Influence of Instructional Approach on Initial Letter-Speech Sound Binding: Effects of Explicit and Implicit Instructions on Learning
Jesse Lemsom
Student Number 11699973
Institute University of Amsterdam
Department Clinical Developmental Psychology Supervisor Dr. Jurgen Tijms
Daily Supervisor C.T. Verwimp, Msc Submission Date February 12th 2021 Number of words ± 6400
Abstract
This study examined the influence of instructional approach on the initial L-SS binding in 107 primary school children (Mean = 106.8 months). They received either explicit or implicit instructions, after which they had to learn eight L-SS associations within an artificial orthography. Learning progress was assessed by accuracy and reaction time during the learning task. In addition, letter knowledge and reading rate within the artificial script were assessed right after the task and after one day of retention. Our results showed that explicit instructions led to a better knowledge of the newly learned script and a better application of this new knowledge in a time-limited reading task compared to implicit instructions. It is thought to be caused by more goal-directed behavior in the explicit condition, which would have a positive effect on L-SS binding and automatization. This study suggests that letter-speech sound binding is more than merely mapping letters onto speech sounds and sheds a light on attentional processes that are involved in the initial phases of L-SS binding.
Keywords: Letter-Speech Sound Binding, Implicit Learning, Explicit Learning, Fluent Reading, Accurate Reading, Artificial Learning Task, Artificial Orthography, Goal-directedness.
Influence of Instructional Approach on Initial Letter-Speech Sound Binding: Effects of Explicit and Implicit Instructions on Learning
Reading is a recently developed skill and relies on decoding visual information in order to find the associated speech sounds and word meanings (Dehaene & Cohen, 2007). Mapping letters onto speech sounds in alphabetic languages is essential for learning to read (Ehri, 2005; Brem et al., 2010). The letter-speech sound (L-SS) associations in more transparent alphabetic languages, such as Dutch, can be established within approximately one year of reading instruction (Wentink & Verhoeven, 2003; Siegel & Faux, 1989; Snowling, 1980), but seem not to be sufficient for developing accurate and fluent reading. For fluent reading, L-SS associations have to be automatized which extends over a period of four years, yet is still not adult-like by then and keeps developing long after (Froyen, Bonte, Van Atteveld & Blomert, 2009).
People with developmental dyslexia have difficulties with accurate and fluent reading, which is thought to be caused by problems with the automatization process, due to a L-SS binding deficit (Fletcher & Lyon, 2008; Blomert, 2011). A better understanding of L-SS binding will contribute to the prevention and remediation of reading disabilities.
Instructional Approach
Teaching reading skills can be done with different instructional approaches, which can have a substantial influence on learning outcome and related brain changes (McCandliss, 2010). For example, verbal instructions increase sustained concentration on task goals (Kirkham, Breeze & Marí-Beffa, 2011). In addition, prompting learners to develop specific skills or master new knowledge (i.e., mastery-goal instructions) promotes more active processing and deeper learning, compared to instructions that prompt learners to be better than others (i.e., performance-goal instructions; Erhel & Jamet, 2016). This means that clear verbal instructions lead to more goal-orientated behavior, which improves learning.
In recent years there has been a shift away from explicit learning strategies and towards more implicit learning strategies (Lovio, Halttunen, Lyytinen, Näätänen, & Kujala, 2012; Aravena & Tijms, 2009). This shift should be received with caution, as it has been proven that the most effective treatments for developmental dyslexia include implicit and explicit instruction (Singleton, 2009). This is in line with the idea that slow, effortful, controlled processing proceeds fast, low-effort, automatic processing in skill acquisition (Chein & Schneider, 2005; Siegler, 2005). This suggests that explicit instructions are necessary to increase cognitive control during the task, to increase focus on the goal of the task, in order to improve learning. Knowledge of the influence of different instructional approaches on learning outcome can contribute to current teaching methods, including treatments for disorders such as developmental dyslexia.
Current Research
Previous research, by Aravena, Snellings, Tijms & van der Molen (2013 suggested that explicit instructions combined with implicit techniques resulted in the most efficient technique for initial L-SS binding. It also showed that first accuracy would improve and later on there would be a shift to reaction time. In addition, the results showed a difference between dyslectic and non-dyslectic children and it is suggested that similar tools can be applied for early
detection of dyslexia since it is independent of phonological or reading instructions. However, in the study of Aravena and colleagues (2013), different game designs were used in the different conditions, which means that the differences in learning between the conditions might have been caused by the different game designs. In addition, it is questionable how reliable their follow-up measurement was, as it was executed at home by the parents of the participants.
To address these issues, the current study used an artificial learning paradigm similar to that of Aravena et al. (2013), but the training methods for both conditions were identical. The two conditions only differed in the instructions that were given prior to the learning phase. In addition to Aravena et al. (2013), the present study measured the reaction time and accuracy during the training. This allows for the mapping of the learning curve, in order to investigate reaction time and accuracy during all phases of learning. The reaction time and accuracy allow for a more sensitive measurement for the letter-speech sound binding, since this reflects the integration of letter-speech sound associations in the brain (McCandliss, 2010). Furthermore, the training length was reduced to 20 minutes since this seemed sufficient for identifying difficulties with L-SS binding. Another addition was a congruency task after the training session in which reaction time and accuracy were measured again. Moreover, these tests were executed a day later by the researcher in the school building for a more standardized method compared to Aravena, et al. (2013). This study served as a pilot for a future EEG study with dyslectic and non-dyslectic children. An artificial script was used that allowed for mimicking the initial steps in learning a novel script without possible interference of previously learned associations.
