Composite targets in the Attentional Blink: New insights into the neural substrates of expectations, interference, switching & consolidation by means of MEG.

Mohammed, Sarah (2013) Composite targets in the Attentional Blink: New insights into the neural substrates of expectations, interference, switching & consolidation by means of MEG. PhD thesis, University of Glasgow.

Abstract

Top-down expectations allow us to quickly filter relevant events from the continuous stimulation by the environment. The Attentional Blink (AB) paradigm simulates such taxing situations in the laboratory (Raymond et al., 1992). For the first time, I employed composite events as targets in an AB paradigm. A first target (T1) was defined as 3 consecutive digits, while a second target (T2) was an ‘X’. In this way, unfulfilled expectations regarding T1 could be elicited by presenting partial targets, containing only 1 or 2 digits. My paradigm was specifically developed to allow the manipulation of global and local expectations independently. While global expectations are induced by relative trial frequencies and modulate target expectations prior to a trial, local expectations are induced by the consecutive presentation of digits that raise the expectation for a full target as a trial unfolds.

I predicted that, when expectation for the full target was globally high and locally raised to a maximum, disappointing these expectations would result in an AB in itself without a full T1 being presented (i.e. with 2 digits). Possibly, this AB could be even larger in magnitude than when expectations are met (with 3 digits). This was indeed the case when 3 digit trials were the most frequent event (Experiments 1, 3, 4 and 5). When global expectations for 3 digits were high, I observed the strongest AB with 2 digits, greater even than with the full T1. This suggests that maximally raised but violated expectations (with 2 digits) have a more detrimental effect on subsequent T2 processing than fulfilled expectations for the full target (3 digits).

I subsequently showed that this pattern could not simply be explained by enumeration of the digits (Experiment 2). Counting the digits was associated with the same blink magnitude with 1, 2 and 3 digits, suggesting that counting involves consolidation of digits into working memory as targets, a process unaffected by global expectation of trial types or the quantity of digits.

I then investigated the role of bottom-up signals in switching from raised expectations for the T1 to T2. I considered the roles of the post digit mask (Experiment 3), and of colour- (Experiment 4) and sound-based (Experiment 5) switching signals which marked the end of the digit event. Performance was improved considerably by removing the lag 1 mask after partial targets (1 digit and 2 digits), but not after full targets (3 digits). For partial targets, the presence of a mask seems to conflict with the expectations for another digit, and the knock-on effect could be a delay in switching to T2. The AB magnitude for full targets is unaffected by the mask, suggesting consolidation of T1 and switching to T2 occurs immediately after the 3rd and final digit. On the other hand, marking the final digit with a colour or sound did not influence the AB in any digit condition. Taken together, this suggests that in this paradigm, top-down expectations play a much stronger role than bottom-up factors.

I then manipulated the relative frequencies of the digit events to investigate how top-down global and local expectations contributed to the AB magnitudes. In Experiments 6 and 7, the 3 digit event became increasingly rarer as the 2 digit event became more frequent. Overall, I found a reversal of the previous pattern: The AB with 3 digits became increasingly stronger than the AB with 2 digits. My findings suggest that global expectations influence the build up of local expectation increments with each presented digit, yet this occurs in a non-linear fashion.

Two MEG experiments further show that raised and subsequently violated expectations for T1 have a distinct neural signature (MEG Experiment 1), which cannot be accounted for by the act of counting the digits (MEG Experiment 2). In particular, beta-band oscillations seem to code for both the target processing and the rapid changes in local expectations during the trial. A key finding from MEG Experiment 1 was the similar widespread beta power decrease for 2 digits and 3 digits until 0.6 s, when the conditions dissociated into a frontal power increase for 2 digits while the 3 digit power decrease continued. Tentatively, I suggest that this could represent the building of expectations for the full target (power decrease) and the subsequent violation of expectations (frontal power increase) in the 2 digit trials.

An M300 to each digit suggests that all digits enter the global workspace and are consolidated into working memory; a finding which is corroborated by the theta-band results. Meanwhile, alpha oscillations were more posterior and appear to code for the processing of full targets only (3 digits in MEG Experiment 1, and 1, 2 and 3 digits in MEG Experiment 2).

To summarise, my key finding is that an AB can be elicited without a T1 actually occurring. Building up expectations about the T1 event within a trial can alone induce an AB. Furthermore, built and subsequently violated expectations for the T1 event seem to be coded in the brain by widespread modulations of beta oscillations. I discuss the findings with relation to several models of the AB. In particular, my behavioural and MEG findings fit well with the “Robust State” hypothesis (Kessler et al., 2005a), set within the wider “Global Workspace” framework (Dehaene et al., 2003a), which accounts for the dynamic interaction of top-down expectations with bottom-up visual processing. I envisage this novel aspect of expectations complementing the growing knowledge of the AB and emphasising the importance of network dynamics as a metaphor for the parallel distributed processing in the brain.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Keywords: Attentional Blink, temporal attention, expectations, MEG, brain oscillations, global workspace, robust state hypothesis, working memory
• Subjects: B Philosophy. Psychology. Religion > BF Psychology; Q Science > QP Physiology
• Colleges/Schools: College of Science and Engineering > School of Psychology
• Supervisor's Name: Kessler, Dr Klaus
• Date of Award: 2013
• Depositing User: S Mohammed
• Unique ID: glathesis:2013-3907
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 29 Jan 2013 10:53
• Last Modified: 29 Jan 2013 10:57
• URI: https://theses.gla.ac.uk/id/eprint/3907