Posted comment on ´Fast and slow rhythms of naturalistic reading revealed by combined eye-tracking and electroencephalography` written by L.Henke, A.G.Lewis and L.Meyers and published in Journal of Neuroscience 2023 vol. 43(24)p. 4461 doi 10.1523/JNEUROSCI.1849-22.2023
SUMMARY
Henke, Lewis and Meyers report in their article their findings relating to rhythmic eye movements and brain activity relating to reading.
The authors used the Zurich Cognitive Language Processing Corpus (ZuCo) which involves simultaneous eye tracking and EEG recordings with naturalistic reading. Twelve participants (average age about 37) performed a natural reading task (original task 2 from the corpus) which was analysed for analytical simplicity and naturalness. This involved 500 sentences to be read and then a comprehension question which had to be answered after some of them (mean accuracy approx. 88%). For 10 participants, the sentences were presented in 6 blocks with recalibration of the eye-tracker in between the blocks. For two participants one block was omitted (249 of 250 red sentences) and in the remaining data of all participants, nine sentences with missing eye-tracker data were excluded.
The measurements of the eye movements were made using an infrared video-based eye tracker at a sampling rate of 500 Hz and EEG data was also acquired at 500HZ. Data was bandpass filtered from 0.1 to 100HZ. Eye-tracking analysis was carried out on fixations and saccades that landed on words. The fixations not associated with reading were removed by preprocessing and the assignment of fixations to the text was improved using a Gaussian Mixture Model. This model was also used to saccade landing positions. Fixations greater than 1000ms or less than 60ms and saccades greater than 80ms were removed by the authors. The saccades and fixations were converted into time series in order to analyse potential periodicity of eye movements during reading. The saccades binary time series was sampled at 1000HZ set to 1 at saccade onset and 0 at all other time points and fixations time series were calculated by the authors using the assumption that chunk endings were accompanied by abrupt changes in fixation durations. The initial fixation after each backward saccade was excluded and then the difference between every fixation (n – each having a mean duration of 220ms) and the fixation directly preceding it was computed (y value). This y value was plotted against the onset time point of fixation-n as the x value and the vector linearly interpolated to a sampling rate of 1000HZ to give a continuous time series.
Spectral analysis was carried out by Henke, Lewis and Meyers using Welch`s PSD with different window lengths. This was carried out to optimise analysis of higher frequencies for saccades and analysis of lower frequencies for fixations. Overlap was given as half of the window length. The window size was increased to accommodate for periodicity within the expected frequency. Then, the mean of the observed spectral variance as an estimate for the effect size was compared against the mean of the spectral variance of a surrogate distribution (raw fixation durations were shuffled whilst keeping the original fixation time points) based on 1000 permutations of each time series. Saccades were permuted by shuffling the binary vector in time and then observed and permuted values were then averaged over participants (statistical significance at over 95% value difference between observed and permuted). Multiple comparisons at different frequencies were controlled for by carrying out a False Discovery Rate (FDR) correction. For the combined analysis with EEG data, the time series were down-sampled to 500HZ.
For the EEG analysis, Henke, Lewis and Meyers used 104 EEG channels and 9 electro-oculogram (EOG) channels for the regression of eye movements. Various adjustments were made to the data obtained. For example: bad channels were removed (flatline for greater than 5 secs); residual drift was removed by high pass filtering at 0.1HZdata; line noise was removed by notch filtering at approx. 50HZ; eye movements were removed by linearly regressing the EOG channels from the EEG channels; and automatic artifact rejection was carried out by using MARA on 1HZ filtered data. Then the data was synchronised with the eye movements using EEGLABs EYE-EEG toolbox based on shared events across acquisition modalities. Confirmation of alignment between eye movements and eye tracking data and whether the periodicity of eye movements during reading was associated with rhythmic neural activity were carried out. The former was carried out by correlating vertical and horizontal EOGs with the corresponding gaze position and the latter was carried out by investigating the phase coherence between the values. Appropriate statistical analyses were carried out on all data.
