How many words can be formed with the help of 3 consonants and 2 vowels such that no two consonants are adjacent?

7201440240480

Answer : B

Solution : There are 5 vowels and 3 consonants in the word 'EQUATION'. Three vowels out of 5 and 2 consonants out of 3 can be chosen in `""^[5]C_[3]xx""^[3]C_[2]` ways. So, there are `""^[5]C_[3]xx""^[3]C_[2]` groups each containing 3 consonants and two vowels. Now, each group contains 5 letters which are to be arranged in such a way that 2 consonats occur together. Considering 2 consonants as one letter we have 4 letters which can be arranged in 4! ways. But two consonants can be put together in 2! ways. Therefore, 5 letters in each group can be arranged in `4!xx2!` ways.
`:.` Required number of words `=[""^[5]C_[3]xx""^[3]C_[2]]xx4!xx2!=1440`.

Introduction

To successfully identify a word, readers must encode both the identities and the positions of the letters within that word. Without encoding letter position information, readers would not be able to discriminate between words, such as tame and mate. Strikingly, readers often are able to identify successfully words in which the positions of letters are altered within a word—transposed text. For example, people can understand the sentence, “it dseno’t mtaetr in waht oerdr the ltteres in a wrod are, the olny iproamtnt tihng is taht the frsit and lsat ltteer be in the rghit pclae.” [//www.dailywritingtips.com/cna-yuo-raed-tihs/]. Despite the sentiment expressed in this statement, recent research has shown that there is a cost associated with reading text in which the order of letters is altered [Christianson, Johnson, & Rayner, 2005; Johnson, 2007, 2009; Johnson & Dunne, 2012; Johnson & Eisler, 2012; Johnson, Perea, & Rayner, 2007; Perea & Lupker, 2003a, b, 2004; Rayner, White, Johnson, & Liversedge, 2006; White, Johnson, Liversedge, & Rayner, 2008]. Although we can understand transposed text, our reading is significantly slowed by such manipulations.

The fact that readers can understand transposed text [despite the cost to processing time] shows that there must be some flexibility in letter position encoding during lexical identification. In the experiments reported here, we aimed to investigate the degree of flexibility in letter position encoding during reading, by examining the effects of two specific manipulations—the within-word distance between transposed letters [TLs, Experiment 1], and whether the TLs are consonants or vowels [Experiment 2]. We also wished to examine the extent to which identification of TL words is supported by the presence of a meaningful sentence context. To date, the majority of research that has examined TL effects [Rayner & Liversedge, 2011] has used isolated word presentation techniques [e.g., masked priming]. Such methods allow extremely strict control over the participants’ perceptual experience and, on the basis of such work, a number of models of letter encoding in lexical identification have been presented: SERIOL [Whitney, 2001, 2008; Whitney & Cornelissen, 2008], SOLAR [Davis, 1999, 2010], open-bigram [Grainger & van Heuven, 2003], and the Overlap [Gomez, Ratcliff & Perea, 2008] models. As argued by Rayner and Liversedge, however, such methods do not necessarily approximate the “normal” reading process, in which words are visually sampled during a series of fixations that may or may not be spatially and temporally contiguous on the target word. Here, we report a series of experiments in which our two key manipulations [the distance between TLs, and the consonant-vowel status of TLs] were implemented in both a natural sentence reading task and a single-word decision paradigm. Our goal was to investigate the flexibility of letter encoding during reading and to examine the extent to which this process is influenced by the presence of a sentence context that may support lexical identification.

Experiment 1

The majority of studies that have examined processing of transposed text, either in single word paradigms or in more natural sentence reading tasks, have transposed two adjacent letters within a word [Christianson et al., 2005; Johnson & Eisler, 2012; Johnson et al., 2007; Perea & Lupker, 2003a, b; Rayner, White, et al., 2006; White et al., 2008]. There are now several models of visual word identification that can account for the fact that TL strings have greater perceptual similarity to their base words than substituted letter [SL] strings, thus allowing some flexibility in letter position encoding. There are two main mechanisms by which this flexibility is realized: 1] open-bigram coding, incorporated in both the SERIOL model [Whitney, 2001, 2008; Whitney & Cornelissen, 2008] and open-bigram model [Grainger & van Heuven, 2003]; and 2] spatial coding, incorporated in the SOLAR model [Davis, 1999, 2010].Footnote 1

In open-bigram coding, letters are identified and encoded as ordered pairs within the word. For example, for the word lunch, the following set of bigrams would be encoded: lu, ln, lc, lh, un, uc, uh, nc, nh, and ch. In the open-bigram model, these are encoded discretely, with activity of either 0 or 1 and, importantly, bigrams are encoded with up to two intervening letters. So, in the lunch example, lu, ln, and lc would be encoded, but not lh. In the SERIOL model, bigram units are activated on a continuous scale from 0 to 1, and this activity is used to encode both identity and relative position within the word. Again, up to two intervening letters are permitted; within this constraint, all possible bigrams are encoded, and these are weighted such that bigrams with closer proximity between the two letters receive a greater weight than those where the two letters are further apart [Whitney, 2008, Davis & Bowers, 2006; Perea, Duñabeitia, & Carreiras, 2008]. So, in open-bigram models, flexibility is achieved by encoding each letter’s positional relation to the other letters in the word, and this mechanism offers an explanation for the perceptual similarity of TL letter strings. For example, the similarity between a word, such as desk, and a TL string, such as dsek, is greater than an SL string, such as dosk [the TL string activates 5/6 bigrams while the SL string only activates 3/6 of the desk bigrams].

In spatial coding, used in the SOLAR model, letter units are position-independent; the unit for a given letter will be activated by the occurrence of that letter in any position within a word. However, the strength of activation for individual letter units is determined by the position of that letter within the word. So, for the word, lunch, the l unit has the highest activation and the h unit has the lowest activation; activation decreases monotonically from left to right within the word. Again, this encoding system offers flexibility that can explain TL effects. For the word desk and the TL string dsek, all the same letter units are activated and it is only the relative strength of activation across these four letter units that varies. In contrast, for the word desk and the SL string dosk, only three of the same letter units are activated. Thus, in comparison to the activation profile associated with desk, the profile for dsek is more similar than that for dosk.

These input coding schemes provide a theoretical account of TL effects in lexical identification: all three models include a mechanism that offers some flexibility for letter position encoding, whereby a word can still be identified despite the transposition of nonadjacent letters. Consistent with these theoretical models, significant priming effects have been found from letter strings containing nonadjacent TLs, where there was one intervening letter between the transposed letters [Perea & Carreiras, 2006; Perea & Lupker, 2004]. Clearly, such TL strings can activate the lexical representations of their base words; however, these studies did not systematically manipulate the distance between transposed letters. This was done in a later study by Perea et al. [2008]. Perea et al. presented participants with masked prime TL letter strings where the transpositions were adjacent, or one or two letters apart within the word [herein referred to as one-space TLs and two-space TLs, respectively]. Critically, they found that there was a large decrease in the magnitude of the priming effect from adjacent to one-space transpositions, but a much smaller difference in the priming effect from one- to two-space transpositions. We wished to examine the flexibility in letter-position encoding during sentence reading and also to determine whether such flexibility might be increased by the presence of a meaningful sentence context.

To date, few studies have examined the influence of nonadjacent letter transpositions during sentence reading [Johnson, 2007; Winksel & Perea, 2013]. In the Johnson study, one-space TL strings were presented in parafoveal preview, using the gaze-contingent boundary paradigm [Rayner, 1975]. Preview strings were either identical to the target word, had the third and fifth letters transposed, or had the third and fifth letters substituted. SL previews led to longer fixation durations than TL previews in single and first fixation durations on the target word. In contrast, a preview effect of TLs was only seen in gaze durations—a relatively late effect. Johnson argued that letter identity is encoded earlier than letter position during lexical identification in natural sentence reading. Comparable results were reported by Winskel and Perea [2013], who also used the boundary paradigm, and found a transposed letter effect in gaze durations [but not in earlier measures] for Thai readers. In the current set of experiments, we examined the pattern of effects when readers were foveally attending to TL manipulations through both an eye-tracking experiment [Experiment 1a], and an isolated word decision experiment [Experiment 1b]. On the basis of the literature, we made three predictions for Experiment 1a: 1] there would be a cost to reading times associated with all letter transpositions; 2] adjacent letter transpositions would cause less disruption to reading than transpositions where there are intervening letters between the two that are transposed; 3] in the case of nonadjacent transpositions, there would be relatively little difference between one- and two-space transpositions. With respect to Experiment 1b, we predicted greater disruption from the TL manipulation due to the lack of a supporting sentence context to facilitate lexical identification.

Experiment 1a: methods

Participants

Participants were 24 native-English speakers with normal, or corrected-to-normal, vision and no known reading difficulties who took part voluntarily. They were students at the University of Massachusetts, Amherst, with an age range of 18 to 30 years. All participants were naïve regarding the purpose of the study and were paid for their participation.

