehearsed. According to Jones (1993), sound containing acoustic variation would automatically evoke a seriation process for perceptual organisation which is similar to the process for retaining the to-be-remembered list. The errors in recall thus arise through the interference between two concurrent processes of seriation: one is for maintaining the to-be-remembered list; another is for organizing the irrelevant sound stream with acoustic variation. In this hypothesis, acoustic change might refer to any appreciable changes in tonality, pitch or timbre. That is, for disruption to occur, each physical unit of the sound stream must be significantly different from its next unit. The key prediction that follows this assumption is thus sound with more acoustical variation would induce a higher level of seriation and in turn produce a greater disruption.
Jones, and his colleagues carefully manipulated the acoustic features of the irrelevant sound and conducted a series of experiments to examine the changing-state hypothesis (Jones, Alford, Macken, Banbury & Trembaly, 2000; Jones, Madden & Miles, 1992). Jones, Madden and Miles (1992) found that repeated irrelevant tone of ‘CCCC’ has no significant disruption to verbal short-term memory, whereas the disruptive effect of sound sequence with tonal variation, ‘CHCH’, is pronounced.
In a later study, Jones, Alford, Macken, Banbury and Trembaly (2000) confirmed the changing-state hypothesis in another way that when acoustic features of a sound sequence were removed, an improvement in serial recall performance was observed. These studies harmonize with the changing-state hypothesis that the greater the acoustic change, the greater the disruptive potency the sound is.
In sum, the irrelevant sound effects of speech and tonal sequence have been well-documented in the literature. Given that music resembles to speech and tonal sequence in the sense that they are both highly patterned sound, it would be reasonably to predict that music might have a similar effect. Studying irrelevant sound in the context of music is moreover of considerable application value as it helps extending the current scope of research beyond speech and tone.
Music as irrelevant sound
Over the years, voluminous study has been conducted to explore the impact of music on a wide range of cognitive performance, for example, reading comprehension (Forgerson, 1973), mental mathematics (Tucker, & Bushmam. 1991), and writing (Ransdell & Gillroy, 2001). As yet, however, only a few have examined the effects of background music in an irrelevant sound paradigm. Among the sparse literature, several researchers have demonstrated an irrelevant sound effect of background music. Pringe and Walker (1994), for instance, found that serial recall performance was lower for participants listening to nursery rhymes than for those in the silence condition. Similarly, a more recent study by Nittono (1997) using an identical paradigm of immediate serial recall task also reported that classical music reduced verbal short-term memory performance. Iwanaga and Ito (2002), moreover, shed light on the subjective experience of disturbance. In the study, subjects reported that the perceived disturbance is higher under the background music condition than in silence. A study by Alley and Greene (2008), furthermore, reported a counter-intuitive finding that the magnitude of background music disruption is independent of its familiarity. It was found that subjects who were unfamiliar with the lyrics were affected by the background music in a similar way as those who were familiar.
Not until recently, researchers have attempted to extend the applications of the phonological loop hypothesis to the context of music. These preliminary applications of the phonological loop hypothesis have demonstrated that vocal music reliably and consistently interfered with verbal short-term memory performance (Boyle & Coltheart 2006; Salame & Baddeley, 1989).
Vocal music and phonological loop hypothesis
As the phonological loop hypothesis holds that speech is the source of interference to verbal short-term memory, the more the auditory stimulus resembles speech, the greater degree of its disruptive potency. Demonstration of a musical irrelevant sound effect to testify the phonological loop hypothesis thus requires a vocal (music that is more speech-like) and non-vocal background music (music that is less speech-like) condition. Previous studies using this experimental design have provided consistent evidence to support the phonological hypothesis that vocal music is more disruptive to memory performance than non-vocal music.
Of which, Salame and Baddeley (1989) were among the first to testify the phonological loop hypothesis by using background music. The authors examined the effect of background music on immediate serial recall performance while non-vocal rock music and vocal operas pieces were played. Confirming the phonological loop hypothesis, music containing speech (vocal opera pieces) significantly reduced performance relative to both non-vocal instrumental music and silence.