The current study focused on the influence of instructional approach on the initial letter-speech sound binding in children. Based on previous research it is hypothesized that explicit instruction will result in a better initial letter-speech sound binding compared to implicit instruction. More specifically it is investigated whether the instructional approach influences (1) the performance during the learning phase, (2) the knowledge of the artificial script right after the learning phase and (3) the knowledge of the artificial script after a retention time of 24 hours. It is expected that during the learning phase reaction will not differ between conditions or over time. Accuracy is expected to have the same starting point for both conditions, but over time will improve more for the explicit condition. Knowledge of the artificial scripts right after the learning phase will be higher for the explicit training condition compared to the implicit. Finally, it is expected that after one day of retention the knowledge of the artificial script will increase for both conditions, but more for the explicit condition (see Figure 1.1).
Expectations of Accuracy and Reaction Time During the Learning Phase
Figure 1.1. It shows the expected accuracy and reaction time during the learning phase. It also shows the expected knowledge of the artificial script right after the learning phase and after one day of retention. Both the explicit (red) and the implicit (blue) condition are shown.
Method Participants
This study was conducted with 107 children, aged between 92 months and 136 months old (Mean = 106.8 months, SD = 8.9 months), and consisted of 53 boys and 54 girls. All participants were following primary education and their native language was Dutch. The recruitment of participants was done through two primary schools in Amsterdam and exclusion criteria were uncorrected hearing or sight disabilities. Participants were also excluded if there was missing data. An informed consent was obtained from all parents whose child participated in the study. The ethics committee of the University of Amsterdam approved this study. Prior to the study the participants were randomly allocated to either the explicit (n = 53) or the implicit (n = 54) condition.
Letter-Speech Sound binding task
The Letter-Speech Sound Binding Task (L-SS Task) was based on prior research of Aravena et al. (2013). The task is supposed to mimic the initial letter-speech sound binding mechanism in children, including the formation of the neurocognitive reading network. The task was executed with PsychoPy (Peirce, et al., 2019). The goal of the task was to learn a set of eight artificial L-SS correspondences in order to read words. The artificial orthography of the task consisted of eight visual symbols (Vidal, Content & Chetail, 2017), known as artificial letters, that were matched to Dutch phonemes; called speech sounds (see Table 2.1). The phonemes consisted of four vocals and four consonants; phonemes with similar articulation were avoided. The L-SS Task consisted of four blocks; three aiming to learn associations and the last to test the learned knowledge.
The goal of the first phase was to learn all L-SS pairs: a symbol that was matched with a phoneme. It consisted of block 1 and 2; each with 72 trials, four L-SS pairs and a break in the middle. Block 3 included all eight L-SS pairs and consisted of 56 trials. There was a break between each of the blocks. In each trial the participant was presented with a fixation cross in the middle of the screen that lasted for 500, 750 or 1000 milliseconds (ms) after which two symbols and one phoneme was presented; the symbols were presented for 1000 ms. The participant had to indicate whether the left or the right symbol corresponded to the presented phoneme, by pressing the left of the right button. Afterwards feedback was presented with a duration of 1000 ms: a happy smiley-face when the right symbol was indicated and a sad smiley-face when the wrong symbol was indicated. A drawing of a snail was presented if the participant did not respond within 4000 milliseconds, this served as encouraging feedback (see Figure 2.1A).
The goal of the second phase, block 4, was to test the L-SS binding and consisted of 96 trials. Each trial presented an L-SS pair that was either congruent (previously learned match) or incongruent (new match). Participants had to respond with the left or right button to indicate if the trial was congruent or incongruent. There was a total of 48 congruent pairs and 48 incongruent pairs. During this block no feedback was given, since it served as a testing block (see Figure 2.1B).
Table 2.1
The Artificial Orthography of the L-SS Task Adapted from the BACS Script (Vidal et al., 2017) Letter
Speech sound (IPA) [ɑʊ] [t] [z] [ɛɪ] [e] [n] [f] [o]
A day later a shortened version of block 4 from the L-SS Task comprising 16 trials, of which eight congruent and eight incongruent, was repeated to test the consolidation of the L-SS correspondences (see Figure 2.1C).
The L-SS task instructions were given verbally by the experimenter and were either explicit or implicit. Both instructions were standardized through a provided protocol. In the explicit condition the participants were explained what the goal of the task was, they were told that they were going to crack a secret code by learning symbols that match with a sound of a letter. The better they learned the symbol sound correspondence, the better they could crack the secret code. The meaning of the happy smiley-face, the sad smiley-face and the drawing of the snail was explained as well. The participants in the implicit condition were told to play a game and that the goal of the game would become clear during the task itself. All participants were told which two keys they had to use to play the task and received the same computerized feedback during the learning phase.