Apparatus

Monocular eye movement recordings from the right eye were taken using an EyeLink 1000 eye tracker [though viewing was binocular]. The position of the participant’s right eye was recorded every millisecond. Sentences were presented on a 21-inch monitor at a viewing distance of 61 cm, and one character space subtended 0.3°. Sentences were presented in white, Courier New font size 16, on a black background. Participants leaned on chin and forehead rests during the experiment to minimise head movements.

Materials and design

Sixty experimental sentence frames were constructed, each containing an eight-letter long target word. These target words were selected so that the second letter could be transposed with either the third, fourth, or fifth letter to alter the word, or the third letter with the fourth, fifth, or sixth; none of these pairs of letters were the same [i.e., office, future, and letter were not suitable as some transpositions involved two of the same letters and so the word would not be altered]. Thus, each word could be presented in each of four conditions: 1] normally, 2] with adjacent letters transposed [letters 2 and 3, or 3 and 4], 3] with one character between the transposed letters [letters 2 and 4, or 3 and 5], or 4] with two characters between the transposed letters [letters 2 and 5, or 3 and 6]. The printed word frequency of the target words ranged from 1 to 289 occurrences per million [M = 73, SD = 107] [values from the CELEX database; Baayen, Piepenbrock, & Gulikens, 1995]. In addition to the target words, all words in the sentences with six or more letters were transposed in the same manner to keep the target word as congruous as possible with the rest of the sentence frame. An example experimental sentence showing the letter transposition manipulations is given in Table 1.

Table 1 Sample experimental sentences containing the transposed letter manipulations from Experiments 1 and 2

Full size table

All sentences were between 55 and 80 characters long. In addition to the 60 experimental items, 5 practise items were presented at the beginning of the experiment. After 20 of the sentences, participants were required to answer a simple comprehension question, making a “yes/no” response using a button box. The transposition conditions were counterbalanced following a Latin-square design, such that each sentence was presented in every condition, but each participant read each sentence only once.

The stimuli were prescreened for comprehension to ensure that readers could understand the transposed letter words, as well as for sentence plausibility and target word predictability. Thirty participants took part in this prescreening, none of whom participated in the main eye movement study.

In the comprehension prescreen, participants were presented with the transposed text sentences and asked to first circle any words that were not spelled correctly, and then to write underneath what they thought really should be the word. In addition to matching comprehension across TL distance conditions, this prescreen allowed us to ensure that the transposed text manipulation did not prevent participants in the eye movement study from being able to identify the words. If at least nine out of ten participants were able to correctly identify what should be the word, then the item was included in the final stimulus set. Importantly, there was no significant difference among conditions for comprehension [p > 0.9].

In the plausibility prescreen, the sentences were presented to participants in their normal format, and the participants were asked to rate each sentence on a 5-point plausibility scale [where higher scores indicate greater plausibility]. Only sentences that scored, on average, at least 3/5 were included in the final stimulus set. As the distance between transposed letters was a within-item manipulation, it was not meaningful to compare plausibility across conditions for this prescreen [the same set of items appeared in all conditions].

Finally, in the predictability prescreen, participants were presented with the beginnings of the sentences, up to but not including the target word, and were asked to fill in the word that they thought was most likely to come next. For 50 of the target words, none had a higher predictability than 0.2. The remaining 10 items were more predictable, between 0.4 and 1.0. The effect of predictability was included in the model for each measure reported. The interaction between predictability and transposition distance was always included in the linear mixed-effects [LME] model but, importantly, did not contribute significantly to the models. Therefore, we removed the interaction from the models reported below.

Procedure

Participants were instructed to read the sentences normally and to answer the questions by pressing a button box to respond “yes/no.” An initial calibration of the eye tracker was performed in which the participant was instructed to look at each of three fixation points, extending in a horizontal line across the centre of the screen, while their fixation position was recorded for each point. Once the eye-tracker had been calibrated with satisfactory accuracy [maximum 0.2° error for any one of the three calibration points], the sentences were presented one at a time in random order. Following each sentence, the calibration was checked for accuracy and the eye tracker was recalibrated whenever necessary. The entire experiment lasted approximately 20 minutes.

Experiment 1b: methods

Participants

Participants were 18 native-English speakers with normal or corrected-to-normal vision and no known reading difficulties, aged 18-21 years, who took part voluntarily. All participants were undergraduate students at the University of Southampton, were naïve regarding the purpose of the study, and were compensated with either cash or course credit for their participation.

Apparatus

Stimuli were presented on a 21-inch monitor at a viewing distance of 61 cm. Words were presented in upper case, white, Courier New font size 20, on a black background. Lexical decisions were made using a button box.

Materials and design

Half of the stimuli presented were real words containing a letter transposition, and half were nonwords. The real words were the 60 target words from Experiment 1a, and each word was presented in the three transposed letter conditions described for Experiment 1a [counterbalanced in a Latin-Square design between participants so that no participant saw the same word more than once]: with an adjacent transposition, with one letter between the TLs, and with two letters between the TLs. These TL words were randomly intermixed with 60, 8-letter pronounceable, orthographically legal nonwords [Frankish & Barnes, 2008]. Each nonword was presented in the three different transposition conditions, counterbalanced between participants, as was the case for the real word stimuli. Thus, in total, each participant saw 120 stimuli; half real words containing a letter transposition, and half nonwords containing a letter transposition.

Procedure

Participants viewed one letter string at a time. They were told that some of the stimuli would be nonwords, and others would be real words with spelling mistakes. They were instructed to press one of two buttons to indicate as quickly and accurately as possible, for each letter string, whether they thought it was a misspelled real word or a nonword [similar to the experiments reported by O’Connor & Forster, 1981]. Each participant had six practise trials before the experimental trials began. The entire experiment lasted approximately 15 minutes.

Results

All of results were analysed using a LME model, with participants and items as crossed random effects. Reading and response time data were log-transformed prior to analysis. The significance and standard errors [SE] reflect variability across both items and participants. The p values were estimated using posterior distributions for model parameters obtained by Markov-Chain Monte Carlo sampling. In order to examine the effect of increasing within-word TL distance for each measure, we compared adjacent and one-space transposition conditions, and one-space and two-space transposition conditions [predicting that as the distance between transposed letters increased, reading/response times also would increase].

Experiment 1a

Fixations of less than 80 ms were combined with the previous/next fixation if they were within 1 character space of each other; fixations of less than 40 ms were deleted if within 3 characters of the nearest fixation. Remaining fixations of less than 80 ms or more than 800 ms were excluded from the analysis. All participants scored at least 70 % on the comprehension questions [mean comprehension was 94 %].

We report four common measures of processing of the target words: single fixation durations, first fixation durations, gaze durations, and total word reading times [Rayner, 1998, 2009]. Single fixation duration is the duration of the initial fixation on the word when only one fixation was made on that word in the first pass, whereas first fixation duration is the duration of the initial fixation on the word irrespective of the number of first pass fixations the word receives. Gaze duration is the sum of fixations on a word before the eyes leave that word. Total word reading time is the sum of all fixations on a word during the entire trial. In these analyses, the normal condition was used as the baseline in the LME model [i.e., the intercept was the mean of the normal condition], and subsequent contrasts were performed to compare adjacent and one-space transpositions and to compare one-space and two-space transpositions. We initially included interaction terms between our experimental manipulation and both word frequency and predictability. In most cases, these interaction terms did not improve the fit of the models to the data and were removed. In the case of gaze duration, the condition by frequency interaction term improved the fit of the model but did not have a significant effect within the data. For this reason, and for clarity when comparing results across measures, we report the model without the interaction term.

The reading time data are summarised in Table 2. Single fixation durations were longer in the adjacent condition compared with the normal condition [b = 0.06, t = 2.39, SE = 0.03, p = 0.02]. The difference between the adjacent and one-space conditions was not significant [b = 0.03, t = 1.05, SE = 0.03, p = 0.3], nor was the difference between one and two-space conditions [b = 0.03, t = 1.12, SE = 0.03, p = 0.26]. Thus, single fixation durations were longer for TL words than for correctly spelled words, but the differences between TL conditions were not significant. In first fixation duration, the difference between the normal and adjacent TL conditions was significant [b = 0.05, t = 2.12, SE = 0.02, p = 0.03]. The difference between adjacent and one-space TLs also was significant [b = 0.05, t = 2.17, SE = 0.02, p = 0.03]. As was observed in single fixation durations, the difference between one- and two-space transpositions was not significant [b = 0.01, t = 0.40, SE = 0.02, p = 0.7]. Thus, in very early measures of processing, there was a basic cost of letter transpositions [compared with the control condition], and some suggestion of an additional cost associated with the presence of an intervening letter between TLs, but no difference between having one and two letters between the TLs.

