Our results demonstrate that regions of the cortex are engaged in directing attention to acoustic features even before the sounds begin; moreover, different regions are engaged more strongly depending on what feature is directly selective auditory attention: left FEF when attending location and left posterior STS when attending pitch. Previous fMRI studies demonstrate that activity in a left dominated frontoparietal network is enhanced during attentionally demanding trials compared to fixation trials, whether subjects attend to a spatial or a non-spatial feature, and in both visual and auditory tasks. During a visual attention task, anatomical regions proximal to bilateral FEFs were more active during spatial attention, while the left ventral occipital cortex was more active when subjects attended to color (Giesbrecht et al., 2003
; Slagter et al., 2007
). During an auditory task, left FEF showed enhanced activity for both location and pitch, even on catch trials where there was no acoustic stimulus (Hill and Miller, 2010
), consistent with the anticipatory activity we found in our task. However, in this earlier study, right FEF was more active in location trials, and the inferior frontal gyrus (a region linked to language processing) was more active in pitch trials. While it is difficult to directly compare results of studies using different sensory stimuli and different neuroimaging techniques, especially since the relationship between neural activity measured using MEG and BOLD responses measured using fMRI is not well established, our results add to evidence that FEFs are involved in control of covert spatial attention across different modalities (while other areas may be similarly engaged when attending to non-spatial features). In contrast to previous studies, our results suggest that there is an asymmetry in auditory processing whereby left FEF is more strongly involved in attending to auditory stimuli based on spatial location compared to pitch. We also show that this attention-specific control begins in preparation for upcoming stimuli containing a to-be-attended feature. It is worth noting that the activity observed in preparing to attend to stimuli based on spatial condition may be due to the deployment of both auditory and visual attention to the spatial location of interest, as would likely be the case if auditory and visual spatial attention share a common supramodal network. Teasing this apart could be interesting in future studies that make use of either auditory-only cues or visual cues that come on well before auditory attention must be directed. However, our observations here are unlikely to be due solely to the deployment of visual attention.
It has been shown previously that preparing to attend to a sound likely to originate from a given direction biases cortical activity in auditory cortex contralateral to that direction (Voisin et al., 2006
), indicating that prior to sound onset, listeners “prime” cortical representations to favor upcoming sounds from the direction to be attended. Given this, our results are consistent with our listeners engaging auditory attention to perform our tasks, although we cannot rule out the possibility that listeners co-deploy both auditory and visual spatial attention networks in anticipation of an upcoming sound. Additionally, although some evoked responses are visible in the traces of left FEF and STS (likely due to leakage from primary sensory cortices), the observed differences reported here must be due to the task condition (attend-space versus attend-pitch) since the acoustic stimuli used in the two conditions are identical.
The left lateralization of FEF activity initially seems at odds with past reports of “hemispheric dominance.” It is well established that the right hemisphere processes information in both visual fields, whereas the left hemisphere exclusively encodes the right visual field (Mesulam, 1981
). This raises the question of why left, but not right, FEF is more active in our location trials, regardless of the direction of the target, and why there are no significant differences in FEF activity for “attend left” versus “attend right” trials. It is unlikely that this is due to preparatory motor activation (i.e., preparing to press a button with the right hand), since such activity would be the same for both the space task and the pitch task that were contrasted, yet differences in activation were observed before a sound stimulus was presented (and thus before an appropriate response could be prepared). We believe this left FEF bias may reflect its participation in a dorsal, top-down attention network (Corbetta et al., 2008
), with right FEF involved in top-down attention, exogenous attention, and shifting of attention. Previous auditory fMRI studies that find bilateral FEF activation during auditory spatial attention tasks used paradigms that differ from ours: most either required subjects to explicitly shift their auditory spatial attention (Salmi et al., 2007
), or exogenously cued the auditory location to attend (Wu et al., 2007
). Moreover, because of scanner noise, listeners in these studies may have deployed some form of non-spatial attention to focus on the desired acoustic stimuli instead of or in addition to engaging spatial attention. The one study that required top-down deployment of auditory spatial attention (Hill and Miller, 2010
) yielded poor behavioral performance (especially early in the experiment), suggesting that the subjects were not always successful in deploying attention, and sometimes reoriented attention while trying to perform the task. In contrast, our study presented a brief stimulus (one syllable long), yet subjects performed the task reasonably well, showing that they successfully deployed top-down spatial attention in the preparatory period. Thus, we suggest that the left FEF is differentially more involved in top-down auditory spatial attention, consistent with the supramodal attentional network previously proposed (Corbetta et al., 2008
