NEUROPSYCHOLOGICAL REHABILITATION, 2002, 12 (3), 177–197
Training of developmental surface dyslexia improves reading performance and shortens eye fixation duration in reading Anna Judica1, Maria De Luca1, Donatella Spinelli,1,2 and Pierluigi Zoccolotti1,3 1
Neuropsychology Unit, IRCCS Fondazione Santa Lucia, Rome, 2IUSM, Rome, 3 Department of Psychology, University of Rome “La Sapienza”
Eighteen surface dyslexics were studied. Their reading deficit was evaluated on the basis of two standard test batteries. Nine subjects were submitted to reading training that consisted of reading briefly presented words. Nine dyslexics acted as a control group (receiving delayed treatment). Accuracy and speed of reading improved after therapy; also the performance in a lexical decision task improved. No effect was observed for a task using homophone contrasts. Furthermore, reading comprehension did not change as a function of training. All dyslexics were submitted to two experimental procedures: measurement of vocal reaction times (RTs) to reading single words and analysis of eye movements in reading short passages. Vocal RTs were faster for the treated group after therapy. As for eye movements, mean fixation duration was shorter after training. Other parameters (number and size of rightward saccades and number of regressions) showed small improvement with time, independent of training. When the control group was submitted to therapy in the next school year, similar improvements in reading were obtained. Overall, training affected reading parameters that appear to reflect the speed of extraction of visual information.
Correspondenc e should be sent to Pierluigi Zoccolotti, Neuropsycholog y Unit, IRCCS Fondazione Santa Lucia, via Ardeatina 306 00179, Roma, Italy. Email:
[email protected] t Telefax: (06)5150136 6 This research was supporte d by funds from MURST and from the Department of Health. We like to thank Prof. G. Spanò Cuomo, Dean of the school A. Sacchi in Nettuno, for his cooperation, P. Baratti and C. Cioccari for help in administerin g the therapeutic interventio n and C. Luzzatti for useful comments on the manuscript . Ó 2002 Psychology Press Ltd http://www.tandf.co.uk/journals/pp/09602011.html DOI:10.1080/09602010244000002
178
JUDICA ET AL.
INTRODUCTION Developmental dyslexia is one of the most common learning disturbances and causes severe school consequences. Most available literature is based on English (and French) speaking individuals. Special interest in reading disorders in these languages is understandable in view of the irregularity of the grapheme–phoneme correspondence rules. However, in recent years an interest has grown in comprehending the nature of reading deficits in languages with more regular grapheme–phoneme correspondence (e.g., German, Italian). In these languages, reading deficits tend to be less frequent (Lindgren, De Renzi & Richman, 1985), but evidence indicates that this disturbance is not simply a mild version of the deficits observed in English-speaking children. Wimmer (1993) first noticed that dyslexia among Austrian children was predominantly characterised by a reduction in reading speed while reading errors were considerably less frequent than in languages such as English and French. Similar observations have been made in Italian children (Zoccolotti et al., 1997, 1999). Particularly informative is the analysis of pattern of eye movements in these subjects (De Luca et al., 1999). Italian dyslexics read the text through short and numerous saccades (more than a factor of two with respect to controls), intermixed with long fixations (about 20% longer than controls). Also, the number of regressions (i.e., leftward eye movements) was higher than that recorded in controls. It is of note that these subjects showed normal scanning of non-verbal material, excluding the presence of an oculomotor deficit per se. The eye movement pattern observed in Italian dyslexics was taken to indicate a deficit in the direct access to written lexical representations and the prevalent use of a sub-lexical procedure (De Luca et al., 1999). This view was also supported by the analysis of vocal latencies in reading words which indicates a monotonic increase of reaction times (RTs) from shorter to longer words (Zoccolotti et al., 1999). On the basis of psycholinguisti c and eye movement data, we have proposed that the predominant deficit in Italian children can be described as a surface type of dyslexia, even though the description of the disturbance is different in English and Italian (De Luca et al., 1999; Zoccolotti et al., 1999). Note that in English-speaking countries, the most frequent form of dyslexia is represented by a deficit in the application of the grapheme–phoneme correspondence rules (sub-lexical procedure), known as phonological dyslexia (Snowling & Rack, 1991). Surface dyslexia is also present but considerably less frequent (Castles & Coltheart, 1993). In English, surface dyslexia is revealed most clearly by iper-regularisation errors in reading irregular words (such as yacht) or in discriminating homophones (Patterson, Marshall, & Coltheart, 1985). Deficits in homophone discrimination may be observed in Italian subjects by constructing ad hoc homophonic contrasts (Job, Sartori, Masterton, & Coltheart, 1983). In fact, homophones (as well as irregular words) are nearly