Table 2 Means for single fixation durations, first fixation durations, gaze durations, and total word reading times across the four TL conditions in Experiment 1a, and for response times Experiment 1b

Full size table

In gaze durations, all differences between conditions were reliable [normal vs. adjacent, b = 0.1, t = 3.35, SE = 0.03, p < 0.001; adjacent vs. one-space, b = 0.06, t = 2.11, SE = 0.03, p = 0.04; one-space vs. two-space, b = 0.06, t = 2.07, SE = 0.03, p = 0.04]. Similarly, in total word reading times, all differences between conditions were reliable [normal vs. adjacent, b = 0.13, t = 3.75, SE = 0.03, p < 0.001; adjacent vs. one-space, b = 0.07, t = 2.14, SE = 0.03, p = 0.03; one-space vs. two-space, b = 0.09, t = 2.57, SE = 0.03, p = 0.01]. Clearly, there was a cost associated with letter transpositions in comparison to normally presented text and this cost increased as the distance between TLs increased.

For all four measures, the effect of predictability was either significant or marginally significant [single fixation duration: b = −0.11, t = −2.06, SE = 0.05, p = 0.04; first fixation duration: b = −0.08, t = −1.9, SE = 0.04, p = 0.06; gaze duration: b = −0.15, t = −2.71, SE = 0.05, p = 0.01; total word reading time: b = −0.24, t = −2.37, SE = 0.1, p = 0.02]. Similarly, the effect of word frequency was significant for all measures [single fixation duration: b = −0.02, t = −2.38, SE = 0.01, p = 0.02; first fixation duration: b = −0.01, t = −2.41, SE = 0.01, p = 0.02; gaze duration: b = −0.03, t = −4.43, SE = 0.01, p < 0.001; total word reading time: b = −0.04, t = −2.85, SE = 0.01, p < 0.01].

In order to compare our data with those reported by Perea et al., we examined the effect size among conditions from the model for our gaze duration data [this measure is generally accepted as a reliable index of lexical identification time, see Clifton, Staub, & Rayner, 2007; Rayner, 2009]. The calculated decreases in similarity match value for each of the three models as a function of the distance between TLs, as well as the effect sizes that we observed in our data, are reported in Table 3.

Table 3 The decrease in similarity match values for our target words, generated by the Match Calculator software [Davis, 2001], associated with increasing distance between TLs for SOLAR, open-bigram, and SERIOL models; the effect size between conditions as observed on gaze duration in Experiment 1; and the effect size in lexical decision times reported by Perea et al. [2008]

Full size table

We found that the effect sizes between conditions were of similar magnitude: an increase of 28 ms between adjacent and one-space transpositions, and a further increase of 24 ms between one- and two-space transpositions. Thus, processing difficulty continued to increase substantially as the distance between transposed letters increased. In contrast to Perea et al.’s data, we found two-space transpositions to be significantly more disruptive than one-space transpositions with an effect size that was comparable to that between adjacent and one-space transpositions.

Experiment 1b

We analysed response times and accuracy to the real word stimuli in the three different transposition conditions. Responses where the decision time was more than three standard deviations from the overall mean were excluded from both analyses. These data are summarised in Table 2. For these analyses, the adjacent condition was used as the baseline condition in the LME model, and subsequent contrasts were performed to compare one-space and two-space transpositions. We included the log frequency of the words as a fixed effect in the analyses to account for as much variance as possible within the model [although the transposed letter manipulation was conducted within-items and so could not be confounded with word frequency]. We initially included an interaction term between frequency and experimental condition; however, this was not significant and so was removed from the model. A logistic LME model was run on response accuracy.

Response times

We analysed response times for correct responses to the real word stimuli. There was a significant effect of frequency on response times, such that low-frequency words elicited longer response times than high-frequency words [b = −0.03, t = −3.91, SE = 0.001, p < 0.001]. With respect to experimental condition, response times to words containing one-space TLs were significantly longer than response time to words containing adjacent TLs [b = 0.11, t = 4.7, SE = 0.02, p < 0.001]. However, there was no significant difference between response times to one-space and two-space TL words [b = 0.04, t = 1.52, SE = 0.03, p = 0.13]. We also examined the effect size among the conditions for the response time. Unlike Experiment 1a, the effect size between one-space and two-space TL conditions was much smaller than the effect size between adjacent and one-space TL conditions.

Response accuracy

There was no significant effect of word frequency on response accuracy [p > 0.1]. Response accuracy was high in the adjacent TL condition [94 %] but was significantly poorer in the one-space TL condition [83 %; z = −4.59, SE = 0.29, p < 0.001]. Furthermore, response accuracy was significantly poorer in the two-space TL condition than in the one-space TL condition [68 %; z = −5.89, SE = 0.22, p < 0.001]. We conducted a one-sample t test, comparing response accuracy in the two-space TL condition to 50 % [e.g., performance at chance], and found it to be significantly higher [t 1 [17] = 29.49, p < 0.001; t 2 [59] = 28.52, p < 0.001]. These data clearly show that, when presented in isolation, increasing the distance between transposed letters within a word impacted substantially on the participants’ ability to identify the word. However, performance was always above chance—participants were able to discriminate reliably between real words containing transpositions and nonwords.

Discussion

In summary, the data from Experiment 1a show, first and foremost, that there is a clear disruption associated with reading transposed text; the adjacent TL condition led to a significant increase in all reading time measures compared with the normal text condition. Second, the data also show a clear effect of the distance between transposed letters. As the distance between transposed letters increased, reading times increased, reflecting the reader’s greater difficulty in identifying the words. We calculated the effect size for gaze duration, considered to be a reliable index of lexical identification during reading [Clifton et al., 2007; Rayner, 2009]. These data indicated that the increase in difficulty was proportionally similar between adjacent and one-space TLs and between one and two-space TLs. Our data show the greatest effect size [reflecting the greatest increase in processing difficulty] between the normal condition and the adjacent TL condition.

The data from Experiment 1a are inconsistent with those from the priming study reported by Perea et al. [2008]. They found the difference between one- and two-space TLs to be considerably smaller than that between adjacent and one-space transpositions, and consequently argued that their data were more consistent with the SERIOL model than the SOLAR or open-bigram models. Importantly, the different patterns of results reported here largely stems from the amount of disruption associated with the two-space TL condition compared with the one-space TL condition. Perea et al. found that for a base word, such as question, the two-space TL string quisteon was almost as similar to the adjacent TL form, quetsion, as it was to the one-space TL form, queitson. In contrast, the eye movement data from the present experiment showed that there was almost as great a difference between quisteon [two-space] and queitson [one-space] as there was between queitson [one-space] and quetsion [adjacent]. Note, however, that transposed letter effects in the Perea et al. study were computed by comparison with a substituted letter condition, which was not included in the present experimental design.

In Experiment 1b, we used a second task in which the transposed letter strings were presented in isolation [e.g., with no parafoveal processing preceding the processing during direct fixation, and with no sentence context to support identification], and participants were asked to decide whether these letter strings were based on real words or not, when intermixed with an equal number of nonwords. For this task, we observed a much greater effect size between the adjacent and one-space TL conditions than between the one-space and two-space TL conditions. Thus, the pattern of data for this task was similar to that reported by Perea et al. [2008] and contrasts with the eye movement data from Experiment 1a where the effect size between conditions was more consistent. These analyses clearly indicate, therefore, that while the direction of the effect is very consistent across studies, the magnitude of the effect between conditions is affected by the particular task used. When participants respond to words in isolation [either TL strings, or correctly spelled words following a TL prime], there is a proportionally large increase in disruption to processing when there is a letter in between the two transposed letters, but further increasing the distance between TLs within the word has a much smaller impact. In contrast, when identifying words as part of a natural reading task [where parafoveal pre-processing and processing of a meaningful sentence context are inherently linked to the identification of individual words], then there is a more consistent increase in the disruption associated with TL letter strings as the distance between TLs increases [see General discussion].

To summarise, when presented in isolation, the presence of an intervening letter between TLs appears to cause substantial processing difficulty for participants [this also is seen in the response accuracy data from Experiment 1b]. When presented within a sentence, however, participants are able to process information more incrementally, that is, processing involving the combination of parafoveal pre-processing and information obtained during direct fixation, as well as information from the sentence context to facilitate identification of TL letter strings. In this latter case, the disproportionate increase in difficulty associated with having an intervening letter between TLs is not observed. Overall, the data from the present experiment [from both tasks] showed that as the distance between transposed letters increased, the reader’s difficulty in identifying that word also increased.

Experiment 2

In Experiment 2, we examined the effect of letter transpositions when the manipulated letters were both consonants, both vowels, or one was a consonant and one was a vowel. Much research has concluded that consonants and vowels are differentially processed during lexical identification in a number of languages [Berent & Perfetti, 1995; Buchwald & Rapp, 2006; Caramazza, Chialant, Capasso, & Miceli, 2000; Chetail & Content, 2012; Monaghan & Shillcock, 2003; Taft & Krebs-Lazendic, 2013; Winskel & Perea, 2013].

One study used the Stroop task to examine readers’ processing of the consonant/vowel structure of words [Berent & Maron, 2005]. Participants were presented with printed nonwords and had to name the ink colour that they were printed in. The CV structure of the pseudoword was manipulated in relation to the CV structure of the ink colour. For example, the printed letter string pof [CVC] is congruent when printed in red [CVC] ink, but incongruent when printed in green [CCVVC]. They found task performance was facilitated when the skeletal structure of the letter string was congruent with the colour they were naming and argued that readers automatically process the skeletal structure of a word during lexical identification.