The first set of results reported by Henke, Lewis and Meyers related to the eye-tracking analysis. The authors found that saccades had significant spectral peaks at 4.4HZ and 10HZ with a fixation duration difference time series showing peaks at around 0.49HZ and 4.3HZ. Henke, Lewis and Myers then went on to describe their EEG findings. The intersaccade phase coherence EEG data around saccade onset (ISPC) analysis showed a spatially broad cluster (statistically significant – peak over posterior EEG sensory around entire time window of around plus/minus 200ms around saccade onset, frequency range 2.5 – 9.5HZ). The coherence analysis on the fixation duration difference time series carried out gave a significant cluster over the posterior EEG sensors from 1 to 2.5HZ (cluster sum t(11) equals 528,201; peak coherence at electrode E83 at 4HZ, 130ms after saccade onset FIG3AB) and another one from 3.5 to 5HZ (cluster sum t(11) equals 334.60, maximal coherence at electrode E90 at 4.5HZ). The authors concluded that this observed coherence over the posterior sensors did not likely stem from only the muscular activity of the saccades and fixations because preprocessing had been carried out that removed the activity from the anterior EEG sensors.
The next set of results reported related to the processing of multi-word units seen through periodic changes in fixation durations with EEG coherence. Henke, Lewis and Meyers used phase clustering at sentence boundaries to indicate chunk endings. They found that phase angles of fixation duration differences at the last word of a sentence did not show significant non-uniformity. The authors suggested that this could have resulted from clustering out of randomness of the permuted data due to sparse sampling of the data (ie. one fixation per word). However, there was significant difference from uniformity of the distributed phase angles of the low pass filtered EEG at sentence endings at electrode E83 (has maximal coherence with the fixation duration difference). Henke, Lewis and Meyers then assessed post hoc whether the phase clustering was restricted to the chosen electrode showing the maximal coherence by analysing separately all the electrodes within the significant cluster of coherence at the low frequencies. It was found, following a FDR correction for multiple comparisons over electrodes) that 53 from 68 electrodes within the cluster showed statistically significant phase clustering.
The report of the findings led Henke, Lewis and Meyers onto discussing what impact their results had on views about eye movements during naturalistic reading. They started their discussion with the finding that eye movements were found to show periodicity synchronised with phase clustering of the EEG oscillatory activity above the visual cortex. The synchronicity occurred in two distinct frequency bands: the faster, analogous to the multiplexed auditory sampling of speech possibly reflecting the active sampling of words; the slower possibly reflecting the endogenous chunking mechanism integrating words into larger multi-word units. Hence, the authors proposed that readers actively align sentence endings to specific phase angles of neural oscillations that subserve linguistic chunking. A peak approx. 4HZ in the fixation duration differences and their coherence with EEG was also found. Therefore, this was suggested as mirroring the saccadic rhythm. Fixation duration differences are inserted at fixation onsets each following a saccade with rhythmic electrophysiological activity being an endogenous pacemaker for reading since text does not provide temporal information. The association between saccadic rhythm and comprehension success could not be ascertained by the authors since their experimental set-up did not include word-by-word analysis.
Henke, Lewis and Meyer continued by discussing the approx. 5HZ saccade rhythm observed during reading. Research had already shown theta band oscillations in auditory processing are possibly linked to syllable sampling and therefore, the authors suggested that the theta band oscillations observed could provide optimal sensitivity for word processing during reading. This was indicative of visual attention with each saccade bringing new letters into attentional focus. It has also been suggested that theta band oscillations can modulate saccades optimising input gain by proactively increasing neuronal excitability to amplify upcoming stimuli. The saccades were suggested as possibly aligning the fixations (the times at which information uptake could occur) with time points of optimal sensitivity (ie. a particular phase of the cycle). This ´pacemaker` type functioning suggested was said to be also supported by research showing that the frequency of visual sampling is independent of spatial selection.
One point made by Henke, Lewis and Meyers was the influence of language on fixations. Their study was carried out using the English language where letter-to-sound associations are described as non-transparent resulting in longer fixations. Research by others had shown that in a selection of 14 languages, saccadic periodicity was seen to range from 3.9 to 5.2HZ which suggested an effect of linguistic processing that was more than just perceptual sampling. That research and the observation that saccades during reading are found to be influenced by internal linguistic and cognitive factors led the authors to suggest that future work should investigate whether language itself causes differences in the endogenous rhythms observed.
