When briefly presented with pairs of words, skilled readers can sometimes report words with migrated letters (e.g., they report hunt when presented with the words hint and hurt). This and other letter migration phenomena have been often used to investigate factors that influence reading such as letter position coding. However, the neural basis of letter migration is poorly understood. Previous evidence has implicated the right posterior parietal cortex (PPC) in processing visuospatial attributes and lexical properties during word reading. The aim of this study was to assess this putative role by combining an inhibitory TMS protocol with a letter migration paradigm, which was designed to examine the contributions of visuospatial attributes and lexical factors. Temporary interference with the right PPC led to three specific effects on letter migration. First, the number of letter migrations was significantly increased only in the group with active stimulation (vs. a sham stimulation group or a control group without stimulation), and there was no significant effect on other error types. Second, this effect occurred only when letter migration could result in a meaningful word (migration vs. control context). Third, the effect of active stimulation on the number of letter migrations was lateralized to target words presented on the left. Our study thus demonstrates that the right PPC plays a specific and causal role in the phenomenon of letter migration. The nature of this role cannot be explained solely in terms of visuospatial attention, rather it involves an interplay between visuospatial attentional and word reading-specific factors.
For skilled readers, accurate visual processing of single words occurs rapidly and in parallel (e.g., Braet & Humphreys, 2006). However, when healthy participants are briefly presented with two words that have some letters in common and are asked to report just one of the two words, they will sometimes report a new word with migrated letters (Marcel et al., 2006; Davis & Bowers, 2004; Jordan & Patching, 2003; Van der Velde, 1992; Mozer, 1983; Allport, 1977; Shallice & Warrington, 1977). For instance, when they are briefly presented with the words hint and hurt and are then asked to report the former word, healthy participants may report the word hunt. These letter migration phenomena have been shown to be specific (i.e., they do not simply represent misreading errors). Critically, their occurrence depends on the identity of the context word (i.e., the one of the two words that should not be reported): Letter migrations occur significantly more often when one of the letters of the context word can be migrated to the target word to generate a new meaningful word (e.g., as above, hint and hurt) than when this is not the case (e.g., hint and heat; Davis & Bowers, 2004). Moreover, letter migrations are influenced by reading-specific phenomena, such as orthographic (Van der Velde, 1992) and semantic priming (Van der Velde, Van der Heijden, & Schreuder, 1989).
The nature of the mechanisms leading to letter migration and their neural underpinnings is poorly understood. The erroneous binding between the letters that comprise words and their spatial location within the words may underlie the effect. Letter migrations have been explained on the basis that there is confusion between letters in corresponding positions within the two presented words. The migrating letter from the to-be-ignored word would be erroneously bound to the corresponding position of the target word. In keeping with this account, Marcel et al. (2004, 2006) found interindividual differences in perceptual migration and fusion that were reproducible in different sensory modalities (tactile, auditory, and visual, also using a letter migration paradigm) and interpreted this phenomenon as a disturbance of the binding between perceptual content and spatial location. The authors postulated that perceptual migration mimics the phenomenon of allochiria (i.e., the transposition of sensory information from the contra- to the ipsilesional side of space), which has been observed in patients with damage to the right posterior parietal cortex (PPC).
In fact, the parietal lobes and their contribution to spatial attention are believed to play a crucial role in perceptual binding (for a review, see Robertson, 2003). On the other hand, the right PPC has also been shown to play a role in word recognition; in particular, in the encoding of spatial features of the letters (Pammer, 2012). For instance, data from PET (Mayall, Humphreys, Mechelli, Olson, & Price, 2001) and TMS (Braet & Humphreys, 2006) have implicated the right PPC in word recognition within unfamiliar contexts (i.e., with inconsistent fonts or mixed-case words). Moreover, evidence from magnetoencephalography indicates that there is an early signal in the right PPC during word recognition, which is significantly stronger when the single letters composing the word are spatially shifted (Pammer, Hansen, Holliday, & Cornelissen, 2006). One possible interpretation of these data is that unfamiliar spatial formats require more extensive visuospatial attentional resources, thereby preferentially recruiting the right PPC (Pammer, 2012; Braet & Humphreys, 2006). For the specific case of visual word processing, the above-mentioned results thus suggest that the right PPC has a central role as an interface between the lexical and the visuospatial properties of words. Additional independent evidence consistent with this interpretation comes from studies that have shown impaired function of the right PPC associated with, among others, letter migration errors in patients suffering from reading disorders such as attentional dyslexia (Hari, Renvall, & Tanskanen, 2001).