Outcome Measures
Accuracy during the L-SS task. The accuracy was recorded during the training for each trial. For each block, trials were divided in 4 bins (see Data Analysis for more details), for which the accuracy score was expressed as the total number of correct items divided by the total number of non-missed items (i.e., for which an answer was given) for that particular bin (% correct). The maximum score was 100% and the minimum score of 0% was obtained if all answers from that one bin were incorrect.
Reaction time during the L-SS task. The reaction time (RT) was recorded during the training for each trial. Participants had a maximum of 4 seconds to respond. The mean reaction time was calculated as the mean of all accurate responses that were given within a particular bin (in seconds).
Productive symbol knowledge. The Productive Symbol Knowledge Task (PSK Task) was executed by presenting the participant with all eight symbols on a sheet of paper. The researcher asked the participant to name the corresponding phoneme out loud, while pointing to each of the symbols. The maximum score of this test was eight and the minimum score was zero.
Word reading rate in artificial orthography. The artificial orthography from the L-SS Task was used to compose fourteen Dutch monosyllabic words and fourteen monosyllabic pseudowords that obeyed Dutch phonotactic regularities (see Table 2.2). In all of the composed words all vocals had the same articulation. The One-Minute Test Artificial Orthography (artificial OMT) was executed by the researcher with two sheets of paper, one with the 14 Dutch words (OMT-Dutch) and one with the 14 pseudowords (OMT-Pseudo). The participants had one minute to read as many words as possible per list and were told that if they did not know all the symbols they should try to read the ones they did know. The maximum score for one list was 14 and the minimum score was zero.
Table 2.2
Dutch Monosyllabic Words and Monosyllabic Pseudowords that Obeyed Dutch Phonotactic Regularities, Written in Dutch and in Artificial Orthography
Dutch Words Pseudowords
Dutch Artificial Dutch Artificial
of ot fout foun zon zof zei teif tot fot zout zoun tent nent zijn zijf ton fon fijn nijf net fet zou oun en et zet tef
Word reading rate in Dutch. The One-Minute Test in Dutch consisted of 116 unrelated words of increasing difficulty and was executed by the researcher to assess word reading rate in Dutch (Brus & Voeten, 1973). The participant had one minute to read as many words as possible correctly out loud and this served as a measure for word reading skills in Dutch (r=.79, test-retest). The maximum score was 116 and the minimum score was zero.
Non-verbal IQ. A time-limited version of the Raven’s Coloured Progressive Matrices (CPM) was used to measure non-verbal IQ. It consisted of 36 items and was simultaneously executed by the whole class on paper. Each trial consisted of a patterned rectangle with a missing piece on the same place and with the same shape. Under the rectangles six numbered pieces were presented as answers. All pieces had the same shape, but only one would fit into the pattern. The participant had to choose the right answer and had to note the corresponding number on the scoring form. The raw data was used as a measure of non-verbal IQ, with a maximum of 36 and a minimum of zero.
Procedure
The study consisted of two one-on-one testing sessions for each participant. Before the first testing session, the non-verbal intelligence was tested for each participant by a time-limited version of the Raven’s CPM simultaneously for the whole class that lasted for about 20 minutes. All of these sessions took place in a quiet room inside the school building. The first testing session lasted around 35 minutes and consisted of the L-SS Task, with one of the instructional approaches, which was followed by the PSK Task and the artificial OMT (i.e., real words and pseudowords). The second testing session took place one day later and lasted around 15 minutes. A shortened version of block 4 from the L-SS Task comprising 16 trials (eight
congruent and eight incongruent trials) was executed. This was followed by the PSK Task and the artificial OMT artificial tasks (see Figure 2.1).
Experimental Design
Figure 2.1. The experimental design is shown, including the L-SS task (A and B) and the different outcome measures (C). The experimental design is similar for both conditions, but the instructions will be either explicit or implicit. The goal of the task is to learn a set of L-SS correspondences, with eight artificial symbols from Bacs-1 alphabet (Vidal et al., 2017) that were randomly match with a Dutch phoneme. These learned correspondences can be used to read monosyllabic words written in the artificial symbols. The instructional approach is manipulated to examine its influence on learning. Both reaction time and accuracy will be measured during the L-SS task. The outcome measures will be scored by the researcher. Data analysis
Participants were excluded from the data-analysis if they did not complete both testing sessions (n = 3). The trials without any response from the participant were considered incorrect. The reaction time was excluded if the answer was incorrect and if the reaction time was lower than 0.100 seconds or higher than 3 times the standard deviation of this participant. To execute the data analysis 4 bins were made for each block, this was done for all participants. If a bin did not have at least one reaction time left the participant was excluded from the study (n = 1). For all statistical analyses an alpha of 0.05 was used.
To determine the influence of instructional approach on the performance during the learning phase of the L-SS task a One-Way Repeated-Measures MANOVA was executed for each block. Instructional approach (explicit and implicit) was used as the between-subject factor, Bin as within-subject variable and as continuous dependent variables the reaction time and accuracy of block 1, 2 and 3 were used. All assumptions were checked before executing
this test. If the test showed significance an RM ANOVA was used for each of the dependent variables and if this was again significant a pairwise t-test with an FDR correction was executed. A Factorial MANOVA was executed with Instructional approach (explicit and implicit) as between-subject factor, Time as within-subject factor and as dependent variables the reaction time and accuracy of block 4 (the congruential task), the score of the PSK task and the score of the artificial OMT. This test was used to determine the influence of instructional approach on learning outcome directly after the learning phase and after one day of retention. The assumptions were checked before executing this test. If the test showed significance, an ANOVA was used for each of the dependent variables and if this was again significant a pairwise t-test with an FDR correction was executed.
