Interference by process, not content, determines semantic auditory distraction

Abstract

Distraction by irrelevant background sound of visually-based cognitive tasks illustrates the vulnerability of attentional selectivity across modalities. Four experiments centred on auditory distraction during tests of memory for visually-presented semantic information. Meaningful irrelevant speech disrupted the free recall of semantic category-exemplars more than meaningless irrelevant sound (Experiment 1). This effect was exacerbated when the irrelevant speech was semantically related to the to-be-remembered material (Experiment 2). Importantly, however, these effects of meaningfulness and semantic relatedness were shown to arise only when instructions emphasized recall by category rather than by serial order (Experiments 3 and 4). The results favor a process-oriented, rather than a structural, approach to the breakdown of attentional selectivity and forgetting: performance is impaired by the similarity of process brought to bear on the relevant and irrelevant material, not the similarity in item content.

Introduction

One of the most influential constructs in memory research is interference: the ease with which items are retrieved from memory is dictated, at least in part, by other stimuli or events that are similar in some way to the target (see, e.g., Anderson, 2003, Baddeley, 1986, McGeoch, 1942, Nairne, 1990, Nairne, 2002, Neath, 2000). The classical, structuralist, view has been that such interference-by-similarity-of-content directly causes forgetting, that is, forgetting is a passive side-effect of structural changes that result from the storing of new, similar, events in memory (Anderson, 1983, Cowan, 1999, McGeoch, 1942, Mensink and Raaijmakers, 1988, Oberauer and Lange, 2008, Oberauer et al., 2004, Salamé and Baddeley, 1982). However, an alternative, more functional, view is that ‘forgetting’ (or the impairment of retrieval) reflects the legacy of dynamic and adaptive selective attention processes (such as inhibition; e.g., Houghton & Tipper, 1994) that are designed to resolve conflict during the selection of candidates at retrieval (e.g., Anderson, 2003). Set within this quintessentially attentional approach to forgetting, the present article explores the nature of phenomena relating to impaired retrieval from memory due to distraction from irrelevant auditory events using the structuralist, interference-by-similarity-of-content, approach as a theoretical counterpoint.

One line of research in which a dynamic selective attention framework has been used to reconstrue putatively mnemonic phenomena is that concerned with the disruptive effects of to-be-ignored sound on visual-verbal serial recall whereby a list of around 6–8 verbal items (e.g., letters or digits) is to be recalled in strict serial order (the irrelevant sound effect—hereafter ISE—e.g., Colle and Welsh, 1976, Jones et al., 1992, Jones and Tremblay, 2000, Salamé and Baddeley, 1982). The mere presence of background sound depresses serial recall appreciably, the weight of evidence favoring the view that the effect results from interference-by-process, and is not a passive side-effect of having similar items to remember and to ignore (Hughes and Jones, 2005, Jones and Tremblay, 2000). Specifically, this ‘classical’ ISE is thought to result from the obligatory, preattentive, seriation (or ordering) of sound sequences producing competition for the deliberate process of seriating the to-be-remembered items. Here we examine whether the principle of interference-by-process can be extended to a setting in which the focal memory task involves not serial processing but semantic retrieval strategies: Does the concurrence of similar semantic processing (rather than serial processing) applied to relevant and irrelevant material now dictate the form and degree of distraction? What little evidence there is seems to suggest that the structural accounts seem perfectly adequate in this context, that is, disruption from irrelevant sound in semantic memory seems amenable to a classical, and arguably simpler, ‘interference-by-content’, explanation (Beaman, 2004, Neely and LeCompte, 1999) rendering the ‘interference-by-process’ account rather paradigm-bound to serial short-term memory. The goals of the present series were to revisit the empirical signature of auditory distraction in the context of episodic short-term memory tasks that tap semantic memory processes (particularly given the paucity of studies on the issue) to establish the degree to which it is distinct from that found in serial recall, and to examine thereafter how such distinct phenomena might be reconciled with a dynamic process-oriented approach to interference.