Taken together, these results suggest that the right PPC is thus a good candidate as a cortical region playing a critical role in the phenomenon of letter migrations, since—during visual word processing—it seems to support both visuospatial and reading-specific processes. However, direct evidence of the involvement of this area is, to the best of our knowledge, still lacking.
TMS is a noninvasive technique that has been applied to temporarily interfere with the activity of a cortical region below the site of stimulation. On the basis that interference can be conceptualized as a “virtual lesion,” it is thereby possible to infer the contribution of the affected cortical region to a behavioral task (Pascual-Leone, Walsh, & Rothwell, 2000). Recently developed repetitive TMS (rTMS) protocols, such as continuous theta burst stimulation (cTBS; e.g., Nyffeler et al., 2006; Huang, Edwards, Rounis, Bhatia, & Rothwell, 2005), have been shown to induce long-lasting behavioral effects (in the order of minutes) following short application times (in the order of seconds). Hence, behavioral testing can take place offline (i.e., after the application of cTBS).
The aim of this study was to investigate the role of the right PPC in the phenomenon of letter migration by means of a causal, interference approach involving the application of cTBS. Healthy participants performed a behavioral task in which two words were presented bilaterally for a brief period under two conditions: (1) migration and (2) control context. In both conditions, participants were asked to report one word while ignoring the other. Behavioral performance under migration and control context conditions was assessed by recording the number of letter migrations, the number of other error types, and the RTs. All of these dependent measures were assessed as a function of cTBS applied over the right PPC and, in turn, against sham stimulation and no intervention.
Twenty-four healthy individuals participated in the study. Eight participants (one woman), aged between 22 and 44 years (mean [m] = 30, SEM = 2.93), were assigned to the stimulation group. Eight further participants (one woman), aged between 20 and 42 years (m = 29.5, SEM = 3.15), were assigned to the control group and performed the task without stimulation. The remaining eight participants (four women), aged between 20 and 34 years (m = 25.5, SEM = 1.83), were assigned to the sham stimulation group. There was no significant difference in age between the three groups (F(2, 21) = .83, p = .45).
All participants were right-handed and had normal or corrected-to-normal visual acuity. They were all native English speakers, had at least 13 years of formal education, and did not have a (self-reported) reading disorder. Participants in the stimulation group and in the sham stimulation group were screened by means of a questionnaire that was based on internationally recognized safety standards for the application of TMS (Rossi, Hallett, Rossini, Pascual-Leone, & Safety of TMS Consensus Group, 2009). All participants gave written informed consent before the beginning of the study. The study was approved by the West London 1 Research Ethics Committee and was conducted according to the principles laid down in the latest Declaration of Helsinki.
Materials and Design
The visual stimuli and the presentation parameters applied in our behavioral task closely corresponded to the methodology used in a previous study by Davis and Bowers (2004). Visual stimuli consisted of pairs of words that were four letters in length, in which the two external letters (first and fourth) were identical, whereas the two internal ones (second and third) were different. One word of each pair had to be reported (henceforth referred to as, the target word), whereas the other word was designated as, to be ignored (henceforth referred to as, the context word). The word pairs were chosen such that if one of the two internal letters of the target word was replaced by one of the two internal letters of the context word, this would lead to a new meaningful word.
To maximize the number of stimuli, different types of word pairs were merged from two previous studies (Marcel et al., 2006; Davis & Bowers, 2004). In the 60 word pairs used by Davis and Bowers (2004; Experiment 1), only one of the two internal letters of the context word could replace one of the two internal letters of the target word to generate a new meaningful word. The replacement letter could be in the same position in the target and in the context word (e.g., HINT and HURT, resulting in the new word HUNT) or in the adjacent position (e.g., BLED and BIRD, resulting in the new meaningful word BRED). Davis and Bowers (2004) have shown that the migration of letters between context and target words is equally frequent when the replacement letter is in the same or in the adjacent position. Four additional word pairs were constructed using the same criteria. The second type of word pairs were drawn from those used by Marcel et al. (2006; Experiment 3, same outer letters). In these word pairs, both of the two internal letters of the context word could replace the corresponding letter in the target word, resulting in two possible new meaningful words (e.g., MARE and MOLE, resulting in the new meaningful words MORE or MALE). The inclusion of these word pairs resulted in a total set of 74 word pairs (henceforth referred to as, the migration context set).