A later research by Boyle and Coltheart (2006) studying the effect of vocal music in an irrelevant sound paradigm has replicated Salame and Baddeley (1989)’s findings, further, they reported that vocal music could induce a similar degree of impairment on serial recall performance as that of speech. In sum, the previous researches have demonstrated a consistent pattern that music containing vocals could significantly reduce performance. This therefore dovetails with phonological loop hypothesis’s prediction that speech-like sound would induce an irrelevant sound effect.
Although these preliminary applications of the phonological loop hypothesis suggest a consistent pattern, these findings are remained inconclusive. In most previous studies, the confounding factors between vocal and non-vocal music have not been systematically controlled. Salame and Baddeley (1989), for instance, compared opera pieces in the vocal condition with instrumental rock music in the non-vocal condition to examine the effects of vocals in short-term memory performance. When vocal and non-vocal music were compared, they were however different in genre, rhythm, tempo and many other features. The difference in the degree of interference between vocal and non-vocal music is, therefore, constructed to vary more significantly. Suggested by a vast amount of research evidence that differences in music genre, tempo and modality could contribute to the discrepancies in physiological response (Dilman & Potter, 2007), karate task performance (Ferguson, Carbonneau & Chambliss, 1994) and reading efficiency (Kallinen, 2002 ), raising the concern over these uncontrolled variables might confound the results. Conceivably, previous literature has limited attempt to control these confounding factors and to use vocal and non-vocal music stimulus in an otherwise equivalent basis. Without excluding the aforementioned confounding factors, it is thus difficult to accurately determine whether vocality is the key factor to disruption.
Within the realm of musical irrelevant sound effect, although attempts were made to explore the phonological loop hypothesis, scant attention is being put to examine the changing-state hypothesis in the context of music. Presumably, the changing-state hypothesis suggests that sound with variations in pitch, tonality and timbre is the key determinant of disruption, while music is characterized as a pattern of sound with systematic variations, the application value of the changing-state hypothesis is therefore assumed to be considerable. The changing-state hypothesis could furthermore shed light on the disruptive effects of many music properties that have left unexplored, and one of them is tempo.
Tempo and changing-state hypothesis
Within the changing-state hypothesis, tempo reflects the frequency of acoustic changes of a sound stream. Compared to slow-paced music, the sound stream of fast-paced music contains more acoustic information and more variations per unit of time. The changing-state hypothesis, if applied, implies that background music of faster pace will trigger the seriation process more frequently. As suggested by Jones and Macken (1992), the memory resource is shared between two seriation processes, including maintaining the list and the automatic processing of the irrelevant sounds. If the automatic seriation processing of the irrelevant sound stream is trigger more frequently, fewer resources will be left available for maintaining the to-be-remembered list. Therefore, errors in recall are more likely when listening to fast-paced background music.
Although no existing research has been conducted to examine the effect of music tempo in the irrelevant sound paradigm, evidence from two studies employing more naturalistic settings has indicated linkages between fast-paced music and deficit in memory performance (Hahn & Hwang, 1999; Oakes, 1999). In a study by Hahn and Hwang (1999), participants watched TV advertisements either accompanying with fast tempo background music or slow tempo background music. It is shown that fast background music significantly reduced in the number of advertising message recalled. Another line of research by Oakes (1999) investigating the content recall of radio programme also suggested that the content recall was less accurate when fast tempo background music was presented. Consistent with the changing-state hypothesis, these two lines of evidence suggest that faster background music is likely to have a higher disruptive potency to memory performance. To the extent that recalling of TV advertisement message and radio programme content are similar to the serial recall paradigm that both require temporary retention of information, it is thus reasonable to expect that music tempo might have a similar effect on serial recall task performance.
The present study
To shed further light on the irrelevant sound effect, this study attempted to explore the irrelevant sound effect in a broader context — background music. Specifically, we were interested in extending the applications of the phonological loop and changing-state hypothesis to the context of music. With reference to these two theories, we sought to examine the musical properties that determine the disruptive potency of background music on our verbal short-term memory performance. Of which, there were two musical properties that this study tried to look at. The first musical property is vocality. As potential confounding factors between the vocal and non-vocal music conditions have not been systematically controlled in previous studies, the present study adopted a number of measures to eliminate the possible confounds existed.
