The two studies presented here were designed to demonstrate the local and distributed systems that are involved in attention and cognitive control when listening to a speaker and recalling what has been said. Two experimental methods were designed to capture everyday listening conditions: task demand, namely delayed or immediate recall of what had been said by the attended speaker; and the presence or absence of background competing speech. Based on the results of the two studies, there were three cortical nodes that responded to speech-in-speech masking irrespective of the task demand; the precuneus, the left PT/IPL and the right aI/FOp. I will start by discussing these common regions.
Neuropsychological lesion-deficit analyses of the precuneus to determine the function of this region are sparse due to the rarity of the condition. Functional neuroimaging studies, however, have implicated this region in a number of different functions (Cavanna and Trimble, 2006). These most probably relate to the multiple overlapping components within this region that form parts of anatomically and functionally dissociable networks, as previously shown for the adjacent posterior cingulate cortex (Leech and Sharp, 2014). One function of this region is egocentric spatial orientation, which has often been considered in terms of visuospatial navigation (for review, see Boccia et al., 2014). The precuneus is also a component of the Dorsal Attention Network (DAN), which incorporates the dorsal precuneus and bilateral medial intraparietal sulci, midline supplementary eye field and frontal eye fields, and superior parietal lobules. This network has been most often investigated with regard to its response to visual tasks, becoming active as participants voluntarily focus attention on perceptually distinctive visual stimuli that are salient within the context of a specific task-dependent goal (for reviews, see Corbetta et al., 2008; Corbetta and Shulman, 2011). However, a recent
study has also strongly implicated the precuneus in detecting a target sound in complex acoustic environments (Zündorf et al., 2013). In this present study, the precuneus was more active during the diotic presentation of two speakers (in the absence of spatial cues), compared with attending to a single speaker. This finding is compatible with a top-down role in the detection of the salient speech stream based on non-spatial perceptual cues, such as the fundamental frequency of the voice; but, as in the study by Zündorf and colleagues (2013), activity within this region increased significantly when there were auditory cues indicating a spatial separation of the two speakers. Further, associated with this increased activity in the presence of spatial cues, the results also identified an unexpected increase in activity in regions located in the supplementary eye field and the frontal eye fields. Future studies may choose to investigate this further to determine whether spatial cues during speech-stream segregation are accompanied by automatic eye movements towards the attended speaker.
The study of Zündorf and colleagues (2013), although of different design and employing non- verbal auditory stimuli, also demonstrated an increased response of the left PT to spatial cues, with evidence of some right posterior temporal involvement. Across both studies reported here, activity in the left PT increased in response to one speaker, increased further when there was more than one speaker, and was greatest in the presence of spatial cues. The PT has been proposed to be a computational hub, directing both spectrotemporal and spatial information to wider distributed networks involved in the identification, semantic recognition and auditory stream segregation of sounds, both verbal and environmental (Griffiths and Warren, 2002). These authors proposed that the PT might be a central node in resolving the ‘cocktail party’ effect, and my results presented in this study support this hypothesis. Of note, a clinical study on stroke patients by Zündorf and colleagues (2014),
using the same complex non-verbal sounds of their earlier study (Zündorf et al., 2013), indicated that the right posterior temporal cortex, including the PT, is central to sound localisation. However, it is important to note that patients with lesions that included the left PT were under-represented because such patients were often too language-impaired to participate. Based on my present study, I would argue that segregating one speech stream from others, using all available non-spatial and spatial cues, is dependent on the left PT, although activity was also evident in the right posterior STG, suggesting that this function may be shared between the cerebral hemispheres.
Turning to the role of the right aI/FOp, previous studies have proposed that this region is specialised for initiating response inhibition and task-switching (reviewed in Aron et al., 2004). However, in a more recent study, Hampshire and colleagues (2010) demonstrated that this region, part of the cingulo-opercular network, becomes active during the detection of important cues irrespective of whether this results in the generation or inhibition of a motor response, or even when there is no overt response. In their study, activity was preferentially greater in the cingulo-opercular network for tasks that most depended on working memory, when the range of tasks resulted in activity in both the cingulo-opercular and the bilateral IFS/IPS networks. In the model proposed by Menon and Uddin (2010), the right aI/FOp is a core node involved in the generation of control signals following the perception of salient environmental events. These signals direct working memory, attention and other higher-order control systems towards the mental processing of these events.
The clear difference between the two studies is the functional dissociations between the response of the right dorsolateral prefrontal cortex, centred on the MFG, and inferior parietal
cortex (SMG). The first study demonstrated activity within the ventral right fronto-parietal network associated with the presence of a competing speaker. Therefore, the linguistic and semantic processing of heard speech, the encoding of the information as episodic memories and remembering the information until the end of the scanning session all required minimal involvement of this system. However, perceptual difficulty due to the presence of a competing speaker markedly increased activity in both the frontal and parietal components. One explanation for this is an increased need for sustained attention when attempting to encode the information conveyed by the ‘attended’ speaker on the perceptually difficult trials; this system has been associated with sustained attention (for a review, see Singh-Curry and Husain, 2009). Changing the task demand in Study 2, where an immediate response to what had been heard was required, abolished activity in the right SMG, regardless of perceptual difficulty, and resulted in increased activity in the right MFG across all trials. Therefore, a task that relied on working memory rather than encoding in episodic memory meant that this ventral right fronto-parietal system was no longer influenced by the need for speech-stream segregation. The ICA analysis demonstrated that Component 4, as well as showing activity in the cingulo-opercular and IFS/IPS networks, revealed correlated activity in the left inferior frontal gyrus, posterior inferolateral temporal lobe and inferior parietal cortex. This indicates the operation of a left hemisphere verbal working memory system, which was also demonstrated to be active during the Response trials. Therefore, the task demand had a major influence on ventral left and right parietal networks, with tasks depending heavily on working memory, resulting exclusively in left hemisphere activity, whereas episodic memory encoding requiring increased attention due to perceptual difficulty depended on right hemisphere involvement.
and attention that are active across many different kinds of task (Fedorenko et al., 2013; Hampshire et al., 2012). It was not surprising that they were most active during the Response trials of Study 2, although ICA analyses of both studies demonstrated that they were also active during the attentive demands of the listening conditions. Activity in these systems, however, was not modulated by the perceptual difficulty associated with speech-in-speech masking. The one exception, as previously discussed, was the right aI/FOp component of the cingulo-opercular network, which was strongly influenced by speech-in-masking, indicating a particular role for this region in regulating attention and cognitive control as the perceptual difficulty increases.