Eye movement studies have examined the time course of processing consonants and vowels during fixations in sentence reading [Lee, Rayner, & Pollatsek, 2001, 2002]. Participants read sentences where the presentation of a critical letter within a target word was delayed until either 30 or 60 ms after fixation onset on that word [it was replaced by a dash in the initial portion of the fixation]. The critical letter was either a consonant or a vowel. The results showed that reading was disrupted by the delayed presentation of consonants but not vowels in the 30 ms condition, but by both consonants and vowels in the 60 ms condition [Lee et al., 2001]. A similar pattern of results was found using the fast priming paradigm, where a prime string was briefly presented at fixation onset before being replaced by target [Lee et al., 2002]. Primes had either the same consonants or the same vowels as the target words, or were control primes. The results from these studies suggest that consonants are processed earlier and/or more quickly than vowels in the initial stages of lexical identification.

Other research has indicated that the encoding of consonants is less flexible than the encoding of vowels during lexical identification [Cutler, Sebastián-Gallés, Soler-Vilageliu, & van Ooijen, 2000]. Cutler et al. asked participants to listen to nonwords [such as kebra] and then change them into words by replacing one phoneme. In all cases, a word could be formed by replacing either a consonant [to zebra] or a vowel [to cobra]. Participants were more likely [and faster] to change vowels than consonants, and so Cutler et al. argued that vowels play a more flexible role in lexical activation.

A few studies have examined processing of TL letter strings, specifically manipulating whether the TLs were consonants or vowels. Perea and Lupker [2003a, 2004] used masked priming and lexical decision tasks to compare the [nonadjacent, one-spaced] transposition of two consonants [CC] or two vowels [VV] in both English [Perea & Lupker, 2003a] and Spanish [Perea & Lupker, 2004]. They found that CC transpositions caused significant priming for the base word and that lexical decisions were significantly slowed by CC transpositions. In contrast, they did not find significant priming effects from VV transpositions, and although lexical decisions were slowed by this manipulation, the effect was smaller than in the case of CC transpositions. They concluded that “vowel transposition nonwords are perceptually similar to their base words,” but “they are simply less similar than consonant transposition nonwords” [Perea & Lupker, 2004, p. 240]. Similar results have been reported by New, Araújo, and Nazzi [2008], who also used a masked priming lexical decision task.

Johnson [2007] examined whether readers are sensitive to consonant/vowel status in parafoveal processing during sentence reading. Sentences were presented using the boundary paradigm [Rayner, 1975], where the letter string in a target location contained a letter transposition prior to direct fixation [when it then appeared normally]. The two transposed letters were either two consonants or two vowels. Whereas readers were sensitive to the preview manipulation overall [reading times were longer when the preview had been a TL string compared with the control identity condition], this effect was not modulated by the consonant/vowel status of the two transposed letters. Johnson concluded that there is no differential processing of consonants and vowels in the parafovea during normal sentence reading and that parafoveal TL effects reflect low-level visual word recognition processes.

A slightly different pattern of results from Thai readers was reported by Winskel and Perea [2013]. No differences were found between consonant and vowel transpositions in single or first fixation durations, but there was a difference in gaze durations such that a transposed letter effect was found for consonant manipulations but not for vowel manipulations. The authors argue that this reflects the more critical role of consonants compared to vowels in Thai; a detailed discussion is not relevant here given the present data set relate to reading of English.

To directly examine the temporal locus of consonant/vowel differences in lexical identification, Perea and Acha [2009] compared consonant and vowel TL effects using two different tasks – a masked priming lexical decision task, and a same-different task. In the latter task, the participant must make a same-different decision for a lower-case prime [e.g., faith] and an upper-case target [e.g., FAITH]. This task is argued to reflect very low-level, prelexical, abstract letter encoding. In both tasks, the primes contained a letter transposition of either two vowels or two consonants. In the lexical decision task, significant priming was found from CC transposition primes only. In contrast, in the same-different task, significant [and equal] priming was found from both CC and VV transposition primes. Perea and Acha argued, therefore, that letter position encoding occurs at an earlier stage of lexical identification than does the distinction of consonants and vowels. This is consistent with Johnson’s finding that there is no difference between CC and VV letter transpositions in terms of parafoveal preview benefit, and with the results from several ERP studies showing differences between early and late time windows [Carreiras, Gillon-Dowens, Vergara, & Perea, 2008; Carreiras, Vergara, & Perea, 2008]. They suggest that later, phonological, stages of lexical access might be where differential processing of consonants and vowels occurs.

The literature on processing of consonants and vowels within isolated word recognition tasks clearly shows some significant variability in results. It seems likely that this variability is task-dependent, affected by the stage of processing that is being tapped into by the particular experimental paradigm employed. We wished to examine the influence of consonant and vowel transpositions during a natural sentence reading task. By measuring eye movements during reading, we can obtain an extremely detailed and sensitive index of the psychological processes underlying encoding and identification of words, and the time course of their interpretation in respect to sentential context.

In Experiment 2, we examined reading times on target words that contained transpositions of either two consonants [CC], two vowels [VV], or a consonant and a vowel [CV]. On the basis of the literature, we predicted that CV transpositions would cause the longest reading times, because they disrupt the skeletal structure of the word. It was less straightforward to generate a prediction for the CC and VV conditions in relation to each other. Some studies, including eye movement studies, have suggested that consonants are encoded earlier/faster, and less flexibly than vowels [Lee et al., 2001, 2002; Cutler et al., 2000]. On this basis, we would expect CC transpositions to be more disruptive than VV transpositions, because the CC transpositions would impact early, inflexible processing through which the structure of the word is determined, thus hindering lexical identification. However, the data reported from lexical decision tasks [Perea & Acha, 2009; Perea & Lupker, 2003a, 2004] show better priming for the base word from CC transpositions than from VV transpositions. On this basis, we would expect VV transpositions to be more disruptive than CC transpositions, because the CC transpositions have greater perceptual similarity to, and facilitate identification of, the base word. As in Experiment 1, two different tasks were used: an eye movement task, and an isolated word decision task.

Experiment 2a: methods

Participants

Participants were 21 students at the University of Southampton, with an age range of 18 to 30 years. They were all native English speakers with normal, or corrected-to-normal, vision and no known reading difficulties. All participants were naïve regarding the purpose of the study, and were paid for their participation.

Apparatus

Monocular eye movement recordings [with binocular viewing] from the right eye were taken using an EyeLink 1000 eye tracker. The position of the participant’s right eye was recorded every millisecond. Sentences were presented in white, Courier New font size 14, on a black background, on a 21-inch monitor. For 7 participants, the tower mounted eye tracker was used, with a viewing distance of 57 cm, whereas for 14 participants the desk mounted eye tracker was used with a viewing distance of 70 cm. At 57 cm, one character space subtended 0.3°, whereas at 70 cm, one character space subtended 0.2° deg. There were no significant differences between these two groups of participants [p > 0.1 for all], and so all analyses in the Result section are based on the combined data set. Participants leaned on chin and forehead rests during the experiment to minimise head movements.

Materials and design

Thirty triplets of target words were selected. In each target word, two letters were transposed. These letters were either two consonants, two vowels, or a consonant and a vowel [letter status – CC, VV, or CV]. In 15 triplets, the two transposed letters were adjacent within the word [letters 2 and 3], whereas in the other 15 triplets the two transposed letters were separated by one character [letters 2 and 4]. Thus, the transposed letters either could [adjacent] or could not [one-space] form a single grapheme [the orthographic representation of a phoneme; see Rey, Ziegler, & Jacobs, 2000]. None of the transpositions crossed a morpheme boundary, and they were generally all within a syllable. Importantly, none of the transposed letter stimuli [that were pronounceable nonwords] yielded more syllables than the base word [i.e., the syllabic structure was the same for the control and experimental items]. Furthermore, there were no significant differences between the experimental conditions in terms of the frequency of the target words [p > 0.1]. As a consequence of these tight controls, we were limited in the number of target words that were available for inclusion. For this reason, and to maximise the power of analyses concerning our primary manipulation [consonant and vowel transpositions], we did not include additional control conditions where the target words were presented normally. Note that the basic cost associated with letter transpositions compared with normal text has been replicated many times in the literature [Christianson et al., 2005; Johnson et al., 2007; Perea & Lupker, 2003a, b, 2004; Rayner, White et al., 2006; White et al., 2008], as well as in Experiment 1. Note also that single fixation durations, first fixation durations, and gaze durations for adjacent and one-space TLs are highly similar to those reported in Experiment 1a [Table 2].

Three sentence frames were written for each triplet of target words. An example is shown in Table 1 [adjacent TLs]. Three counterbalanced files were created such that each participant read every target word but only saw each sentence frame once. The sentences were prescreened for comprehension, plausibility, and predictability of the target words. There were no significant differences between conditions for any of these prescreen measures [p > 0.08].