Henke, Lewis and Meyers then went on to discuss the changes in fixation duration exhibiting periodicity and synchronicity with the EEG within the delta band. This phase clustering at the sentence endings was interpreted as readers actively sampling larger units at their preferred electrophysiological processing rate. Therefore, the delta band was said to serve the active segmentation of speech into multi-word chunks. This view was said to be supported by work by others such as: chunk size may be limited by the delta band oscillations consistent with a temporal limitation of multi-word units; readers can impose implicit segment boundaries to facilitate integration; and slowdowns as shown by the authors fixation duration differences may reflect the readers own self-imposed limitations. However, an alternative explanation was put forward in that periodic chunking could instigate proactive allocation of attention. In this case, direct attention to critical information in the stimulus could be obtained by the alignment of delta band cycles to the chunks. Therefore, future research ideas were suggested by Henke and team such as investigating the behavioural impact on information integration during reading and active sampling of entire sentences.
Having described how theta and delta brain wave oscillations serve the different aspects of reading, Henke, Lewis and Meyers went on to state that their observations were not sufficient to claim causality of endogenous activity for reading although it was supported by evidence from others that the practice of reading developed in the presence of those two wave oscillations. They went on to state the limitations of their experiments which would need further study. For example: low spatial resolution of the EEG meaning that more specific cortical substrates could not be identified; lack of proof that the observed rhythms did reflect oscillatory dynamics or a sequence of evoked responses especially since there was no explanation for their periodicity nor temporal regularity of reading; and that there was no specific periodic change in visual input or motor activity related to the chunking rhythm.
However, in spite of the limitations of experimentation given, Henke, Lewis and Meyers continued with their view of the endogenous rhythm demonstrated during naturalistic reading by the delta and theta brain waves which were said to be associated with information selection and structuring. The authors said that their observation of synchronicity between eye movements and EEG could relate to inner speech produced during reading. Evidence for this association was given by reading direct but not indirect speech quotes being linked to increased phase-locking at the theta band frequency. This was additionally modulated by a verbal description of speaking rate and suggested that inner speech is influenced by contextual, meta-cognitive and/or linguistic factors similar to saccades during reading. However, Henke, Lewis and Meyer counteracted this argument by saying that readers are found to generate implicit prosodic contours which may be a mechanism of speech production system to assist the formation of syntactic structure. Examples of this were given as the sentence-final wrap-up effect in reading and pausing and clause-final lengthening in speech that were suggested to reflect the insertion of implicit prosodic boundaries. However, Henke and team did say that these examples may reflect instead a chunking process for information integration. This view was supported by evidence from others that reading involves word skipping and regressive eye movements which indicates that linguistic input is sampled in a sparser way than speech. Speech sampling was said to occur at the rate of syllables giving a different amount of linguistic information per sample whereas sparse sampling would still allow for extracting all necessary information as parafoveal processing allows for accessing a word even before its fixation. Also, reading samples at one fixation per word although parafoveal processing can gather additional information. Therefore, Henke and team confirmed that speech processing is an integral part of reading since during reading linguistic information is extracted by the readers and mapped onto speech. Impaired reading was found by others to be linked to impaired neuronal synchronisation to speech and that individuals suffering from dyslexia showed that: developmental dyslexia is associated with impaired tracking of speech and nonverbal auditory rhythms; and non-rhythmic eye movement patterns are displayed that have an increased fixation frequency and longer fixation durations. Therefore, Henke, Lewis and Meyers suggested that their future work based on their findings of synchronicity of eye movements and neural oscillations shaping reading could benefit this group of individuals.
COMMENT
What makes this article interesting is that it again shows a link between specific brain wave frequency and particular neuronal and cognitive functioning. The prime example of this link is that seen in stages of sleep and information processing and brain memory storage, but here in this article, Henke, Lewis and Meyers show another aspect of brain waves where in this case they are associated with eye movements and specific phases of reading. Henke and colleagues carried out a set of experiments where they performed measurements of eye movements and EEG during a reading task. They found that eye movements during reading exhibited rhythmic patterns which were wave frequency selective. The periodicity of the eye movements above the visual cortex showed two distinct frequency bands, theta and delta. The whole head theta brain wave activity was said to correlate to saccades at a frequency of 4-5HZ with display coherence 130msec after cascade onset. This type of activity is already known to be associated with auditory tracking syllables, is synchronous with lip movements and sign language and guides temporal prediction of speech and for their study set-up Henke and team described it as being associated with word-locking with active sampling of words. Henke and colleagues also suggested that the activity was linked to visuospatial attention changes independent of external cues with the purpose of visual optimisation for processing of subsequent stimuli. With regards to the delta activity, Henke and team associated this with fixation durations of frequencies of approx. 1Hz. This was found to be in coherence with occipital area activity and is already known to be involved in the tracking of periodic phrases, synchronisation with lip movements and sign language. Henke and team found that the delta brain wave activity is found to be phase locked to sentence endings and therefore, the team proposed that individuals actively align sentence endings to specific phase angles of neural oscillations that subserve chunking. Therefore, individuals when reading can impose segment boundaries to facilitate integration and can instigate pro-active allocation of attention to important information contained in the chunks inputted by the visual system.