THERAPY AND EYE MOVEMENT IN DYSLEXIA
179
absent in Italian and the main deficit is reading slowness (Zoccolotti et al., 1999). The present study describes a reading rehabilitation project for Italian children showing surface type dyslexia. Rehabilitation of developmental reading disorders has a relatively long tradition particularly in Englishspeaking countries (for a recent review and meta-analysis see Swanson, 1999). However, only a handful of studies have considered rehabilitation training within a cognitive neuropsychological framework (cf., Weekes & Coltheart, 1996). In particular, few single-case reports have dealt with subjects showing developmental surface, or morphemic, type deficits (Broom & Doctor, 1995; Seymour & Bunce, 1994). In planning a therapy for Italian-speaking children, it must be remembered that this language lacks some types of words (such as irregular words or homophones) that are critical for detecting surface deficits and that comprise the training material for some remediation programmes for English-speaking readers (Broom & Doctor, 1995; Scott & Byng, 1989). Training that focuses on the acquisition of orthographic competence may be considered as an example of “direct approach—treat the problem” to dyslexia (Seymour & Bunce, 1994). Alternatively, rehabilitation may be aimed at dealing with the causes of the reading deficit (in Seymour & Bunce’s, 1994 terms, “indirect approach—treat the underlying causes”). In this vein, in order to plan a therapy for Italian surface dyslexics, we focused on their peculiar scanning of written texts. In particular, we aimed to improve reading speed and accuracy by countering the excessive tendency to fractionate analysis of the visual material and favouring a global analysis of words. For this purpose, we presented single words with brief exposure duration, thus forcing the observer to attempt a unitary analysis of the visual stimulus. Special care was taken in bringing the child to read progressively longer words. The training was given to a group of dyslexic children during the first year of middle school; another group of dyslexics did not receive the training and served as an untreated control group (however, these subjects were treated one year later). The training effectiveness was evaluated using a set of reading tasks, by examining the vocal RTs in reading words and by examining the eye movement pattern.
METHODS Subjects Subjects were selected from a middle school in Nettuno (Rome). Criteria for inclusion in the sample were marked reading delay on two standard reading batteries (see later), WISC IQ performance of at least 80, normal or corrected to normal visual acuity and absence of severe refractive errors. All subjects were
180
JUDICA ET AL.
males and were enrolled in the first year of middle school. No direct information was available describing the method of reading instruction used in elementary school. During the first year of the study, nine subjects fitted the criteria for inclusion and were submitted to training during the same school year (hereafter these subjects will be referred as treated dyslexics). During the next school year, a second group of nine subjects was diagnosed. Due to causes unrelated to the planning of the study (i.e., lack of trained therapists due to funding shortage), it was not possible to treat this group of subjects immediately. Consequently, it was decided to delay their treatment to the next school year and to consider them as an untreated control group (hereafter control dyslexics). Therefore, the assignment of the dyslexic subjects to the treated or control groups depended on their year of enrolment in school. All parents gave their informed consent to the study and, more specifically, parents of children in the control group were aware that their children would not receive immediate treatment. Because of his low level of school achievement, one of the dyslexic subjects in the control group was assigned a “support” teacher in school who provided general help for coping with his school work. No dyslexic subject in either group had received or was receiving any other treatment. Mean age of the treated dyslexics (11 years and 11 months; SD = 9 months) did not differ from that of the control dyslexics (11 years and 9 months; SD = 6 months; t < 1). WISC mean Performance IQ was 94.1 (SD = 8.5) for the treated dyslexics and 94.8 (SD = 8.2) for the control dyslexics (t < 1). All subjects had at least 10/10 (range 10–11) mean binocular visual acuity and absence of severe refractive errors.