Procedure

Same as in Experiment 1a.

Experiment 2b: methods

Participants

The participants were 18 native-English speakers who were undergraduates from the University of Southampton, aged 18-21 years, with normal or corrected-to-normal vision and no known reading difficulties. All participants were naïve regarding the purpose of the experiment and were compensated for their time either through course credit or through cash.

Apparatus

Same as in Experiment 1b.

Materials and design

The stimuli were the 90 target words from Experiment 2a, in their TL format, intermixed with 90 pronounceable, orthographically legal nonwords that were matched for length. The 90 nonwords were matched to the real words in terms of having consonants or vowels [or one of each] in the critical positions: letters 2 and 3 for comparison with adjacent TLs, and letters 2 and 4 for comparison with one-space TLs. In total, each participant saw all 180 letter strings: 90 real words containing transpositions [in the same six conditions as in Experiment 2a], and 90 nonword fillers.

Procedure

Same as in Experiment 1b.

Results

All results were analysed using a LME model, with participants and items as crossed random effects. The p values were estimated using posterior distributions for model parameters obtained by Markov-Chain Monte Carlo sampling. Reading time and response time data were log-transformed prior to analysis. The consonant-consonant TL condition was used as the baseline [i.e., the intercept was the mean of this condition], and subsequent contrasts were performed to compare the consonant-vowel and vowel-vowel TL conditions. Again, as in Experiment 1a, we initially included interaction terms between our experimental manipulations and word frequency. In most cases, these interaction terms did not improve the fit of the models to the data and were removed. In the case of gaze duration, the condition by frequency interaction terms improved the fit of the model but did not have a significant effect within the data. For this reason, and for clarity when comparing results across measures, we report the model without the interaction term.

Experiment 2a

Data elimination and trimming procedures were the same as those used in Experiment 1a. All participants scored at least 78 % on the comprehension questions [mean comprehension was 91 %]. Again, we examined single fixation durations, first fixation durations, gaze durations, and total word reading times on the target words [Table 4]. On all four measures, there was a significant effect of transposition distance; one-space transpositions disrupted reading more than adjacent transpositions [t > 2, p < 0.01]. Also, on all four measures the effect of word frequency was significant [t > 3, p < 0.01].

Table 4 Means for single fixation durations, gaze durations, total word reading times, and regressive fixation durations, and regression in probability, across the six TL conditions in Experiment 2a, and for response times Experiment 2b

Full size table

The differences among CC, CV, and VV transpositions were not significant for single fixation durations [t < 2, p > 0.09] or first fixation durations [t < 2, p > 0.3]. Furthermore, for neither of these measures did the interaction between condition and transposition distance contribute to the model [t < 2, p > 0.2]. Thus, for these early measures of processing, there was no difference between consonant and vowel transpositions. In gaze duration, there were no significant main effects of consonant-vowel condition [t < 2, p > 0.1]. There was, however, a significant interaction between transposition distance and letter status in both the main LME model [comparing the CV transposition against the baseline CC condition; b = −0.11, t = −2.01, SE = 0.06, p = 0.04] and in the subsequent planned contrast [comparing CV and VV conditions; b = 0.11, t = 1.95, SE = 0.06, p = 0.05]. As described, there was an overall effect of transposition distance such that adjacent transpositions resulted in shorter total fixation times than one-space transpositions. These interaction terms indicate, however, that there was a minimal cost associated with 1-space TLs in the CV condition compared with the increased gaze durations on 1-space TLs in both the CC and VV conditions [Table 4; CC = 41-ms increase from adjacent to 1-space TLs, VV = 56-ms increase, CV = 12-ms increase]. Thus, the data from first pass reading times indicate some differential level of processing difficulty associated with CV transpositions; this pattern became more substantial in later reading time measures.

There were overall differences between consonant-vowel conditions in total word reading times. Longer total word reading times resulted from CV transpositions than from CC transpositions [b = 0.18, t = 3.37, SE = 0.05, p < 0.001] or from VV transpositions [b = −0.23, t = −4.4, SE = 0.05, p < 0.001]. The difference between CC and CC transpositions was not significant [b = −0.05, t = −0.99, SE = 0.05, p = 0.3].

As observed in gaze durations, there was again a significant interaction between letter status [the CV condition] and transposition distance in both the main LME model [compared with the baseline CC condition; b = −0.35, t = −4.39, SE = 0.08, p < 0.001] and in the planned contrast [compared with the VV condition [b = 0.39, t = 5.05, SE = 0.08, p < 0.001]. As in gaze duration, these significant interaction terms show that there was a disproportionately small increase in reading times associated with one-space transpositions in the CV condition. In total fixation times, the increase from adjacent to one-space TLs was more than 100 ms in both CC and VV conditions but was just 4 ms for CV transpositions. As a consequence of this, the difference between consonant-vowel conditions also was reduced in the one-space condition [Table 4].

The difference between consonant and vowel TLs was most evident in total word reading time—a late measure of processing. This is consistent with previous research, which suggests that the effects of consonant/vowel status appear later in processing. We investigated the source of this effect further by examining both the probability of making a regression onto the target words as well as the durations of those regressive fixations. A logistic LME model was run on regression probability. In both cases, including the interaction terms between our experimental manipulations and word frequency significantly improved the fit of the models to the data [p ≤ 0.05] and had significant effects within the models; the models reported include, therefore, these interaction terms.

Regressions back to adjacent CC target words were 1 % less likely than to CV transpositions, and this difference was not significant [z = 0.66, SE = 0.34, p = 0.5]. In contrast, VV transpositions led to 4 % fewer regressions back onto the target word, and this was significant [z = −2.91, SE = 0.37, p < 0.01]. There was no overall difference between adjacent and one-space transpositions [z = −0.53, SE = 0.41, p = 0.6]. There was a significant effect of word frequency on regression probability [z = −2.74, SE = 0.1, p < 0.01], as well as a significant interaction between letter status [the VV condition] and word frequency [z = 2.1, SE = 0.15, p = 0.04], such that fewer regressions were made back to high-frequency target words than to low-frequency target words, but this effect was attenuated in the VV condition compared with the CC and CV conditions.

Regressive fixations made on words containing a VV transposition were shorter than those made on words containing either a CC transposition [b = −0.48, SE = 0.19, t = −2.61, p = 0.01] or a CV transposition [b = −0.75, SE = 0.21, t = −3.64, p < 0.01]. There also was an effect of transposition distance, such that one-space TLs resulted in longer regressive fixations than did adjacent TLs [b = 0.44, SE = 0.19, t = 2.34, p = 0.02]. Finally, there was a significant interaction between the VV condition and word frequency, such that once again the frequency effect was attenuated on words containing VV transpositions [main LME model: b = 0.17, SE = 0.08, t = 2.24, p = 0.03; planned contrast: b = 0.19, SE = 0.08, t = 2.41, p = 0.02]. No other effects were significant [all t < 2, all p > 0.05].

Thus, the effect observed in total word reading times was related to both the probability of making a regression back to the target word and the durations of regressive fixations. Consonant-vowel transpositions received more regressions, and longer regressive fixations, than consonant-consonant or vowel-vowel transpositions.

Experiment 2b

We analysed response times and accuracy to the real word stimuli in the six different transposition conditions. Responses where the decision time was more than three standard deviations from the overall mean were excluded from both analyses. These data are summarised in Table 4. For these analyses, the adjacent condition was used as the baseline condition in the LME model. We included the frequency of the words as a fixed effect in the analyses to account for as much variance as possible within the model [although the transposed letter manipulation was conducted within-items and could not be confounded with word frequency]. We initially included an interaction term between frequency and experimental condition; however, this was not significant and was removed from the model [this is unsurprising, given that the experimental conditions were controlled so that they did not differ in terms of target word frequencies]. A logistic LME model was run on response accuracy.

Response times

There was a significant effect of frequency such that high-frequency words elicited shorter response times than low-frequency words [b = −0.03, t = −3.09, SE = 0.01, p < 0.01]. With respect to the experimental manipulations, response times to words containing a VV transposition were not significantly different from response times to words containing a CC transposition [b = 0.004, t = 0.1, SE = 0.05, p = 0.9]. In contrast, response times to words containing a CV transposition were significantly longer than both CC transpositions [b = 0.15, t = 3.05, SE = 0.05, p < 0.01] and VV transpositions [b = −0.15, t = −2.89, SE = 0.05, p < 0.01. Thus, consistent with the eye-movement data, response times showed that CV transpositions were most disruptive. In addition to these effects of consonant-vowel status, there also was an effect of the distance between transposed letters: one-space transpositions resulted in longer response times than adjacent transpositions [b = 0.21, t = 3.69, SE = 0.06, p < 0.01]. This effect is consistent with Experiments 1a, 1b, and 2a showing that the presence of an intervening letter between TLs causes increased processing difficulty. Here, the interactions between consonant-vowel status and the distance between TLs were not reliable for response times [t < 1, p > 0.3]. Thus, in contrast to the eye movement data [where there were differences between conditions for adjacent transpositions, but there was a minimal cost associated with increasing the distance between TLs in the CV condition, resulting in a smaller difference between one-space consonant-vowel conditions], we observed the same pattern for both adjacent and one-space transpositions in response times to the TL words presented in isolation. Consistent across data from both tasks was the observation that CV transpositions caused the most processing difficulty.