First, background music was carefully chosen on the basis of uniformity in musical characteristics. Second, the music excerpts were also reconstructed into different sound conditions to eliminate possible confounds existed. Specifically, the vocal and non-vocal music conditions were manipulated in an otherwise equivalent basis by using computer software. Third, a pink noise condition was introduced to serve as a placebo control to reflect any expectation-based effects due to the mere presence of acoustical distracter.
Another factor that the present study tried to examine is tempo. As mentioned, there was no known study established today explores the effects of music tempo on serial recall performance. Thus, we were left to guess whether music tempo is also a critical factor in determining an irrelevant sound effect. Examining the effects of tempo would moreover provide a fuller picture in the understanding of the applications of changing-state hypothesis on background music. It was therefore the purpose of this study to fill this research gap and to understand how verbal short-term memory performance is impacted by music tempo.
To investigate these two factors, we examined the serial recall performance of six conditions. They are vocal-slow music, vocal-fast music, non-vocal slow music, non-vocal fast music, silence and pink noise. Based on previous researches investigating the applications of phonological loop hypothesis on musical irrelevant sound effect, it was hypothesized that vocal music would produce a greater interference to serial recall performance. We also predicted that if fast-paced music would evoke seriation process more frequently as the changing-state hypothesis implies, then fast tempo conditions would disrupt serial recall performance more than would the slow tempo conditions.
The present study aimed to explore the irrelevant sound effect of music. The major objective was to investigate how different kinds of background music would impact serial recall performance and give rise to an irrelevant sound effect. With reference to the implications of the phonological loop and changing-state hypothesis, two musical properties were studied: vocality and tempo. By examining the musical properties concerned in the present study, we sought to identify the feature of background music that determines its disruptive potency on serial recall performance.
Effects of Vocality
The overall pattern of results indicates that vocality is a critical factor that determines the irrelevant sound effect. Specifically, this study demonstrated an irrelevant sound effect of vocal music, whereas performance under non-vocal music did not differ from silence. We found that, in the presence of vocal music, serial recall performance was degraded by 16% when compared to silence and 13.5% relative to non-vocal music.
The phonological loop hypothesis (Salame & Baddeley, 1982), which emphasizes that verbal content of sound is the key to interference, provides a good theoretical foundation for understanding this pattern of results. In line with the assumption made by Salame and Baddeley (1982) that speech-like material would gain automatic access to the phonological loop and disrupt serial recall performance, we found that vocal music significantly reduced serial recall performance and produced the greatest decrement among the six auditory conditions.
Adding further support to the phonological loop hypothesis, the results indicate that non-vocal had no significant interference effect to performance. It is therefore consistent with the view that sound containing limited vocal content does not gain the similar access as speech to the phonological loop and reduce serial recall performance.
These findings are consistent with previous research (Boyle & Coltheart, 2006; Salame & Baddeley, 1989) in demonstrating an irrelevant sound effect of vocal music. Whereas the results of previous studies may possibly be confounded by the uncontrolled factors between the vocal and non-vocal condition, the current results extend these findings by clarifying that the cause of interference is due to vocal content. To the author’s knowledge, the present study is the only study that had vocal and non-vocal stimuli carefully matched in an otherwise equivalent basis to eliminate the possible confounds. By showing that an equivalent non-vocal version of music failed to produce any significant impairment, our study therefore elucidates the cause of this damaging effect of vocal music is due to its vocal content.
Regarding the effects of non-vocal music, the results of the current studies and past literature were inconsistent which leave the effects somewhat unclear. On the one hand, the present study showed performance in the non-vocal condition did not differ from silence. Past literature, on the other hand, reported a small yet significant disruption of non-vocal music. For example, Iwanaga and Ito (2002) found that non-vocal music, despite with a lesser interference, also reduced serial recall performance. Similarly, Nittono (1997) found an intermediate disruptive effect of instrumental classical music.