Therefore, what can we deduce from their findings? The first deduction is that brain waves mean groups of synchronised neuronal cell firing are associated with initial reading phases. What does this mean? Brain waves are neural oscillations of particular frequencies, eg. delta, beta, gamma found in parts of the brain. The oscillation refers to the rhythmic and/or repetitive electrical activity of groups of neurons and are generated representing an endogenous rhythm, a rhythm that is generated spontaneously or in response to stimuli. In the latter, activation of neuronal cells leads to action potentials when firing reaches the threshold. However, action potentials are followed by periods of cell recovery (the refractory period) where ´household` functioning takes place so that the cell can fire again. Synchronicity of firing means that group action functioning could result and/or there could be a transmission of firing signal leading to the formation of pathways if more than one area is involved. This is how brain cell mechanisms are linked to cognitive functioning and it could involve multiple areas such as that seen in memory formation for example, or word processing. The value of synchronised firing of neural oscillations comes from it strengthening firing and therefore, when it occurs in response to active stimulation of cells then neural representations will give priority to those features that cell group represents. This is important for example in decision-making where an option chosen depends on the greatest strength of firing shown by a particular neuronal cell group amongst others.
Therefore, can we deduce from Henke`s findings of two types of brain wave frequency in reading that there is separation of function and/or brain area? As said above, Henke found eye movements during reading associated with two rhythmic EEG patterns above the visual cortex, that of theta and delta band activity. (It should be noted here that although Henke attributes these to visual input only, there must have been implicit word identification (understanding of meaning and context) as well since the participants of the experiment were told that possibly a comprehension test would follow at the end of the test session. This might have influenced the brain wave frequency observations although this would be easy to confirm.) The theta activity observed was said to be saccades at 4-5HZ frequency with display coherence 130msec after cascade onset. This was said to be involved in word-locking and active sampling of words. A further explanation was given in that it was also said to be associated with visuospatial attention changes which would optimise the input of information. The delta brain wave activity observed was associated, according to Henke, with fixation durations of approx. 1Hz. This activity was found to be in coherence with occipital area activity and was found to be phase locked to sentence endings. Therefore, Henke and team said that individuals could actively align sentence endings to specific phase angles of neural oscillations that subserve word chunking. Therefore, the imposition of segment boundaries facilitate integration and instigates the proactive allocation of attention to important information contained in the chunks being observed.
Therefore, how do Henke and team`s findings fit in with what is known about brain input and functioning related to reading? First of all, it should be noted how important language is not only to the individual, but also to society. In the case of the individual, then language gives the individual the ability to think and process information for example as well as within oneself using inner speech, it can be used to spur, motivate and encourage behaviour. Language is also used to represent things that are not actually existing in real-time. For the society, language is used as a communication tool such as communicating ideas, thoughts, plans, warnings and expression of emotions. Again, it can be used to communicate something that does not exist in real-time or perceived by sensory systems. Therefore, there is a need to have good language skills and these skills, both vocabulary and grammar, have to learnt. The skills of speaking, reading, writing and listening are applied to the four areas of language described by Shaffer (1993) and these are: phonology – the sound system of language; semantics – the meaning conveyed by words and sentences; syntax – the set of grammatical rules indicating how words may or may not be combined to make sentences; and pragmatics – the principles determining how language should be modified to fit the context.
Unlike some motor skills, eg. riding a bike, language skills demand life-long learning and adaptation since language is ever-developing, for example as seen with the rise in the past decades of vocabulary connected to the computer/mobile/digital world. Therefore, it is suggested here that language should be considered as a ´tool` like attention and requires a number of basic components to make it work, eg. sensory input, information processing, memory storage and recall as well as general skills of concentration, ability to focus and process information. This means that a large number of brain areas are involved in the language capability. For example, in general:
Right brain hemisphere cortical function – left side of body, left field of vision – focuses on colour/tones, intonation and emphasis – shape such as spatial skills, spatial orientation, sense of space – movement includes spatial skills, spatial orientation, rhythm, imagination, day-dreaming – language such as general aspects of speech, intonation and emphasis) includes Wernicke area. Damage causes effect on words but has no effect on grammar.