Results
A total of three participants were excluded due to missing data for a testing session. After the removal of outliers and dividing the trials into bins for each participant, one other participant was excluded due to lacking data in one bin. No significant differences were found between the two conditions for age (p = .398), IQ (p = .760) or reading skills (p = .943). All assumptions were tested before executing the analyses.
Performance During the Learning Phase
To investigate the effects of the instructional approach on performance during the learning phase a One-Way Repeated Measure MANOVA was executed for each of the blocks. For each block both reaction time and accuracy were measured and the means were calculated for four bins for all participants. All means and standard deviations of reaction time are shown in Table 3.1, all means and standard deviations of accuracy are shown in Table 3.2 and all line graphs for both reaction time and accuracy are shown in Figure 3.1.
Block 1. The One-Way Repeated Measures MANOVA for block 1 showed a significant effect of instructional approach (F[2] = 14.600, p = .048), time (F[6] = 47.185, p < .001) and the interaction of instructional approach and time (F[6] = 19.573, p = .002).
Reaction time. Univariate Repeated Measures ANOVA including only reaction time showed no effect of instructional approach (F[1,212.573] = .211, p = .647) and a non-significant interaction effect of instructional approach and time (F[2.453,Inf] = 1.341, p = .261). It also showed a significant effect for time (F[2.453,Inf] = 17.317, p < .001).
Multiple pairwise T-test with FDR correction showed significant results for reaction time in the explicit group between bins 1 and 3 (p = .022), but not between any other bins (all p’s > .059). In the implicit group it showed significant differences between bins 1 and 2 (p = .010), 1 and 3 (p = .002) and between 1 and 4 (p = .003), but no significant differences were found between bins 2, 3 and 4 (all p’s > .648) (see Figure 3.1A).
Accuracy. A univariate RM ANOVA including only accuracy as dependent variable showed a significant effect of instructional approach (F[1,236.341] = 6.203, p = .013), time (F[2,Inf] = 14.478, p<.001) and the interaction of instructional approach and time (F[2,Inf] = 6.408, p = .001).
The pairwise T-tests with FDR correction for accuracy showed significant results in the explicit group between bins 1 and 3 (p < .001), 1 and 4 (p < .001), 2 and 3 (p = .024) and
Table 3.1
Mean Reaction Time of the L-SS Task and Standard Deviations (SD) for Bin 1, Bin 2, Bin 3 and the Total for Each Block for Both the Explicit and Implicit Condition
Condition Explicit Implicit
Mean (Sec) SD (Sec) Mean (Sec) SD (Sec) Block 1 Bin 1 0.893 0.377 1.000 0.465 Bin 2 0.772 0.354 0.770 0.384 Bin 3 0.739 0.299 0.707 0.383 Bin 4 0.701 0.299 0.734 0.387 Total 0.776 0.339 0.803 0.420 Block 2 Bin 1 0.685 0.348 0.623 0.266 Bin 2 0.639 0.327 0.596 0.289 Bin 3 0.632 0.295 0.559 0.285 Bin 4 0.632 0.255 0.573 0.225 Total 0.647 0.307 0.588 0.266 Block 3 Bin 1 0.678 0.353 0.676 0.316 Bin 2 0.646 0.288 0.680 0.334 Bin 3 0.608 0.276 0.612 0.297 Bin 4 0.659 0.278 0.603 0.254 Total 0.648 0.300 0.642 0.302
between 2 and 4 (p = .0011), but no significant differences were found between bins 1 and 2 (p = .069) or between 3 and 4 (p = .278). For the implicit group it showed no significant differences between any of the bins (all p’s > .30).
Pairwise T-tests were also executed between the two groups for accuracy. It showed significant differences for bin 3 (p = .006) and 4 (p < .001). No significant differences were found for bin 1 (p = .650) and 2 (p = .890) (see Figure 3.1D).
Block 2. The One-Way RM MANOVA for block 2 showed a significant effect of instructional approach (F[2] = 15.386, p = .076), time (F[6] = 35.541, p < .001) and the interaction of instructional approach and time (F[6] = 15.568, p = .006).
Reaction time. Univariate RM ANOVA including only reaction time showed no significant effects for instructional approach (F[1,148.309] = 1.382, p = .244), time
(F[2.782,Inf] = 3.028, p = .066) and no significant interaction effect of instructional approach and time (F[2. 782,Inf] = .174, p = .914) (see Figure 3.1B).
Accuracy. A univariate RM ANOVA including only accuracy as dependent variable showed a significant effect for instructional approach (F[1,188.687] = 4.043, p = .045), time (F[2.948,Inf] = 23.573, p<.001) and the interaction of instructional approach and time (F[2. 948,Inf] = 5.255, p = .005).