The debate between the structuralist and process-based standpoints can be observed in microcosm in a body of research showing that the presence of irrelevant, to-be-ignored, sound markedly increases forgetting in a (usually visually-presented) serial recall task (e.g., Colle and Welsh, 1976, Jones et al., 1992, Salamé and Baddeley, 1982). The conventional viewpoint, that forgetting can occur as a direct and passive consequence of the structural similarity between to-be-remembered and irrelevant episodes or stimuli (e.g., McGeoch, 1942), is evident in several theoretical accounts of the ISE that view it as a mere consequence of auditory stimuli gaining access to the same representational space as the to-be-remembered items (e.g., phonological store, Burgess and Hitch, 1992, Gathercole and Baddeley, 1993, Salamé and Baddeley, 1982; primary memory, Neath, 2000). Although these accounts differ in their detail of how interference arises, the important point for present purposes is that they are all examples of an interference-by-content approach: recall is impaired as a result of the similarity in identity (i.e., content) between to-be-remembered and to-be-ignored items.

Several strands of evidence converge to weaken the interference-by-content approach. First, non-speech sounds such as tones—which bear little or no resemblance to the to-be-remembered items—produce disruption similar in degree and kind to that from irrelevant speech (e.g., Jones and Macken, 1993, Neath and Surprenant, 2001). Second, the magnitude of disruption is unrelated to the degree of phonological similarity between to-be-remembered and to-be-ignored items (Jones and Macken, 1995, LeCompte and Shaibe, 1997; but see Hughes & Jones, 2005) thereby disconfirming the predictions of an early account based on the concept of a phonological store (Gathercole and Baddeley, 1993, Salamé and Baddeley, 1982). As a result of these findings, the phonological store account of the ISE has been modified and expressed computationally such that irrelevant speech disrupts a representation of order within the passive store rather than interfering with item representations (Norris et al., 2004, Page and Norris, 2003). However, problematic for any account that views irrelevant speech as disrupting the phonological store is recent evidence showing that rehearsal is a precondition for its expression (Jones, Macken, & Nicholls, 2004).

Third, the interference-by-content approach fails to acknowledge adequately the critical importance of the nature of focal task processing, the impairment of recall being chiefly determined by the co-existence of similar to-be-recalled and to-be-ignored items within a store. That is, they cannot account for why the ISE is only found if the focal task necessitates or tends to encourage a seriation process (e.g., serial rehearsal) and why the mere presence of similar content between the memory material and the sound is not sufficient (or necessary) for the effect (Beaman and Jones, 1997, Farley et al., 2007, Henson et al., 2003, Hughes et al., 2007, Perham et al., 2007).

Whilst the preoccupation of the interference-by-content approach is with item identity, on the interference-by-process account, the key determinant of the disruption in serial recall is the extent to which both the irrelevant sound and the focal memory task share similar seriation (or ordering) processes (Jones, 1993, Jones and Tremblay, 2000). A key observation underpinning this account is the changing-state effect (e.g., Jones et al., 1992) whereby a sound sequence—regardless of whether it comprises speech or non-speech—that exhibits abrupt changes in acoustic properties (e.g., “k v h q…”, or a sequence of tones changing in frequency) is invariably more disruptive than a continuous or repeating stimulus (e.g., “k k k k…”, or a repeated tone). On the interference-by-process account it is assumed that the preattentive perception of acoustic changes between segmentable elements in the sound yields cues as to the order of those elements as a by-product of primitive, acoustic-based, perceptual organization processes (cf. Bregman, 1990). These irrelevant order cues compete for—and hence impair—the deliberate seriation process (serial rehearsal) supporting ordered recall of the to-be-remembered items (Hughes and Jones, 2005, Jones, 1993). In support of this view, the ability to encode the order of stimuli in an attended changing-state auditory sequence predicts the degree to which that sequence is disruptive when presented as irrelevant sound during serial recall (Macken, Phelps, & Jones, in press).