And left brain hemisphere cortical function – right side of body, right field of vision, focuses on details, sequences and logical tasks (logic numbers, sequences, linearity analysis), language (grammar and word production includes Broca region which is linked to Wernicke area). Damage in left temporal lobe and Broca area as seen with subject Tan who suffered with conduction aphasia (could understand, but not speak); left hemisphere temporal lobe damage leads to decreased verbal performance.
Activity is also identified in areas related to specific and other aspects of reading and these include:
Prefrontal cortex – decision-making, working memory, attention
Left inferior frontal gyrus (Broca) – speech comprehension (damage to angular gyrus – loss of ability to read, as seen with Alzheimer patients)
Temporal areas – temporoparietal circuit – working memory, attention
Ventral temporal areas – WHAT P pathway – object recognition with origin in occipital cortex V1 (damage – apperceptive agnosia – can describe features but not name or function)
Parietal areas – working memory, attention
Dorsal parietal areas – WHERE M pathway –– perception/action – origin V1 occipital cortex (damage to right – cannot remember complex shapes)
Left lateral inferior parietal lobule – highly sensitive to recency, not repetitive input (Buschbaum)
Left superior parietal lobule – spatial orientation, sensory information from hand (writing) visual input received
Left ventral lateral occipital complex, right ventral lateral occipital complex – visual information
Hippocampus – working memory, decision-making
Superior colliculus – attention, eye movements
And cerebellum activity – motor sequencing corresponding to mouth movements during mouthing.
Therefore, just like with other cognitive skills, language skills develop with increasing age and increasing cognitive capabilities. For example, listening to the spoken word develops before reading and writing. Restricting this comment to reading, language development in the early stages, requires young individuals to have the visual ability to recognise the presence of shapes, patterns, spaces etc. and with practice and learning, these visual shapes are associated with auditory sound and meaning. Therefore, the mechanism behind this early stage of language development means that shapes and sounds are linked to definitive information. Sequences of auditory information (sounds) and visual information (syllables) are learnt according to motor sequencing rules and linked to other sensory information (real-time or stored) to add meaning. Practice equivalent to motor sequencing practice leads to expertise due to strengthening of these neural connections. The later stages of language development where language is pro-actively used, requires a familiarity with syllables, words and grammar as well as an appreciation and perception of incoming visual information. The inner voice or audible interpretation of shapes may also be required. This stage also requires efficient memory recall via activation of the relevant neuronal cell assemblies.
Therefore, if we consider the mechanisms required for reading, these can be divided roughly into two groups: sensory visual input and learning for novel words; and sensory visual input and recall for known words. The visual system of the reader is required to distinguish the stimuli presented and this is, for readers of the English language, a system of lines, curves and dots that make up the letters of the Germanic language and an appreciation of spaces. These are then grouped into syllables, then words and sentences which must be visible to the reader. Readers must also be capable of moving the visual apparatus from left to right (English text goes left to right) and top to bottom (direction of text on a page). Therefore, activation of the visual system in reading a word begins with a group of letters being placed in the visual field. The visual field size depends on the individual and can be more than one word, ie. words separated by spaces and/or punctuation.
From a neuroscience perspective, the input of the information begins with the eye with cells activated at this lowest level specific for the most characteristic features (the reference points – letters, syllables). The signal then goes up the visual pathway with convergence of the output from lots of photoreceptors onto fewer bi-polar cells leading to excitation or inhibition of retinal ganglion cells. Activation of these cells then leads to direct stimulation of the brain´s superior colliculus area or activation of the optic nerve. Stimulation of the latter leads to activation of the optic chiasm, then optic tract, which activates the thalamus or pineal gland. From the thalamus, parts of the visual cortex and higher cortical areas (the Wernicke area and Broca area are the cortical areas supposed to be associated with language) are stimulated and therefore, the incoming visual features are represented by a group of firing cells (the neuronal cell assembly).
