Posted comment on ´Interactions between conscious and subconscious signals: selective attention under feature-based competition increases neural selectivity during brain adaptation` written by Y. Kikuchi, J. Ip, G. Lagier, J.C. Mossom, S. Kumar, C.I. Petkov, N.E. Barraclough and Q.C. Vuong and published in Journal of Neuroscience July 2019 39(28) p. 5506 doi.org/10.1523/JNEUROSCI.3052-18.2019
SUMMARY
The article by Kikuchi and team begins by describing efficient perception as a neural interaction between voluntary cognitive processes such as attention and automatic and subconscious processes such as repetition suppression due to adaptation from predictable input. The authors examined in their study how attention interacts with visual and auditory neural adaptation from the perspective of whether it counteracts adaptation reduction in neural gain as previously observed, or not and if it did whether attended stimulus features were in competition with each other.
Kikuchi and team`s experiments were based on the manipulation of attentional focus on visual stimuli (ie. faces) or auditory stimuli (ie. voices) and assessment of neuronal responses by comparing fMRI images. It was expected that the BOLD responses would be strongest for the repetition of identical stimuli and therefore, the face and voice sensitive regions of interest (ROIs) would demonstrate increases in activities for the second manipulated stimuli in comparison to the first un-manipulated ones. This observation was believed to demonstrate stimulus-related adaptation function. Increases of equal amounts equivalent to the changing level of stimulus manipulation would, according to the authors, indicate that attention affected only neural gain and not selectivity whereas differential increases would indicate that attention affected neural selectivity. In their experiments, auditory stimuli were taken to be emotionally neutral Ah-like utterances and 12 samples were paired into six lots of one male-one female voice combinations. Each voice pair was manipulated using recording equipment until over 5,000 stimuli were possible. The visual stimuli used in the experiments were 6 male and 6 female faces with neutral expressions. In all 6 pairs were created with one male face and one female and these were ´morphed` using software into over 120 different facial images. Spatial location was manipulated by presenting the facial images in different parts of the screen.
In Experiment 1, spatial selective attention was compared to non-spatial selective attention. Seven spatial locations on the display monitor for the faces (or in the virtual acoustic space for voices) were selected for the behavioural experiment and 3 perceptual spatial distances were selected for the fMRI experiments. The test subjects had to discriminate the identity or spatial position of the presented face or voice pairs and had to attend to one of the required stimuli judging whether it had changed or not. The other presented feature had to be ignored. The identity and spatial location were manipulated during the course of the experiment and the experiment was calibrated in order to control for the participant`s expectancy of change. Therefore, only 50% of the stimulus repetitions were changed. The participant`s indicated their decision by pressing a computer key during the fixation period between the images. In Experiment 2, attention was manipulated to voice identity compared to sound loudness. Different participants took part in this experiment to those in Experiment 1, but the participants were presented with the same voice stimuli as that of Experiment 1. Noise bursts were generated by pink acoustic sounds with 21 variations in intensity to simulate changes in loudness. In Experiment 2, subjects were asked to judge whether the identities of the voices were the same and were also asked to judge the loudness of two noise stimuli. The other stimuli were ignored. In both experiments fMRI images of the BOLD responses in the face and voice ROIs located in the temporal lobe were taken, processed and analysed.
Kikuchi and team found in Experiment 1 that attention to face or voice identity whilst ignoring the stimulus location increased only the neural gain of the face or voice stimuli ie. cortical firing was increased. The response increased as the identity level increased for both faces and voices (ie. as the morphing difference between the two stimuli increased there was a release from the adaptation response). There appeared to be no significant interaction between the types of stimulus. In the case of attention to spatial position, responses increased as the spatial position level increased for both stimulus types (ie. as the spatial displacement between the two stimuli increased and attention was focused on spatial information). However, no spatial adaptation effect was observed in the face or voice ROIs in the cortex. The main effect of stimulus type occurred as expected (face and voice ROIs are more involved in processing face or voice features than spatial location), but although attention to spatial position increased the gain of neural responses in the ROIs, it did not change the selectivity of these responses.
The results of Experiment 2 showed that attention to voice identity whilst ignoring stimulus loudness increased neural selectivity. The proportion of varying responses increased as the morphing differences increased only when the participant attended to and discriminated voice identity, but not loudness. This showed the authors that attention stopped the decrease caused by adaptation. However, there was significant interaction between task and identity levels which indicated a change in selectivity during attention to voice identity. Further analysis showed that attention to voice identity whilst ignoring loudness increased both gain and selectivity of the neural responses in the voice ROIs. Therefore, the subjects were able to discriminate identity and loudness. In this case, the proportion of different responses increased as the loudness level rose when subjects attended to loudness. Again, therefore, there appeared to be no adaptation effect for loudness. The significant interaction between task and loudness was driven by a quadratic trend and the authors interpreted this as although voice ROIs are insensitive to changes in noise intensity, attention to stimulus features including loudness moderated the selectivity of neural responses within the ROI. Further analyses showed a three-way interaction (attend identity, attend other – loudness or position, plus stimulus level). Therefore, the authors proposed that their experiments showed that the effect of focus of attention on adaptation responses is significantly different between their two experiments on auditory stimuli ie. gain effects from Experiment 1 and selectivity exhibited in Experiment 2.