In sum, results based on research using the serial recall paradigm favor a dynamic process-based approach (Jones & Tremblay, 2000). However, the phenomenon of interference-by-process seems highly specific to a particular process (seriation) and little evidence is available with respect to whether such conflict occurs between other types of processes. In the present study, therefore, we addressed whether the phenomenon extends to auditory distraction in the context of a focal task that is likely to be dominated by semantic-based, rather than seriation, processes. This is a particularly pertinent issue in light of the fact that the little evidence there is on semantic auditory distraction effects suggests that the interference-by-content approach offers a perfectly adequate explanation.

On the interference-by-process account of auditory distraction in serial recall, it is the processing of the precategorical, acoustic, attributes of the sound that is key to the disruption (e.g., Jones & Macken, 1993). Consistent with this view, neither the lexical-semantic content of the irrelevant sound (when speech is used) nor the similarity in terms of semantic content between the speech and the to-be-remembered list has any bearing on the magnitude of disruption (Buchner et al., 1996, Jones et al., 1990, LeCompte et al., 1997, Surprenant et al., 2007; but see Buchner, Rothermund, Wentura, & Mehl, 2004). Such findings are easily explicable on the intereference-by-process account: The key process supporting serial recall is an articulatory-based seriation process, not a semantically-based one (e.g., Jones et al., 2004). Thus, the lexical-semantic attributes of irrelevant speech would not be expected to conflict with the focal seriation process. However, the absence of such lexical-semantic effects in the context of serial recall can also be accommodated within the interference-by-content approach. In the typical serial recall task, the to-be-recalled items (e.g., digits, letters) are relatively impoverished in terms of semantic content. Thus, the representations of to-be-recalled items—devoid of rich semantic content—may not be susceptible to degradation or retrieval-confusion as a result of activated semantic representations of the irrelevant speech items.

In line with the interference-by-content approach, the results of a small number of studies suggest that when the items are semantically rich, recall is indeed impaired by the semantic attributes of the irrelevant sound (Beaman, 2004, Jones et al., 1990, Martin et al., 1988, Neely and LeCompte, 1999, Oswald et al., 2000). For example, in a category-exemplar recall task, in which a list of, say, 16 semantically-rich items (nouns) taken from a single semantic category are presented for free recall, the semantic similarity between the relevant and irrelevant items impairs performance (Beaman, 2004, Neely and LeCompte, 1999). The free recall of relatively low-dominance category-exemplars (e.g., “avocado”) is disrupted (as reflected in omission errors) more by related, high-dominance, irrelevant category-items (that are not included in the to-be-remembered list; e.g., “apple”) than by high-dominance, categorically-unrelated, irrelevant items (e.g., “hammer”). Such results seem to be readily explained within an interference-by-content approach: The semantic representations of the to-be-recalled items may be degraded or otherwise made less accessible as a function of their semantic similarity to the irrelevant items (e.g., Anderson, 1983, Oberauer and Lange, 2008, Oberauer et al., 2004, Rundus, 1973).

However, our central contention in this paper is that these semantic auditory distraction effects may also be amenable to, and indeed be better explained by, an interference-by-process analysis (cf. Marsh et al., submitted for publication, Marsh et al., 2008). The starting point for this analysis is that, compared to serial recall, the semantic richness of the items in category-exemplar recall, as well as the longer list-length, demotes the likelihood of a seriation strategy and instead promotes the use of semantic-based organization processes. It is this shift in the nature of the dominant process/strategy used to support performance in the focal task, not simply the semantic richness of the to-be-recalled items, that renders such tasks vulnerable to competition from the processing of the semantic content of the irrelevant sound. Thus, in contrast to the interference-by-content approach, it is not the mere co-activation or co-registration of relevant and irrelevant semantic representations that impairs performance. Rather, it is the integrity of dynamic, semantic-based, organizational processes engaged in support of retrieval that is compromised by the semantic processing of the irrelevant sound. In this way, semantic auditory distraction may be regarded as an extension of the general case of interference-by-process: It may reflect a difficulty in selecting amongst two sets of semantic representations both of which represent plausible candidates for populating the semantic-organizational skill used in the focal task, just as with two sets of serial representations in the classical irrelevant sound effect.