In order that future encounters of this group of letters is recognised as a word and/or with meaning, the visual information has to be stored. It is the end-of the road cortical cells (neuronal cell assembly) fired according to the complexity of the image observed that will form the first temporary sensory stores. By keeping the object within the visual field (fixation), there is sustained activation of the relevant cells/group of cells and pathways so that the temporary stores are shifted to the more stabile visual short-term memory stores (last up to 10 minutes). On average 5 to 9 pieces of information can be held in the short-term memory and repetition or rehearsal of these pieces of information will lead to consolidation into long-term memory (lasts longer than 30 minutes). This involves physiological changes to the cells, eg. the modification of gene expression of pyruvate dehydrogenase complex along the firing route so that re-activation results in stronger firing.
Several peculiarities of the storage system of words appear to exist. It has been suggested that since the storage of the visual and auditory information relating to words occurs with concurrent early attempts at mimicking the sounds heard, that along with the long-term visual and auditory memories are also memories of a series of muscle movements of the jaw, position of the tongue and sounds expressed, reminiscent of the word if spoken. Therefore, in the case of reading, words are learnt by the formation of long-term memories of the sensory information (visual) experienced and the corresponding muscle movements relating to loud speech, inner speech or mouthing. It should also be noted that since recall of written words is independent of several things such as writing styles, typefaces, letter size and capitals that the brain memory storage process does not regard these as core features and probably only generic shape and form are stored.
The second group of mechanisms required for reading is for known words and this involves sensory visual input and memory recall. It is responsible for word identification and the association of meaning leading to comprehension. The measurement of brain waves in Henke, Lewis and Meyers experiments did not cover this area, but the effect of implicit firing associated with it probably cannot be ruled out since the participants were warned that a test on comprehension would be made. However, since only eye movement measurements were made by Henke and research methods associated with reading and word identification were not (eg. lexical decision task, naming task) it has to be assumed that the reading phases studied by Henke and team were mainly confined to visual system word fixation and shifting to new word/groupings (saccades). However, visual presentation of the letters/words would have led to unconscious processing and memory recall processes would have been activated. In this case, memory recall would mean that exposure to the same visual information, ie. seeing the letters/word would lead to neuronal firing along the same firing pattern originally experienced. However, since the previous exposure had made these neuronal cells supersensitive, the response would be quicker. The firing of the neuronal assembly cells in the cortex, recreates the characteristics that caused firing of the cells at the sensory level in such a way that the person re-experiences the event or object and remembers the word.
Therefore, with reference to Henke, Lewis and Meyers experiments and findings the mechanisms involved in the performance of their reading task would be associated with visual input plus possibly implicit word identification and meaning. More specifically, the method used, ie. recording of eye movements would allow the fixation of the visual input sample via the lack of eye movement and the shift of visual input sample via the change in visual field (the eye movement itself). Eye movement studies are commonly used for this type of study since reading is known to involve rapid jerking movements across the page (saccades) and the method is unobtrusive and provides a detailed online record of the reading processes. Using this method, Henke, Lewis and Meyers identified two brain wave bands associated with two phases of the visual input associated with reading and these were:
Delta brain waves associated with fixation durations of approx. 1Hz and found to be in coherence with occipital area activity – shown to be phase locked to sentence endings and hence, linked to ´chunking` of the visual information;
And theta brain wave activity associated with saccades at 4-5HZ with display coherence 130msec after cascade onset – shown to be word-locked with active sampling of words.
In the case of delta brain waves, chunking can be considered as a ´snapshot` of the visual field and it allows binding of visual information within a neuronal cell grouping (the neuronal cell assembly). This concept is common and allows multiple visual features of a single event to be bound together in a common group of actively firing cells. In the case of reading, letters make up words which are separated by gaps and/or punctuation and therefore, the eye focuses on the text and in any one single time period (the fixation period), the features of that field activate the necessary neuronal cells. According to Henke and team, this fixation period of the eye movements was associated with brain waves of frequency of approx. 1HZ and identified as delta brain waves. Henke found that the delta waves were phase locked to sentence endings and therefore, stated that individuals actively align sentence endings to specific phase angles of neural oscillations that subserve the chunking of the text presented. This would mean that specific boundaries are imposed and attentional resources can be allocated to the content of the chunk.
