Distinct neural processes are thought to support the retrieval of semantic information that is (i) coherent with strongly-encoded aspects of knowledge, and (ii) non-dominant yet relevant for the current task or context. While the brain regions that support readily coherent and more controlled patterns of semantic retrieval are relatively well-characterised, the temporal dynamics of these processes are not well-understood. This study used magnetoencephalography (MEG) and dual-pulse chronometric transcranial magnetic stimulation (cTMS) in two separate experiments to examine temporal dynamics during the retrieval of strong and weak associations. MEG results revealed a dissociation within left temporal cortex: anterior temporal lobe (ATL) showed greater oscillatory response for strong than weak associations, while posterior middle temporal gyrus (pMTG) showed the reverse pattern. Left inferior frontal gyrus (IFG), a site associated with semantic control and retrieval, showed both patterns at different time points. In the cTMS experiment, stimulation of ATL at ∼150 msec disrupted the efficient retrieval of strong associations, indicating a necessary role for ATL in coherent conceptual activations. Stimulation of pMTG at the onset of the second word disrupted the retrieval of weak associations, suggesting this site may maintain information about semantic context from the first word, allowing efficient engagement of semantic control. Together these studies provide converging evidence for a functional dissociation within the temporal lobe, across both tasks and time.

Deteriorated phonological representations are widely assumed to be the underlying cause of reading difficulties in developmental dyslexia; however, existing evidence also implicates degraded orthographic processing. Here, we used event-related potentials whilst dyslexic and control adults performed a pseudoword-word priming task requiring deep phonological analysis to examine phonological and orthographic priming, respectively. Pseudowords were manipulated to be homophonic or non-homophonic to a target word and more or less orthographically similar. Since previous ERP research with normal readers has established phonologically driven differences as early as 250 ms from word presentation, degraded phonological representations were expected to reveal reduced phonological priming in dyslexic readers from 250 ms after target word onset. However, phonological priming main effects in both the N2 and P3 ranges were indistinguishable in amplitude between groups. Critically, we found group differences in the N1 range, such that orthographic modulations observed in controls were absent in the dyslexic group. Furthermore, early group differences in phonological priming transpired as interactions with orthographic priming (in P2, N2 and P3 ranges). A group difference in phonological priming did not emerge until the P600 range, in which the dyslexic group showed significantly attenuated priming. As the P600 is classically associated with online monitoring and reanalysis, this pattern of results suggest that during deliberate phonological processing, the phonological deficit in reading may relate more to inefficient monitoring rather than deficient detection. Meanwhile, early differences in perceptual processing of phonological information may be driven by the strength of engagement with orthographic information.

Often, as we read, we find ourselves thinking about something other than the text; this tendency to mind-wander is linked to poor comprehension and reduced subsequent memory for texts. Contemporary accounts argue that periods of off-task thought are related to the tendency for attention to be decoupled from external input. We used fMRI to understand the neural processes that underpin this phenomenon. First, we found that individuals with poorer text-based memory tend to show reduced recruitment of left middle temporal gyrus in response to orthographic input, within a region located at the intersection of default mode, dorsal attention and frontoparietal networks. Voxels within these networks were taken as seeds in a subsequent resting-state study. The default mode network region (i) had greater connectivity with medial prefrontal cortex, falling within the same network, for individuals with better text-based memory, and (ii) was more decoupled from medial visual regions in participants who mind-wandered more frequently. These findings suggest that stronger intrinsic connectivity within the default mode network is linked to better text processing, while reductions in default mode network coupling to the visual system may underpin individual variation in the tendency for our attention to become disengaged from what we are reading.

Grapheme-to-phoneme mapping regularity is thought to determine the grain size of orthographic information extracted whilst encoding letter strings. Here we tested whether learning to read in two languages differing in their orthographic transparency yields different strategies used for encoding letter-strings as compared to learning to read in one (opaque) language only. Sixteen English monolingual and 16 early Welsh-English bilingual readers undergoing event-related brain potentials (ERPs) recordings were asked to report whether or not a target letter displayed at fixation was present in either a nonword (consonant string) or an English word presented immediately before. Bilinguals and monolinguals showed similar behavioural performance on target detection presented in words and nonwords, suggesting similar orthographic encoding in the two groups. By contrast, the amplitude of ERPs locked to the target letters (P3b, 340-570 ms post target onset, and a late frontal positive component 600-1,000 ms post target onset) were differently modulated by the position of the target letter in words and nonwords between bilinguals and monolinguals. P3b results show that bilinguals who learnt to read simultaneously in an opaque and a transparent orthographies encoded orthographic information presented to the right of fixation more poorly than monolinguals. On the opposite, only monolinguals exhibited a position effect on the late positive component for both words and nonwords, interpreted as a sign of better re-evaluation of their responses. The present study shed light on how orthographic transparency constrains grain size and visual strategies underlying letter-string encoding, and how those constraints are influenced by bilingualism.