A second set of word pairs was introduced as a control condition to test the specificity of the letter migration phenomena. The target words of this set were the same as in the migration condition. The context words had the same external letters as the respective target words and different internal ones. For this second set, however, if one of the two internal letters of the target word was replaced by one of the two internal letters of the context word, this did not result in a new meaningful word (henceforth referred to as, the control context set). For the word pairs used by Davis and Bowers (2004), control context words were provided in the original paper (e.g., HINT and HEAT, resulting in no new meaningful word by letter migration). For the word pairs used by Marcel et al. (2006), control context words were constructed in accordance with the same criteria as above (e.g., MARE and MUSE, resulting in no new meaningful word by letter migration).
Hence, there were 148 word pairs overall, 74 in migration and 74 in control context. Every participant saw each target word twice during the experiment, that is, once in the migration context and once in the control context. The order in which the participants saw the words in the two contexts (i.e., migration or control context first) was randomized.
The words were capitalized and presented in a black Arial font on a light gray background. Each word subtended 1.5° visual angle in the horizontal dimension and 0.4° visual angle in the vertical dimension from central fixation. One word was presented on the left, and the other was presented on the right of the center of the screen. The two words were separated on the horizontal dimension by 0.3° visual angle, as calculated from the center of the screen.
Two different versions of the behavioral task were generated, in which a specific target word was presented on the left or on the right, respectively. Half of the participants within each group (i.e., stimulation, control, or sham stimulation group) were presented exclusively with one of these two counterbalanced versions of the behavioral task.
As described above, there was one between-subject factor with three levels: the stimulation group, the behavioral control group, and the sham stimulation group. In all three groups, the experiment was performed in two blocks, each of which was comprised of 74 trials; that is, 37 word pairs in migration context (32 of these word pairs had only one possible replacing letter—16 in same and 16 in adjacent position—and 5 word pairs had two possible replacing letters) and 37 word pairs in control context (with the same target words). The first block of the experiment was administered as a baseline to assess performance before intervention (pre). The second block was administered to assess the effects of the intervention (none, stimulation, or sham stimulation) in the three groups (post; see Figure 1). A short break of approximately 5 min was introduced between pre and post blocks to allow the application of stimulation or sham stimulation in the respective groups.
The timeline of each trial involved the following five stages: (1) presentation of a black central fixation cross (0.3 × 0.3° visual angle) on a light gray background for 1000 msec; (2) presentation of a blank screen, with a light gray background, for 500 msec; (3) presentation of the word pair for 71 msec; (4) presentation of a (backward) mask (a series of black hash signs on a light gray background, subtending 4.1° visual angle on the horizontal dimension and 0.4° visual angle on the vertical dimension) for 200 msec in the region previously occupied by the word pair; and (5) presentation of a cue (a black horizontal line on a light gray background, subtending 1.5° visual angle on the horizontal dimension), located below the area previously occupied by one of the words from the word pair. The cue indicated which word to report and remained on the screen until the participant provided a verbal response. After each response, the experimenter initiated the next trial by pressing a button.
The participants were instructed to keep their gaze on the central fixation cross, attend to the word pair, and verbally report the cued word, as accurately and as quickly as possible, spelling the word afterwards. In order not to influence the responses of the participants, the experimenter referred to the stimuli as “letter strings” (and not as “words”) during the whole experiment.
Vocal responses were recorded by means of a microphone and stored in separate audio files for offline analysis.
To familiarize the participants with the behavioral task, we administered 20 practice trials before the start of the two blocks of trials. The word pairs used in the practice session were constructed according to similar criteria as the experiment proper but were not used elsewhere in the experiment.