Table 3.2
Mean Accuracy of the L-SS Task and Standard Deviations (SD) for Bin 1, Bin 2, Bin 3 and the Total for Each Block for Both the Explicit and Implicit Condition
Condition Explicit Implicit
Mean (%) SD (%) Mean (%) SD (%) Block 1 Bin 1 54.21 14.75 52.88 15.19 Bin 2 59.82 11.80 59.44 15.09 Bin 3 66.94 15.91 57.97 16.72 Bin 4 70.13 16.93 57.30 19.65 Total 62.78 16.11 56.90 16.82 Block 2 Bin 1 56.73 14.82 57.20 15.37 Bin 2 65.08 15.85 61.71 16.75 Bin 3 73.49 17.69 63.59 19.87 Bin 4 73.16 17.66 63.29 19.92 Total 67.12 17.82 61.45 18.13 Block 3 Bin 1 70.29 20.37 67.19 18.85 Bin 2 74.86 19.21 65.50 21.73 Bin 3 76.25 17.11 65.20 21.73 Bin 4 73.87 15.95 64.96 19.31 Total 73.82 18.24 65.71 20.32
Pairwise T-tests with FDR correction were also executed between the two groups for accuracy. It showed significant differences for bin 3 (p = .009) and 4 (p = .009). No
significant differences were found for bin 1 (p = .870) and 2 (p = .300).
The pairwise T-tests with FDR correction for accuracy showed significant results in the explicit group between bins 1 and 2 (p = .016), 1 and 3 (p < .001), 1 and 4 (p < .001), 2 and 3 (p = .016) and between 2 and 4 (p = .016). No significant difference was found between bins 3 and 4 (p = .921). For the implicit group it showed no significant differences between any of the groups (all p’s > .270) (see Figure 3.1E).
Block 3. The One-Way RM MANOVA for block 3 showed a significant effect of instructional approach (F[2] = 18.019, p = .044) but no significant effect for time (F[6] = 3.637, p = .158) and the interaction of instructional approach and time (F[6] = 3.762, p = .136).
Reaction time. Univariate RM ANOVA including only reaction time showed no significant effects for instructional approach (F[1,177.96] = .011, p = .913), time
(F[2.696,Inf] = 2.77, p = .081) and no significant interaction effect of instructional approach and time (F[2.696,Inf] = 1.055, p = .315) (see Figure 3.1C).
Performance During the Learning Phase
Figure 3.1. This figure shows line graphs with standard error of the reaction time and accuracy for each block of the L-SS binding task. The red color is used for the explicit condition and the blue color for the implicit condition. Significance is shown, with ‘*’ is p < 0.05, ‘**’ is p < 0.01 and ‘***’ is p < 0.001.
Accuracy. A univariate RM ANOVA including only accuracy as dependent variable showed a significant effect for instructional approach (F[1,178.907] = .012, p = .013). No significant effect was found for time (F[2.963,Inf] = .659, p = .651) or for the interaction effect of instructional approach and time (F[2. 963,Inf] = 2.158, p = .092).
Pairwise T-tests with FDR correction were executed between the two groups for accuracy. It showed significant differences for bin 2 (p = .022), 3 (p = .005) and 4 (p = .012). No significant difference was found for bin 1 (p = .430) (see Figure 3.1F).
In sum, these results show that reaction time during the learning phase is overall not affected by instructional approach. Only during the first few trials of the experiment the reaction time of both conditions was significantly higher. Accuracy, however, is affected by the instructional approach. For block 1 and 2 an interaction effect was found for instructional approach and time. During block 1 and 2 for the explicit condition the first two bins are significantly higher than the last two bins, but no significant differences were found within the implicit conditions. The accuracy for the explicit condition is significantly higher for the last two bins, compared to the implicit condition. During block 3 children in both conditions did not significantly improve, however the differences between the conditions become apparent in bin 2, 3 and 4.
Learning Outcome
A Factorial MANOVA was executed to investigate the effects of the instructional approach on learning outcome directly after the learning phase and after one day of retention. For each participant the letter knowledge was measured by means of reaction time and accuracy during the fourth block, and the score of the PSK task. Word reading rate was assessed using the artificial OMT with Dutch words and with pseudowords. All means and standard deviations of all dependent variables are shown in Table 3.3 and violin plots with boxplots of all dependent variables shown in Figure 3.2.
The Factorial MANOVA showed a significant effect of instructional approach (F[2,398] = 44.431, p < .001) and time (F[1,198] = 3.629, p = .004). No interaction effect of instructional approach and time (F[1,198] = .740, p = .595) was found.
Reaction time. The first univariate RM ANOVA including only reaction time showed no significant effects for instructional approach (F[1,112.470] = .014, p = .907), time
(F[1,Inf] = .495, p = .489) and no significant interaction effect of instructional approach and time (F[1,Inf] = .088, p = .774) (see Figure 3.2A).
Accuracy. A second univariate RM ANOVA including only accuracy showed a significant effect for instructional approach (F[1,111.504] = 15.441, p < .001). No significant effect was found for time (F[1,Inf] = .619, p = .441) or for the interaction effect of
instructional approach and time (F[1,Inf] = .037, p = .848).