Assessment of reading Two standard reading evaluation instruments were used, one to assess reading level and the other to qualify further the nature of the reading disorder. Reading level was examined with a standard reading achievement battery (MT Reading Test: Cornoldi, Colpo, & Gruppo MT, 1981). Two meaningful passages were presented. In the first, the subject had to read aloud (with a 4 min time limit); speed (time in seconds per syllable read) and accuracy (number of errors, adjusted for the amount of text read) were scored. A second passage was given without a time limit; the subject had to read it and respond to 10 multiplechoice questions (measure of comprehension). Stimulus materials (and related reference norms) varied depending on school level, with different stimuli at the beginning and end of each school year. The MT Reading test was administered twice to the treated and control dyslexics, at the beginning (circa December) and at the end (circa June) of the first year of middle school. The nature of the reading disturbance was examined by means of the Developmental Dyslexia and Dysorthography (DDD) battery (Sartori, Job, &
THERAPY AND EYE MOVEMENT IN DYSLEXIA
181
Tressoldi, 1995). A screening sub-test (Graphemes) required the subject to read aloud 21 single letters. Two sub-tests checked the reading of Words and Pseudo-words. A list of 112 words was presented and the subject was required to read them aloud. A similar list of 48 pseudo-words was presented. Number of errors and speed of reading were scored. Unlike English, Italian has very few homophone words; to approximate homophonic contrasts (same pronunciation with different graphic characteristics) specific stimuli were used. In the Correction of homophonic words sub-test, the subject was given a list of 20 words, half correct and half incorrect. Incorrect words are produced either by the insertion of an apostrophe (e.g., “l’ametta” instead of “lametta”; “razor blade”) or by inserting a space between two parts of the word (“di vano” instead of “divano”; “sofa”). The subject is requested to say whether or not the words are correct and the number of errors is computed. The incorrect versions have no meaning but sound like their correct counterparts; therefore, to solve the task, reference to a sight vocabulary is required. Finally, the Lexical Decision subtest was given. The subject was required to cross out real words on a list that included 48 items, half of which were words and half pseudo-words with a single letter different. Both time and errors were measured. The timing of the administration of the DDD battery was the same as that of the MT Reading Test during the first year of middle school. The test was also administered twice in the second year of middle school to the control dyslexics.
Vocal reaction time Words of 2, 3, 4, and 5 letters were used (54 for each length for a total of 216 stimuli) (see Di Pace et al., 1995). The first letter of each word was consistent across groups; for example, there were equal numbers of words in each word length group beginning with the letter “a”. In order to obtain a sizeable proportion of short words, both function and content words were used, controlling only for median frequency value. This was 7105 (IBM Italia, 1989). Word frequency did not vary among different length words, Kruskal Wallis H (4) = 2.82, n.s. Words were presented on the centre of the screen of an Apple II computer following the presentation of a fixation dot (displayed for 500 ms). Each letter subtended 0.4 cm horizontally. At a distance of 57 cm, this corresponds to 0.4 deg of visual angle. A chin-rest and a headrest prevented the subject from making head movements. One block of 30 practice stimuli and 6 experimental blocks of 36 stimuli were administered, interspersed with brief pauses. Word length was randomised within each block. The subject’s task was to read the words aloud into a microphone as quickly as possible. The stimulus remained on the screen until the subject responded. Performance was measured by the mean vocal reaction time (RT) (in ms) to correctly identified words.
182
JUDICA ET AL.
Vocal RTs were recorded in eight treated dyslexics, and nine control dyslexics. The timing of the administration of the Vocal RTs test was the same as that of the MT Reading Test.
Eye movement recordings Apparatus and general procedure. Eye movements were recorded by means of an infrared pupil reflection system (AMTech ET3 Eye Tracking System). This enables measuring movements in both horizontal and vertical directions for one eye. The sampling rate was 200 Hz. The system had a maximal accuracy of 5 min arc. The subject sat on a chair; a headrest and a chinrest were used to keep his head immobile during recording. He was requested to stay as still as possible and to try not to blink during the recording period. Stimuli were presented on the screen of a PC computer. The viewing distance was 60 cm. A calibration procedure was carried out before each reading trial. The subject was requested to fixate a target (a little square) which was displayed in nine different successive positions on the screen according to a 3 × 3 matrix spanning 21.4 deg horizontally and 6.2 deg vertically. After fixation of the first square, the target disappeared and a new target appeared in the next position. The subject’s task was to saccade to the next position, etc. The reading task (see later) was performed immediately after calibration. The portion of the trace contaminated by artefacts due to blinks was signalled by the computer and rejected. Artefacts due to occasional head movements were rejected. Only artefact-free recordings were used. Stimuli and procedure. Stimuli were short passages (black letters on a light background). Fourteen passages containing an average of 30.6 (SD = 1.6) words were presented. Passages were composed in such a way that the frequency of content words was homogeneous (median frequency = 3808) (IBM Italia, 1989). There was no difference in frequency values in the 14 passages, Kruskal Wallis H (13) = 7.34, n.s. Mean character width (centre-to-centre letter distance) was 0.5 deg. Each passage contained 4 lines (7.6 words per line, SD = 1.3; centre-to-centre line distance = 1.4 deg) and was fully displayed on the screen without a time limit. The text was 21.4 deg wide and 6.2 deg high. The subject’s task was to read the passages silently. After reading each passage, the children were asked a few questions to evaluate their comprehension. The timing of the recordings was the same as that of the MT Reading Test. Analysis. Eye movements were recorded in nine treated dyslexics, and seven control dyslexics. Eye movements of the dominant eye were analysed. Four parameters of eye movement in reading were considered: number of rightward saccades per line, amplitude of rightward saccades (in deg), number
THERAPY AND EYE MOVEMENT IN DYSLEXIA
183
of regressions per line and fixation duration (in ms). As noted above, the subjects read 14 passages. However, only those passages for which a reliable recording was obtained were submitted to further analysis. In the case of the treated dyslexic, an average of 5.6 passages (SD = 1.1) before treatment and of 6.8 passages (SD = 1.6) after treatment were considered; for the control dyslexics, an average of 5.0 passages (SD = 1.1) and of 6.1 (SD = 0.9) passages in the first and second session, respectively.