). Note that although left FEF shows greater activity during spatial rather than pitch-based attention trials, left FEF also may well play a role in non-spatial attention as well, as evidenced by a significant (before a multiple-comparisons correction) correlation between activity in left FEF and behavioral performance on attend-pitch trials. The pre-auditory-stimulus left FEF activity observed here is also similar to the anticipatory activity in FEFs reported in past visual studies that is linked to top-down control of spatial attention (Awh et al., 2006
We found that left posterior STS, which was not chosen a priori
as an ROI, was recruited in attend-pitch trials during the preparatory period, a result that mirrors previous findings of other cortical regions showing attention biases for non-spatial features: ventral occipital cortex for attention to color (Giesbrecht et al., 2003
; Slagter et al., 2007
), inferior frontal regions for attention to spectral features in a language-related task (Hill and Miller, 2010
), and preparatory activity in auditory cortex contralateral to the expected location of an upcoming sound (Voisin et al., 2006
). Several studies have associated the left STS with the identification or categorization of sounds based on non-spatial attributes (Möttönen et al., 2006
; Liebenthal et al., 2010
), especially for people with absolute pitch (Schulze et al., 2009
). Although this area is likely involved in performing categorization in both the spatial location and pitch tasks, our results suggest that the cortical region associated with categorization of pitch information becomes more active and helps listeners prepare to select a target stimulus based on pitch. These results support the existence of different pathways for processing “what” and “where” sound attributes (e.g., Rauschecker and Tian, 2000
; Ahveninen et al., 2006
); however, since the posterior location of the left STS activation observed here is more consistent with the previously reported “where” pathway, additional experiments will be necessary to provide the spatial resolution required to definitively tease apart the contributions of these areas. Moreover, we were relatively conservative in our data analysis (e.g., Bonferroni correction) to decrease the likelihood of false positives; however, this increases the chance that additional areas are significantly involved during tasks like those used here; experiments that relax constraints or that employ other statistical approaches might expose such other activity (e.g., Singh et al., 2003
; Maris and Oostenveld, 2007
). For example, the use of masking noise could obscure differences in activity that the two tasks might have evoked in auditory cortex for presentations in quiet, especially given the conservative analyses we adopted. It is also possible that there is an underlying activity difference in left STS during the stimulus (“post” period; as seen in Figure C) as well; future studies with better SNR or less strict thresholding may well observe significant left-biased activity differences during the stimuli when attending based on pitch.
Finally, our results show that left FEF is involved both before and after the onset of sound while activity in left posterior STS is significantly enhanced only prior to the onset of sound. These changes were correlated across subjects, as if the degree of attentional modulation in both attend-location and attend-pitch trials depends on some common signal, regardless of the feature attended. However, these activity differences are not significantly correlated with behavioral performance, even though listener ability varies widely across subjects. In our neural analysis, we contrasted two conditions, each of which engages selective auditory attention; thus, any differences in the strength with which listeners engage cortical regions that are common to both attend-location and attend-pitch trials is invisible in our analysis; we only see the indirect effects of such common control in the strength of modulation of the feature-specific areas left FEF and left STS. Combined with the observation that across subjects, performance is strongly correlated in the attend-location and attend-pitch trials, our results suggest that overall selective attention performance depends on the degree of engagement of neural areas that are employed both when attending to location and when attending to pitch, and/or on individual differences in the fidelity of sensory encoding of the basic acoustic information needed to compute auditory features like location and pitch (Ruggles et al., 2011
). Here, we also found insignificant but trending correlations between activity in attend-space and attend-pitch trials. However, the estimates of neural activity normalized to baseline used here are influenced by the signal-to-noise ratio and the number of valid trials for each subject, which could contribute to the lack of observed significant correlations. Future experiments thus could be undertaken to explore the degree to which the overall activity of a general “attention network” helps to predict individual ability on this kind of selective auditory attention task. Additionally, although we did not have a sufficient number of incorrect trials to perform a meaningful analysis here, future studies could also look at activations in error trials to also examine the auditory selective attention network.
Taken in the context of previous psychoacoustical and neuroimaging work, our findings support the conclusions that (1) left FEF is involved in both directing and sustaining auditory spatial attention and (2) the left STS aids object selection based on its pitch feature prior to the onset of sound.