Pairwise T-tests with FDR correction were executed between the two groups for accuracy. It showed significant differences between the explicit and the implicit group for Test moment 1 and 2 (p’s < .001) (see Figure 3.2B).
Table 3.3
Mean and Standard Deviations (SD) for reaction time, accuracy, score of the PSK task, score of the OMT with Dutch words and the score of the OMT with pseudowords for both testing moments: 1) directly after the learning phase and 2) after one day of retention
Condition Explicit Implicit
Mean SD Mean SD
Reaction time
1 0.872 Sec 0.218 Sec 0.871 Sec 0.284 Sec
2 0.888 Sec 0.254 Sec 0.878 Sec 0.314 Sec
Accuracy
1 79.59 % 13.82 % 66.63 % 17.28 %
2 78.70 % 17.46 % 66.09 % 19.50 %
PSK Task
1 6.058 letters 2.146 letters 3.922 letters 3.019 letters 2 6.250 letters 2.076 letters 4.333 letters 3.011 letters OMT-Dutch
1 5.481 words 4.705 words 3.353 words 4.534 words 2 8.077 words 5.295 words 4.725 words 5.441 words OMT-pseudo
1 4.692 words 4.377 words 3.059 words 4.532 words 2 6.788 words 5.403 words 4.100 words 5.434 words
PSK-Task. A third univariate RM ANOVA including only the PSK-task showed a significant effect for instructional approach (F[1,95.464] = 16.114, p < .001) and time (F[1,Inf] = 10.009, p = .002). No significant effect was found for the interaction effect of instructional approach and time (F[1,Inf] = 1.321, p = .253).
Pairwise T-tests with FDR correction were executed between the two groups for PSK. It showed significant differences between the explicit and the implicit group for Test moment 1 and 2 (p’s < .001).
Pairwise T-tests with FDR correction were executed between the two Test moments for PSK. It showed no significant differences for the explicit group (p = .640) and the implicit group (p = .490) (see Figure 3.2C).
OMT-Dutch. A fourth univariate RM ANOVA including only OMT-Dutch showed significant effects for instructional approach (F[1,178.802] = 8.467, p = .004), time
(F[2.696,Inf] = 44.924, p < .001) and the interaction effect of instructional approach and time (F[2.696,Inf] = 4.270, p = .044).
Pairwise T-tests with FDR correction were executed between the two groups for OMT-Dutch. It showed significant differences between the explicit and the implicit group for Test moment 1 (p = .021) and 2 (p = .002).
Pairwise T-tests with FDR correction were executed between the two Test moments for OMT-Dutch. It showed significant differences for the explicit group (p = .010), but not for the implicit group (p = .170) (see Figure 3.2D).
Learning Outcome
Figure 3.2. This figure shows violin plots with a boxplot of the reaction time, accuracy, score of the PSK task, score of the artificial OMT with Dutch words and the score of the artificial OMT with pseudowords for both testing moments: 1) directly after the learning phase and 2) after one day of retention. The red color is used for the explicit condition and the blue color for the implicit condition. Significance is shown, with ‘*’ is p<0.05, ‘**’ is p<0.01 and ‘***’ is p<0.001.
OMT-Pseudo. The last univariate RM ANOVA including only the OMT-Pseudo showed a significant effect for instructional approach (F[1,119.523] = 5.429, p = .022) and time (F[1,Inf] = 25.914, p < .001). No significant interaction effect of instructional approach and time was found (F[1,Inf] = 2.945, p = .090).
Pairwise T-tests with FDR correction were executed between the two groups for OMT-Pseudo. It showed no significant difference between the explicit and the implicit group for Test moment 1 (p = .066), but there was a significant difference for Test moment 2 (p = .013).
Pairwise T-tests with FDR correction were executed between the two Test moments for OMT-Pseudo. It showed significant differences for the explicit group (p = .032), but not for the implicit group (p = .300) (see Figure 3.2E).
Firstly, the results of directly after the learning phase will be discussed. It is shown that reaction time was not influenced by instructional approach, but accuracy was
significantly higher for the explicit condition than the implicit condition. The score of the PSK task showed a similar significant difference, as well as the score of the OMT with Dutch words. The OMT with pseudowords, however, did not show any significant difference
directly after the learning phase.
The results after one day of retention showed the same significant differences as directly after the learning phase, except the score of the OMT with pseudowords in the explicit condition was significantly higher than in the implicit condition. For the explicit condition both scores of the OMT with Dutch words and with pseudowords were higher after one day of retention, compared to directly after the learning phase. Lastly, an interaction effect of instructional approach and time was found for both OMT with Dutch words and with pseudowords.
Discussion
The current study focused on the effects of instructional approach on initial letter-speech sound binding in children, aiming to give insight into the cognitive processes involved in the development and integration of L-SS associations. Our results showed that children that received goal-directed instructions (i.e., explicit) performed better on almost all tasks on most time points, compared to those that received implicit instructions. No significant differences were found for reaction time between the two conditions. The only interaction effects were found for accuracy in block 1 and 2, and the OMT with Dutch words. The results show that L-SS binding is more than just mapping letters onto speech sound using associative processes. Explicit instructions influence attentional processes that result in a more accurate L-SS binding and more importantly, a better ability to use this knowledge in an artificial reading task. It was concluded that goal-directed instructions result in better L-SS binding and a better automatization of L-SS correspondences in children, compared to implicit instructions.