The idea of fixations, visual information chunking and eye movements controlling them is not new. It is known that the size of the perceptual span (the field of view) means that parafoveal information (that coming from the surround of the central or foveal region) is used in reading and that there are three different perceptual span widths (Raynor). These are: total perceptual span – the longest; letter-identification span – the area in which information about letters is obtained; and word identification span – the shortest span and the area from which information relevant to word-identification processes. The perceptual span, although affected by the difficulty of the text and print size, often extends 3 or 4 letters to the left of the fixation and up to 15 letters to the right. This means that the most informative text lies to the right of the fixation point. This gives an idea of the perceptual span with regards to reading and can be compared with working memory of objects which is limited to 5-9 pieces of information.
Using the original E-Z Reader Model of the patterns of eye movements during reading (Reichle 1998), the content of the chunks in the field of view was investigated. Therefore, it was found that there was: a preference for the fixation of words with 80% of the content being nouns, verbs, adjectives; only 20% of function words (eg. articles, conjunctions, prepositions, pronouns) were fixated; words which were predictable in the sentence context were fixated quicker; and rare words were fixated for longer than common words. However, this model was developed further in 1998 to take into account time on shifts of focus. The original 1993 model said that readers fixate on a word until they have processed it sufficiently after which they immediately shift the field of view and fixate on the next word. Therefore, it would take 150-200 ms to execute an eye movement program. With this hypothesis, time would be wasted waiting for the eyes to process and then move plus words could not be skipped because there would be no knowledge of it until the next word had been fixated. Therefore, in 1998 Reichle argued that the next eye movement is programmed after only part of the processing of the currently fixated word has occurred. Therefore, efficiency would be improved since there would be a reduction in time between completion of processing on the current word and movement of the eyes to the next word. Readers would check the familiarity of the word currently being fixated and completion of the frequency checking of the word (first stage of lexical access) would be the signal to initiate the eye movement. Processing of this second word would immediately proceed and if processing occurs rapidly enough it would be skipped. Processing of the first word would continue (second stage of lexical access) and this would mean accessing the learnt semantic and phonological information until identification is made. Both first and second stages of lexical access would be completed quicker in the case of predictable words than for unpredictable (Reichle and supported later by McDonald 2003). Therefore, it was said that cognitive processes determine when to move the eyes whereas low level processes (eg. length of word) determine where to move the eyes. Further work by Reichle in 2003 was carried out identifying the brain regions responsible for the different reading fixation stages. For example: primary visual cortex – responsible for processing the features making up the orthographic form about 90ms after the word is fixated; left extrastriate cortex – responsible for the integration of individual letters; left extrastriate cortex and left inferior temporal gyrus – responsible for the assembly of the word`s orthographic form occurring 150-200 ms after fixation; and the left angular gyrus – responsible for the word`s phonological representation.
Therefore, it is plausible that fixation of letters/words in the perceptual span (field of view) occurs in reading and that chunking of visual information occurs just like object features so that further processing can occur in higher brain areas. However, Henke, Lewis and Meyer`s experiments went further and associated this stage with delta brain wave activity. This at first glance appears less convincing since higher levels of processing and memory activity are normally associated with the faster frequency waves of alpha, beta and gamma and not delta. Research mainly linked to sleep show that delta waves originating in the thalamus, are associated in Sleep Stage 3 with slow, large amplitude waves with some appearance of sleep spindles. This is mirrored in the nucleus reticularis thalami neurons (NRT) and thalamocortical (TC) neurons where during the slow, sleep waves there are high frequency bursts of action potentials mediated by low-threshold calcium spikes due to T-type Ca2+ channel activation and in the hippocampus, sharp -wave/ripple (SPW/R) complexes in sleep occur which are short episodes of increased activity with superimposed high-frequency oscillations. It is thought that the slow, large amplitude delta waves have the responsibility of coordinating interregional cortical communication so that particular cognitive/processing activity occurs in sharp bursts. Therefore, in the case of reading and saccades the delta waves may not be involved in the information processing of the word chunks but solely in maintaining the visual field.