Reading training The training was given for about 5 months during the school year. Each training session lasted 1 hour and was repeated twice a week. The remediation programme was administered individually by a speech therapist. Overall, each child had about 35 sessions. The training was based on brief (60 to 150 ms) presentation of single words on the screen of a PC at 45 cm viewing distance. The stimulus was displayed at the centre of the screen, in white letters on a blue background. Mean character width (centre-to-centre letter distance) was 0.6 deg. The computer programme was developed by Morchio, Ott, Pesenti, and Tavella (1989). Three parameters of the stimulus were varied: category (nouns, verbs, adjectives, etc.), frequency (high and low), and length (from 2 to 8 letters). Frequency was based on De Mauro and Moroni’s (1996) dictionary: This is an educational instrument that codes words with relatively similar use within the lexicon in various categories. The “high” frequency words in the present study were selected from a pool of 5000 words comprising items at the core of the Italian language. The “low” frequency words were derived from a list of 10,000 words considered important for mastering comprehension in more formal contexts such as newspapers, contemporary literature, history, etc. About 2000 words were used in the present study to generate the lists of stimuli. Ten-to-fifteen blocks of 20 words per session were administered. In total, about 9000 stimuli were presented to each child. The subject’s task was to read the word aloud or to read it silently and print it on the keyboard. These two response methods were alternated, after each block of words, within the same session. Before training, individual performance was assessed by counting the number of errors using one 20-word block for each frequency level and wordlength (except for 8-letter words that proved too difficult for all subjects at this stage) with presentation time 100 ms. Training difficulty was adjusted for each child to start with stimuli yielding optimal performance (e.g., three-letter high frequency words with a 150 ms presentation time). The difficulty of the exercises was then increased by using shorter presentation times and longer and less frequent words, in order to obtain 60–70% correct responses in each
184
JUDICA ET AL.
session. At the end of each session results obtained and improvement were discussed with the subject. The control dyslexics did not receive any form of reading training, except for the reading activity part of the normal school programme.
Statistical treatment of data To evaluate the effect of the training, ANOVAs comparing treated and control dyslexics were performed. Whenever appropriate, Duncan a posteriori comparisons were carried out. As for the MT Reading test, raw scores were converted to z-scores based on Italian normative data (Cornoldi et al., 1981) to allow a pre- vs. post-treatment comparison, where different stimulus materials were used. Similar analyses using z-scores based on Italian normative data were used for the DDD battery (Sartori et al., 1995). For vocal RTs, the ANOVA was performed on mean latencies. For eye movements, separate ANOVAs were made for each of the four parameters considered (number of rightward saccades, amplitude of rightward saccades, number of regressions, and fixation duration). The effect of delayed treatment in the control dyslexics was evaluated by means of ANOVA and paired t statistics. Only the DDD battery was considered in this case; second year middle school Italian normative data were used (Sartori et al., 1995).