Response accuracy

There was no significant effect of word frequency on response accuracy [z = 1.22, SE < 0.01, p = 0.22]. Within the adjacent transposition stimuli, response accuracy was reasonably high; 86 % for CC transpositions. This increased slightly, but not significantly, for VV transpositions [91 %; z = 0.73, SE = 0.68, p = 0.46]. There was, however, a significant decrease in response accuracy for CV transpositions [69 %; z = −2.43, SE = 0.63, p = 0.02]. There also was a significant decrease in response accuracy associated with having an intervening letter between TLs compared with adjacent transpositions, such that in all three conditions [CC, VV, and CV] response accuracy was less than 50 % [z = −4.72, SE = 0.64, p < 0.001]. As was observed in the response time data, there were no significant interactions between the decrease in response accuracy associated with one-space TLs and the consonant-vowel status of the transposed letters [all z < 2, all p > 0.1]. For the CC condition [the baseline condition in our LME model], we conducted one-sample t tests to compare response accuracy for one-space TL words to 50 %. There was no significant difference [t 1 [17] = 0.82, p = 0.42; t 2 [14] = 0.35, p = 0.73], indicating that participants were performing at chance in their efforts to distinguish between real words with transposed letters and nonwords when there was an intervening letter between the TLs. Overall, these data show very clearly: 1] that CV transpositions caused more disruption to word identification than either CC or VV transpositions; and 2] that when presented in isolation, the presence of an intervening letter between TLs causes extreme difficulty for participants in terms of their ability to discriminate those real TL words from nonwords.

Discussion

In the eye-movement data, for adjacent transpositions, the most disruption was associated with CV transpositions with significantly less disruption to reading resulting from CC or VV transpositions. While there was some indication in gaze duration of a differential processing cost associated with CV transpositions, the effect was clearest in total word reading times, and resulted from both the probability of regressing back to the target words and the duration of regressive fixations. In contrast, although there were minimal differences between letter status transpositions when the TLs were one space apart within the word. This seems to have been driven by the fact that there was a cost of more than 100 ms associated with increasing the distance between TLs for both CC and VV transpositions, but virtually no cost at all [just 4 ms] for CV transpositions.

When the target words were presented in isolation [Experiment 2b], we found a similar pattern in that CV transpositions caused the most processing difficulty. Here, the greater difficulty of processing CV TL strings was maintained in the one-space condition, as all three consonant-vowel conditions showed a cost associated with increasing the distance between TLs. Strikingly, in the one-space conditions, response accuracy was reduced to chance; participants appeared to be unable to discriminate between real words with transposed letters and nonwords. This is consistent with the lack of a difference that we observed between one-space consonant-vowel conditions in the total fixation time data from Experiment 2a, indicating that participants were unable to identify the TL words alone and were reliant on the sentence context to work out what the word should be.

The finding that consonant-vowel transpositions caused the greatest disruption was as predicted, given that such transpositions changed the basic, CV structure of the word. As far as we are aware, this is the first experiment to have included this condition. With respect to the CC and VV transpositions, our data differ slightly from those from lexical decision tasks [Perea & Acha, 2009; Perea & Lupker, 2003a, 2004]. These studies found CC transpositions to have greater perceptual similarity to their base words and, hence, caused greater priming than VV transpositions. In contrast, we did not find a significant difference between CC and VV transpositions. As with Experiment 1, it seems most likely that this difference in results is attributable to the particular task employed [see Discussion for Experiment 1].

An additional factor that may contribute to these consonant-vowel differences is whether or not the transposed letters formed a single grapheme. It is possible that, as a consequence of the letter transpositions, participants identified these words through grapheme-phoneme decoding strategy [“sounding out” the words]. If this were the case, then our consonant-vowel manipulations may have impacted differentially upon that decoding process. Two adjacent vowels, such as ea in healthy, almost always formed a single grapheme [corresponding to a single phoneme; 14/15 target word transpositions, whilst an adjacent consonant and vowel, such as ec in secret, never formed a single grapheme [0/15 target word transpositions]. It is possible that when the transposed letters are within a grapheme, the transposition causes minimal disruption and that transpositions between graphemes are more disruptive to lexical identification. In the nonadjacent conditions, none of the transpositions were within a single grapheme [e.g., i and e in cinema, k and t in sketch, or e and l in health]. Supporting this suggestion was the finding that, within the adjacent conditions, consonant-vowel transpositions resulted in the longest reading times, whereas in the nonadjacent conditions, where all transpositions were between-graphemes, all letter status transpositions were equally disruptive. However, we did not find a significant difference between CC and VV transpositions in the adjacent conditions. Two adjacent consonants, such as th in ethics, sometimes formed a single grapheme [3/15 target word transpositions] but mostly did not, such as nt in antler, where each letter is an individual grapheme [thus, each one corresponded to a separate phoneme]. Whilst there was a numerical trend for adjacent VV TL words to have shorter reading times than adjacent CC TL words [a 79-ms difference in total fixation times, Experiment 2a], this was not significant. The data indicate, therefore, that grapheme-phoneme decoding processes are not solely responsible for the differences that we observed between our consonant-vowel transposition conditions.

Our finding that the relative differences between consonant and vowel transpositions influenced later measures of processing was consistent with other studies [Carreiras, Gillon-Dowens et al., 2008; Carreiras, Vergara, et al., 2008; Johnson, 2007; Perea & Acha, 2009]. Note, however, that the basic finding of a cost to reading times associated with TLs was reflected in first fixation duration, single-fixation duration and gaze duration data. These data were highly similar to the adjacent and one-space TL data reported from Experiment 1 [which included the control, nontransposed condition; Tables 2 and 4]. So, our data show a general influence of letter transpositions on early measures of processing, as is well established in the literature, but it was the relative differences between consonant and vowel transpositions that affected later measures of processing.

Compared with the control condition of Experiment 1a, whereas single fixation and gaze duration data were highly similar, total word reading times in Experiment 2a were inflated. This is likely due to the design of these experiments. In Experiment 1a, the critical manipulation was made such that every target word appeared in each condition. Thus, the sentence frames were written for each target word individually; whereas the target words were low predictability, the sentence for each was plausible with respect to the identification of that particular target word [for example, “The athlete tore a weak ligament in his shoulder and was angry.”]. In contrast, in Experiment 2a, the critical manipulation was made such that target words were arranged in triplets depending on whether their second and third/fourth letters were consonants or vowels, and so each target word could only belong to one condition. Thus, each sentence frame was deliberately written to be as neutral as possible, so that all three target words from any given triplet were equally plausible within the sentence [for example, “Sam was learning about the new system/feature/format at college today.”]. It seems likely that the sentence frames in Experiment 1a better supported identification of the TL target words than the sentence frames in Experiment 2a [although note that, ultimately, the prescreen for comprehension indicated that all words in both experiments could be successfully identified when presented within their sentence contexts]. In support of this argument, target word predictability was 0.19 in Experiment 1a, while it was less than 0.01 in Experiment 2a. We suggest that this is the reason for the increased total word reading times in Experiment 2a compared with Experiment 1a.

In the second task of each experiment, where the TL words were presented in isolation, intermixed with nonwords, we also observed an overall difference in terms of response accuracy. In Experiment 1b, our experimental manipulations affected response accuracy but the poorest performance observed was 68 % [in the two-space condition] and was significantly above chance. In Experiment 2b, we also observed effects of our experimental manipulations on response accuracy; however, here response accuracy decreased to less than 50 % in the one-space condition and was not significantly different to chance. There was, therefore, a disproportionate decrease in response accuracy for one-space TL words presented in isolation in Experiment 2 compared with in Experiment 1. The most likely cause of this discrepancy is the length of the target words used. In Experiment 1, the target words were all eight letters long, and so transposing two letters disrupted 25 % of the orthographic input. In contrast, in Experiment 2, the target words were six letters long and so transposing two letters disrupted 33 % of the orthographic input. The higher proportion of letters that were affected by the transposition in Experiment 2 seems to have resulted in particularly poor response accuracy when participants were presented with those TL letter strings in isolation and asked to discriminate them from nonwords.

As discussed in the Introduction, models of visual word identification account for transposed letter effects in reading through one of two main mechanisms: 1] open-bigram coding, incorporated in both the SERIOL model [Whitney, 2001, 2008; Whitney & Cornelissen, 2008] and open-bigram model [Grainger & van Heuven, 2003], or 2] spatial coding, incorporated in the SOLAR model [Davis, 1999, 2010]. None of these models of lexical identification differentiate between consonants and vowels in their current implementations. Our data from Experiment 2 show that transpositions, which alter the CV structure of the word are significantly more disruptive than those that maintain the CV structure. As yet, however, this distinction is not incorporated in models of word recognition.