This discrepancy could possibly due to the different musical stimuli used. It has been suggested by Salame and Baddely (1989) that instrumental music might also contain a sub-sample of features that are similar to speech and might potentially disrupt serial recall performance. As the present study removed vocals from vocal music to reconstruct the otherwise equivalent version of non-vocal music, it might well because of this procedure also removed the sub-sample speech-like features, thus making the non-vocal music no longer disruptive to verbal short-term memory. In regard to this inconsistent finding, future research should study the effects of instrumental music in more comprehensively through examining and comparing the effects of different kinds of instrumental music on serial recall performance. Nonetheless, combining the results of the present study with the previous findings, it seems safe to conclude that non-vocal music is relatively less disruptive than vocal music.
Overall, the above findings were consistent with the implications of phonological loop hypothesis that vocal music has a disruptive effect on serial recall performance and in turn induce an irrelevant sound effect. The present findings, moreover, clarify the disruption of vocal music is due to its vocal content by showing that non-vocal music produce no significant impairment to serial recall performance.
Effects of tempo
The present study also looked at the effects of music tempo on serial recall performance. Contrary to the hypothesis, tempo did not prove to be associated with the interference of music on performance. The result showed fast-paced music, even though containing a larger amount of acoustic variations, was not more disruptive than slow-paced music. It therefore appeared that music tempo might have no effect on the irrelevant sound effect. There are three possible explanations could be put forward to interpret this finding. Firstly, the difference in the amount of acoustic variation between the fast-paced and slow-paced music might not be sufficient enough to elicit a significant difference on serial recall performance. Therefore, this notion suggests that music tempo does have an effect on influencing verbal short-term memory performance, but the difference between the two tempos was too small to have an effect. Given that there is an absence of empirical research to suggest the ‘threshold’ of acoustic change that could evoke additional impairment to serial recall performance, it thus provides no way to verify this interpretation. Clearly, further study is needed to explore the threshold of acoustic change that could potentially induce an irrelevant sound effect, and this would moreover enrich the changing-state hypothesis and advance the present understanding on the irrelevant sound effect.
In the light of the changing-state hypothesis, the second possibility is that the degree of acoustic variation is defined by its qualitative but not absolute term. Generally, the changing-state hypothesis has been understood as the larger the amount of acoustic variation, the larger the degree of impairment. However, this acoustic variation could be defined into two ways: the qualitative variation between each physical unit of the sound stream; and the quantitative or absolute number of variation occur across the sound stream. As reviewed earlier, past examinations on the changing-state hypothesis have consistently confirmed the former notion by showing that sound stream with more abrupt changes in pitch, timbre and tonality was particularly disruptive (eg. Jones & Macken, 1995; Jones, Madden & Miles, 1992). The present study, on the other hand, disconfirmed the later notion by demonstrating that increasing the absolute amount of variation across the sound stream did not increase the degree of disruption to serial recall performance. Taken these two lines of evidence together, a more compelling explanation to the inability of tempo to determine serial recall performance may relate to the nature of changing-state cue that it is not evoked by the absolute amount of variation but the qualitative variation between each successive physical unit. A tentative conclusion that drawn from this notion would thus be increasing the absolute amount of acoustic variation does not produce a higher degree of disruption but when the variation is irregular, the disruption would increase markedly. Our study therefore delineates the properties of acoustic variation and clarifies that the increasing the variation across a regular sound stream does not impact serial recall performance.
Thirdly, this result may possibly account for the induced arousal change by music tempo. According to the Yerkes-Dodson law, the relationship between arousal and cognitive performance follows an inverted U-shaped function. This is also the case for memory performance. Considering the level of arousal, optimal memory performance occurs in the intermediate levels of arousal, whereas very low or very high levels of arousal result in diminished memory performance (Sarason, 1980). Past studies have been conducted showing that music tempo of 110 – 120 bpm was the optimal level of arousal in memory task performance (Mayfield & Moss, 1989; Webster and Weir, 2005). It is thus possible that participants in the slow tempo (60 bpm) condition were under-aroused while were over-aroused in the fast tempo (140 bpm) condition. Therefore, performance in both conditions were suppressed, making the difference in serial recall performance might not be significant between the two conditions. While the optimal music tempo condition in serial recall performance has no