The first question of the study focused on the effect of instructional approach on L-SS binding during the learning phase, by measuring reaction time and accuracy. Aravena et al. (2013) showed that first accuracy would improve and later on there would be a shift to reaction time. In our study, the reaction time was not affected during the learning phase, which suggests that more time is needed before this shift takes place. However, bin 1 of block 1 did differ
compared to all three bins for the implicit condition and only to bin 3 for the explicit condition. During the first trial the participants of both conditions were potentially less familiar with the task and consequently responded slower. The participants that received implicit instructions responded slower due to limited instructions. These differences disappeared later in the experiment, which strengthen this explanation. The results are not completely in line with the expectations, but the results are easily explained by previous research. In line with Aravena et al. (2013) accuracy did improve during the learning phase. When the symbols were presented for the first time the participants had to guess which of the two symbols was correct, which led to an accuracy of around 50% for both instructions. Afterwards only the accuracy of the participants that received goal-directed instructions increased over time. This was found for both the first and the second set of symbols. When all eight symbols were presented the first few trials did not seem to differ between the conditions, but differences between the two conditions became more apparent towards the end of the third block. However, accuracy did not increase over time when all eight symbols were presented, suggesting a ceiling effect of L-SS binding. In sum, the results show that explicit instructions influence goal-directed behavior and lead to a better L-SS binding, compared to implicit instructions.
The second question focused on the effect of instructional approach on L-SS binding and automatization directly after the learning phase, by measuring letter knowledge and word reading rate. Letter knowledge was higher for the explicit condition than the implicit condition, however reaction time did not differ between the conditions as was also seen in the previous question. These results are in line with the expectations of this research and previous research of Aravena et al. (2013). The word reading rate was only significantly higher in the explicit condition for Dutch words, but not for the pseudowords. Even though the difference between the conditions was expected, the difference between pseudowords and Dutch words was not. Interestingly, the same letter knowledge results in a difference in word reading rate for pseudowords and Dutch words. This would suggest that there is a (partly) different process underlying the reading of pseudowords and Dutch words. An important difference between them is that pseudowords do not have any meaning for the participant, contrary to the Dutch words which carry more meaning; like a sound, a feeling or even a picture (Purves et al., 2013). This would cause a different level of processing between the Dutch and pseudowords which would influence the level of integration. This is also in line with the word superiority effect (Reicher, 1969; Wheeler, 1970). The automatization of L-SS binding, which is important for reading words, is likely to be influenced by the difference in the level of integration of the words. The results concerning the L-SS binding are in line with the previous question. It is also shown that a deeper level of integration results in more automatization, but only when goal-directed behavior is present.
The last question of this research focused on the effect of instructional approach on L-SS binding and automatization after one day of retention, by measuring letter knowledge and word reading rate. For letter knowledge the same results were found as directly after the learning phase. This shows that sleep did not lead to a significant increase in knowledge, however it could be possible that sleep influenced the consolidation of the previously learned knowledge. The consolidation of L-SS mapping might be better due to goal-directed instructions, which would be reflected by an increase in the word reading rate. The word reading rate after one day of retention for the explicit condition was not only higher than for the implicit
condition, but also higher than directly after the learning phase. This goes for both the Dutch words and the pseudowords. This increase cannot be attributed to an increase in letter knowledge, since that is absent. However, it can be explained by consolidation that took place during the retention time. It is known that consolidation takes place during sleep and that information with higher salience is consolidated better (Purves et al, 2013). This would also mean that information is stored more securely inside the brain and thus more easily to remember and apply, which is exactly what is needed for automatization. Since the explicit condition showed goal-directed behavior toward the L-SS associations it became more salient, which led to the possibility to remember and apply it more easily. This caused an increase in L-SS automatization. The lack of goal-directed behavior in the implicit condition would explain the lack of improvement for this condition. Thus, after one day of retention the L-SS automatization increased due to goal-directed behavior.
This study has shown that explicit instructions can have a positive effect on L-SS binding and automatization, by influencing goal-directed behavior. On a theoretical level this means that explicit instructions have clear positive effects on the acquisition of knowledge already after little exposure, which are still present after one day. On a practical level this means that explicit instructions are essential in the learning environment and should not be replaced by game-based interventions. For future research it would be interesting to study what would happen after more exposure, or a longer retention time. It would also be very interesting to investigate if the same differences arise in a sample of dyslectic children and how this would compare to non-dyslectic children. For more insight in attentional processes and its influence on initial L-SS binding EEG could be used. A better understanding of how attentional processes influence L-SS mapping might contribute to the development of more individualized remediation programs for children with reading disabilities.
In summary, our findings contribute to the understanding of L-SS binding and fluent reading. We were able to assess the initial steps of mapping letters onto speech sounds and the automatization of these correspondences with the use of an improved experimental design with an artificial orthography. Our results show that goal-directed behavior results in better learning and a better consolidation overnight, which cannot be attributed to differences in prior exposure to the information. Importantly, we found that goal-directed instructions lead to a higher reading rate compared to implicit instructions. This study shows that reading is more than merely mapping letters onto speech sounds. We hope that our design and results will inspire future research to investigate the nature of mapping letters onto speech sound and how this relates to the development of fluent reading.