General discussion

These experiments build on the existing literature showing a cost to reading times associated with reading transposed text [Christianson et al., 2005; Johnson, 2007; Johnson & Eisler, 2012; Johnson et al., 2007; Perea & Lupker, 2003a, b, 2004; Rayner, White et al., 2006; White et al., 2008]. In both experiments, we reported different patterns of results between the sentence reading task and the isolated word recognition task. In Experiment 1, the cost associated with increasing the distance between TLs was more linear in the sentence-reading task than in the isolated-word task. In Experiment 2, increasing the distance between TLs increased processing time in all consonant-vowel conditions for the isolated-word task, whereas there was a minimal influence of having an intervening letter between CV TLs in the sentence-reading task [compared with CC and VV TLs]. This task difference also may account for the discrepancy between our reported eye-movement data from the sentence-reading task and the effects reported in the literature [e.g., Perea et al.]. In our isolated-word task [Experiments 1b and 2b], participants were required to decide for each trial whether the presented string of letters was a real word or a nonword. This lexicality judgement is not part of normal, silent reading, which was the task employed in Experiments 1a and 2a. While some researchers have found some correspondences between eye-fixation times during silent reading and lexical decision times in the form of moderate correlations [Schilling, Rayner, & Chumbley, 1998; Williams, Perea, Pollatsek, & Rayner, 2006], other studies have found these two measures to be poorly correlated [Everatt & Underwood, 1994; Kuperman, Drighe, Keuleers, & Brysbaert, 2013].

There are three key differences between isolated word decision tasks and more normal sentence reading, which are likely to have affected the pattern of results observed [for a detailed discussion, see Rayner & Liversedge, 2011]. First, in sentence reading the participant is able to obtain information about the target word in parafoveal preview as well as when it is being directly fixated. Parafoveal pre-processing is a critical component of skilled adult reading [Rayner, 1975; Rayner, Liversedge, & White, 2006]. No such opportunity is available to the participant during a lexical decision task where single words are presented in isolation. Second, the sentence context surrounding the target word will inevitably have supported the identification of the TL letter string. While our target words were all low predictability, they were still semantically plausible within the sentence frames. Such supporting information is, again, unavailable to the participant during a lexical decision task. It seems likely that the letter string quisteon can be more easily identified as the base word question when presented as part of a meaningful sentence than when presented in isolation. Third, in lexical decision tasks [although not the task we report in Experiments 1b and 2b here], the participant responds to the correctly spelled target; it is the prime that contains the transposed letter manipulation. In contrast, in the sentence reading task reported here, the participant must identify the word on the basis of the transposed letter string. The different patterns of results reported between tasks indicate that the opportunity for parafoveal preview, and the semantic context provided by a sentence frame can have a substantial effect on the participants’ processing of a TL string. Furthermore, whether the participant responds to/fixates a TL word [as in all the experiments reported here] or whether they respond to a correctly spelled word [as in a standard lexical decision task where the TL manipulation affects the prime rather than the target] also may account for differences between the present data sets and those reported in the literature [Perea et al., 2008; Perea & Acha, 2009; Perea & Lupker, 2003a, 2004]. With respect to this final point, we also note that the processing demands associated with our task are much greater than those in a standard lexical decision. Specifically, we required participants to determine, for each nonword letter string, whether or not it was a misspelled real word. The processing demands for this task could be considered somewhat similar to proofreading, and research on proofreading has demonstrated substantial differences in lexical processing associated with the particular task instructions and error types [Schotter, Bicknell, Howard, Levy & Rayner, 2014].

The novel task reported here, where participants were required to determine whether a TL letter string was a misspelled real word or not, when presented in isolation, provides valuable insight into the cost associated with processing such letter strings. Comparison with the eye movement data demonstrates the importance of a meaningful sentence context for successfully identifying TL letter strings. Both the distance between transposed letters, and the consonant-vowel status of the transposed letters, affects the ease with which the word can be identified. If the transposed letter manipulation affects letters that are both nonadjacent and that change the consonant-vowel structure of the word, then the reader is largely dependent on the sentence context to determine what the word should be; in the absence of the supporting context, readers cannot distinguish these TL strings from nonwords. These data are consistent with Norris’ [1986] criterion bias model, in which lexical access and word recognition are separate processes with an intervening checking process in which context is used to bias recognition of candidates within the set that are more plausible within the context. Such mechanisms could account for the finding of differences between consonant and vowel transpositions that appear exclusively in relatively late measures. Such mechanisms also might explain differences between the effects in our sentence reading and isolated word tasks, demonstrating participants’ reliance on sentence context for recognition of the correct lexical candidate on the basis of the misspelled letter string that was presented. Indeed, Norris argues that a strength of this model is its capacity to account for identification of misspelled words and also specifies that the time course of the checking process is dynamic and will be lengthened by any factor, such as a misspelling that increases the time between the generation of a candidate set and the recognition point [pp. 112-113].

With respect to theoretical models, the data from these experiments are relevant to both 1] models of lexical identification and 2] models of eye movement control during reading. First, to reiterate, current model implementations do not distinguish between processing of consonants and vowels, yet the data from Experiments 2a and 2b add to the literature showing that the skeletal structure of a word is processed during lexical identification. Second, the data from both Experiments 1a and 2a were entirely consistent with current models of eye movement control during reading. In such models, a primary determinant of when the eyes move during reading [e.g., fixation times on a word] is the ease of lexical processing. In both reading experiments reported here, we demonstrated that experimental manipulations that were hypothesized to be more disruptive to lexical processing did in fact increase processing times of those words.

In summary, the present experiments used isolated word recognition methodology and eye tracking to investigate how people process words with transpositions. Our results showed that readers can identify words containing letter transpositions, that the presence of an intervening letter between TLs is costly to identification, and that transpositions that change the CV structure of a word are more difficult to process than those that do not. Overall, the results provide insight into the nature of orthographic encoding, particularly during normal reading, and demonstrate the important role of sentential context in relation to flexible letter encoding in lexical identification.

References

• Baayen, Piepenbrock, & Gulikers. [1995]. The CELEX Lexical Database. [CD-ROM]. Philadelphia: Linguistic Data Consortium, University of Pennsylvania.

  Google Scholar 

• Berent, I., & Maron, M. [2005]. Skeletal structure of printed words: Evidence from the Stroop task. Journal of Experimental Psychology: Human Perception and Performance, 31, 328–338.

  PubMed  Google Scholar 

• Berent, I., & Perfetti, C. A. [1995]. A rose is a REEZ: The two-cycles model of phonology assembly in reading English. Psychological Review, 102, 146–184.

  Article  Google Scholar 

• Buchwald, A., & Rapp, B. [2006]. Consonants and vowels in orthographic representations. Cognitive Neuropsychology, 23, 308–337.

  PubMed  Article  Google Scholar 

• Caramazza, A., Chialant, D., Capasso, R., & Miceli, G. [2000]. Separable processing of consonants and vowels. Nature, 403[6768], 428–430.

• Carreiras, M., Gillon-Dowens, M., Vergara, M., & Perea, M. [2008a]. Are vowels and consonants processed differently? Event-related potential evidence with a delayed letter paradigm. Journal of Cognitive Neuroscience, 21, 275–288.

  Article  Google Scholar 

• Carreiras, M., Vergara, M., & Perea, M. [2008b]. ERP correlates of transposed-letter priming effects: The role of vowels versus consonants. Psychophysiology, 46, 34–42.

  PubMed  Article  Google Scholar 

• Chetail, F., & Content, A. [2012]. The internal structure of chaos: Letter category determines visual word perceptual units. Journal of Memory and Language, 67, 371–388.

  Article  Google Scholar 

• Christianson, K., Johnson, R. L., & Rayner, K. [2005]. Letter transpositions within and across morphemes. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31, 1327–1339.

  PubMed  Google Scholar 

• Clifton, C., Jr., Staub, A., & Rayner, K. [2007]. Eye movements in reading words and sentences. In R. van Gompel [Ed.], Eye movements: A window on mind and brain [pp. 341–372]. Amsterdam: Elsevier.

  Chapter  Google Scholar 

• Cutler, A., Sebastián-Gallés, N., Soler-Vilageliu, O., & van Ooijen, B. [2000]. Constraints of vowels and consonants on lexical selection: Cross-linguistic comparisons. Memory & Cognition, 28[5], 746–755.

• Davis, C. J. [1999]. The self-organising lexical acquisition and recognition [SOLAR] model of visual word recognition. Unpublished doctoral dissertation, University of New South Wales, Sydney, Australia.

• Davis, C. J. [2001]. Match calculator. [Computer software available at //www.pc.rhul.ac.uk/staff/c.davis/Utilities/]. Accessed 21 Sept 2010

• Davis, C. J. [2010]. The spatial coding model of visual word identification. Psychological Review, 117[3], 713–758.

  PubMed  Article  Google Scholar 

• Davis, C. J., & Bowers, J. S. [2006]. Contrasting five different theories of letter position coding: Evidence from orthographic similarity effects. Journal of Experimental Psychology: Human Perception and Performance, 32, 535–557.