RESULTS MT reading battery Major results for reading short texts are presented in Table 1 (raw scores) and in Figure 1 (z-scores). Thus, for example, values for Accuracy indicate that subjects are circa 3–5 SDs below the normal mean (indicated by the dashed line). An inspection of the figure indicates the following:
Both groups of dyslexics showed a very severe deficit in both accuracy and speed; although below average, text comprehension was considerably less affected. When examining the effect of the time factor, the control dyslexics showed a worsening in performance in the post-session for all reading parameters. This does not necessarily indicate a lower performance in absolute terms but rather that the control dyslexics did not change their performance over time. For example, they spent on average 0.63 s per syllable in the pre-test and 0.60 s in the post-test (see Table 1). This performance stability was associated with a larger difference compared to normative reference values (i.e., lower z-scores in the second test, reported in the last two columns of Table 1). A similar pattern concerns
TABLE 1 MT battery: mean values and standard deviations obtained at pre- and post-testing for treated and control dyslexics Pre Group/Parameter
Normsa
Post
Normsb
Mean
(SD)
Mean
(SD)
Mean
(SD)
Mean
(SD)
Comprehension (correct response/10)
5.0
(2.6)
3.8
(2.2)
8.5
(1.8)
7.9
(1.9)
Speed (s per syllable)
0.85
(0.52)
0.66
(0.39)
30.1
(12.4)
25.1
(9.5)
Treated dyslexics
Accuracy (number of errors
44.0
(20.4)
21.8
(13.4)
9.0
(7.8)
6.0
(5.6)
Comprehension (correct response/10)
6.3
(1.3)
4.3
(1.5)
8.5
(1.8)
7.9
(1.9)
Speed (s per syllable)
0.63
(0.22)
0.60
(0.19)
30.1
(12.4)
25.1
(9.5)
9.0
(7.8)
6.0
(5.6)
Control dyslexics
Accuracy (number of errors
31.8
(18.7)
25.2
(10.0)
Normative values from Cornoldi et al. (1981) are presented for comparison ( a beginning of the first year of middle school. b end of the first year of middle school).
Performance (z score)
0
control dyslexics treated dyslexics
-2
-4
-6
pre-
post-
Correctness Accuracy
pre-
Speed
post-
pre-
post-
Comprehension
Fig.1
Figure 1. Reading performance on the MT battery. Mean values (expressed as z-scores) and SEs obtained at pre- and post-testing are presented for treated and control dyslexics separately for accuracy, speed, and comprehension .
185
186
JUDICA ET AL.
the comprehension data of the treated dyslexics; for the other two parameters, the training either improved performance (accuracy) or prevented it from worsening (speed) as compared to controls. Separate ANOVAs with group (treated vs. control dyslexics) as nonrepeated factor and time (pre- vs. post-testing) as repeated factor were carried out on the accuracy, speed and comprehension parameters (z-scores). As for accuracy, no main effect for the group, F < 1, n.s., and time, F(1, 16) = 2.8; n.s., factors. The interaction was significant, F(1, 16) = 9.17, p < .05: a posteriori comparisons indicated an improvement from a mean z-score of –4.75 to one of –2.97 in the treated group, p < .01 and a non-significant decrement in performance for the control group (pre: –2.91; post: –3.43). The analysis of the speed parameter showed no main effect for group, F < 1, n.s. The time factor was marginally significant, F(1, 16) = 3.22, p = .09. The interaction was significant, F(1, 16) = 6.18, p < .05: the treated group did not change performance (pre = –4.69; post = –4.54) while the control group performed more poorly in the second test (–3.67) than on the pre- test (–2.71, p < .01). The analysis of the comprehension parameter indicated no main effect of the group factor, F(1, 16) = 1.25, n.s. The time factor was significant, F(1, 16) = 6.11, p < .05: performance measured at the second test (–2.06) was worse than at the pre-test (–1.54). The interaction was not significant, F < 1. Comments. At the pre-therapy evaluation, all subjects were severely affected in reading for both speed and accuracy, as expected on the basis of previous evidence for surface dyslexics (Zoccolotti et al., 1999). Comprehension was less affected in both groups, indicating that, if tested without a time limit, these subjects can reach an almost sufficient identification of the main information in the text. Clear reading improvement in terms of accuracy was observed after treatment. The treated dyslexics showed no increase in reading speed; however, this result must be considered in light of the increasing decrement in speed over time shown by the untreated controls. In contrast, text comprehension did not improve as a function of training. This general pattern appears consistent with the characteristics of the training that emphasised practice in the correct identification of isolated words; in contrast, no attempt was made to train higher levels of cognitive processing. The finding that performance improvement was present in reading meaningful passages indicates that improvement due to training on single words generalised to some extent to functional reading.
DDD battery All subjects were in the normal range in the Graphemes sub-test (i.e., only one subject made one error), indicating preserved reading at the letter level.