Visual stimuli were presented on a 21-in. cathode ray tube display (Sony Trinitron E530, Tokyo, Japan), set at a resolution of 1024 × 768 pixels, 32-bit color depth, and a refresh rate of 85 Hz. The viewing distance was fixed at 60 cm, and the head of the participants was stabilized by means of a chin-and-head rest. Central fixation was monitored online by the experimenter by means of a video-based eye-tracking system (EyeLink 1000, SR Research, Mississauga, Canada).
cTBS was applied using with a Magstim Rapid stimulator, in a configuration with four booster modules (Booster Module plus, Magstim, Whitland, UK), and connected to a commercially available figure-of-eight coil (diameter of 70 mm; Magstim). Sham stimulation was delivered with a figure-of-eight coil connected to a commercially available placebo coil system (Magstim). In this setup, the coil generated the same discharge noises as during real stimulation and produced a mild cutaneous sensation but did not stimulate the underlying cortical tissue. The stimulator was controlled by a laptop (Dell Latitude, Round Rock, TX), running a customized script programmed to implement the cTBS protocol parameters described below (E-Prime 2.0 Pro, Psychology Software Tools, Pittsburgh, PA).
TMS Site and Protocol
TMS was delivered by means of a cTBS train of 300 pulses for a duration of 20 sec (100 bursts of 3 pulses each, delivered at 50 Hz, with an interburst interval of 200 msec). This TMS protocol has been shown to exert inhibitory effects on the human cortex lasting about 20 min (Huang et al., 2005). cTBS was applied over P4 (according to the 10–20 International EEG system), which has been shown to overlie the PPC in the proximity of the intraparietal sulcus (Herwig, Satrapi, & Schönfeldt-Lecuona, 2003; Hilgetag, Théoret, & Pascual-Leone, 2001). The choice of this stimulation site is consistent with previous studies that have sought to stimulate the right PPC to investigate lateralized attentional mechanisms (e.g., Cazzoli, Müri, Hess, & Nyffeler, 2009; Cazzoli, Wurtz, Müri, Hess, & Nyffeler, 2009; Rosenthal, Roche-Kelly, Husain, & Kennard, 2009; Nyffeler et al., 2008; Müri et al., 2002; Hilgetag et al., 2001; Pascual-Leone et al., 1994). The coil was held in position by the experimenter for the duration of the cTBS, with the handle pointing backwards and at an angle of 45°, with respect to the participants' sagittal plane. cTBS was delivered at 80% of the participants' individual resting motor threshold (rMT) of the left small hand muscles. The rMT was defined as the minimum stimulator output intensity of TMS single pulses that was able to consistently elicit motor responses (i.e., in at least three of five consecutive trials) in the small hand muscles of the left relaxed hand. The mean rMT was 54.50% of the maximum stimulator output (SD = 8.49; range = 41–66).
Audio files obtained from each trial were scored to classify the vocal response associated with each trial as either correct or incorrect. Errors were subdivided into three different categories (see also Davis & Bowers, 2004): (1) letter migrations/pseudo letter migrations; (2) word migrations; and (3) other errors. Letter migrations were defined as the cases in which—in a migration context—participants reported the target word with a changed letter, migrated from the context word (e.g., target word HINT, migration context word HURT, participants report the word HUNT). In a control context, migrations represent unexpected errors, in which the participants reported the target word with a changed and migrated letter, which could not have come from the context word (e.g., target word HINT, control context word HEAT, participants report the word HUNT). These errors are thus referred to as pseudo letter migrations. Comparison between the number of letter migrations and pseudo letter migrations enabled us to establish whether the source of letter migrations was the context word, rather than misreading. Word migrations were defined as the cases in which participants reported the context word instead of the target word. Other errors refer to the cases in which the word reported by the participants did not match the target word, the context word, the response with a migrated letter, and those cases in which the participants gave no response.
The number of letter migrations and pseudo letter migrations, word migrations, and other errors were examined by means of three separate mixed-model repeated-measures ANOVAs, with Stimulation as the between-subject factor (levels: cTBS over the right PPC, control [no CTBS], and sham stimulation over the right PPC) and Time point (levels: pre, post) and Context (levels: migration, control) as the within-subject factor. To follow up on the effect of the presentation side of the target on letter migrations, we performed separate, additional mixed-model ANOVAs only on the data obtained in the migration context, for target words presented on the left and the right, with Stimulation (levels: none, cTBS over the right PPC, sham stimulation over the right PPC) and Time point (levels: pre, post) as between- and within-subject factors, respectively.