In the studies that follow, we scrutinize further the nature of semantic auditory distraction and attempt to clarify the extent to which semantic-organizational processing in the focal task—as opposed to the mere semantic richness of the to-be-recalled items—is responsible for disruption by the lexical-semantic attributes of irrelevant speech. We use a setting in which a relatively long list of 32 exemplars (e.g., “strawberry”, “pigeon”, etc.) drawn from a smaller set of 4 semantic categories (e.g., “Fruit”, “Birds”, etc.) are presented for recall. It is well established that under such conditions participants tend at test to cluster the randomly presented exemplars according to their category at a greater-than-chance level even without instruction to do so (Bousfield, 1953, Smith et al., 1981). This semantic category-clustering (henceforth termed “semantic-categorization”) implies secondary organization whereby participants bring to bear pre-existing conceptual relationships or semantic associations to guide encoding and retrieval of episodic information which is distinct from primary organization whereby the organization corresponds to the serial order of the list (Tulving, 1968).

In another study (Marsh et al., submitted for publication, Marsh et al., 2008), we have used lists of words drawn from a single semantic category to investigate the role that source-monitoring processes play in governing the false recall of items presented as irrelevant sound. Single category lists being ‘blocked’ by semantic category were ideally suited to this purpose because this method of list presentation increases false recall (Brainerd, Payne, Wright, & Reyna, 2003). The research reported here, though conceptually similar, follows a distinct empirical line. In the following experiments, we are not interested in false recall (indeed random presentation of exemplars drawn from several semantic categories attenuates false recall), but rather in how meaningful irrelevant sound may impair the recruitment of semantic category knowledge as an organizing principle. Recall of semantically categorizable lists, as compared with lists of words drawn from single semantic categories, yields measures of semantic processing in the form of the semantic organization of responses and the probability of producing each category at test (Burns & Brown, 2000). The purpose of the present study was thus to examine whether the meaningfulness of irrelevant speech interferes with semantic organization during free recall of categorizable word-lists.

Assuming quite distinct, semantic-based, processing in the category-exemplar recall task as compared with the seriation-based processing in serial recall, what predictions does the interference-by-process approach make? The first is that the empirical signature of auditory distraction in semantic task settings will be qualitatively distinct from that found in serial recall: the semantic properties of the sound should be endowed with disruptive power but the acoustic, changing-state, properties of the sound should prove relatively impotent. Whilst this first prediction also flows from the interference-by-content approach, the second is unique to the interference-by-process approach: the impairment should be a product not only of the processing of the lexical-semantic attributes of the sound but also of the deployment of semantic-based processes as a means of supporting retrieval in the focal task. That is, the mere presence of semantic content within the to-be-recalled and irrelevant material should not be sufficient to produce disruption. Thus, if the task does not necessitate or encourage the use of semantic processes but instead seriation processes, the semantic attributes of the sound should not have disruptive potency even when the to-be-recalled items are rich in semantic content. The same prediction holds for between-sequence semantic similarity effects: the greater disruption found with greater semantic similarity between to-be-recalled and irrelevant items is not a passive by-product of their greater overlap within some semantic-psychological space (e.g., Oberauer et al., 2004). Rather, such similarity exacerbates the difficulty of coupling the correct (i.e., task-relevant) set of semantic representations to the semantic-organization processes being engaged to support retrieval.

The present series of experiments begins by seeking to establish whether or not the action of irrelevant sound in the context of semantically-driven episodic tasks—using a category-exemplar task—is indeed distinct from that in the standard serial recall setting. Later experiments in the series investigate the nature of between-sequence semantic similarity effects in this setting and explore the key hypothesis—based on the interference-by-process framework—that when the same type of lists are subject to primary organization (seriation) their recall should become immune to semantic auditory distraction.