187
THERAPY AND EYE MOVEMENT IN DYSLEXIA
Major results for accuracy and speed parameters in the Words and Pseudowords sub-tests are presented in Table 2 (raw scores) and in Figure 2 (z-scores). An inspection of the figure indicates the following:
Both groups were severely affected on all parameters. As predicted in subjects showing a surface type of dyslexia (e.g., Castles & Coltheart, 1993), performance in reading pseudo-words was somewhat less impaired than that in reading words; baseline scores tended to be lower for the treated than the control group. Examining the effect of time, note that the stimulus material in the DDD battery did not vary from pre- to post-testing and that there was a single TABLE 2 DDD battery: Mean values and standard deviations obtained at the pre- and post-testing for the treated and for the control dyslexics Pre Group/Stimulus
Post
Norms
Mean
(SD)
Mean
(SD)
Mean
(SD)
List of words Speed (s per word) Accuracy (% of errors)
2.01 12.41
(1.02) (8.3)
1.68 6.07
(1.1) (3.5)
0.74 1.25
(0.3) (1.6)
List of pseudo-words Speed (s per word) Accuracy (% of errors)
2.95 30.83
(1.60) (11.3)
2.62 17.50
(1.8) (5.8)
1.30 7.5
(0.4) (6.0)
Correction of homophonic words Accuracy (% of errors)
24.50
(11.5)
19.00
(11.0)
4.0
(6.5)
Lexical decision Speed (s per word) Accuracy (% of errors)
3.11 14.58
(1.1) (4.4)
2.46 8.33
(1.1) (4.8)
1.52 6.25
(0.4) (6.0)
List of words Speed (s per word) Accuracy (% of errors)
1.60 9.02
(0.6) (5.6)
1.59 6.61
(0.6) (4.9)
0.74 1.25
(0.3) (1.6)
List of pseudo-words Speed (s per word) Accuracy (% of errors)
0.99 11.70
(0.2) (5.0)
1.13 12.23
(0.7) (4.5)
1.30 7.5
(0.4) (6.0)
Correction of homophonic words Accuracy (% of errors)
14.50
(10.0)
16.00
(11.0)
4.0
(6.5)
Lexical decision Speed (s per word) Accuracy (% of errors)
2.75 11.88
(1.2) (4.4)
3.07 7.71
(1.2) (4.6)
1.52 6.25
(0.4) (6.0)
Treated dyslexics
Control dyslexics
Normative values modified from Sartori et al. (1995).
188
JUDICA ET AL.
Performance (z score)
Words
Pseudo-words
0
-2,5 control dyslexics
-5
t reated dyslexics
-7,5 Accuracy Correctness
Speed
Accuracy Correctness
Speed
-10 pre-
post-
pre-
post-
pre-
post-
pre-
post-
Fig.2
Figure 2. Reading performance on the DDD battery. Mean values (expressed as z-scores) and SEs obtained at pre- and post-testing are presented for treated and control dyslexics separately for words and pseudo-words.
normative value per year (see last column of Table 2). Thus, in this case, a stable performance was indicated by stability of z-scores. The control group showed an improvement limited to accuracy in reading words remaining stable or worsening in the other cases. The treated group of dyslexics showed improved performance for both accuracy and speed in reading words and pseudo-words. An ANOVA with group (treated vs. control dyslexics) as non-repeated factor and type of stimulus (words vs. pseudo-words), reading measure (speed vs. accuracy) and time (pre- vs. post-testing) as repeated factors was run on zscores. The analysis indicated a main effect of type of stimulus, F(1, 16) = 7.06, p < .05: performance on pseudo-words (–3.32) was better than performance on words (–4.11). The main effect of time was reliable, F(1, 16) = 10.27, p < .01: performance on post-testing (–3.17) was better than performance on the pretesting (–4.26). Two first-order interactions were significant. The group by time interaction, F(1, 16) = 7.89, p = .01, indicated that the treated group improved passing from –5.16 to –3.10, p < .01; the control dyslexics showed no change between the pre- (–3.36) and post- (–3.22) test. The type of stimulus by reading measure interaction, F(1, 16) = 4.54, p < .05 indicated that the difference between words and pseudo-words depended predominantly on accuracy: Accuracy values for words were lower (–4.41) than those for pseudo-words (–3.07; p < .01); no difference was present in speed between words (–3.80) and pseudo-words (–3.56). The type of stimulus by time interaction approached significance, F(1, 16) = 3.87, p = .06: performance in reading words improved with time passing from –4.94 to –3.28 (p