RTs were measured as time between the offset of the backward mask to the onset of the word utterance, as assessed with commercially available audio software (WavePad, NCH Software, Canberra, Australia). In the cases in which the word was preceded by a hesitation (e.g., filler words such as “uhmm”), the onset of the utterance of the word proper was defined as the RT. To obtain a sufficient number of trials for each response type (i.e., correct or incorrect, according to the different categories of errors), we analyzed the RTs only for the migration context (this issue was particularly relevant for the letter migration errors, which were nominal in the control context).
The median RT was computed for each participant for each of the respective response categories (i.e., correct responses, letter migrations, words migrations, and other errors), target word sides (left, right), and time points (pre, post). The results were analyzed by means of separate mixed-model ANOVAs for target words presented on the left and the right, with Stimulation as between-subject factor (levels: none, cTBS over the right PPC, sham stimulation over the right PPC) and Response type (levels: correct responses, letter migrations, words migrations, other errors) and Time point (levels: pre, post) as within-subject factors.
All subsequent post hoc tests were performed by means of least significant difference corrected t tests.
All participants tolerated the cTBS well and did not report any side effects. Participants virtually never (i.e., in 1.18% of cases on average, SEM = 0.34) reported nonwords as responses.
Letter Migrations and Pseudo Letter Migrations
The mixed-model ANOVA performed on the number of letter migrations and pseudo letter migrations (see Figure 2) revealed a main effect of the Context (F(1, 21) = 113.35, p < .001); that is, the number of letter migrations was significantly higher in the migration context than the number of pseudo letter migrations in the control context. There was no significant main effect of Time point (F(1, 21) = 3.15, p = .09) or stimulation (F(2, 21) = .69, p = .51); that is, the stimulation condition and the time point of the measurement (pre or post), per se, did not modulate the number of letter migrations and pseudo letter migrations. Two-way interactions between Context × Stimulation (F(2, 21) = .40, p = .67) and Context × Time point (F(1, 21) = 2.33, p = .14) were not significant. By contrast, the interaction between Time point and Stimulation was significant (F(2, 21) = 3.72, p = .04). Post hoc tests revealed that—irrespective of the context—there was a significantly higher number of letter migrations after cTBS over the right PPC than in the two other stimulation conditions (none or sham). More relevantly, however, a significant three-way interaction between Context × Time point × Stimulation (F(2, 21) = 5.63, p = .01) indicated that this effect was context specific. Importantly, post hoc tests revealed that there was a significant pre–post increase in the number of letter migrations only after cTBS application over the right PPC and only in the migration context (see Figure 2).
The number of letter migrations was not significantly different in word pairs with replacement letters in the same or in the adjacent position (t(23) = .84, p = .41, two-tailed).
Additional analyses intended to assess the effect of the presentation side on the number of letter migrations revealed a significant interaction between Time point and Stimulation (F(2, 21) = 4.17, p = .03) for the cases in which the target word was presented on the left side, but no main effects of Stimulation or Time point (F(1, 21) = 1.17, p = .33; F(1, 21) = 1.33, p = .26). For the cases in which the target word was presented on the right side, there were no significant main effects of Stimulation or Time point (F(1, 21) = .11, p = .99; F(1, 21) = 1.09, p = .31), and the corresponding interaction was also not significant (F(2, 21) = .42, p = .66). As also assessed by subsequent post hoc tests, the only significant increase in letter migration errors was observed in the case of a left-sided presentation of the target word and after the application of cTBS over the right PPC (see Figure 3). Moreover, a direct comparison revealed that the number of migration errors was significantly higher when target words were presented on the left rather than on the right side after application of cTBS over the right PPC (t(7) = 2.50, p = .04, two-tailed), but not in the pre cTBS condition (t(7) = .73, p = .49, two-tailed).