In conclusion, Kikuchi and colleagues proposed that their experiments showed that neural populations responsible for visual and auditory features are separate and that voice and face ROIs will adapt to stimulus features and reduce their responses more to repetition of stimuli with similar features as a form of neural selection. Attention was found to amplify and counteract the adapted neural signal by a constant factor, ie. top-down attention affects adaptation depending on the level of feature-based competition. However, the authors also showed that attention does more than just affect neural gain in that it can also influence selectivity. In the case of auditory vocal feature and loudness, the voice sensitive cortex demonstrated no adaptation to loudness, but attention to voice or loudness alone led to a greater increase than expected by releasing the adaptation. It was therefore concluded that attention selectively boosts the reduced adaptation of stimuli that is crucial for discriminating differences in identity. The authors then went on to explain their findings in terms of mathematical models. They also considered in their discussion expectancy even though this was a feature controlled against in their experiments. (Expectation suppression is said to occur when a repetition is expected and when this is violated, an expectation prediction error results). Attention and expectation are believed to work together and can change feature selectivity, ie. can dampen or sharpen neural responses. Kikuchi and colleagues stated that further work is required on this topic even though in their experiments perceptual expectations were minimised by the test set-up. They reiterated to conclude their article the proposal that their experiments showed that the properties of the event that are attended determine whether attention in interaction with adaptation modulates the signal gain alone or also effects selectivity.
COMMENT
What makes this paper interesting is that confirms two observations associated with attention. The first is that deliberate focused attention on specific features can alter positively the extent of neural firing against the neurochemically expected reduction in firing associated with repeated events (Experiment 1). And the second is that it can also affect the selectivity of certain features where there is competition (Experiment 2). The experimental set-up of Kikuchi and colleagues to demonstrate these factors was quite simple: visual versus auditory stimuli in the form of faces and voices and hence, features associated with common person identity mechanisms; manipulation in the form of computer generated morphing of facial images and mechanical sound adjustment for changes to vocal characteristics and volume intensity; and fMRI imaging, processing and analysis of BOLD responses for quantitative assessment of the extent of neural firing. With such a set-up, the neurochemical mechanisms required included sensory input and processing and cognitive decision-making (eg. is this face different to the previous?), but did not require mechanisms involved with long-term memory formation and recall and those associated with emotions (eg. assessment of stimulus value). Both of which are also affected by attentional capability.
The first conclusion from Kikuchi and colleagues was that deliberate attention counteracts the neural firing decrease observed with adaptation, eg. their fMRI ´fatigue` model. This conclusion leads to a consideration of the effects of top-down attention (ie. conscious awareness and deliberate focusing of attention) versus subconscious/unconscious attention. Such a consideration already demonstrates the fact that attention and conscious awareness are not from a neurochemical perspective the same thing. Attention brings to and maintains objects in the working memory state whereas awareness brings objects into the personal SELF/cognitive thinking and processing arena. There can be consciousness without some form of attention eg. objects can be attended to that are perceptually invisible and vice versa there can be top-down attention without consciousness eg. one can become conscious of isolated objects or a gist of a scene in the virtual absence of top-down attention (van Boxtel). This is because attention itself can come in different forms aside from those relating to emotional status. The ´thinking` form, top-down attention means personal, cognitive involvement and requires frontoparietal connectivity and alpha brain wave firing synchrony (van Schouwenburg). Greater levels of attention mean greater awareness of selected objects, but it does not mean greater awareness of distractors (Oriet). Therefore, it results in memory benefits not just by reducing the cognitive load, but also by enhancing the level and quality of the task-relevant information. On the other hand, unconscious attention means that characteristics of an event may be attended ie. inputted and processed, but they do not reach conscious awareness. Top-down attentional modulation may bring those features to conscious awareness and in doing so, the information normally becomes attended.
This is where the connection to adaptation of neural firing due to repeated input comes in. Attended information relating to a single event involves only features that occur within a temporal window and there may be a shift from unattended information to attended in order to expand the neural representation to optimise processing. Repeated firing of the same cells to maintain this neural representation leads to the cells going into their refractory periods in order for them to neurochemically recover and therefore, firing adaptation (the fMRI ´fatigue` model) is said to have occured. This results in other cell firing to dominate in this particular temporal window and hence, the neurochemical ´priority to the unattended` and ´inhibition of return` rules apply. Other features that may have been unattended during the initial firing phase may then reach conscious awareness through being attended. This is a natural process and is different to deliberate selection and suppression of feature input as carried out in Kikuchi and team`s experiments where the participants are told specifically to select or ignore certain sensory features. In this case, the instructions lead to top-down modulation of attention brought to particular features and therefore, the neurochemical firing returns to the cells representing these original features once their neurochemical recovery is completed in order that the event representation is restored. In this case, as described by Kikuchi, neural adaptation (ie. ´priority to unattended` and ´inhibition of return`) is overruled.