And so, what about the appearance of theta brain waves? According to Henke and colleagues these brain waves are associated in the case of reading with the visual system saccades at 4-5HZ with display coherence 130msec after cascade onset. This fits in with other researchers also using eye movement studies. It is known that in reading the eyes move in rapid jerks (saccades) across the page separated by fixations as described above where the information is extracted from the text. This means that the saccade is associated with active sampling of letters/words/spaces which has been determined to be about 8 character/spaces long and take about 20-30 ms to complete. Henke`s work led to association of theta brain waves with these saccades. From a neurochemical perspective, it is likely that saccades occur because of action of the attentional system seeking to overcome the refractory periods of the firing neuronal cells representing the visual features of the fixation period by causing a shift to unattended features. This is known as giving priority to the unattended. Unlike normal visual features of objects where the shift can be to other features in a relatively ´random` fashion, it appears that in reading the visual field of view shift occurs in a particular direction (ie. to the right in reading the English language). It should be noted that eyes can move backwards in regressions (about 10% of all saccades), but this is probably top-down controlled.
Both Henke, Lewis and Meyers and Reichle (2003) suggested this association of visuospatial attention and saccades. Henke, Lewis and Meyers contribution to the knowledge was the association of theta brain waves to the saccade periods and Reichle contributed the different brain areas involved and their roles. For example: parietal area – the disengagement of attention once the current word is partly identified; the pulvinar nucleus of thalamus – shift of attentional focus forward resulting in the frontal eye fields and superior colliculus using some information (eg. word length) to start programming a saccade to the next word; and Wernicke area and various associative cortex regions – continued processing of the fixated word during the shift until the meaning is accessed.
Therefore, it is plausible that the saccades are associated to attentional focus shifts in field of view during reading. The association with theta brain wave activity is also likely since although theta brain wave is known to be associated with deep sleep, it is also linked to memory, perception and neural consciousness in the wake state. In the case of memory, it is associated with the transformation of working memory into long-term memory along with high frequency stimulation (Zhu, Wang) as well as remote memory recall involving the mPFC, hippocampus and entorhinal cortex (Vetere), and visual memory involving the V4 and PFC (Liebe). The link with ´shifts of focus` as in the case of saccades are confirmed by its involvement in working memory where shifts of focus according to strength of firing has to occur. For example, items in working memory have number and temporal order and prefrontal theta oscillations increase during temporal order maintenance whereas alpha brain waves increase over the post-parietal and lateral occipital areas for item maintenance (Hsieh); and theta activity is known to indicate the amount of interference by irrelevant information in the cingulate cortex (Staudigl).
Therefore, to summarise Henke, Lewis and Meyers work shows another example where brain wave activity is indicative of groups of synchronised firing neuronal cells associated with specific cognitive functioning. The eye movements measured and the observed simultaneous brain wave activities seen during the reading task set up by Henke and team show that low frequency delta and theta brain waves are associated with reading fixations and saccades respectively and it is likely that these waves are accompanied by the faster frequency gamma, alpha bursts for the pushes up the brain hierarchy in order that cognitive processing (eg. in reading word identification and determination of meaning) can be carried out. Although these latter waves are likely to achieve more research time and recognition than the slower delta/theta types, there is value in the knowledge that they are present in certain visual phases during reading since they exist for longer and are more easily measured by the method for example Henke and team used. Since changes in reading ability are observed with for example stroke and illness (Alzheimer sufferers have difficulty with irregular verbs which may reflect damage to areas of brain responsible for explicit memory whereas Parkinson sufferers have difficulty with regular past tenses, -ed endings, but not irregular ones and this may reflect damage to brain areas responsible for procedural memory) then this type of measurement would be suitable for diagnostic tests or assessment of disease progression.
Since we`re talking about the topic ……………………………
…………………if Henke and team`s experiments were repeated, can it be assumed that brain wave activity during visual saccades and fixations would also show as well expected alpha/beta/gamma type activities indicative of processing if the conditions of the experiment were altered to for example, make the participant read the words aloud, have to answer questions about the text, or use words with distinct emotional inferences?
……..since the administration of ketamine is known to lead to increased theta activity in the medial frontal cortex (Muthukumaraswamy) and administration of cannabinoids disrupts theta brain wave activity in the hippocampus (Robbe), would the pre-administration of either drug cause changes to attentional shifts (saccades) as demonstrated by Henke`s eye movement measurements?
……… the amount of text from which useful information can be obtained in each fixation can be studied by using the ´moving window` technique (Raynor). If this is used with Henke`s experimental set-up, would it be possible to show precisely when the switch to the next word occurs?
………sufferers of dyslexia demonstrate reading difficulties which have been associated in some cases with attentional problems (Miles). Could Henke`s experimental set-up be used as a method for gauging success or progress as individuals undertake therapies for their dyslexia?
neurochitchat 