A mixed-model ANOVA performed on the number of word migrations revealed no significant main effects of factors Context, Time point, and Stimulation (F(1, 21) = 1.11, p = .30; F(1, 21) = .04, p = .85; F(2, 21) = .21, p = .81; respectively). Similarly, there were no significant interactions between these factors (Context × Stimulation: F(2, 21) = .92, p = .42; Time point × Stimulation: F(2, 21) = .88, p = .43; Context × Time point: F(1, 21) = 2.47, p = .13; Context × Time point × Stimulation: F(2, 21) = .16, p = .85). That is, the number of incorrectly reported context words was equivalent in the three stimulation groups, irrespective of the migration or control context and of the time point (pre or post).
A mixed-model ANOVA performed on other errors revealed no significant main effects of Context, Time point, and Stimulation (F(1, 21) = 1.77, p = .19; F(1, 21) = .15, p = .70; F(2, 21) = .29, p = .75; respectively). In addition, none of the interaction terms were significant (Context × Stimulation: F(2, 21) = 1.64, p = .22; Time point × Stimulation: F(2, 21) = .68, p = .82; Context × Time point: F(1, 21) = .003, p = .96; and Context × Time point × Stimulation: F(2, 21) = .01, p = .99).
RTs were examined using mixed-model ANOVAs, and the results revealed that there was a significant main effect of Response type (correct responses, letter migrations, word migrations, and other errors) both for target words presented on the left (F(3, 63) = 29.36, p < .01) and on the right (F(3, 63) = 33.45, p < .01). No other main effects or interaction terms were significant, neither for target words presented on the left (time point: F(1, 21) = .56, p = .46; stimulation: F(2, 21) = .05, p = .96; Time point × Stimulation: F(2, 21) = .41, p = .67; Response type × Stimulation: F(6, 63) = .78, p = .59; and Time point × Response type: F(3, 63) = 1.59, p = .20; Time point × Response type × Stimulation: F(6, 63) = .49, p = .81), nor for target words presented on the right (time point: F(1, 21) = .44, p = .52; stimulation: F(2, 21) = 1.06, p = .37; Time point × Stimulation: F(2, 21) = .73, p = .50; Response type × Stimulation: F(6, 63) = .48, p = .82; Time point × Response type: F(3, 63) = 1.57, p = .21; and Time point × Response type × Stimulation: F(6, 63) = .14, p = .99).
As confirmed by post hoc tests, both for target words presented on the left and on the right, errors were associated with significantly longer RTs than trials on which the correct word was reported, irrespective of the stimulation condition and the time point (pre or post intervention). However, the RTs for letter migrations were significantly faster than those associated with the word migrations and the other errors (see Figure 4).
In this study, we examined the role of the right PPC in the phenomenon of letter migration using an interference-based approach involving the application of cTBS. To this end, healthy participants were briefly presented with two words that were four letters in length, with identical external letters and different internal letters, and were asked to report one of the two words (the target word), while ignoring the other word (the to-be-ignored, context word). cTBS was applied over the right PPC to examine the effect on the baseline rate of the erroneously reported words due to letter migrations. Both the basic behavioral letter migration effect and the effect of cTBS over the right PPC on letter migration were shown to be context dependent: first, letter migrations occurred significantly more often when the target word was in a migration context (i.e., when one letter of the to-be-ignored word could replace one of the target word) than in a control context (i.e., when this replacement could not result in a new meaningful word); and, second, cTBS only modulated letter migration in the migration context and only in the group undergoing active stimulation (i.e., letter migration was not modulated in the sham stimulation group nor in the control group without stimulation). These latter data demonstrate that the observed modulation cannot be explained by a general change in arousal or by a general disruption of behavioral performance, because neither RTs nor the number of word migrations (i.e., the cases in which the participants reported the to-be-ignored instead of the target word) or other errors were affected by the stimulation condition. Furthermore, the cTBS effect was lateralized; that is, the increase in letter migrations after cTBS application over the right PPC was significant when the target word was presented on the left, but not when the target word was presented on the right. Finally, the results revealed that words reported as a result of letter migrations were associated with significantly longer RTs than correctly reported words, but significantly shorter RTs than word migrations and other errors.