The second observation about attention confirmed by Kikuchi`s experiments is that attended stimuli and features compete and that focused attention can bring about selectivity of features (Kauramaeki). In Kikuchi`s study the team found that attention to voice/face with the instruction to ignore location leads to increased face/vocal responses; attention to loudness with an instruction to ignore voice leads to increased loudness; and attention to voice with an instruction to ignore loudness leads to increased selectivity, ie. voice. From a neurochemical point of view, in a natural situation it is clear that the strongest firing would come from the most dominant event feature and/or if appropriate the feature that is task-relevant. Neuronal populations responding to sensory input are localised separately and therefore, the event representation can consist of many independent characteristics eg. two primary features of auditory perception, pitch and timbre, are processed in overlapping auditory cortex regions, but are separable (Allen). The event representation may also contain other information eg. auditory processing has been shown to encorporate information not only relevant to sounds such as frequency, level and amplitude modulation over time, but also unrelated aspects of the experience such as arousal level, past experience and motor planning (King). Top-down modulation of sensory input and processing by deliberate direction of attention would shift the normal selection of features, ie. a strong visual input may be suppressed by being told to ignore visual features and concentrate on the loudness of the voice. It would result in the natural order of competition relating to strength of firing to be overruled and is likely to involve suppression at the primary cortical level responsible for the sensory input (Jacob). Such a direction has been shown to be advantageous in general in some cases. For example, it has been found that: spatial attention brings about an increase in visual processing (Connor); sustained attention on one auditory object in a complex scene leads to improved attentional selectivity over time (Best); attention will expand the attended features at the cost of the unattended to optimise processing (Cukor, Kuo); and the precision of the feature itself can be increased (Lim – deliberate attention to a auditory feature leads to decreased working memory load and increased task-relevant feature precision; Andersen – selection of one visual feature will optimize the firing of another; and Bartsch – sharpened selectivity for a particular colour arises from feedback processing at the V1 level). This correlates to the fMRI ´sharpening` model where changes in selectivity or timing increases some neuronal responses whilst decreasing others.
Therefore, we conclude that deliberate attention is just as important as attention elicited through bottom-up mechanisms and influenced by the emotional state. It too provides focus, timing, and conflict assessment to sensory input and working memory content and information processing. Although deliberate attention can have a negative effect on task performance, eg. by leading to focussing on task-irrelevant information, its value is that it can have a positive influence on feature strength and feature selectivity. Therefore, the amount of task-relevant information can be increased, distraction can be prevented and task performance aided. The use of deliberate attention can counteract not only the natural neurochemical mechanisms associated with firing adaptation, but also the defective personal cognitive capabilities associated with inappropriate object value, memory impairment as well as the cognitively detrimental physiological factors such as tiredness, glucocorticoid level and hormonal status.
Since we`re talking about the topic…………………………….
….. Blair found that the ability to decrease task irrelevance is suppressed with age and Gazzeley found that not being able to decrease task irrelevant information leads to a detrimental effect on working memory performance. Can we assume that if Kikuchi and team`s experiments were repeated using participants of a senior age that although firing of the ROIs would be located the same as their younger counterparts the intensity of the firing and the specificity would be lower? In a similar manner, Weltman and Wegbreit for example showed that anxiety like ageing decreases information relevance. Therefore, would the same type of experiments as indicated above produce the same effects as ageing?
….. Caeyenberghs showed that adaptive working memory training led to improvement on working memory tasks and generalization to tasks of reasoning and inhibition. Would a period of attentional training before taking part in the experiment promote an increase in the fMRI BOLD responses, increase the speed of responses and/or improve task performance as expected if working memory capability is positively influenced by the training?
…. Carlson found that fearful facial expressions conveyed threat-related information and automatically captured spatial attention through increased attentional orienting to threat. Therefore, if Kikuchi and colleagues´ experiments were repeated but the faces had fearful expressions and the voices projected fear, would there be a change in the firing patterns observed and in the ability to follow the instructions to ignore certain stimuli?
…. Verleger and Garcia-Perez found that normally people are equally aware of events in both hemispheric fields, but when two streams of stimuli are rapidly presented left and right containing two targets, the second target is better identified in the left than in the right visual field. If Kikuchi and colleagues` experiments are repeated with the first stimuli placed in both hemispheres, but the second in either the right or left, would the fMRI images observed be affected?
neurochitchat 