  PubMed  Google Scholar 

• Everatt, J., & Underwood, G. [1994]. Individual differences in reading subprocesses: Relationships between reading ability, lexical access, and eye movement control. Language and Speech, 37, 283–297.

  PubMed  Google Scholar 

• Frankish, C., & Barnes, L. [2008]. Lexical and sublexical processes in the perception of transposed letter anagrams. The Quarterly Journal of Experimental Psychology, 61, 381–391.

  PubMed  Article  Google Scholar 

• Gomez, P., Ratcliff, R., & Perea, M. [2008]. The Overlap Model: A model of letter position coding. Psychological Review, 115, 577–601.

  PubMed Central  PubMed  Article  Google Scholar 

• Grainger, J., & van Heuven, W. J. B. [2003]. Modeling letter position coding in printed word perception. In P. Bonin [Ed.], The mental lexicon [pp. 1–23]. New York: Nova Science.

  Google Scholar 

• Johnson, R. L. [2007]. The flexibility of letter coding: Nonadjacent letter transpositions in the parafovea. In R. van Gompel, M. Fischer, W. Murray, & R. Hill [Eds.], Eye movements: A window on mind and brain. Oxford: Elsevier.

  Google Scholar 

• Johnson, R. L. [2009]. The quiet clam is quite calm: Transposed-letter neighborhood effects on eye movements during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35, 943–969.

  PubMed  Google Scholar 

• Johnson, R. L., & Dunne, M. D. [2012]. Parafoveal processing of transposed-letter words and nonwords: Evidence against parafoveal lexical activation. Journal of Experimental Psychology: Human Perception and Performance, 38, 191–212.

  PubMed  Google Scholar 

• Johnson, R. L., & Eisler, M. E. [2012]. The importance of the first and last letter in words during sentence reading. Acta Psychologica, 141, 336–351.

  PubMed  Article  Google Scholar 

• Johnson, R. L., Perea, M., & Rayner, K. [2007]. Transposed letter effects in reading: Evidence from eye movements and parafoveal preview benefit. Journal of Experimental Psychology: Human Perception and Performance, 33, 209–229.

  PubMed  Google Scholar 

• Kuperman, V., Drieghe, D., Keuleers, E., & Brysbaert, M. [2013]. How strongly do word reading times and lexical decision times correlate? Combining data from eye movement corpora and megastudies. The Quarterly Journal of Experimental Psychology, 66, 563–580.

  PubMed  Article  Google Scholar 

• Lee, H.-W., Rayner, K., & Pollatsek, A. [2001]. The relative contribution of consonants and vowels to word identification during reading. Journal of Memory and Language, 44, 189–205.

  Article  Google Scholar 

• Lee, H.-W., Rayner, K., & Pollatsek, A. [2002]. The processing of consonants and vowels in reading: Evidence from the fast priming paradigm. Psychonomic Bulletin & Review, 9, 766–772.

  Article  Google Scholar 

• Monaghan, P., & Shillcock, R. C. [2003]. Connectionist modelling of the separable processing of consonants and vowels. Brain and Language, 86, 83–98.

  PubMed  Article  Google Scholar 

• New, B., Araújo, V., & Nazzi, T. [2008]. Differential processing of consonants and vowels in lexical access through reading. Psychological Science, 19, 1223–1227.

  PubMed  Article  Google Scholar 

• Norris, D. [1986]. Word recognition: Context effects without priming. Cognition, 22, 93–136.

  PubMed  Article  Google Scholar 

• O’Connor, R. E., & Forster, K. I. [1981]. Criterion bias and search sequence bias in word recognition. Memory & Cognition, 9, 78–92.

  Article  Google Scholar 

• Perea, M., & Acha, J. [2009]. Does letter position coding depend on consonant/vowel status? Evidence with the masked priming technique. Acta Psychologica, 130, 127–137.

  PubMed  Article  Google Scholar 

• Perea, M., & Carreiras, M. [2006]. Do transposed-letter similarity effects occur at a prelexical phonological level? The Quarterly Journal of Experimental Psychology, 59, 1600–1613.

  PubMed  Article  Google Scholar 

• Perea, M., Duñabeitia, J. A., & Carreiras, M. [2008]. Transposed-letter priming effects for close versus distance transpositions. Experimental Psychology, 55, 397–406.

  Article  Google Scholar 

• Perea, M., & Lupker, S. J. [2003a]. Transposed-letter confusability effects in masked form priming. In S. Kinoshita & S. J. Lupker [Eds.], Masked priming: State of the art [pp. 97–120]. Hove, UK: Psychology Press.

  Google Scholar 

• Perea, M., & Lupker, S. J. [2003b]. Does judge activate COURT? Transposed-letter similarity effects in masked associative priming. Memory & Cognition, 31, 829–841.

  Article  Google Scholar 

• Perea, M., & Lupker, S. J. [2004]. Can CANISO activate CASINO? Transposed-letter similarity effects with nonadjacent letter positions. Journal of Memory and Language, 51, 231–246.

  Article  Google Scholar 

• Rayner, K. [1975]. The perceptual span and peripheral cues in reading. Cognitive Psychology, 7, 65–81.

  Article  Google Scholar 

• Rayner, K. [1998]. Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124, 372–422.

  PubMed  Article  Google Scholar 

• Rayner, K. [2009]. The Thirty Fifth Sir Frederick Bartlett Lecture: Eye movements and attention in reading, scene perception, and visual search. The Quarterly Journal of Experimental Psychology, 62, 1457–1506.

  PubMed  Article  Google Scholar 

• Rayner, K., & Liversedge, S. P. [2011]. Linguistic and cognitive influences on eye movements during reading. In S. Liversedge, I. Gilchrist, & S. Everling [Eds.], The Oxford Handbook of Eye Movements [pp. 751–766]. UK: Oxford University Press.

  Google Scholar 

• Rayner, K., Liversedge, S. P., & White, S. J. [2006a]. Eye movements when reading disappearing text: The importance of the word to the right of fixation. Vision Research, 46, 310–323.

  PubMed  Article  Google Scholar 

• Rayner, K., White, S. J., Johnson, R. L., & Liversedge, S. P. [2006b]. Raeding wrods with jubmled letetrs: There is a cost. Psychological Science, 17, 192–193.

  PubMed  Article  Google Scholar 

• Rey, A., Ziegler, J. C., & Jacobs, A. M. [2000]. Graphemes are perceptual reading units. Cognition, 75, B1–B12.

  PubMed  Article  Google Scholar 

• Schilling, H. E. H., Rayner, K., & Chumbley, J. I. [1998]. Comparing naming, lexical decision, and eye fixation times: Word frequency effects and individual differences. Memory & Cognition, 26, 1270–1281.

  Article  Google Scholar 

• Schotter, E. R., Bicknell, K., Howard, I., Levy, R., & Rayner, K. [2014]. Task effects reveal cognitive flexibility responding to frequency and predictability: Evidence from eye movements in reading and proofreading. Cognition, 131, 1–27.

  PubMed  Article  Google Scholar 

• Taft, M., & Krebs-Lazendic, L. [2013]. The role of orthographic syllable structure in assigning letters to their position in visual word recognition. Journal of Memory and Language, 68, 85–97.

  Article  Google Scholar 

• White, S. J., Johnson, R. L., Liversedge, S. P., & Rayner, K. [2008]. Eye movements when reading transposed text: The importance of word beginning letters. Journal of Experimental Psychology: Human Perception and Performance, 34, 1261–1276.

  PubMed Central  PubMed  Google Scholar 

• Whitney, C. [2001]. How the brain encodes the order of letters in a printed word: The SERIOL model and selective literature review. Psychonomic Bulletin & Review, 8, 221–243.

  Article  Google Scholar 

• Whitney, C. [2008]. A comparison of the SERIOL and SOLAR theories of letter-position encoding. Brain and Language, 107, 170–178.

  PubMed  Article  Google Scholar 

• Whitney, C., & Cornelissen, P. [2008]. SERIOL reading. Language & Cognitive Processes, 23, 143–164.

  Article  Google Scholar 

• Williams, C. C., Perea, M., Pollatsek, A., & Rayner, K. [2006]. Previewing the neighborhood: The role of orthographic neighbors as parafoveal previews in reading. Journal of Experimental Psychology: Human Perception and Performance, 32, 1072–1082.

  PubMed  Google Scholar 

• Winskel, H., & Perea, M. [2013]. Consonant/vowel asymmetries in letter position coding during normal reading: Evidence from parafoveal previews in Thai. Journal of Cognitive Psychology, 25, 119–130.

  Article  Google Scholar 

Download references

How many words of 3 consonants and 2 vowels can be formed?

How many words can be formed with 3 consonants and 2 vowels so that no two consonants remain together?

How many different words each containing 2 vowels and 3 consonants can be formed using all the vowels and 17 consonants?

How many words can be formed each of 2 vowels and 3 consonants from the letters of the given word mathematics?