The basic behavioral effect—the induction of letter migration by means of the bilateral presentation of word pairs—observed in our study is in line with the results reported in multiple previous studies that used similar behavioral tasks (e.g., Marcel et al., 2006; Davis & Bowers, 2004; Jordan & Patching, 2003; Van der Velde, 1992; Mozer, 1983; Allport, 1977; Shallice & Warrington, 1977). One candidate mechanism underlying letter migration is the erroneous binding between the letters composing the words and their spatial location within the latter, which has been considered as an analogue of the phenomenon known as allochiria, which is sometimes observed in patients with damage to the right parietal lobe (e.g., Marcel et al., 2004, 2006). This putative mechanism—mediated by the parietal cortex—potentially explains the significant increase in letter migrations observed in this study after the application of inhibitory cTBS over the right PPC. In fact, it is conceivable that temporary interference with the right PPC disturbed correct binding of letters to words, giving rise to illusory conjunctions between letters and their spatial position, and thus to perceived words with migrated letters.
However, perceptual binding alone does not fully account for the results obtained in this study. First, the letter migrations occurred significantly more often in the migration than in the control context, and participants reported meaningful words in the great majority of letter migrations. If the phenomenon of letter migration was solely perceptual and due to impaired binding, one would also predict a greater number of migrations in the control context and the report of meaningless nonwords, which virtually never occurred. Second, the number of letter migrations did not differ between word pairs that were based on replaceable letters in the same position versus in the adjacent position. This result replicates Davis and Bowers (2004). It thus seems insufficient to explain the observed selective effects on letter migrations solely in terms of an erroneous binding of the migrating letter from one word with the homologous, specific position in the other word (see also Mozer, 1983). Third, the observed lateralization of letter migrations does not follow an allochiria-like pattern, as observed in patients with attentional disorders after a lesion of the right parietal lobe. Typically, patients suffering from allochiria transpose sensory information from the left, contralesional side toward the right, ipsilesional side of space. After cTBS application over the right PPC, the significant increase in the number of letter migrations was lateralized to target words presented on the left side of space; that is, letters putatively migrated from the right-sided to-be-ignored word toward the left-sided target word. This pattern of lateralization is also consistent with a previous study in which words presented on the right often influenced the reported words on the left, but not the other way around (Jordan & Patching, 2003).
It is also important to exclude other potential explanations for the observed selective effects of cTBS on letter migration. In particular, cTBS over the right PPC has been shown to trigger visual extinction (i.e., the report of an unilateral, right-sided stimulus when bilateral stimuli are presented; e.g., Cazzoli, Müri, et al., 2009; Dambeck et al., 2006; Müri et al., 2002; Hilgetag et al., 2001; Pascual-Leone et al., 1994) and neglect-like phenomena (e.g., Cazzoli, Wurtz, et al., 2009; Nyffeler et al., 2008; Ellison, Schindler, Pattison, & Milner, 2004; Bjoertomt, Cowey, & Walsh, 2002; Fierro et al., 2000) in healthy controls. However, it is very unlikely that the cTBS effects observed in this study solely reflected extinction-based phenomena. In fact, the latter would cause the target words presented in the left hemifield to go unperceived and unreported, thus increasing the number of other errors. The same can be stated for a neglect-like pattern: if words presented on the left side or if the left part of the words (i.e., first two letters) would be influenced solely by a neglect-like effect, one should again expect an increase in the number of other errors. Such increases in these types of errors were not observed after cTBS.
By extension, letter migrations and the role of the right PPC in this phenomenon cannot be solely explained by visuospatial attention and binding mechanisms operating at the perceptual level. In fact, letter migrations have been shown to be influenced by semantic priming (Van der Velde et al., 1989) and by orthographic priming (Van der Velde, 1992), even when the target and the to-be-ignored words were not presented simultaneously but successively. This suggests that letter migrations can operate at a higher processing level, “after the translation of the physical identity of a word (i.e., its actual visual form) into abstract information” (Van der Velde, 1990, p. 315), possibly reflecting concurrent preactivations of the target and the to-be-ignored words (Van der Velde, 1992).