References
Aravena, S., & Tijms, J. (2009). Reading fluency and dyslexia: Innovative
developments in the role of associative learning and repetitive exposure in skill acquisition. Educational psychology: Cognition and learning, individual differences, and motivation, 113–141. Hauppauge, NY: Nova Science.
Aravena, S., Snellings, P., Tijms, J., & van der Molen, M. W. (2013). A lab-controlled simulation of a letter-speech sound binding deficit in dyslexia. Journal of Experimental Child Psychology, 115(4), 691–707. https://doi.org/10.1016/j.jecp.2013.03.009
Blomert, L. (2011). The neural signature of orthographic–phonological binding in successful and failing reading development. NeuroImage, 57, 695-703.
Brem, S., Bach, S., Kucian, K., Kujala, J. V., Guttorm, T. K., Martin, E., Lyytinen, H., Brandeis, D. & Richardson, U. (2010). Brain sensitivity to print emerges when children learn letter speech sound correspondences. Proc Natl Acad Sci USA, 107, 7939–7944.
Brus, B., & Voeten, M. J. M. (1973). Eén-Minuut-Test [one-minute test]. Nijmegen, The Netherlands: Berkhout
Chein, J. M., & Schneider, W. (2005). Neuroimaging studies of practice-related change: fMRI and meta-analytic evidence of a domain-general control network for learning. Cognitive Brain Research, 25, 607–623.
Dehaene, S. & Cohen, L. (2007). Cultural recycling of cortical maps. Neuron, 56, 384–398. doi:10.1016/j.neuron.2007.10.004
Ehri, L. C. (2005). Development of sight word reading: Phases and findings. In M. J. Snowling & C. Hulme (Eds.), The science of reading: A handbook (pp. 135–154). Oxford: Blackwell Publishing.
Erhel, S. & Jamet, E. (2016). The effects of goal-oriented instructions in digital game-based learning. Interactive Learning Environments, 24(8), 1744-1757.
DOI:10.1080/10494820.2015.1041409
Fletcher, J. M. & Lyon, G. R. (2008). Dyslexia: why precise definitions are important and how we have achieved them. Perspectives on Language and Literacy, 27-31.
Froyen, D. J. W., Bonte, M. L., Van Atteveldt, N. & Blomert, L. (2009). The long road to automation: Neurocognitive development of letter-speech sound processing. Journal of Cognitive Neuroscience, 21(3), 567–580. https://doi.org/10.1162/jocn.2009.21061
Kirkham, A. J., Breeze, J. M. & Marí-Beffa, P. (2011). The impact of verbal instructions on goal-directed behavior. Acta Psychologica, 139, 212-219.
Lovio, R., Halttunen, A., Lyytinen, H., Näätänen, R. & Kujala, T. (2012). Reading skill and neural processing accuracy improvement after a 3-hour intervention in preschoolers with difficulties in reading-related skills. Brain Research, 1448, 42–55.
McCandliss, B. D. (2010). Educational neuroscience: The early years. Proceedings of the National Academy of Sciences of the United States of America, 107, 8049–8050.
Peirce, J. W., Gray, J. R., Simpson, S., MacAskill, M. R., Höchenberger, R., Sogo, H., Kastman, E. & Lindeløv, J. (2019). PsychoPy2: experiments in behavior made easy. Behavior Research Methods, 51, 195-203. doi:10.3758/s13428-018-01193-y
Purves, D., Cabeza, R., Huettel, S. A., LaBar, K. S., Platt, M. L. & Woldroff, M. G. (2013). Principles of Cognitive Neuroscience (2th ed.). Sunderland: Sinauer Associates, Inc.
Reicher G. M. (1969). Perceptual recognition as a function of meaningfulness of stimulus material. Journal of Experimental Psychology, 81, 275–280.
Siegel, L. S. & Faux, D. (1989). Acquisition of certain grapheme–phoneme correspondences in normally achieving and disabled readers. Reading and Writing, an Interdisciplinary Journal, 1, 37–52.
Siegler, R. S. (2005). Children’s learning. American Psychologist, 60, 769–778. Singleton, C. (2009). Intervention for dyslexia: A review of published evidence on the impact of specialist dyslexia teaching. Paper commissioned by the Steering Committee for the ‘‘No to Failure’’ project and funded by the Department for Children, Schools, and Families, United Kingdom. Retrieved from
http://www.bdadyslexia.org.uk/files/Singleton%20Report.pdf.
Snowling, M. J. (1980). The development of grapheme–phoneme correspondence in normal and dyslexic readers. Journal of Experimental Child Psychology, 29, 294–305.
Vidal, C., Content, A. & Chetail, F. (2017). BACS: The Brussels Artificial Character Sets for studies in cognitive psychology and neuroscience. Behavior Research Methods, 49(6), 2093–2112. https://doi.org/10.3758/s13428-016-0844-8
Wentink, H. & Verhoeven, L. (2003). Protocol leesproblemen en dyslexie. Nijmegen: Expertisecentrum Nederlands.