We thus reason on the basis of these foregoing considerations that the letter migrations associated with interference with the right PPC occurred as the result of an interaction between attentional and word reading-specific processes. Our results seem, in fact, to correspond with phenomena observed in patients suffering from attentional dyslexia (Saffran & Coslett, 1996; Warrington, Cipolotti, & McNeil, 1993; Shallice & Warrington, 1977), who can read single words but have difficulties in reading a word in the presence of other words or letters and present with letter migrations (Coslett, 2003). Interestingly, letter migrations in patients with attentional dyslexia after bilateral lesions of the PPC are not random but conform to lexical and orthographic constraints (Shalev, Mevorach, & Humphreys, 2008). Some authors have also drawn parallels between attentional dyslexia and hemispatial neglect. Hari and Renvall (2001) postulate a “sluggish attentional shifting” in dyslexic patients, which is particularly accentuated for the left side of space—sometimes referred to as minineglect—and may reflect a diminished function of the right PPC (Hari et al., 2001).
Alternatively, but not exclusively, one could also hypothesize that the increased number of right-to-left letter migrations after cTBS over the right PPC was facilitated by a change in the deployment of visual attention in space. The temporary interference with the right PPC by means of cTBS could have caused a biased attentional allocation toward the right hemispace (Mesulam, 1999; Kinsbourne, 1987, 1993; Posner, Walker, Friedrich, & Rafal, 1984). The words presented in the right hemifield may have thus benefited from greater attentional processing, even when they were not targets and should have been ignored. In turn, these processes may have interacted with hemispheric differences in word recognition: The left hemisphere seems to perform word recognition in the right visual field by means of a parallel analysis, deploying attention across the word as a whole. In contrast, the right hemisphere seems to perform word recognition by means of a serial analysis, deploying attention over one single letter at a time (Lindell, Nicholls, Kwantes, & Castles, 2005; Lindell, Nicholls, & Castles, 2003). The words presented in the left hemifield may have thus been more prone to the intrusion of migrating letters due to this putative serial processing.
Finally, our results showed that the RTs for the different types of responses were significantly different. All the erroneous responses were associated with significantly slower RTs than the correct ones. For tasks in which participants are asked to report presented words without feedback on their performance—such as, for instance, in word stem completion tasks—it is common to observe slower RTs for erroneous responses than for correct ones (e.g., Gabrieli, 1998; Schacter & Buckner, 1998). However, the RTs associated with letter migrations were significantly faster than the ones for other types of errors. One simple explanation for this observation would be that words with migrated letters may be more plausible than other erroneously reported words, because the former still conform to orthographic and lexical constraints. This is, however, also the case for migrated words (i.e., when the participants report the to-be-ignored word), which nevertheless are characterized by significantly slower RTs. If one interprets the significant lengthening of RTs due to erroneous responses in terms of interference, then one could hypothesize that letter migration errors present a qualitatively different form of interference to that associated to the other error types.
Future studies should further investigate factors potentially influencing letter migrations and their neural underpinnings. An interesting question would be to investigate the role of the contralateral homologue cortical area (the left PPC) in letter migrations and its interplay with the right PPC by means of an interference approach such as TMS. Furthermore, studies with carefully designed and controlled stimuli could investigate the influence of the potential migration of visual features of single letters, further clarifying the role of regions such as the PPC in processing words as hierarchical stimuli (i.e., stimuli composed of letters, which are, in turn, composed of a defined number of visual features).
In conclusion, our results show—with a causal, interference approach involving the application of cTBS—that the right PPC plays a central role in the phenomenon of letter migrations. The effects were lateralized, that is, letter migrations occurred significantly more often from the right, ipsilateral hemifield toward the left, contralateral hemifield. The results suggest that the letter migration phenomena can be interpreted as resulting from an interaction between visuospatial attentional factors and reading-specific (lexical, orthographic) factors, and support the idea of a critical contribution of the right PPC to this process.
D. C. was supported by the Swiss National Science Foundation (Grant PBBEP3_134978), the Holcim Foundation, and the Janggen-Pöhn-Foundation. C. R. R. was supported by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre based at Oxford University Hospitals NHS Trust and University of Oxford.
Reprint requests should be sent to Dario Cazzoli, Nuffield Department of Clinical Neurosciences, University of Oxford, John Radcliffe Hospital, West Wing, Level 6, OX3 9DU Oxford, United Kingdom, or via e-mail: firstname.lastname@example.org.